Adam, VA7OJ/AB4OJ's IC-756Pro/Pro II User Review

In July 2000, I sold my IC-756, and bought a 756Pro from a local dealer. I amdelighted with the Pro. I find the Pro a big improvement over its predecessor, the 756.The Pro receiver seems much quieter than that of the 756 - probably due to a cleaner DDSLO implementation.. The DSP IF filters have much steeper skirts than the analogue crystalfilters in the older rig, and are much more effective against adjacent-channel QRM than analogue filters. George, W5YR's IF Filter Page dramatically illustrates this point. Also, read George's "Notes on roofing filters" (below).

The manual pre-AGC IF notch filter (70 dB deep) isdynamite. It makes an S9+20 undesired tone disappear off the S-meter. The DSP-IFfiltering, including a tuneable notch filter, is all inside the AGC loop (unlike theIC-756). The combination of the DSP-NR and noise blanker renders night-time 40m listeningmuch more pleasant and less fatiguing. I observe significant artifacts under strong-signalconditions only when the noise blanker is enabled. These are clearly due to the NB gatingon signal peaks, and are eliminated by switching the NB out.

Overall, the Pro pulls the "weak ones" out of the noise noticeably betterthan the 756 (or any of its other predecessors in my shack) did. The measured sensitivityon 20m with Pre-amp 1 on, and 500 Hz bandwidth, is 0.1 µV for 10 dB S+N/N (using an HP8640B generator). I am able to copy easily SSB signals which do not move the S-meter.Those signals would have been barely intelligible on the 756. The ability to tailor thefilter passband to the received signal (using the Twin PBT or the filter tables) also provides a superb tool for pulling out the "weak ones". The manual notch is alsohelpful in improving the SNR of the received signal.

With the Pro, you can optimise the IF bandwidth by tailoring it to the occupiedbandwidth of the received signal, thus yielding optimum S/N ratio. Also, the DSP IFfilters are inside the AGC loop, so strong signals outside the DSP filter bandwidth willnot swamp the receiver. The fact that all the DSP IF filters, including the Manual Notch(but excluding the Auto-Notch) are inside the AGC loop sets the Pro (and the late,lamented Kachina 505) apart from all other amateur HF transceivers on the market.

The vertical sensitivity of the 756Pro Spectrum Scope is significantly higher than thatof the 756. A signal of less than 1 µV is visible, whilst the 756 requires at least 20 µVto produce a spike. The only alignment procedure for the IC-756Pro spectrum scope isvertical (amplitude) alignment and calibration. The horizontal (frequency) display is inthe digital domain, and thus never goes out of cal. The CAL control on the lower rightside of the chassis will center the marker correctly. Incidentally, you can observe thespectral content and occupied bandwidth of your transmitted signal by setting "Scopeduring Tx = ON" in the "Scope Set" menu.

The transmit speech amplifier of the Pro has a little less gain than that of the 756,requiring a slightly higher MIC GAIN setting when using a Heil microphone with a dynamicinsert (HC-4 or HC-5). The new Heil HM-i electret will fully drive the Pro with Mic Gainat around 9 o'clock.

The infinitely-variable DSP IF filters have far better shape factors than classicalanalogue filters. The tuneable IF notch filter is 70 dB deep. And once you have gotused to the spectrum scope, you will never wish to be without one again. The AGC voltageis also derived from the DSP.

The IC-756Pro and an Icom amplifier - PW-1, IC-4KL or IC-2KL/AT-500 - make an excellent combination; the amplifier tracks the radio. The Pro can also be interfaced to a Yaesu Quadra. My first impressions of the Quadra are documented here.

The IC-756Pro/Pro II Monitor is excellent. A sampleof the 36 kHz analogue transmit IF at the output of the main DAC isdown-converted to baseband in a mixer whose LO is the 36 kHz  ADC/DACclock. The resulting audio is fed to the speaker/headphone output and the accessoryaudio output (ACC1 Pin 5).  At the sampling point, the IF parameters are those of thetransmitted signal; the next step in the main signal path is the analogue up-conversion and power-amplification chain.

Using the Monitor and a good pair of headphones, you can set up the Pro for the desiredtransmit audio quality with very little trouble. Try switching between NAR, MID and WIDE TX occupied bandwidth. View George,W5YR's IC-756Pro Monitor page.

The recommended microphone for the 756Pro is the Heil HM-i. The HM-i is plugged directly intothe front-panel [MIC] socket. 756Pro Settings: [MIC GAIN] at 9 o'clock, Treble +5 dB, Bass -2dB, compression OFF and Tx occupied bandwidth = MID (COMP OFF MID). No auxiliaryequipment is interposed between the microphone and the radio. If compression is used, set COMP ONMID, and adjust [COMP] for 5 to 10 dB compression, no more. This will avoid overdrive.

Notes on the IC-756Pro II:

In May 2002, I purchased an IC-756Pro II at the Dayton Hamvention. Upon returning home, I installed the Pro II in my station and began evaluating it.

One week later, I can report that the verdict is very favourable. I have been using the Pro II, and haveobserved quite an improvement in the receiver performance compared to the Pro.So far, I have noticed superior strong-signal handling, DSP IF filteringand DSP noise reduction (NR). 

The adjustable noise-blanker (NB) threshold is also a big advantage over the fixed NB level in the old Pro. I find that by increasing theNB threshold from 50% (default) to 75%,  I can almost completely eliminatelocal HV power-line noise.

The improvement in strong-signal handling on the IC-756Pro II is dramatic. My nearest hamneighbour is 1 km down the street from me, and has a 3-element quad at a heightof 18m. When he transmits SSB on 20m with approximately 1.2 kW PEP (S9 + 60 dB atmy QTH), the peak valuesof received artifacts are as follows:

Front-end settings for this test are Preamp OFF, ATT off, RF Gain 12 o'clock.Spectrum Scope settings are Span 12.5 kHz, ATT off. For offset > 20 kHz, theartifacts are barely audible, and do not degrade the intelligibility of weak SSB signals (S1 ~S2), despite the increase in scope "grass" level due to this powerfulsignal.

By contrast, my neighbour's transmissions overloaded the IC-756Pro front endso severely as to render the entire 20m band unusable.

The Manual Notch is at least as good as that of the Pro. TheAuto-Notch is more effective in suppressing multiple tones, and thereceived audio seems to me to be totally free of DSP artifacts and "munge". There was occasionally abarely-perceptible trace of such artifacts in the Pro. Note: TheAuto-Notch is not selectable in CW mode, as it would function to notch out thevery CW signal that one is trying to receive. This applies to the IC-756Pro andIC-756Pro II.

The Manual Notch can also be effective in removing the most disturbingcomponent of a complex interfering signal. During a recent sked on 15m with my friend Matt KK5DR, a band-limited noisespectrum about 2 kHz wide popped up on our frequency at about S9. Matt's signal at that time was runningS6, and the interference obliterated it. By engaging MN and adjusting the Manual Notch control, I was able to pull Matt's signal rightback out of that stuff with optimal articulation and 100% copy. I defy any analogue radio to emulate this.

Using the Twin PBT with the 250 Hz CW filter selected, one can crank the CWIF filter bandwidth down to 50 Hz (as in the Pro). The CW Pitch control will not put theCW signal out of the IF filter passband, even at 50 Hz bandwidth.

Selectable filter shape factors: On SSB, the recovered audio sounds a little more "mellow" to my earwith the "SOFT" shape factor selected. On CW, the effect is moresubtle. If the signal is in the middle of the filter bandpass, one will probablynot notice much change. The "SOFT" CW shape factor has noticeably widerskirts than the "SHARP" setting. Try this test: Tune in asingle-tone signal in "SHARP", with 250 Hz BW. Tuneit off (up in frequency) until the signal disappears. Switch to"SOFT". The signal will reappear.

The CW "SOFT" setting is useful in cases of adjacent-channelinterference caused by a nearby signal with severe key-clicks. Thekey-click sidebands are visible on the spectrum scope, but in some cases the"SOFT" shape factor will eliminate them from the recovered audio.

From my perspective, weak-signal handling appears to be superior to that ofthe earlier Pro. Very subjectively, I would say that the Pro II receiver isapproximately "3 dB quieter" than that of the Pro. This is a function of an even quieter DDS LO,more effective DSPnoise-reduction and improved DSP IF filters.

The "grass" (baseline noise) level of the Pro II spectrum scopeis  lower than that of the Pro. I also find the display somewhat sharper and crisper than on the earlier Pro.In addition, the white meter backlight is a bit more legible.

The transmitter seems to perform at least as well as the Pro. I have hadexcellent audio reports with the Heil GoldLine/HC-5.

All in all, I am delighted with the Pro II. I feel that the purchase was wellworth the outlay - especially at the Dayton price! As I "put morehours" on the radio, I plan to add to these notes.

I finally sold the earlier Pro to an old friend in July 2003. It had servedme flawlessly for 3 years.

Notes on the BPF Indicator and SSB Digital-Mode Filters:

The IC-756Pro II also extends the functionality of the BPFIndicator in the top line of the display beyond that of the IC-756Pro. The BPF Indicator now operates inSSB and CW modes (CW only in the IC-756Pro). Whenthe IF bandwidth is set to 500 Hz or less via the FILTER menu, the BPF Indicator lights,and the CW shape factor is selected. When the IF bandwidth is greater than 500Hz, the BPF Indicator is dark, and the SSB shape factor is selected. Note:This feature is also implemented in the IC-7800,  IC-746Pro, IC-7400 and IC-7000, although the displays are totally different in the latter 3 radios.

• The BPF Indicator in the IC-746Pro (the IC-7000 BPF display is similar)

If the IF bandwidth is set to, say, 600 Hz, then reduced to 500 Hz via theTwin PBT controls, the BPF Indicator remains dark, and the SSB shape factor isselected. Try this test: Tune in asingle-tone signal in USB-D mode, with 500 Hz BW (BPF lit). Tuneit off (up in frequency) until the signal disappears. Set BW = 600 Hz via theFILTER menu, then set BW = 500 Hz, SFT = 0 Hz via Twin PBT (BPF dark). The signal willreappear. Press DEF softkey to restore default 500 Hz filter (BPF lit). Thesignal will disappear again.

These two alternate methods for setting the IF bandwidth offer the operatora choice of filter shape factors for CW and the digital modes. The additional"SHARP" and "SOFT" settings discussed earlierfurther enhance this choice. With BPF lit in CW mode, selecting"SHARP" further tightens the filter skirts by comparison to"SOFT". In CW mode, with BPF not lit, switching from "SOFT"to "SHARP" does not appear to make any difference. The converse is true forUSB-D/LSB-D (on the Pro II); switching shape factors is effective for non-BPF,but not for BPF.

Typical 500 Hz CW filter passbands
BPF		Non-BPF

In USB-D and LSB-D modes, the IC-756 Pro II FILTER menu now presents the threeCW filter choices, 1.2 kHz, 500 Hz and 250 Hz. The 500 Hz and 250 Hz filtersettings default to the CW shape factor (BPF Indicator lit), although the SSBshape factor can be selected as described above. This improvement in the SSBdigital-mode filter selections renders the IC-756Pro II ideal for PSK31and other narrowband digital modes.

In February 2003, I tried outvarious 100 Hz CW filter settings on my IC-756Pro II. The following table shows the results:

The optimum filter settings to use in any given case are a function of bandconditions and personal taste. 

Note on the Audio Image, and Filter Bandwidth Settings, in the 756Pro/Pro2: 

Recent discussion of the received audio image in conjunction with the CWoffset provided an incentive for a further investigation: What happens if theDSP-IF filter bandwidth is set equal to, or greater than, twice the differencebetween virtual carrier and passband centre? The investigation, and itsresults, are detailed here.

Note on Dual Watch:

Q. With Dual Watch, can one send the DX station to one ear only,and the hounds chasing the DX to the other ear, and vary the volume of the twosignals independently of each other?

A. In a word, not quite. As the Pro2 uses a common IF chain (includingthe DSP), and a common receive audio chain, for both Dual Watch channels,separate audio in L & R headphones is not possible. To provide thiscapability, the radio would need two independent IF chains, each with its ownADC, DSP and DAC. This would have a major impact on the price of the radio.Icom's product-management people probably felt that the demand would not justifythe increased cost.

Dual Watch permits reception of two signals in the same mode, IF bandwidthand band (the B channel can be in an adjacent band e.g. A on 20m and B on 17mwithout too much loss of sensitivity). When combined with Split, Dual Watchenables the operator to maintain a "listening watch" on the distantstation's listening (receive) frequency. This is very useful when working a DXstation which is operating split. The operator hears both signals mixed at thecommon receiver audio output. The relative A and B signal levels are adjustedvia the BAL control. 

Additional notes on Dual Watch are presented here.

Note on headphone level:

In the IC-756Pro, there are 330Ω resistors in series with the tip & ring leadsof the front-panel [PHONES] jack (one resistor in each lead). The corresponding values forthe IC-756 and IC-756Pro II/III are 100Ω and 180Ω, respectively. As a consequence, the headphone level will be approximately 10 dBlower on the Pro (or 5 dB on the Pro II/III) than on the 756 (with 8Ω phones).

You can increase headphone volume by using higher-impedance headphones, by connecting a2:1 or 4:1 (turns ratio) step-down audio matching transformer between the phone plug and theheadphones or (least desirable of all) by disassembling the radio and changing out theresistors (R1 & R2 on the PHONE board).

Note on transverter operation:

There appears to be a misprint on pages 12 and 16 of the IC-756Pro user manual. Themanual shows the left-most RCA socket on the rear panel (1) as RX ANT, and the socket toits right (2) as XVERTER. On the radio, the reverse is true; (1) is marked X-VERTER, and(2) as RX ANT. This error was finally corrected in the IC-756Pro II user manual, 2nd edition, A-6072H-1EX-(2), and in the IC-756Pro III user manual.

Setting transmitter output to 5W for QRP on the IC-756Pro andIC-756Pro II:

(Contributed byGeorge, W5YR)

Press and hold the METER button for 2 sec. The digital bar-graph meter scales will appear in the space normally occupied by the Spectrum Scope. RF power output is shown on the top bar. When the RF POWER control is set such that the first two bars are lighted, the RF output is 5W.There is a slight hysteresis in the bar graph; it increases from 1 to 2 bars at 5W, and decreases from 3 to 2 bars at 5.5 to 6W.

Note on CW netting (spotting):

The netting procedure on the IC-756Pro and Pro II is very straightforward.Here are five methods:

1: In BK-IN FULL or SEMI, depress the Morse key, and note the sidetone pitch. Then unkey, and tune the received signal to the same pitch.

2 (from John Rippey, W3ULS): Select100 Hz CW filter. Tune desired station to centre of filter passband. DepressMorse key, and note sidetone pitch. Then unkey, and tune the received signal to the same pitch.The 100 Hz filter eliminates the signals of many of theother stations which are also calling the desired station.

3 (Most accurate): Select BK-IN OFF, depress the Morse key, and tune the main VFO knob to zero-beatthe sidetone to the received signal. Then select BK-IN FULL or SEMI as desired.

The receive and transmit frequencies will now be exactlyequal.

4 (Visual method): In SCOPE/SET,activate "Scope on during Tx". Select SPAN = 12.5 kHz. Tune thereceived signal so that the peak of the spike is on the scope centre line.

5 (from Jim Shaw, WA6PX): Pressand hold the CW/RTTY button for 1 sec. to switch to CW-R mode. Verify thatthe pitch is the same as for CW mode. Tune the main VFO knob as needed, toequalise the pitch. Then switch back to CW mode.

The receive frequency will now be within a few Hzof the transmit frequency.

Note 1: On the IC-756Pro/Pro II, the transmit CW sidetone tracks the CW PITCH control. However, the transmit offset is fixed at 600 Hz, and does not change as the Pitch control is varied. (IC-756Pro2 user manual, p. 29).

Note 2: The default CW Pitch setting (12 o'clock) is 600 Hz.

Note on CW element truncation in the 756Pro, 756Pro II and746Pro: 

(Based on recent comments byGeorge, W5YR)

All the radios in the Icom Pro family suffer from this deficiency. As a functionof the radio's overall receive/transmit transition time, the leading edge of thefirst code element following Rx/Tx switchover is truncated (clipped) byapproximately 10 mS.

This truncation affects the initial element in semi-break-in operation. In fullbreak-in (QSK), every elementis clipped. The effect is most disturbing at sending speeds significantly above20 wpm. The problem is negligible to minor at speeds up to around 20 wpm. Above thatspeed,  the dits begin to sound a little light; by 30-40 wpm, they are clearlyfar out of balance with the spaces.

A definitive "fix" calls for the CW logic to continue to send each code element for about 10mS after the key is opened. That is what an external keyer such as the Logikey does.Several CW operators have successfully used the Logikey in conjunction with theIC-756Pro II as a work-around for this issue.

Update (May 2005): This defect has been corrected in the IC-756Pro3.

Notes on the low-frequency rumble in the 756PRO:

A small group of IC-756Pro owners, including myself, have confirmed that under certainspecial conditions a low-frequency rumble appears on theexternal speaker output. This rumble is at a very low level, and is completely masked atnormal listening levels. 

The apparent cause of this rumble is a low-frequency ripple which the NB superimposeson its own +8V supply rail. This +8V rail also powers low-level audio stages following theDAC. There is a suspicion that the ripple is modulating these stages.

A "fix" for the rumble has just been published. Read "The Icom756PRO - A Cure for the Rumble", by Tony Brock-Fisher K1KP and Jim JarvisN2EA. This article appears in "Technical Correspondence", QST,June 2002, page 68. Read a condensedversion of this article.

It is entirely possible that other artifacts observed by some IC-756Pro users are dueto inadequate opposite-sideband suppression at the transmitting station. With someoperators "tweaking" the carrier set point in the quest for "hi-fi"audio, there's a lot of that going around these days - not to mention degraded carriersuppression. Fortunately for the rest of us on the bands, the IC-756Pro has no carrier setpoint adjustment, as the DSP performs all modulation tasks. (By the way, the 756Pro alsogenerates mathematically near-perfect AM, FM and CW signals, as well as"textbook" SSB. If you have access to a spectrum analyzer, you can check thatout.)  Read "Noteson Carrier Set Point and the IC-756Pro" (below).

Note on residual speech-amplifier noise, March 2003:

There has been some recent reflector discussion on this topic. Onecorrespondent wrote: "There is a small amount of white noise with the mic disconnected and the mic gainwide open, the compression on, and the compression wide open. Connecting the mic shows that the mic pickup from the background is considerably more noise than produced internally in the radio. The internal noise in the rigis insignificant unless one desires to whisper into the mic and have full output power."

To verify this, I ran a test on my IC-756Pro II. Transmitting at 28.5 MHz USB, with a 10W 25 ~ 60 MHz Bird element and the mic unplugged, I noticed no deflection on the power meter with Po set to 100W, for any Mic Gain and settings with Comp on or off. (TX BW = MID for all tests).However, I did observe a 5 dB spike on the Spectrum Scope with � 12.5 kHz span. 

I then ran the same test at 14.1 MHz USB using an HP 853A/8558B spectrum analyzer. With100W = 0 dBc reference on the analyzer screen, I observed the following: (TX BW = MID)

1. Mic Gain at 0%, Comp at 50%, Comp off: -75 dBc spike at 14.1 MHz

2. Mic Gain at 0%, Comp at 50%, Comp on: -65 dBc spike at 14.1 MHz

3. Mic Gain at 50%, Comp at 50%, Comp off: -65 dBc spike at 14.1 MHz

4. Mic Gain at 50%, Comp at 50%, Comp on: -55 dBc spike at 14.1 MHz

The -6dB occupied bandwidth of the observed spike was approx. 3 kHz.

The spike appeared to contain low-frequency products (possibly AC mains harmonics.) It did not look like phase noise. I did not perform a full-blown phase-noise test, but the transmitted composite noise at 100W output in RTTY mode appeared to be below the threshold of my test lash-up. 

The spike described above appears to be due to "garbage" picked up by the unterminated speech-amplifier input modulating the transmitter. In any event, the observed power level is considerably less thanthe 2W observed by an earlier correspondent. For the worst-case measurement (4), -55 dBc = 55 dB below 100W = -5 dBm = 0.3mW. 

With Mic Gain at 0%, I could not hear any rise in noise floor on a nearby Sony ICF-SW7600G portable receiver with Comp off or on. Increasing Mic Gain to 50% with Comp off raised the noise floor on the portable slightly, and a little more with Comp on. The portable was 50 cm away from the dummy load.

It is safe to conclude that with the speech-amplifier input correctlyterminated by a low-impedance microphone, the ambient noise input will beconsiderably in excess of the residual noise.

Notes on the Digital Voice Recorder (DVR) feature, May 2003:

Q: Is it possible to configure the DVR toretransmit audio clips previously recorded over the air?

A: In a word, No!According to information received from Icom Japan, no variant ofthe IC-756Pro or Pro II offers this capability. There is no "secretmod" either. The DVR can only retransmit audio previously recorded via themicrophone.

Icom Inc. in Osaka has officially stated that over-the-air playback oftraffic recorded off-air on the IC-756Pro/Pro II voice-recorder function is notpossible in any version of the radio.

Note on AM operation with the IC-756Pro or Pro II:

The DSP modulation process in the IC-756Pro or Pro II generates a mathematically near-perfectdouble-sideband AM signal. The Spectrum Scope greatly simplifies AM setup. (Beforestarting, select "Scope ON during Tx" in the Spectrum Scope menu.)

To set the Pro or Pro II correctly for AM, key the transmitter and set [RFPOWER] for 25W resting  (unmodulated) carrier output (seebox below). Then whistle into the microphone, andadjust [MIC GAIN] so that the peak sideband amplitude is exactly 6 dB below thecarrier amplitude, as displayed on the scope. Thetransmitter is now correctly set up for AM operation.

The transmitted occupied bandwidth (TOBW) in AM mode is approx. 5.6 kHz. Thefollowing receive IF filter bandwidths are available: 9, 6 and 3 kHz. The innerTwin PBT knob functions as an IF Shift control. The Compressor [COMP] isdisabled in AM mode.

Note on FM operation with the IC-756Pro or Pro II:

As with AM, the DSP modulation process in the IC-756Pro or Pro II generates a mathematically near-perfectFM signal. The Spectrum Scope greatly simplifies FM deviation setting. (Beforestarting, select "Scope ON during Tx" in the Spectrum Scope menu.)

Note: As the compression/TOBW softkey function (COMP-ON-WIDE etc.) is disabled in FM mode, the FILTER menu also selects the peak deviation when the FM mode is selected.

The DSP IDC function limits FM peak deviation to �5 kHz when the 15 kHz FM filter is selected via the FILTER menu. To set the Pro or Pro II correctly for FM (16K0F3E) on 6m or on 10m above 29.0 MHz, select the 15 kHz FM filter in the FILTER menu, then key the transmitter and whistle into themicrophone. Set [MIC GAIN] to the point where deviation just limits at TOBW = 16kHz at -26 dBc, as displayed on the scope. (Thesymmetrical sideband pairs are clearly visible on the display.) Thetransmitter is now correctly set up for FM operation.

Similarly, FM peak deviation is limited to �2.5 kHz when the 10 or 7 kHz FM filter is selected. To operate NBFM (11K0F3E) on the HF bands below 29.0 MHz,  select the 10 kHz FM IF filter, then set [MIC GAIN]for TOBW = 11 kHz at -26 dBc using the above procedure. This will comply withFCC Part 97.307(f)(1): No angle-modulated emission may have a modulation index greater than 1 at the highest modulation frequency.  The 10 kHz IF filter will also yield an improved S/N ratio on receive, as IF bandwidth now more closely matches TOBW.

I also recommend setting RF/SQL in the SET/OTHERS menu to "Auto". This will allow the[RF/SQL] control to function as an RF Gain control in SSB, CW and RTTY modes, and as a squelch in AM and FM modes.

Ref. 24 is an excellent discussion of FM in amateur radio.

Notes on receiver sensitivity and noise figure:

It is very refreshing to see the quotation from the ARRL reviews to the effect that atHF, the external (antenna, man-made and sky noise) is many dB higher than the receiver'ssystem (internal) noise. This is discussed in more detail here.

Recently, I measured the sensitivity of my new IC-756Pro (S/N > 2600) at 28.500 kHz,with 500 Hz BW and Pre-amp 2 engaged, as 0.1 µV (-127 dBm) for 10 dB S+N/N. The comparablefigure for my previous rig, an IC-756, was 0.13 µV (-125 dBm). The signal source was arecently calibrated HP 8640B signal generator.  

An examination of the RF Unit schematic reveals that the Pre-amp 1 and 2 circuits inthe two radios are very similar. In both radios, Pre-amp 1 uses a pair of 2SK2171 JFET'sin parallel; the two circuits are almost identical, except for RF output transformer typeand a few minor differences in component values. Again, in both front-ends, Pre-amp 2 isan NEC uPC1658G MMIC. Judging from differences in connections and external componentvalues, this device is being run in the Pro at a lower Vcc, with lower absolute poweroutput, but slightly higher gain, than for the 756 case. This suggests a lower noisefigure and better IP3 characteristics for the 756Pro case. (Refer to Pages 11-6 and10-8 of the 756 and Pro service manuals, respectively). All of this suggests that thesystem noise figure will be comparable for the two transceivers.

According to the ARRL tests,  the transmitted composite noise at fo+ 2 kHz (fo = 14100 kHz) is -115 dBc/Hz for the IC-756, compared to -125 dBc/Hzfor the 756Pro. This suggests a quieter DDS implementation in the Pro. Assuming acomparable system noise figure for both radios, it should be possible to predict betterMDS (and sensitivity) for the Pro. In this light, the 6 dB difference in measured MDS infavour of the 756 is difficult to comprehend, unless the insertion loss of the RF band-passfilters is 6 dB or so higher in the Pro. It is interesting that measurements performed bysome other reflector members, and by myself, tend to favour the Pro.

It is well known that the noise figure (NF) of the first stage of a receiver is a closeapproximation of the system noise figure, assuming substantial gain in the firstRF stage andnegligible insertion loss ahead of that stage. (Note:To ensure that AGC action in the presence of strong signals does not degradesystem NF, the AGC gain-control point isnot placed ahead of the first RF stage. Each dB of loss inserted ahead of thefirst RF stage increases NF by 1 dB.) So we can say that the pre-amp (actuallythe RF amplifier - I do wish the amateur HF industry would use that term!) sets the systemnoise figure. That said, the external noise will always be at least 10 dB above the systemnoise figure at HF. This can be seen rather graphically on the 756 or Pro, by connectingan antenna and observing the sharp rise in "grass" level on the spectrum scope.

Typically,  MDS = -132 dBm at 14 MHz (preamp off, no attenuation). Thetypical broadband power level delivered by a resonant HF antenna to a receiveris around -120 dBm. The resultant 12dB increase in noise level upon connectingthe antenna amply demonstrates the above point. The operator canexploit this 12dB margin for quieter reception on the lower HF bands.

The "bottom line" is that small differences in NF, MDS or 10 dB (S+N)/Nsensitivity amongst HF receivers will, in most cases, be swamped by the external noise -certainly on HF, and often also on 6 metres. We can all take comfort in that as weanxiously hook our new rigs up, first to a signal generator, then to our antenna systems.The sudden roar of noise, and the band of "grass" on the scope, when the antennais plugged in, should make us feel very good about our receivers!

Note on S-meter calibration:   Refer to the IC-756Pro IIService Manual, page 4-14, for the S-meter calibration procedure. Acorrectly-calibrated S-meter will read S9 at a 50 µV (-73 dBm) RF input level.The RF input level for S9+60 dB is 50 mV (-13 dBm). Setup conditions forS-meter calibration are: Preamp OFF, ATT OFF,  AGC MID, Filter = 2.4 kHz, Frequency = 14151.5 kHz.

Note on AGC, and the AGC gain-control point:

It will be seenfrom the IC-756Pro II front-end block diagram that AGCvoltage is applied to the 64.455 MHz first-IF amplifier (Q721) following the roofingfilter. The AGCgain-control point is thus well downstream of the point which determines thesystem NF (1st mixer or RF preamp). This indicates that Icom's designers haveensured that the linear RF/IF stages upstream of the gain-control point aresufficiently linear (over a wide range of RF signal levels) to be run "wideopen". Operational experience with the IC-756Pro/ProII has confirmed thevalidity of this approach. The IC-7800employs  a similar AGC scheme. With Preamp off, the AGC threshold is -96dBm  (3.5 µV). Recent measurements confirmthis value. See also Reference 14.

In the IC-756Pro series, the AGC derivation point is a DSP process following the IF selectivity and variable stopband (manual notch) processes. This means that as long as the ADC is nowhere near "all-1's" (over-range) level, the AGC will respond only to signals within the passband of the DSP IF selectivity process.

If an IMD product generated upstream happens to fall within the DSP IF bandpass, it will develop AGC voltage (as long as its power level exceeds the AGC threshold.) The roofing filter will reduce the statistical likelihood that this will occur. There is a crossover point in roofing-filter bandwidth where the risk of an in-band IMD product increases, rather than decreasing, with diminishing roofing-filter bandwidth due to IMD in the filter and its associated circuits (driver amplifier, transformers etc.)

The -96 dBm AGC threshold was chosen such that at low RF signal levels,before any AGC is applied, sufficient IF signal power is present at the ADCinput to ensure that the quantisation noise of the ADC does not degrade systemNF. As the RF signal level increases above the AGC threshold, the AGC voltageapplied to Q721 levels the signal power at the ADC input  to hold the ADCwell below the "all ones" point. This is discussed in Reference 12,pp. 4, 5.

The AGC attack time is a trade-off. If it is too long, there is a risk that a fast-rising signal wavefront or spike will over-range the ADC. As the AGC loop is DSP-derived, there will then be no AGC action until the spike disappears. If the signal is of constant level, the entire receiver will remain locked up until the signal is removed, as there is now no AGC action as long as the ADC is driven to or beyond its "all 1's" point.

Conversely, if the AGC attack time is too short, the AGC will clamp on a fast-rising signal wavefront or spike, thus desensing the receiver for the duration of the AGC hang/decay time.

The "RF Gain" control applies an adjustable bias voltage tothe AGC line, thus effectively reducing the gain of the 64.455 MHz first-IFamplifier, and raising the AGC threshold. If the RF signal at the receiver inputis sufficiently strong to develop an AGC voltage higher than the preset bias,the AGC will reduce the IF-amplifier gain still further. As the control does notaffect the gain of the receiver's RF stage (preamp/first mixer) for the reasondiscussed above, the term "RF Gain" is a misnomer  here. 

Here is a procedure for using the bandnoise/noise floor margin to your advantage, so as to obtain quieterreception.

Note on front-end AGC and attenuation, Jan. 2004:

The IC-765 and IC-781 both have an AGC-controlled PIN diode attenuator between theRF BPF bank and the preamp input switch. The IC-756 series, starting with the base 756, does not use the PIN RF attenuator; these radios employ a gain-controlled 1st IF amplifier after the first mixer. The IC-736 and IC-738 have AGC-controlled IF amplifiers.

It should be noted here that PIN diodes can cause high levels of IMDproducts at intermediate attenuation levels; this is especially true at low frequencies. For this reason, newer designs employ a carefully-selected voltage-controlled amplifier (VCA) as the gain-controlled 1st IF amplifier.

Whilst the PIN-diode front-end AGC scheme raises the 3rd-order intercept point (IP3) by the amount of attenuation inserted, it also admittedly degrades the system NF by the same amount. Under single-strong-signal conditions, this NF degradation is not a concern, as there is C/N to spare. Now, what happens if two or more strong signals outside the IF passband produce an IMD product in the passband? If the absolute power level of this product is below the AGC threshold, it will not desense the receiver. As the 20 kHz IMDDR3 is already 100 dB or so, the two undesired signals will need to be pretty strong - around +5 dBm each - to cause the RF attenuator to desense the receiver sufficiently to cause a noticeable S/N degradation to a weak signal in the passband over and above the idle-channel noise (ICN) which the offending signals will generate in any event. Presumably some equilibrium point will be reached where the reduction inthis ICN due to the attenuator will offset system NF degradation caused by the same attenuator. Perhaps this is why some commercial HF-radio manufacturers still use this form of RF AGC.

The ambient noise at a typical HF site is 10 to 12 dB above the receiver noise floor. Thus, adding 6 or 12 dB of attenuation ahead of the first mixer will reduce C/N by the corresponding amount, and bring up the receiver noise floor to the point where it is at or just below ambient "grass" level. (We discount impulse or man-made noise for the moment.) This is still acceptable, as the system NF isnot yet the limiting factor in the system's ability to discern a signal just above "grass" level. This is an acceptable price to pay for a 6 or 12 dB improvement in IP3, and could well be another reason why some designers still favour front-end AGC.

To sum up: If an amount of attenuation is inserted at thereceiver's RF input which reduces the input signal power whilst ensuring thatthe antenna noise is still above the receiver's noise floor, the system willgain at least 10 to 12 dB of "IP3 from Heaven" without compromisingsystem NF. When a receiver is connected to an antenna, the system NF isdetermined by the band (antenna) noise, and not by the NF of the front endterminated in 50Ω. (We assume that the stages following the attenuatorhave sufficient gain to overcome the inserted loss.)

Comment by George, W5YR: Front-endattenuation is an effective means for increasing dynamic range as long as itdoes not compromise the NF of the receiver relative to the band noiseenvironment, and there remains sufficient gain margin downstream to compensate.

We know that you can select a band and observe the noise level in terms ofS-meter reading and audio output. Insert attenuation until disconnection andreconnection of the antenna shows, say, a 3 dB increase on the S-meter of bandnoise over set noise. Now, reduce the RF gain until there is just a perceptiblechange in the S-meter with the antenna connected. Now turn up the audio gainuntil the band noise audio output is the same as when you started.

Result: you have increased IP3 by whatever attenuation you introduced. Youhave increased dynamic range by reducing both signal and spurious inputs in thefront end and in the post-mixer stages. The audio noise output of the receiveris the same as before and its sensitivity has decreased only perhaps a smallamount, but this is arguable depending upon the nature of the band noise amongother factors.

Truly "dBm from Heaven" by thesimple expedient of changing the gain distribution within the receiver.

Why is this very basic concept not more widely known and employed? Do radiomanufacturers think that it is too complex to put in their documentation?

Notes on tube amplifier interfacing:

To protect the keying circuit of the 756Pro, use an auxiliary keying relay,such as the Yaesu FRB-757A, or an small open-frame relay with a 12V coil drawing 100 mA orless. Connect a diode (e.g. 1N4001) across the coil with cathode to +12V. The relaycontacts must be rated to carry the keying voltage and current of the amplifier. Here is a goodexample. 

The ALC line should always be connected.  When using a solid-state amplifier, ALC ismandatory, as it is the amplifier's first line ofdefence against abnormal operation. Setthe ALC to limit the drive power so as to drive the amplifier to no more than its ratedoutput. This will hold IMD3 to an acceptable level - a kindness to one's neighbours on theband. A side benefit is that of keeping the Radio Inspector happy.

Note on transverter drive levels, April 2005:

The rear-panel XVERT RCA jack is bidirectional. In response to a reflector request, I measured the transverter RF drive level (in transmit) on my IC-756Pro II. Here are the data:

Test Setup: IC-756Pro II, S/N 26XX, with XVERT jack connected via coax jumper to R&S URV-Z4 Insertion Unit terminated in 50Ω resistive load. URV-Z4 probe cable connected to R&S URV-4 (Ver. 03) RF Millivoltmeter.(IC-756Pro II User Manual, p. 19).

7-pin DIN plug with Pins 6 (XVRT) and 7 (+12V) bridged via 27KΩ 0.25W resistor, plugged into Pro II ACC (2) socket to activate XVERT mode.Pro II set to 28200.000 kHz, CW mode, semi-break-in.

In receive, the signal applied to the XVERT  jack is routed to the RF BPF input. The transverter IF output power at this point should not exceed -13 dBm.

The ALC RCA jack is active in XVERT mode. A transverter equipped witha negative-going ALC line (0 to -4V DC output) will be compatible with theexciter's ALC jack.

Note on RTTY IF and Twin Peak filters, May 2005:

I recently investigated the RTTY IF filters on my IC-756Pro2, using the internal Cal Marker as a signal source. This has led to a somewhat clearer understanding of  the RTTY filters. Like all IF filters in the Pro series, these are implemented in the DSP.

The RTTY filter set ("Other digital filters", Ref. 15, p. 13) is a pre-AGC DSP IF filter, similar to the CW and SSB filters, but optimised for RTTY. I verified this by tuning across the Cal Marker and noting that the S-meter reading fell off sharply as the carrier moved down the filter skirts. When RTTY FILTER ON is set, pressing the FILTER softkey (F-4) displays the current RTTY FILTER setting, but does not allow choices. To change the RTTY filter setting, press and hold the RTTY FILTER softkey. The RTTY filter selections are discrete; the inner PBT knob functions as an IF shift. SSB SOFT/SHARP shape-factor selection is not available for the RTTY filters.

The Twin Peak Filter (TPF) is a post-AGC, post-product-detection DSP filter (loosely, an "audio filter"). When I tuned the Cal Marker across it, I heard the recovered audio level drop slightly in the trough between the peaks (2125 and 2295 Hz), but the S-meter reading did not drop correspondingly. This confirms that TPF is a post-AGC filter. TPF can be activated only when RTTY FILTER ON is set.

The IF filter set selected in RTTY mode with RTTY FILTER OFF is, in fact, the "regular" SSB IF filter set. I confirmed this by determining that the SSB SOFT/SHARP shape-factor selection works for this case.If RTTY FILTER ON is set, it forces the IF filter selection to the RTTY filter set. The abovementioned SSB settings are no longer effective. One can see this by pressing the FILTER (F-4) softkey; only the current "special" RTTY filter BW value will be displayed.

Comments made one month after I bought the 756Pro:

I have had my Pro a month now, and am really enjoying it. It does a far better job ofpulling out weak SSB signals than the 756 did. I use the MID occupied-bandwidth setting(on the COMP soft-key; hold the key in to change NAR-MID-WIDE) with 2 dB treble boost, and0 dB bass boost. I am using the Heil Gold-Line mike with the HC-5 and wide-range elements. 

You can judge the audio quality via the Monitor function. The 756Pro monitor isexcellent; it decodes the digital IF bit-stream generated by the DSP in transmit mode, anddown-converts the resulting 36 kHz analogue IF to base-band. The monitor is thus a truerepresentation of the transmitted SSB signal.  

There is really no need to use any external audio processing hardware with the Pro. Thethree occupied-bandwidth selections, together with the boost/cut menu, offer 121 differenttransmit audio settings.

I sometimes reconfigure the 3 kHz SSB filter to 2 kHz for enhanced adjacent-channelrejection, in cases where the 1.8 kHz filter is a little too narrow. All IF filtering inthe Pro is done by the DSP. I run the NR at about 12 o'clock, and use Pre-amp 1 most ofthe time (Pre-amp 2 for very weak signals). The combination of NR and Noise Blanker isvery effective against impulse noise; however, strong signals cause the noise blanker toproduce artifacts.  

The Manual Notch is about 70 dB deep, and so narrow that it does not degrade thereceived audio. Manual Notch is also inside the AGC loop (unlike Auto Notch). This allowsyou to remove a heterodyne without swamping the receiver.

I was aware that there were some differences in BDR, IMDDR3 and other"numbers" in favour of the earlier IC-756, but I have a feeling that Icom mayhave gone through a rev. level change since the release of the early unit which the ARRLLab tested. My radio is in the S/N 2600 serial number range, vs. 1300 range for the unittested by the ARRL. In any event, I measured the sensitivity with 500 Hz BW and Pre-amp 2engaged, as 0.1 µV for 10 dB S+N/N. The comparable figure for my IC-756 was 0.13µV.

Operationally, by using the Twin PBT or by adjusting the filter bandwidth to match theoccupied bandwidth of the signal, I am able to copy easily signals which wereunintelligible on the 756. Also, I observe that the vertical sensitivity of the spectrumscope on the Pro is at least 20 dB better than that of the 756. Although 40m is quieter atnight out here in the Pacific Northwest than on the East Coast, I have not observed any ofthe typical overload on that band with the Pro. With the 756, I occasionally heardartifacts from adjacent-channel broadcast stations on 40m.

On the Pro, one can adjust the filter to eliminate "splatter" artifacts. Ialso observe that the NB-NR combination is more effective on the Pro than on the 756 ineliminating repetitive impulse noise.

To my way of thinking, the architecture of the 756Pro represents a true paradigm shift;the DSP chipset does all filtering, modulation, demodulation, noise reduction, and audioprocessing. Even the transmit monitor is a digital loop-back from the DSP modulationprocess, so you are listening to the transmit IF, rather than a sample from the speechamplifier as in the earlier 756. All filtering, including the manual notch (but not theauto-notch) is pre-AGC; this eliminates swamping due to strong signals outside the IFpassband. Also, there are no filters to buy!

Clarification regarding "IF-DSP", July, 2000:

Ah...but there are true IF-DSP's.  Both the Icom IC-756Pro and theKachina down-convert their penultimate IF (455 kHz) to a low final IF (36 kHz in the756Pro and 40 kHz for the Kachina). The low IF is then fed to an ADC (analogue/digitalconverter), followed by the DSP chipset. The DSP performs all filtering, signalenhancement and demodulation tasks, including AGC derivation. A DAC (digital/analogueconverter) following the DSP then converts the DSP output to AF base-band.

Perhaps a definition is in order:   IF-DSP means that the DSP operates atan intermediate frequency (prior to demodulation). The exact frequency involved is not theissue here.   AF-DSP means that the DSP operates at audio frequency(post-demodulation). A typical example of this is the UT-106 DSP module for the IcomIC-706 MkII/G.   The only reason why IF-DSP has not yet reached the 455 kHz (orhigher) IF stage in an amateur HF transceiver is the cost of DSP chipsets at higher speedlevels. I am sure that some military HF equipment already has 455 kHz (if not faster)IF-DSP. As with all aspects of the semiconductor industry, it is only a matter of timebefore the price of such devices comes down to the point where they will show up in our HFrigs.

Another point to be made is that to prevent strong out-of-band signals from swamping areceiver, DSP IF filtering must be within the AGC loop. This is only feasible if theend-to-end group delay across the DSP (ADC input to DAC output) is sufficiently short toprevent instability in the AGC loop. We are already at this point with the Kachina and the756Pro; DSP-derived AGC will only get better as the chipsets get faster.

In the current, popular competing HF transceiver, the AGC is derived by envelopedetection of the analogue IF signal before the ADC, and the DSP is entirely post-AGC. Thisis rather disappointing in a newly-designed radio.  

Comments on Blocking Dynamic Range  (BDR):

Blocking dynamic range (also called cross-modulation dynamic range) isdefined as follows:

Blocking dynamic range (BDR) is the difference, in dB, between the noise floor and anoff-channel signal that causes 1 dB of gain compression in the receiver. It indicates thesignal level, above the noise floor, that begins to cause desensitization.  BDR iscalculated by subtracting the noise floor from the level of undesired signal that producesa 1-dB decrease in a weak desired signal. It is expressed in dB. The greater the dynamicrange, (expressed in dB), the better the receiver performance. It is usual for the dynamicrange to vary with frequency spacing.

Key Test Conditions: If possible, AGC is normally turned off; thereceiver is operated in its linear region. Desired signal set to 10 dB below the 1-dBcompression point, or 20 dB above the noise floor in receivers whose AGC cannot bedisabled. The receiver bandwidth is set as close as possible to 500 Hz.

To summarise, the blocking (or cross-modulation) dynamic range is the level (relativeto the noise floor) of an undesired signal offset 20 (or 50) kHz from the desired signal,at which the desired signal is compressed by 1 dB. Example: The noise floor is -127 dBm(0.1 µV in 50Ω). Two signal generators (desired & undesired signal) areconnected to the receiver input via a combiner. "Desired" is cranked up to thepoint where 1 dB compression is seen, then backed off 10 dB. "Undesired" isoffset 20 kHz, then turned up until 1 dB compression is seen. The "Undesired"level is +10 dBm.  Blocking dynamic range is +10 - (-127) = 137 dB.

Now we consider the receiver issues which cause blocking: The first stage in thereceiver's RF/IF chain to go into overload is the one responsible for the blocking. Theblocking can occur in the RF amplifier (preamp), so the test is conducted with preamp offand on. Now, any stage from the first mixer, through IF amplifiers and downstream mixers,to the demodulator, can go into overload. The overall gain distribution is a significantfactor in determining just which stage will hit its non-linear region and overload first.

The IC-765 had particularly good BDR (> 150 dB with preamp off). I believe that thiswas due to careful balancing of stage gains between first mixer, second mixer and IFchain. The 765 was a conventional analogue radio.

A DSP radio such as the IC-756Pro or Kachina 505 is a totally different animal. Thelimiting factor for dynamic range is the ADC (analogue/digital converter) whichdigitizesthe IF for delivery to the DSP. The dynamic range of the ADC is determined by the numberof bits per sample. One bit (a power of 2) = 2 x amplitude (i.e.2 x voltage*), or 6 dB. A 16-bit DSP has a theoretical dynamic range of 96 dB;a 24-bit ADC, 144 dB. The 756Pro uses a 24-bit ADC and DAC (digital/analogue converter,and a 32-bit DSP. The absolute maximum amplitude which the ADC can encode corresponds to an encoded binary value of "all ones" , i.e. a hex value of  FFFF (16-bit)or FF FF FF (24-bit). Note: Representation ofnegative values can result in maximumpossible positive and negative values, rather than the positive maxima shown here.

At this point, please read George W5YR's BriefOverview of A-D Conversion and Reference 8.

In practice, dithering (uncertainty of resolution of Bit 0, the least significant bit)limits the achievable dynamic range to 138 dB - still pretty good. Thus, the receiverdesigner needs to distribute the gain of the analogue stages preceding the ADC so that nostage overloads before the ADC reaches it own overload point (the upper limit of itsresolution). The DSP will then be fully exploited.

The Pro derives its AGC voltage via a secondary DAC, which decodes the processed IFbitstream prior to demodulation (in the DSP). The AGC can do a very good job of protectingthe ADC from overload; there are a few software tricks to minimize swamping by strongout-of-band signals. ADC overload is catastrophic, as the DSP now no longer has a usefulsignal to process.

I am not sure to what extent the conventional method for measuring BDR is valid for aDSP receiver, considering that the "1-dB compression point" may have no meaningat the ADC overload point, when the digital output abruptly ceases to represent the analogue input. Thus, the apparently "poor number" for the Pro in the ARRL testsuite may not be a true index of receiver performance. This seems to be borne out byoperational history, in which the Pro handles strong out-of-band signals very well - aslong as the ADC is not driven to overload.   (Note that as the maximumencoded amplitude value that the DSP data bus can present to the DAC is "allones", the DAC can never be overdriven. However, the analogue circuits downstream fromthe DAC output must have sufficient headroom to accommodate this maximum output amplitudewithout distortion or clipping.)

Another factor is that the radio tested by the League was an early production unit. Usually, the ARRL purchases a radio from a dealer shortly after product release, thereby acquiring a first-run unit for test. The testsuite often identifies performance issues which are corrected in the next series. Thisoccurred with the earlier IC-756 and several other HF radios of various makes.

The final arbiter of receiver (and transmitter) performance is operationalexperience. The IC-756Pro has been used very successfully on a number of DXpeditions,and has performed flawlessly.

Whilst the Icom Pro line, like all radio equipment, has its faults, DSPtechnology for the first time provides the operator with tools to minimize theharmful effect of these faults on receiver performance.

*The input signal being digitized could be incurrent rather than voltage form, althoughthis is unlikely.

Comments from DXpeditions and Field Day:

In July 2000, I was at the West Coast DX Convention in Burnaby, BC, and chatted with acouple of the operators who had participated in the Radio-sport Team Championship inSlovenia earlier this month. They had run two IC-756Pro's at the DXpedition, and reportedthat the radios operated flawlessly. They even acquitted themselves extremely well on 40mat night, with the European broadcasters almost next door. This sentiment was echoed byone of the Kingman Reef operators, Steve Wright, VE7CT, who gave a presentation on the K5Koperation at our local club last February. K5K used a number of 756Pro's, and theoperators were very happy with them. They stood up under very arduous operatingconditions, in a very hostile RF environment, and no contacts were lost due tounsatisfactory strong-signal handling.

In my station, I have not observed any overload, even during contests or Field Day.During K5K, I saw some nasty 20m signals, 10 to 15 kHz wide, on the spectrum scope of thePro. I suspected cross-mod in the radio; however, an on-air check with an HP 853A/8558Bspectrum analyzer confirmed that the signals were indeed that wide.

Notes on "IC-756Pro vs. the competition", February 2001:

DSP is the way of the future. It offers a cost-effective way to design, implement andoptimise radio designs which are as close as possible to the limit of what can be achievedtheoretically. Those who continue to disparage DSP radio architecture and digital filterswill eventually go the way of the proponents of the horse-'n'-buggy, the manual telephoneexchange, the vacuum tube, the vinyl LP, etc. A radio such as the IC-756Pro is the way ofthe future; the concept can only improve as 32-bit ADC/DAC chips, and faster DSP's, becomeavailable at prices we hams can afford. A 32-bit ADC has a theoretical dynamic range of192 dB, and will be limited by the noise floor of the receiver front end.

When you are "shopping" for a high-end HF transceiver, I would recommend thatyou consider the Pro. In comparing the Pro to its popular competitors, consider this; inthe Pro, the DSP does everything - IF filtering, IF notch (70 dB deep!), allmodulation & demodulation, noise reduction, transmit audio equalisation and AGC. TheIF-processing part of the DSP (filtering and notch) is inside the AGC loop. By contrast,the competing product is just another analogue radio with limited DSP tacked onto the backend. The DSP is wholly outside the AGC loop, so the DSP filters cannot remove undesiredsignals before they develop AGC voltage and swamp the receiver.

And then there is the spectrum scope - difficult to do without once you have got usedto it. The visual display of signals in the portion of the band that one is tuned to, orworking, is a very powerful operating tool.

The inherent 6-metre coverage of the Pro is an added bonus. The receiver noisefigure is excellent, and no $1000 transverters are required!

Thoughts on "IC-756Pro vs. IC-756", March 2001:

In the choice between an IC-756Pro and an IC-756, my recommendation is to get a 756Pro.Failing that, the 756 is a very acceptable alternative (or "backup" rig).

The two most significant deficiencies in the 756 which I had prior to the Pro were thelack of an IF notch filter inside the AGC loop, and the relatively poor verticalsensitivity of the scope. I had fitted the FL-222/223 narrow SSB filter combination; thesefilters greatly improved adjacent-channel selectivity, and tightened up the Twin PBT. Thenarrow filter pair is essential if you operate mostly SSB; the filters will cost $250 to$300 (new), on top of your $1200~1300 radio. Use of the Twin PBT to narrow the IFbandwidth comes with a price; the overall shape factor is degraded. Cascaded narrowfilters avoid this. To ensure correct Twin PBT operation, the optional filters should havecomparable bandwidths and shape factors. View the bandpasscurves of popular Icom filters. 

By the time you accumulate filters for SSB and digital modes, you will be close to theprice of a 756Pro. And then there is the hassle of opening up the 756 to swap out filtersaccording to changes in operating practice (e.g. mostly-SSB vs. mostly-digital).

Notes on Twin Passband Tuning, July 2003:

I have found that the DSP Twin PBT in the IC-756Pro/Pro II is much more effective than the analogue PBT inthe IC-756, or in my old IC-781. The 756Pro/Pro II's DSP filters have a minimum-6/-60 dB shape factor of 1.5; in practice, measurements suggest shape factors closer to 1.2. The Twin PBT controls shift the upper and lower flanks of the selected filter independently of one another without degrading the shape factor (unless a500 Hz or narrower CW filter is selected, and the Twin PBT adjustment causes theBPF Indicator to go dark.)

In contrast to this, the Twin PBT controls in the IC-756, IC-781 and IC-775 shift the passbands of two cascaded crystal filters relative to each other and to virtual carrier. This is coveredhere. If the two crystal filters have identical or similar-6 dB and -60 dB bandwidths, the composite shape factor will be better than that of either filter alone withboth Twin PBT controls at 12 o'clock. If either control is offset, the shape factor will degrade to that of a single filter. 

In a stockIC-781, the two SSB filters (FL-80 & FL-96) are mismatched. The FL-80 has a-6 dBbandwidth of 2.4 kHz, compared to 2.8 kHz for the FL-96. Thus, the composite shape factor will be about the same as that of the FL-80 (1.6). This is perceptibly worse than the SF of the756Pro II's DSP filters - confirming the operator's perception of the difference. 

The narrow SSB filters for the IC-756 offer another example. The FL-223 (9MHz) has a -6 dB bandwidth of 1.9 kHz and a shape factor of 1.9. Thecorresponding values for the FL-222 (455 kHz) are 1.8 kHz and 1.66. Withboth Twin PBT controls at 12 o'clock, the composite shape factor isapproximately 1.4. Offsetting either control degrades the shape factor to somevalue between 1.7 and 1.9, depending on the amount of offset.

Comparing filter shape factors in the IC-756Pro, Pro2 and Pro3:

Originally, the -60 dB bandwidth for the 2.4 kHz SSB filter was stated as 2.8 kHz. For the 500 Hz CW filter, the corresponding figure was 700 Hz.

When I purchased my IC-756Pro in July 2000, I recall that there was an erratum sheet in the user manual. This sheet restated the -60 dB bandwidth for the 2.4 kHz SSB setting as 3.6 kHz. The -60 dB figure for 500 Hz CW was unchanged. Unfortunately, I did not keep the sheet - it went to the new owner of the radio. It is puzzling that Icom never updated the IC-756Pro brochure, service or user manual to reflect this change.

The figure of 3.6 kHz at -60 dB applies to the IC-756Pro, the IC-756Pro2, the IC-746Pro (IC-7400) and the IC-7800. This equates to a -60/-6 dB shape factor of 1.5.

The selectivity specifications appear to be worst-case values. In recent tests, the following approximate -60/-6 dB shape factor values were obtained for the 2.4 kHz SSB filter:

Notes on "Carrier Set Point and the IC-756Pro", December2001:

Q. Is there any way to adjust the carrier point oscillator in the 756Pro?

A. The IC-756Pro has no carrier insertion oscillator, and therefore no set pointadjustment. The DSP performs all modulation and demodulation tasks. In SSB mode, the DSPmodels an idealised phasing-type exciter and demodulator, for transmitting and receivingrespectively. The DSP algorithms are hard-coded in firmware, and are not user-modifiable.

In SSB transmit, the frequency relationship between the virtual carrier and thegenerated LSB or USB spectrum is fixed. Three occupied-bandwidth settings (NAR 2.2 kHz, MID 2.4 kHz,WIDE 2.8 kHz) are available. George,W5YR's occupied bandwidth measurementsconfirm these values.

The WIDE transmit AF response at the front-panel MIC socket is approximately 90 Hz ~ 2.8 kHz @ -6 dB. It ispossible to "stretch" the low-end response down to approx.50 Hz @ -6 dB byapplying the transmit audio at the Pin 4 of rear-panel ACC1 socket, with WIDEselected.

In receive, the IF filter bandwidth is infinitely variable, in the range50 Hz ~ 3.6 kHz. In addition, the lower and upper flanks of the IF bandpass can be shiftedby up to � 600 Hz relative to virtual carrier via the Twin PBT controls.

The IC-756Pro and Pro II, and the IC-746Pro (IC-7400) are a "wholly differentanimal" compared to conventional radios. The DSP chipset is actually a fully-functionalfixed-frequency transceiver operating at 36kHz.  RF/IF amplifiers, up/down converters and DDS frequency generators interfacethis transceiver to the operating frequency domain. In a restricted sense, this36 kHz transceiver is a software-defined radio (SDR).

Notes on the transmit audio inputs, ACC1 Pin 4 vs. front-panel MICsocket:

Input levels: ACC1 Pin 4 expects 100 mV rms (280 mV p-p) for fulldrive. The high input impedance will not load down the conventional outputcircuits of accessory devices; these tend to have source impedances in the 500 ~2000Ω range. The MIC input expects 10 mV rms (28 mV p-p) with the MIC GAINcontrol at 12 o'clock (IC-756Pro service manual, p. 4-10). A typicaldynamic microphone has 3 ~ 5 mV output at 94 dB SPL (2.5 cm from talker'smouth).

Measurements performed by George, W5YRconfirm that for the WIDEoccupied-bandwidth setting, the low-end audio response is lower at ACC1 Pin 4(50 Hz at -6 dB) than at the MIC input (85 Hz at -6 dB). These measurements also show that for the MID and NAR settings, thedifference in low-end response between ACC1 Pin 4 and the MIC input isnegligible. The explanation for this is as follows:

From a look at the transmit speech-amplifier circuit (IC451, a µPC5023 ASIC, on the MAIN UNIT board), there is an RC feedback network in the first stage (AMP). This networkis a high-pass filter (HPF) whose -6 dB cutoff frequency is approx. 90 Hz. The HPF accounts for the steeper low-frequency roll-off at the MIC input compared to the ACC1Pin 4 input (significant only in the WIDE setting). The second stage (VCA, voltage-controlledamplifier) has a flat response. The Mic Gain control sets the VCA gain via the MIGV line. (Refer to IC-756Pro servicemanual, Page 10-4.)

The design intent in this circuit is  to cut off low-frequency components of the microphone output (below150 ~ 200 Hz). These components contribute virtually nothing to articulation at the receiver, but rob transmitter powerthat can be more usefully employed to transmit the midband and upper frequencysub-bands which determine the articulation index at the receiver. Note that if bass-boost is applied to the microphone audio fed to the front-panelMIC socket, IC451 will severely attenuate all components below 150 ~ 200 Hz. 

The AMOD (ACC1 Pin 4) input bypasses IC451 entirely. The outputs of IC451 and the AMOD buffer aresummed, and fed to the ADC via the DTAF line. The design intent here is probably to pass AFSK signalsinjected via AMOD without distorting the frequency transitions, which have a low-frequency component.(Note: The transmit audio signal pathdoes not feed a "balanced modulator". There is no such"animal" in the IC-756Pro or Pro II.)

Icom's engineers probably felt that this analogue solution would be more cost-effective and less risky (in terms of potential bugs) than separate, selectable DSP modulation algorithms for voice and data transmission.

Now, in the DSP, the low and high cutoff frequencies of the PSN modulationalgorithm are hard-coded in the DSP firmware. This applies for all three occupied-bandwidth settings: WIDE, MID and NAR.  As the audio signal pathfrom ACC1 Pin 4 to the ADC is flat, its frequency response is determined solelyby the DSP.

We have already seen from the above discussion that the audio signal pathfrom the MIC socket to the speech amplifier (IC451) is 6 dB down at 150 Hz. Inthe WIDE setting, the DSP "modulator" is 6 dB down at 90 Hz. 

Thus, it is possible to extend low-frequency transmitted audio response byselecting the WIDE occupied-bandwidth setting, and injecting microphone audio at linelevel via ACC1 Pin 4.  The above caveatconcerning articulation still holds true.

Note that the Mic Gain control  has no effect on the ACC1 Pin 4 audio input. The reason for this is that Pin 4 bypasses IC451, including the VCA referred to above. VOX is also inactive when transmit audio is fed to ACC1 Pin 4.

Note also that the MIC input and Compressor (and the TX Tone controls onthe IC-746Pro and IC-7400) are disabled when USB-D or LSB-D is selected. The IC-756Pro2 and IC-756Pro3 allow separate filter definitions for USB-D and LSB-D which canbe different from those for USB/LSB. In addition, with USB-D and LSB-D, it ispossible to set BPFfilters with bandwidths of 500 Hz or less.

Note on "Software-Defined Radios":

The term "software-defined radio" (SDR) has been bandied aboutrather a lot lately in the amateur community. This isa misnomer in HF radio design; as long as hardware components such as filters,amplifiers, mixers and oscillators are encountered in the RF signal path, withhardware changes required to modify some or all of their operating parameters,the radio is not "software-defined" - even if it issoftware-controlled. We will not see true SDR in HF equipment until acost-effective ADC, running at the signal frequency, can be placed right at thehead of the RF chain.  Taken as a whole, the IC-756Pro, IC-756Pro II andIC-746Pro are not SDR's, for the reasons given above. In a restrictedsense, though, the 36 kHz DSP core of all these radios is anSDR, as its operating parameters are almost entirely defined in the DSPfirmware. Read "Basic Concept of Icom IF-DSP"and References 12, 13.

Notes on roofing filters and spectrum scope architecture in theIC-756Pro (May 2002):

(Contributed by George T. Baker, W5YR. Refer also toGeorge's Notes on Filtering & DSP.)

Roofing filters:

At the first IF of 64.455 MHz, there is a crystal 15 KHz roofing filter at the input ofthe IF amplifier. At the second IF of 455 KHz, a 15KHz ceramic filter follows the secondmixer. This filter feeds the Noise Blanker Gate which in turn feeds a second 15 KHzceramic filter preceding the third mixer to the final (DSP) IF of 36 KHz. 

Admittedly, these are not  “narrow filters” by classical analogue-radiostandards. Although their primary purpose is image rejection, they are adequately narrow to ultimately prevent overload of the DSPA/D converter(s) and that is all that is required. Ina linear system, the ultimate band-limiting can take place at any point provided thatsystem operation is linear up to that point. Proper distribution of stage gains andappropriate AGC loop operation virtually ensure linearity in the PRO up to the DSP stage.

It is also of dubious merit to ascribe performance ills to a lack of"classical" narrow IF filters - i.e. typical 2.5 - 3.0 KHz IF crystal filters -in the PRO.

While it is “conventional wisdom” that such narrow filters are all to thegood, despite their phase delay, uneven amplitude response and other negative effects,recall that we once had radios built to the “conventional wisdom” that two r-famplifier stages are better than one, since “sensitivity” was the principalmeasure of a good receiver. 

Furthermore, the properties of conventional crystal and mechanical filters have now metwith such intense competition from the DSP approach that one can readily project that in afew years there will be little or no need or requirement for crystal filters since theirperformance will be - and already is - inferior to the DSP equivalent. Have another lookat the IF Filter Page and compare. Here is a test which you can run on your PRO or PRO II to put the DSPIF filters through their paces.

Further comments by Adam, VA7OJ (December 2006):

The up-converting architecture, with a roofing filter at a first IF above the highest RF frequency, allows the designer to limit the bandwidth presented to the first IF chain and second mixer. The bandwidth of this filter is a trade-off. Its 3 dB BW must be sufficient to pass the widest emission the receiver is required to handle, but not so narrow that IMD and temperature-drift effects in the filter become a concern.

Reference 23 has revealed that the 6 kHz roofing filter in the IC-7800 degrades IP3 at 2 kHz offset by 16 dB, as compared to the 15 kHz filter. The most likely reasons for this surprising degradation are:

1. IMD generated in the filter;

2. IMD due to saturation of the IF amplifier driving the filter as a result of mis-termination caused by a sharp increase in filter Z_in outside the filter passband;

3. both these effects in combination.

Furthermore, it can be shown that the limiting factor in close-in HF operation is more likely to be the poor spectral purity of the distant transmitter than any limitations in the local receiver. This is as much an operating and regulatory issue as it is a technical one. 

Spectrum scope topology:

An examination of the IC-756Pro or IC-756Pro II block diagram reveals that  the PRO has a 15 kHzcrystal roofing filter at the first IF - immediately after the 1st-mixer output combiner.

The scope system in the PRO/PRO II is fed from the combiner output (beforethe 64.455 MHz roofing filter) at an IF of 64.455 KHz via a PIN diode attenuatorand separate IF stages to a mixer,which with a 77.8 MHz LO, down-converts to 13.345 MHz for the scope circuitry. There aretwo 13.35 MHz ceramic bandpass filters in cascade in the mixer output, which feeds anothermixer with a 12.89 MHz ± 100 KHz LO to produce an output which via a 455 kHz ceramicbandpass filter (CFJ455K8) drives the scope IF system. Note that all of the above circuitry for thescope operation is completely independent of the circuitry following the first mixer whichdrives the actual receiver IF and subsequent circuitry, including the DSP. No compromisesin receiver IF design are required to support scope operation. (Note: The CFJ455K8 filter defines the scope's resolution bandwidth as 1 kHz at -6 dB.) To increase its dynamic range, the scope incorporates an internal AGC loop derived from the vertical amplitude signal.

As the IF pickoff point for the scope is also before the AGCgain-control point, neither the RF Gain control setting nor AGC actionaffects the vertical sensitivity of the scope.

There is a detailed description of the scope in Section 3-5 of the service manual (all IC-756Pro models).

With Preamp 2 enabled,  0 dB attenuation (RF and scope) and autotunerdisengaged , the Pro II spectrum scope will display a visible spike with 0.12 µV(-125 dBm) at the antenna socket. (This was measured with a CushmanCE-31A service monitor in CW mode, calibrated against a Rohde & Schwarz URV4RF power meter.) The radio was also in CW mode, with the 500 Hz filter selected.

Thus, it can be seen:

1. that there are three roofing filters in the main receiver channel and three more in the scope channel.

2. that none of the main receiver's roofing filters are involved in the scope operation, since the scope operates as a completely independent receiver.

3. that the presence of the scope has no negative effect whatsoever on the design and architecture of the “main” channel of the receiver.

Notes on IC-756Pro2 DSP Noise Reduction, April 2004:

(Contributed by George T. Baker, W5YR.)

Q. Is it desirable to "reinitialize" the NR periodically bytoggling it on and off, or by momentarily pulsing the PTT line?

A. This is largely unneeded since the NR logic is only doing its job.NR in the Pro2 is a time-varying adaptive digital filtering operation whichalters its response according to the statistics of the noise it encounterswithin the IF passband. The statistics can and do vary with time and thenoise reduction algorithms react accordingly.

When first activated, NR appears to consider everything "noise" and thuslowers gain across the entire IF passband considerably giving the perceptionof a very quiet receiver . This can be interpreted as effective NRoperation, but that is probably a premature conclusion...

As time goes on, however, the NR logic has more statistical history to workwith and begins to alter its response accordingly. The actual noise withinthe passband that meets the rejection criteria of the NR logic continues tobe suppressed while other frequency components are restored approximately totheir original level.

One result is that the perceived "noise" (audio) level increases and onecould conclude that the NR is no longer functioning. This may or may not bethe case, depending upon the statistical nature of the noise.

I find it difficult to accept the premise that Icom failed to execute the NRfunction properly in the Pro2 and earlier models. If one examines theIF/audio spectral response of the Pro2 with the NR operating, the change innoise response of the passband can be seen to vary with time.

Initially the whole passband is diminished in level and then usually (perthe type of noise I have locally) the low and high ends of the spectrumquickly "grow" in level while the midrange tends to diminish. The increasedsignal levels at the extremes of the passband appear to account for theapparent increase in audio level, which some are interpreting as a loss ofeffectiveness of the NR function.

As to "defeating" this effect with periodic PTT operation, one couldprobably simulate the same outcome by periodically decreasing the AF gainand then slowly bringing it back up. The PTT operating restores the NR logicto its original "starting" point so it begins anew "thinking" that most ofwhat is in the passband is noise and lowering the overall response level.Continued PTT operation merely restarts the process and keeps the NR logicfrom reaching and maintaining its steady state response condition.

I apologize if any of this sounds negative or even argumentative - that is not myintent - but I think that efforts to circumvent the programmed operation ofthe Pro2 NR logic, while subjectively appealing, actually accomplish littletoward improving the noise reduction capabilities of the receiver.

An interesting exercise would be to bury various signal waveforms in "noise"and to determine the S/N ratios for each as a function of time with NRactivated. The final arbiter is whether NR improves S/N to make a weaksignal more readable, not whether the receiver sounds "quieter" with NRactivated. (IF I can ever find time, I would like to conduct such a test.)

A final note: the NR function is most effective with the wider IFbandwidths. As bandwidth is reduced, the noise statistics are changed andthe NR has less to work with. At very small bandwidths, NR is essentiallyineffective since even pure noise appears to be a randomly modulated sinewave whose frequency is related to the bandwidth. Using NR with a 100 Hzfilter for CW produces little useful effect, for example.

Notes on internal autotuners, July 2004:

Q: I only use resonant antennas. Should I enable my internal autotuner?

A: The purpose of the autotuner is to keep the PA happy by ensuring that it is always presented with a 50Ω resistive load for optimum power transfer and best linearity. (See Reference 19).

It goes without saying that the low-pass filters must also be correctly terminated in 50Ω resistive to fulfil the above requirement.

Even a resonant antenna will present precisely 50 + j0Ω to the transmitter at only one frequency. The mission of the autotuner is to ensure that the transmitter sees 50 + j0Ω over a reasonable VSWR excursion, i.e. a reasonable frequency band. Another advantage of enabling the autotuner is that it remains in the RF signal path on receive, thus providing additional RF preselection.

Q: But if I activate the tuner to match a slightly elevated VSWR, say around 1.5:1, won't the tuner's insertion loss exceed any mismatch loss in my feedline?

A: This will be true for some cases. However, it should be noted that the reflectometer at the PA LPF output (which reports forward and reflected power to the control processor) will start folding back the RF drive at VSWR > 1.5:1. This is yet one more reason for ensuring that the PA is terminated in a 50 + j0Ω load.

Note on "RF" vs. "IF" DSP (March 2005):

An interesting comparison of two approaches to DSP has emerged from a recent discussion with Bob, W4ATM. It is tempting to think of a direct-sampling RF-level DSP design such as the RF Space SDR-14 as the be-all and end-all. However, its 14-bit ADC which is fed directly with the entire MF/HF band 0.5 ~ 33 MHz, and samples at 66.7 MHz, has an ENOB (effective number of bits) of approx. 12.3. Each bit (power of 2) equates to 6 dB of dynamic range; thus, 12.3 X 6 = 74 dB dynamic range (from MDS to ADC "all 1's" point). The only analogue RF circuitry at the ADC input is a wideband variable-gain preamp followed by an LPF. (Read Reference 10, and comments by Pieter, N4IP).

A digital down-converter (DDC) following the ADC processes the digital ADC output into I/Q format, and translates the effective bandwidth down to 150 kHz (typical) within the 33 MHz band. The resulting DDC processing gain* increases the overall dynamic range by 10 log₁₀ (33/0.15) = 23 dB. The resulting dynamic range is now (74 + 23) = 97 dB.

*Processing gain is not "gain" in the sense of amplification; it is a reduction of noise, and thus an increase in S/N ratio, due to the bandwidth reduction.

The overall dynamic range can be improved further by decreasing the bandwidth of the DDC output band segment. The limiting case is determined by the mathematical precision of the post-processor. The I/Q outputs of the DDC drive a PC sound-card via a USB 1.1 interface. The PC performs all demodulation tasks.

By contrast, the lower-cost 24-bit, 36kHz-sampling ADC used in the Icom Pro series has an ENOB of at least 20 (120 dB dynamic range). This is close to the effective blocking dynamic range of the RF/IF chain ahead of the ADC. Thus, Icom has done a trade-off; the engineers in Osaka realised that they would be better off at this stage of the game by not allowing a dynamic-range bottleneck at the ADC, and by designing the analogue RF/IF chain for a dynamic range as close as possible to that attainable in the ADC.

The direct-sampling architecture eliminates the traditional mixer, LO, roofing filter and analogue IF chain. Phase noise is determined principally by the  ADC clock source. The ADC also adds a small increment to the phase noise due to aperture jitter, which the ADC data sheet gives in fs_rms. With careful design, phase noise values of -140 dBc/Hz at 1 kHz offset, and -150 dBc/Hz at 10 kHz offset, are possible. These values have been achieved in the R&S EM510.

Therefore, when considering close-in dynamic range, a properly-designed direct-sampling receiver is substantially better than any conventional receiver, because its phase noise sets the limit for dynamic range.

Fast ADC and DDC IC's are becoming more cost-effective, as are post-processors with high mathematical precision. These trends may  allow RF-DSP receiver designs to eclipse the current analogue front end/IF-DSP topology in a few years. However, there is still the concern that the composite power of a number of strong signals in the sampling bandwidth of a wide-band ADC will drive the converter  beyond its "all 1's" point (over-ranging). Work is currently underway to address this concern by placing a bank of bandswitched BPF's between the receiver's RF input and the ADC input, to limit the RF bandwidth to which the ADC is exposed.

The following comment (from the ADAT data sheet) is significant: "The two-tone measurement procedure is generally not applicable for A/D converters , as the IP3 of an A/D converter decreases proportionally to the input signals. On the other hand, due to dithering, the IM-free dynamic range is in practical operation substantially higher than at the laboratory two tone measurement."

Dithering converts low-level IMD products into quantisation noise, raising the noise floor by a very small amount. The power in the IMD products is spread all over the sampled bandwidth. In a hybrid superhet/IF-DSP design, this increase in noise floor is of little consequence as it occurs at a high signal level.

I suspect that dithering induced in the ADC of a hybrid superhet/IF DSP receiver such as the 756Pro series by the presence of noise and multiple strong signals improves the ADC's linearity. Perhaps this effect would cause the receiver to behave better in actual on-air operation than on the lab bench, where only two pure tones are present. I suspect that this dithering is the key to the apparent disparity between lab test results and actual on-air performance of IF-DSP receivers, as noted by several friends of mine and myself. This is an intriguing thought.

Until high-resolution ADC's with high sampling rates become cost-effective at amateur price levels, the hybrid topology (incorporating a roofing filter to reduce the statistical likelihood of ADC overload by strong undesired signals) will remain in favour among amateur HF equipment manufacturers.

Icom is using ADC's and DAC's designed for use in wireless phones and DVD/CD players. The ADC used in the IC-756Pro II and IC-756Pro III has 120 dB dynamic range, compared to 112 dB for the chip used in the IC-746Pro.

Latest news on "RF-DSP" (direct sampling) (January 2007):

Rohde & Schwarz have announced the EM-510. This is a self-contained, direct-sampling HF receiver using a 16-bit ADC. Abrochure is available.[mirror] 

Visit HB9CBU's ADAT direct-sampling transceiver site (in English). Download ADAT documents. View an ADAT presentation.

More direct-sampling SDR news (January 2010):

The Perseus direct-sampling SDR receiver from Microtelecom S.p.A. of Italy is taking the SWL community by storm, and is also making a significant impact in the amateur community. It features phase noise < 147 dBc/Hz at 10 kHz offset. The Perseus uses a fast 14-bit ADC preceded by an RF attenuator, a preselector with 10 switched BPF's and an RF preamp with +30 dBm IP₃.

The four evolutionary phases of the "DSP" HF amateur transceiver:

1. The audio DSP add-on, such as the Ten-Tec Omni V and VI series. These radios incorporated an OEM audio DSP board manufactured for Ten-Tec by JPS. Other examples are the IC-706 Mk II, IC-703 and R-75 receiver, with optional DSP module.

2. The limited post-AGC final-IF DSP implementation. Examples in historical order: IC-775DSP, Yaesu FT-1000MP, IC-756, Yaesu FT-1000MP MkV. In these radios, the DSP typically performs modulation, demodulation, NR, some audio filtering, TX EQ and auto-notch. The DSP, consisting of an ADC, DSP processor and DAC, is in the final IF but post-AGC. In the IC-775, and the Yaesu radios, an analogue back end is fitted, and touted as an "alternative" to the DSP. (Good marketing propaganda; the fact is that leaving the analogue circuitry in was cheaper than designing it out!) The final IF is in the range 10 ~ 15 kHz.

3. The "true IF-DSP" radio, in which DSP now does all IF filtering (including a tuneable notch), AGC derivation and TX compression, as well as the functions mentioned in (2). Here we have the Kenwood TS-870, the Kachina 505, the IC-756Pro, the IC-7800, the Ten-Tec RX-340, Pegasus and Jupiter, the IC-7400 (IC-746Pro), the IC-756Pro II, the IC-756Pro III,  the IC-7000, the IC-R9500 receiver and the IC-7700.  Other examples are the VS/Yaesu FTDX-9000, FT-2000 and FT-450. The IF filtering functions are now within the AGC loop. Also, due to advances in chip speed, the final IF is now typically 36 to 48 kHz (the earliest implementation, the TS-870, used 11 kHz).

4. The "RF-DSP" radio, in which the ADC is clocked at a frequency above the highest RF operating frequency, and processes the entire HF band. This approach is at the leading edge; it is dependent on the speed and price/performance ratio of fast ADC, DSP and DAC chips. Refer to Note on RF vs. IF DSP above.

The Kachina 505 uses 16-bit ADC/DAC, 24-bit processing and a DSP IF of 40 KHz. It hasanalogue AGC with selectable time constants, etc. plus a Digital AGC which works in tandemwith the analogue AGC. The Ten-Tec product line uses 16-bit ADC/DAC. The Icom IC-756Pro, ProII and IC746 Pro all use 24-bit ADC/DAC and a 32-bit DSP processor.

Concluding thoughts:

Times change, and so does the architecture and design of our radios. Conventionalradios such as the earlier Kenwoods are vastly different in many respects from the PROs,the Kachina 505DSP, and even the Ten-Tec Pegasus and Jupiter. I don’t think that itis any accident that most of the really high-line commercial and military radios (e.g.Rohde & Schwarz, Rockwell-Collins, Harris etc.) have used extensive digital filteringand signal processing for the past several years.

The above article is now also available inRussian.

Read my IC-756Pro III User Review.

Further reading:

1. "HF Radio Systems & Circuits", Sabin & Schoenike, editors. Noble,1998. This textbook was written by members of the Engineering Staff, CollinsDivisions, Rockwell Corporation. Chapter 8 is an exhaustive treatment of DSP radio designconcepts, of which this excerpt gives an example.

2. "The ARRL Handbook for the Radio Amateur", 2001 Edition,Chapter 18, Digital Signal Processing.

3. "TheScientist's and Engineer's Guide to Digital Signal Processing", bySteven W. Smith, Ph.D.

4. "A High-Performance Digital TransceiverDesign, Part 1",  by James Scarlett KD7O, QEX, July/August2002.

5. IC-756Pro III vs. Pro II - a presentation given at the North Shore ARC in January 2006 (PDF)

6. Leif �sbrink SM5BSZ Home Page    mirror site

7. A general discussion on radio receivers, by Leif �sbrink SM5BSZ

8. Receiver dynamic range measurements, by Leif �sbrink SM5BSZ

9. The third-order intercept point, by Leif �sbrink SM5BSZ

10. Dynamic range observations for the SDR-14, by Leif �sbrink SM5BSZ

11. Thinking of buying a "down-converting" HF rig with a mid-band 1st IF? Read this first!

12. Software-Defined Radio White Paper, Wipro Technologies, August 2002

13. Software-Defined Radio Comes of Age, NTIA, Fall 2001

14. "HF Radio Systems & Circuits", Sabin & Schoenike, editors. Noble, 1998, pp. 145-148 (AGC Design); pp. 341-342 and Fig. 8.10 (Digital AGC Methods). Get excerpts (PDF)

15. "IC-756Pro II Technical Report", Icom Inc., 2003. (PDF) mirror site

16. "IC-756Pro II Receiver IMD and DSP Filter Performance",  by J. Saito JA7SSB, CQ Ham Radio, January 2002. Summary by N. Oba JA7UDE (PDF)  Appendix (PDF)

17. "A History of Crystal Filters", by R.E. Kinsman. IEEE International Frequency Control Symposium, 1998.

18. "Intermodulation testing of high performance receivers", by John Thorpe. AOR (UK) Ltd.

19. "HF Radio Systems &Circuits",  Sabin & Schoenike, editors, Noble, 1998, Chapter 12, .Get excerpt (PDF)

20. "Understanding Noise Figure" by I. Rosu, VA3IUL.

21. Receiver Sensitivity Metric Converter - Receiver Noise Figure, Noise Floor & Sensitivity Calculator

22. DSP in HF Radios - a presentation given at the North Shore ARC in February 2006 (PDF)

23. "Test von Intermodulationsfestigkeit" by W. Schnorrenberg, DC4KU, CQ-DL, 8-2005, p. 544. Translation

24. Analog-to-Digital Converter Considerations, by Andrew Roos ZS6AA.

25. Amateur Radio Application of Frequency Modulation, by VK1OD.

Copyright � 2000-2011 A. Farson VA7OJ/AB4OJ. All rights reserved. (Copyright incontributed items reposes with the respective contributors.)

Last revised: October 18, 2019