FT2K: As there are two distinct AGC loops which sense conditions at different points in the receiver signal path (as discussed above), a race condition could develop in response to a fast-rising noise impulse. This could produce a "ticking" sound at the DAC output (baseband). The issue is that a strong signal falling outside the selected IF-DSP filter bandwidth, but within the roofing-filter window, will develop AGC voltage because the DSP code is too "dumb" to detect this condition and weight the AGC loop appropriately. Excessive gain in the RF/IF chain prior to the ADC could also be involved. What is needed is a routine in the DSP AGC process which reduces (delays) the developed AGC voltage if a signal is detected after the roofing filters, but not after the DSP-IF filters. There is no AGC on the second mixer, as the AGC on the 1st IF amp provides sufficient control range. I suspect that the DSP does not do much processing on the primary loop. The primary AGC loop samples the signal amplitude at the output of the 1st IF roofing filter. Strong signals appearing in the roofing-filter window, but outside the DSP IF filter bandpass, will develop AGC voltage and cause swamping unless the DSP algorithm recognises this condition (signal in the roofing filter passband but not in the DSP IF passband) and compensates. I am not sure to what extent, if at all, the secondary AGC loop (wholly within the DSP) influences the primary loop. The wider the roofing filter, or the poorer its shape factor, the greater is the statistical likelihood of swamping. Here is a reference on AGC design: https://www.qsl.net/icom/agc.pdf With a roofing filter having a narrower -6 dB BW and better shape factor, AGC swamping should be less frequent. But if the DSP algorithm does not weight input from the primary loop (whose derivation point is right after the roofing filter) against that from the secondary loop (whose derivation point is after IF selectivity in the digital domain), swamping will still occur. The trick is for the AGC to know when to disregard signals which are in the roofing-filter window, but outside the IF filter bandpass. I do not think VS' programmers are up to such sophisticated coding. I may be able to explain the reported AGC swamping (pumping). The simplified receiver block diagram in the FT-2000 brochure shows a primary and a secondary AGC loop. The primary AGC loop controls the gain of the 1st IF amp (Q1105, MAIN). This loop appears to pick off AGC voltage at the digital output of the ADC, prior to the IF selectivity process in the DSP. This means that a strong signal falling outside the DSP IF filter passband, but within the roofing-filter window, will develop AGC voltage and swamp a weak signal inside the DSP IF passband. There is a secondary AGC loop derived after DSP IF filtering, but it only controls the gain within the DSP. Digging a little deeper into the DSP-Unit schematic, Q7504 is a dedicated DSP which drives the primary AGC line (AGC OUT). Q7504 receives SDA and SCL signals from the corresponding pins on the DSP IC (Q7514). Per the TI TMS320C6713 DSP data sheet, SDA (serial data) and SCL (serial clock) are auxiliary control lines (as distinct from the main SDIN, SCLK and SDOUT lines which are routed to Q7503, the main codec or ADC/DAC.) If my surmise is correct, the bitstream fed to Q7504 is derived prior to the IF selectivity process, as suggested by the simplified diagram. This will account for the swamping, as a strong signal falling within the roofing-filter passband will develop AGC voltage and reduce the 1st IF amp gain. A narrower roofing filter can be selected to reduce the swamping. As there are two distinct AGC loops which sense conditions at different points in the receiver signal path (as discussed above), a race condition could develop in response to a fast-rising noise impulse. This could produce a "ticking" sound at the DAC output (baseband). Comparing this with the IC-7800 AGC scheme, which was also mentioned: The IC-7800 AGC subsystem has two loops. The primary is an analogue loop which rectifies the 36 kHz IF at the output of the 2nd mixer, and applies the resulting DC voltage to one of the ADC's inputs. The other input is fed by the 36 kHz IF output of the 2nd mixer. The secondary is a digital loop, within the DSP, which adjusts the gain prior to the IF selectivity process and also develops a DC AGC voltage via a DAC. This voltage controls the gain of the 1st IF amplifier. The digital loop is similar to the one in the IC-756Pro3. I presume that the analogue and DSP AGC inputs are compared in the DSP AGC process. I believe that if the primary loop detects a signal which has passed through the roofing filter but the secondary loop does not "see" that signal in the DSP-IF filter passband, the primary loop will be inhibited from pulling the AGC line. The purpose of this is to prevent swamping. The ticking artefact may be due to very short attack time in one of the AGC loops (or possibly due to a race condition between the loops), and the designers can correct it. I suspect that excessively short attack time in the primary loop may be the culprit, as the secondary loop may be immune - as suggested in the next paragraph. This artefact does not exist in the IC-756Pro series, as they do not have an analogue loop. In the IC-756Pro3, with which I am very familiar, the AGC signal is derived within the DSP, after the IF selectivity process but prior to the demodulation process. This signal varies the gain ahead of the demodulation process (in the digital domain). The associated bitstream also feeds a dedicated DAC which drives the analogue AGC line controlling the gain of the 1st IF amplifier. The main DAC and the AGC DAC are driven by a common serial data output from the DSP IC on which the baseband and AGC bitstreams are multiplexed. The upshot of all this is that signals outside the DSP IF passband do not develop AGC voltage. The single-loop topology may also explain why the 756Pro series is free of the "ticking" problem. The AGC will respond only to processed and band-limited signals presented to it by prior processes in the DSP. You can test for swamping with a signal outside the IF passband, but within the roofing-filter window, by selecting a 500 Hz CW filter, injecting a signal at about -90 dBm from a signal generator into the main receiver, noting the S-meter reading and then tuning the signal slowly across the DSP-IF passband. The S-meter reading should fall to zero, and the recovered audio tone should disappear. Now increase the generator output until the signal deflects the S-meter again. Ideally, there should be no S-meter deflection until the test signal level reaches a point where reciprocal mixing noise is at a sufficiently high level to develop AGC voltage (around -20 dBm or higher). 756Pro3 and 7800 AGC: I did a bit of comparison between the IC-756Pro3 and IC-7800 AGC systems. The Pro3 has a single-loop DSP-derived AGC circuit which samples the signal in the digital domain, after IF selectivity but before demodulation. https://www.qsl.net/ab4oj/icom/pro_dsp.html I believe the AGC process in the DSP does a bit of signature analysis, and can distinguish between a fast-rising noise impulse and the leading edge of a valid signal waveform. Thus, the AGC will not clamp on a noise or static spike. The IC-7800 AGC has primary and secondary AGC loops. AGC derivation is wholly in the digital domain. https://www.qsl.net/ab4oj/icom/ic7800/7800fe.html The primary loop samples the IF signal amplitude prior to the selectivity process. It regulates the gain of a digital IF amp before the selectivity process and also the gain of the analogue 1st IF amp following the roofing filter. The secondary loop samples the IF signal amplitude after the selectivity process, but prior to the demodulation process. Go to this link: http://www.icom.co.jp/world/products/amateur/7800/index.htm and click "Feature 1" for a concise description of the 7800 ADC. Swamping could occur if a strong signal within the roofing-filter window, but outside the DSP-IF bandpass, developed AGC voltage in the primary loop. I believe that to prevent this, the DSP does a comparison of primary and secondary AGC "voltage" and disregards the primary loop if the same IF signal is not present in the secondary loop. "Ticks" and AGC clamping caused by very short noise impulses can be prevented if the AGC algorithm is sufficiently "smart" to do a simple signature analysis of the noise spike, and change the AGC profile dynamically (longer attack, shorter decay) if it recognises an impulse (as opposed to the leading flank of a signal waveform.) The optimum attack time is a trade-off between a value sufficiently long to reduce response to short transients, and sufficiently short to respond correctly to a CW or digimode signal. Another issue is protecting the NB. If the NB is implemented as a DSP process (which is the case for the 7800, the 7000 and the newly-announced 7700), the AGC must be carefully calibrated to ensure that noise impulses do not over-range the ADC. Driving the ADC past "all-1's" will kill the NB and all other DSP processes. The FT-2000 AGC subsystem superficially resembles that of the 7800, but I suspect that the DSP does not do much processing on the primary loop. The primary AGC loop samples the signal amplitude at the output of the 1st IF roofing filter. Strong signals appearing in the roofing-filter window, but outside the DSP IF filter bandpass, will develop AGC voltage and cause swamping unless the DSP algorithm recognises this condition (signal in the roofing filter passband but not in the DSP IF passband) and compensates. The wider the roofing filter, or the poorer its shape factor, the greater is the statistical likelihood of swamping. Further reading: Sabin et al., "HF Radio Systems & Circuits", Section 8.6 (Digital AGC Methods), p. 341. This section covers the topic very thoroughly. See also ch. 4, pp.145-148, AGC Design. To quote from 8.6: Strong signals that fall outside the narrowband digital filter bandwidth, but inside the analog IF translator bandwidths, can overload or saturate the A/D converter. This results in the generation of in-band IMD products and can result in significant degradation of the desired signal. If large signal levels are detected at the A/D converter, the receiver gain may have to be re-distributed by reducing the pre-conversion analog gain and increasing the digital gain to maintain the desired signal output level. This will, however, reduce the desired signal-to-quantization noise ratio. The IC-7000 AGC scheme is single loop. To quote the service manual: The AGC (Automatic Gain Control) circuit adjusts IF amplifier gain to keep the audio output at a constant level. The receiver gain is determined by the voltage on the AGC line from the DSP circuit. The AGC voltage is detected at the AGC detector section inside the DSP ICs (LOGIC unit; IC301, IC2201) and the AGC voltage is applied to the D/A converter (IC2155). The converted AGC voltage is applied to the IF amplifiers (Q702, Q902, Q1001) after being amplified at IC1201 (pin 12) and sets the receiver gain with [RF/SQL] control. When receiving strong signals, the detected voltage increases and the AGC voltage decreases. As the AGC voltage is used for the bias voltage of the IF amplifiers (Q702, Q902, Q1001), IF amplifier gain is decreased. There are 4 AGC menu settings: FAST, MID, SLOW and OFF. The values for the first three settings are individually configurable. These settings vary the time constant (hang/decay time.) The attack time is fixed. Interestingly, I do not recall any owner complaints on the IC-7000 forum about the AGC "folding up" or otherwise misbehaving. (On a semi-related topic: The 8 kHz tone leakage in early IC-7000's was caused by leakage of a sub-multiple of the DAC clock, due to improper decoupling.) In the commercial HF sphere, designers test for AGC control range, attack transient response, hang time and decay time. The only ITU-R requirement I could find was in ITU-R Recommendation SM.1055, where impairment to AGC must be prevented if the interference burst time is greater than the AGC attack time in a "conventional digital receiver) or greater than 1 mS in an AM voice receiver. I would think that no AGC system should ever respond to a transient shorter than 100 nS. But as I mentioned in an earlier post, the optimum AGC attack time is a trade-off between "too long", leading to instantaneous ADC over-ranging on noise impulses, and "too short" causing AGC clamping on RF spikes. The attack time concerns only the leading edge of a signal envelope or noise event. The hang time should in all cases be longer than the symbol rate (CW), syllabic rate (voice) or the typical interval between noise impulses. Static is a bear to deal with, as it has no predictable pattern. Sabin: The desired receiver gain distribution, or analog versus digital receiver gain, is selected on the basis of a system level analysis. The maximum analog translator gain must be great enough so that the quantizing noise of the A/D converter does not degrade the receiver noise figure or sensitivity below the desired limit. As the signal level increases above the sensitivity level, the analog gain should not be decreased by AGC action until an adequate signal-to-noise ratio is obtained. As the signal level increases still further, the analog AGC should hold the signal level at the A/D converter essentially constant. In addition, adequate headroom at the A/D converter and D/A converter must be provided to avoid saturation during normally high peak-to-average voltage ratio periods of the desired signal. Strong signals that fall outside the narrowband digital filter bandwidth, but inside the analog IF translator bandwidths, can overload or saturate the A/D converter. This results in the generation of in-band IMD products and can result in significant degradation of the desired signal. If large signal levels are detected at the A/D converter, the receiver gain may have to be re-distributed by reducing the pre-conversion analog gain and increasing the digital gain to maintain the desired signal output level. This will, however, reduce the desired signal-to-quantization noise ratio. 1. The AGC attack time in a DSP-based radio is a trade-off. If it is too long, there is a risk that a fast-rising signal wave-front will over-range the ADC. If 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 it's "all 1's" point. Contrastingly, if the AGC attack time is too short, the AGC will clamp on a fast-rising signal wave-front or spike (as we have seen.) 2. As for inserting RF attenuation, it is perfectly acceptable on 7 MHz and below, as there is 10 to 12 dB of headroom to throw away. http://www.qsl.net/ab4oj/icom/dbmheaven.html This is derived from the well-known CCIR-670 curves. http://www.ab4oj.com/icom/nf.html In a receiver in which the NB and AGC are both DSP processes (7000, 7700, 7800) the NB can be used to advantage in certain cases to blank fast-rising spikes which might otherwise provoke AGC action and clamping. Pro3: The AGC (Automatic Gain Control) circuit reduces IF amplifier gain and attenuates IF signal to keep the audio output at a constant level. To realize a digital AGC, it is necessary to obtain the adjustment range for AGC gain internally in the DSP (*3), and to input both the desired signal and the interference signal into the A/D converter without them distorting (*4). For these reasons, Icom decided on a dynamic range for the A/D converter of at least 110dB, and approximately 120dB when the margin is taken into consideration. *3 To control the AGC attack response properly, it is necessary to adjust the gain even after the completion of IF filter processing. If the adjustment range of gain within the DSP is set to 60dB, it is necessary to obtain a wider dynamic range, as the noise floor is raised 60dB under full-gain conditions where AGC is not applied. *4 If the signal is distorted before entering the A/D converter, a distortion component may be mixed in the band. If it is mixed in the band, it is extremely difficult to remove it by post processing. The receiver gain is determined by the voltage on the AGC line (IC2461, pin 4). The D/A converter (IC2461) for AGC supplies control voltage to the AGC line and sets the receiver gain with the [RF/SQL] control. The 3rd IF signal from the level converter (IC2051) is detected at the AGC detector section in DSP IC (IC2001), and is applied to the D/A converter for AGC via the level converter (IC2052). The AGC voltage is amplified at the buffer amplifier (IC2471b), and is then applied to the MAIN-A unit via J2001 (pin 16) to control the AGC line. When receiving strong signals, the AGC voltage decreases via the buffer amplifier (IC2471b). As the AGC voltage is used for the bias voltage of the IF amplifier (RF-B unit; Q1751), and IF amplifier gain is decreased. IC-7800/7700: AGC loop management using the DSP unit The IC-7800 has two types of AGC loops. One of the AGC loops detects the AGC voltage at the BPF input in the DSP unit and feeds back to the 1st IF amplifier. This AGC loop prevents the saturation of the 1st IF amplifier from strong signals out of the BPF bandwidth, and improves the dynamic range against adjacent signals. The other AGC loop detects the AGC voltage at the digital IF filter output which has only passed the intended signal and draws the full potential from the digital IF filter. Combining the digital IF filter, manual notch, and the 1st IF stage, these are all controlled by the DSP unit. 110dB of ultra wide dynamic range in the receiver means the IF amplifier is distortion free from strong signals.