Re: [Elecraft] K3 hardware AGC experiment shows promise
Wayne Burdick wrote: Joe Subich, W4TV wrote: Remember, IMD is driven by total *peak* signal level. The peak signal level is not simply the scalar sum of the levels ... it is a vector sum as the individual signals add in phase. And those peaks could be slightly modulating the AGC loop. Maybe that should say modulating the AGC loops (plural)? The HAGC loop and the DSP AGC loop interact strongly - especially around the HAGC threshold - and this creates the possibility for each loop to modulate the other. When the HAGC begins to act, we know that the firmware automatically adjusts the DSP AGC to give consistent audio levels and correct S-meter readings. The firmware monitors for HAGC action by digitizing the voltage on the HAGC1 line, but there has to be some delay before it can become aware of any change. The dynamics of this process create the possibility for the HAGC loop and the DSP AGC loop to modulate each other. As far as I'm aware, this feature is unique to the K3, which may help to explain the rig-specific nature of the problem. If this is a valid model, the effects would predictably be worst around the HAGC threshold where the response is inherently nonlinear. The dynamic interaction between the two loops would be quite sensitive to the rate(s) of amplitude variation in the composite incoming signal, as individual components are keyed on and off. It would also be affected by added noise, due to its dependency on the instantaneous vector sum of *all* incoming signal, noise and interference components hitting the HAGC detector (as identified by W4TV). The mathematics of two cascaded feedback loops are way beyond me, but the benefit of increasing the HAGC time constant by a very large factor (10-20) sounds at lot like avoiding competition between the dominant poles in the two loops. But the same model also explains why this wouldn't be a complete cure. This is all speculation, because much depends on hidden detail in the DSP AGC firmware... but might it make some sense? -- 73 from Ian GM3SEK http://www.ifwtech.co.uk/g3sek __ Elecraft mailing list Home: http://mailman.qth.net/mailman/listinfo/elecraft Help: http://mailman.qth.net/mmfaq.htm Post: mailto:Elecraft@mailman.qth.net This list hosted by: http://www.qsl.net Please help support this email list: http://www.qsl.net/donate.html
Re: [Elecraft] K3 hardware AGC experiment shows promise
Joe, You may recall that the subject of K3 Mush was discussed on the K3 Contesters Forum earlier this year, and that I suggested that the Group Delay Variations introduced by the narrower bandwidth roofing filters might well be the cause of the Mush problem. The recordings of K3's audio sent to me by two owners of the K3 supported this thought, as did the plots made by another K3 owner of the Delay vs. frequency introduced by roofers of differing bandwidths. A second thought, which I could not post before leaving Scotland, was that the termination of the roofing filter would seem to vary when the HAGC becomes activeThis raises two further points: 1) What effect does an active HAGC have on a roofer's Group Delay Variations? 2) What effect does an active HAGC have on a roofer's Output Odd Order Intercepts (in and outside of the filter's passband), and does this increase the level of filter generated IMD products enough to create audible mush? Unfortunately adding a simple resistive pad to isolate the roofer's output from the HAGC's IF circuitry could increase the overall Noise Figure of the receiver too much. It is my understanding that not all K3s exhibit this mush problem. This does not surprise me because the Delay characteristics of the roofers are probably not specified to keep their cost within affordable limits - i.e. some K3 roofers of x bandwith might exhibit fairly constant delay vs. frequency, others might be quite wild. 73, Geoff LX2AO On Monday, December 05, 2011 3:36 AM, Joe Subich W4TV wrote: Remember, IMD is driven by total *peak* signal level. The peak signal level is not simply the scalar sum of the levels ... it is a vector sum as the individual signals add in phase. W8FN explained some of the constraints in terms of peak to average power of multiple CDMA transmit signals and IMD ... the same applies to receive signals and the peak voltage handling capability in the IF amplifier/ADC. Six to 10 S4-S5 signals can create a greater peak voltage (if just for the shortest time) than a single S9+40 signal. __ Elecraft mailing list Home: http://mailman.qth.net/mailman/listinfo/elecraft Help: http://mailman.qth.net/mmfaq.htm Post: mailto:Elecraft@mailman.qth.net This list hosted by: http://www.qsl.net Please help support this email list: http://www.qsl.net/donate.html
Re: [Elecraft] K3 hardware AGC experiment shows promise
wayne burdick wrote Hi all, The K3's hardware AGC time constant is a compromise between recovery time and IMD due to modulation of the loop. C238, on the bottom of the RF board near the front, sets this time constant. We did a little experiment today . . . I think it would be informative to monitor whether the HAGC is indeed kicking in during the variety of pileup mush type events not involving very strong signals in the passband. For example, could a very visible LED be set up external to the rig that would flash when HAGC kicks in? A few of us could operate like this in an upcoming contest (e.g., next weekend 10M contest) and confirm whether there's correlation between HAGC and pileup mush. Barry N1EU -- View this message in context: http://elecraft.365791.n2.nabble.com/Fw-CQ-WW-DX-Contest-CW-2011-Unofficial-claimed-scores-tp7060148p7062683.html Sent from the Elecraft mailing list archive at Nabble.com. __ Elecraft mailing list Home: http://mailman.qth.net/mailman/listinfo/elecraft Help: http://mailman.qth.net/mmfaq.htm Post: mailto:Elecraft@mailman.qth.net This list hosted by: http://www.qsl.net Please help support this email list: http://www.qsl.net/donate.html
Re: [Elecraft] K3 hardware AGC experiment shows promise ( KE7X config)
Hi Wayne, I'd be happy to try the modification. I can do it with one K3 and not with the other. For reference, here are my configurations and general operation settings when in C6. AF GAIN = high. (I'm thinking maybe to set it low) AF LIM 20 AGC DCY SOFT AGC HLD 0.30 AGC PLS NOR AGC SLP 0 AGC THR 8 AGC-F 120 AGC-S 020 SER NUM 1148 RX EQ 1 -16 RX EQ 2 -16 RX EQ 3 0 RX EQ 4 +6 RX EQ 5 +6 RX EQ 6 0 RX EQ 7 -10 RX EQ 8 -16 When operating, ATT is ALWAYS ON. I don't think I ever took it off except maybe when 15 was dead. PRE never on. RF gain - all over the lot - say between 9 and 3. When there were lots of big signals I'd crank it back and then forget to turn it up again and would CQ in the face of weak callers :-( AF Gain - all over the lot too. I have 250, 400, 1000 and 2.7 kHz filters. DSP _usually_ in the 500 - 700 range, sometimes a little wider, sometimes a little narrower. I don't think I ever went below 300 Hz. AFX never on. I don't' like it but it might be interesting to try. I use Bose QC2 headphones. I wear hearing aids. Most mods have been done to the K3s including the stiffeners on the synthesizers and I have swapped out the DSP boards. I have not done the HAGC mod. 73, Fred KE7X Fred Cady fcady at ieee dot org -Original Message- From: elecraft-boun...@mailman.qth.net [mailto:elecraft- boun...@mailman.qth.net] On Behalf Of Wayne Burdick Sent: Sunday, December 04, 2011 6:59 PM To: Elecraft Reflector Subject: Re: [Elecraft] K3 hardware AGC experiment shows promise By the way, we'll be happy to send out these surface-mount 1 uF capacitors (or whatever we determine is the best value) to anyone who would like to try the experimental mod, immediately and at no charge. These are huge as SMD parts go. Anyone with a fine-tip iron would have no trouble putting one in. There's no need to take the original, smaller one off; you can just stack the new one on top. If you're in a hurry, you could use any type of capacitor that will fit, including an electrolytic. Just be sure to put the (-) end at ground if you use a polarized capacitor. 73, Wayne N6KR On Dec 4, 2011, at 5:50 PM, Wayne Burdick wrote: The K3's hardware AGC time constant is a compromise between recovery time and IMD due to modulation of the loop. C238, on the bottom of the RF board near the front, sets this time constant. We did a little experiment today __ Elecraft mailing list Home: http://mailman.qth.net/mailman/listinfo/elecraft Help: http://mailman.qth.net/mmfaq.htm Post: mailto:Elecraft@mailman.qth.net This list hosted by: http://www.qsl.net Please help support this email list: http://www.qsl.net/donate.html __ Elecraft mailing list Home: http://mailman.qth.net/mailman/listinfo/elecraft Help: http://mailman.qth.net/mmfaq.htm Post: mailto:Elecraft@mailman.qth.net This list hosted by: http://www.qsl.net Please help support this email list: http://www.qsl.net/donate.html
Re: [Elecraft] K3 hardware AGC experiment shows promise ( KE7X config)
Just for comparison and I don't have the K3 that we used unpacked yet so I don't know the agc settings. I use AFX all the time. ATT was almost never engaged. I never use RX equalization. RF gain was used mostly on the low bands. When I had Bose headphones I noticed that they were quite prone to pickup RF. Filters: 200 400 1000 1.8 2.8 Sub 400 and 2.8 On 12/5/11 6:09 AM, Cady, Fred wrote: Hi Wayne, I'd be happy to try the modification. I can do it with one K3 and not with the other. For reference, here are my configurations and general operation settings when in C6. AF GAIN = high. (I'm thinking maybe to set it low) AF LIM 20 AGC DCY SOFT AGC HLD 0.30 AGC PLS NOR AGC SLP 0 AGC THR 8 AGC-F 120 AGC-S 020 SER NUM 1148 RX EQ 1 -16 RX EQ 2 -16 RX EQ 3 0 RX EQ 4 +6 RX EQ 5 +6 RX EQ 6 0 RX EQ 7 -10 RX EQ 8 -16 When operating, ATT is ALWAYS ON. I don't think I ever took it off except maybe when 15 was dead. PRE never on. RF gain - all over the lot - say between 9 and 3. When there were lots of big signals I'd crank it back and then forget to turn it up again and would CQ in the face of weak callers :-( AF Gain - all over the lot too. I have 250, 400, 1000 and 2.7 kHz filters. DSP _usually_ in the 500 - 700 range, sometimes a little wider, sometimes a little narrower. I don't think I ever went below 300 Hz. AFX never on. I don't' like it but it might be interesting to try. I use Bose QC2 headphones. I wear hearing aids. Most mods have been done to the K3s including the stiffeners on the synthesizers and I have swapped out the DSP boards. I have not done the HAGC mod. 73, Fred KE7X Fred Cady fcady at ieee dot org __ Elecraft mailing list Home: http://mailman.qth.net/mailman/listinfo/elecraft Help: http://mailman.qth.net/mmfaq.htm Post: mailto:Elecraft@mailman.qth.net This list hosted by: http://www.qsl.net Please help support this email list: http://www.qsl.net/donate.html
Re: [Elecraft] K3 hardware AGC experiment shows promise
Nice idea, Barry, but probably continuously flashing something on the display should be an easier solution from a firmware mod point of view. Maybe flashing the 'F' of the AGC on the display, or something similar. That should be straightforward and doesn't require any physical modification to the rig. Ciao, Fabio IZ4AFW - NZ1W - HI9/IZ4AFW Il 05/12/2011 13.29, Barry N1EU ha scritto: I think it would be informative to monitor whether the HAGC is indeed kicking in during the variety of pileup mush type events not involving very strong signals in the passband. For example, could a very visible LED be set up external to the rig that would flash when HAGC kicks in? A few of us could operate like this in an upcoming contest (e.g., next weekend 10M contest) and confirm whether there's correlation between HAGC and pileup mush. Barry N1EU __ Elecraft mailing list Home: http://mailman.qth.net/mailman/listinfo/elecraft Help: http://mailman.qth.net/mmfaq.htm Post: mailto:Elecraft@mailman.qth.net This list hosted by: http://www.qsl.net Please help support this email list: http://www.qsl.net/donate.html
Re: [Elecraft] K3 hardware AGC experiment shows promise ( KE7X config)
Fred's configuration struck me as quite surprising. With DSP at 5-700, his K3 would be using the 1000 kHz roofing filter, so there would be a substantial frequency range blocked only by the DSP. And ATT on -- why would this be attractive? I'd think to hear weak signals close together, you'd turn off ATT, turn up RF gain, and squeeze down to a narrow roofing filter. Wouldn't that make the signals sharper and more legible? I wonder if this choice of operating settings is the reason not everyone hears the mushy effect. Peter W0LLN On Mon, Dec 5, 2011 at 7:09 AM, Cady, Fred fc...@ece.montana.edu wrote: When operating, ATT is ALWAYS ON. I don't think I ever took it off except maybe when 15 was dead. PRE never on. I have 250, 400, 1000 and 2.7 kHz filters. DSP _usually_ in the 500 - 700 range, sometimes a little wider, sometimes a little narrower. I don't think I ever went below 300 Hz. __ Elecraft mailing list Home: http://mailman.qth.net/mailman/listinfo/elecraft Help: http://mailman.qth.net/mmfaq.htm Post: mailto:Elecraft@mailman.qth.net This list hosted by: http://www.qsl.net Please help support this email list: http://www.qsl.net/donate.html
Re: [Elecraft] K3 hardware AGC experiment shows promise ( KE7Xconfig)
I wish I had a 700 Hz filter because I like the DSP in the 400-700 region. My reason for this is that when a narrower filter is in use, it tends to have many signals with about the same tone. At least for my hearing, I need and like to have a spectra of audio signals. Sometimes I get the lower tone ones and sometimes the higher ones. Also, notwithstanding the reverse beacon phenomenon, there are usually callers who are off frequency enough that a narrow filter will miss them. Incidentally, some of us suspect the reverse beacon spots are not the issue that people think. We observed the pileups got much bigger when clusters spots showed up. We suspect the RBN spots are used by contesters and general dxers are still using and favoring the DX clusters. Fred Cady fcady at ieee dot org -Original Message- From: elecraft-boun...@mailman.qth.net [mailto:elecraft- boun...@mailman.qth.net] On Behalf Of Peter Wollan Sent: Monday, December 05, 2011 11:45 AM To: Elecraft Reflector Subject: Re: [Elecraft] K3 hardware AGC experiment shows promise ( KE7Xconfig) Fred's configuration struck me as quite surprising. With DSP at 5-700, his K3 would be using the 1000 kHz roofing filter, so there would be a substantial frequency range blocked only by the DSP. And ATT on -- why would this be attractive? I'd think to hear weak signals close together, you'd turn off ATT, turn up RF gain, and squeeze down to a narrow roofing filter. Wouldn't that make the signals sharper and more legible? I wonder if this choice of operating settings is the reason not everyone hears the mushy effect. Peter W0LLN On Mon, Dec 5, 2011 at 7:09 AM, Cady, Fred fc...@ece.montana.edu wrote: When operating, ATT is ALWAYS ON. I don't think I ever took it off except maybe when 15 was dead. PRE never on. I have 250, 400, 1000 and 2.7 kHz filters. DSP _usually_ in the 500 - 700 range, sometimes a little wider, sometimes a little narrower. I don't think I ever went below 300 Hz. __ Elecraft mailing list Home: http://mailman.qth.net/mailman/listinfo/elecraft Help: http://mailman.qth.net/mmfaq.htm Post: mailto:Elecraft@mailman.qth.net This list hosted by: http://www.qsl.net Please help support this email list: http://www.qsl.net/donate.html __ Elecraft mailing list Home: http://mailman.qth.net/mailman/listinfo/elecraft Help: http://mailman.qth.net/mmfaq.htm Post: mailto:Elecraft@mailman.qth.net This list hosted by: http://www.qsl.net Please help support this email list: http://www.qsl.net/donate.html
Re: [Elecraft] K3 hardware AGC experiment shows promise
Should we call ordering to have a cap or two sent out? Fred Cady fcady at ieee dot org -Original Message- From: elecraft-boun...@mailman.qth.net [mailto:elecraft- boun...@mailman.qth.net] On Behalf Of Wayne Burdick Sent: Sunday, December 04, 2011 6:59 PM To: Elecraft Reflector Subject: Re: [Elecraft] K3 hardware AGC experiment shows promise By the way, we'll be happy to send out these surface-mount 1 uF capacitors (or whatever we determine is the best value) to anyone who would like to try the experimental mod, immediately and at no charge. These are huge as SMD parts go. Anyone with a fine-tip iron would have no trouble putting one in. There's no need to take the original, smaller one off; you can just stack the new one on top. If you're in a hurry, you could use any type of capacitor that will fit, including an electrolytic. Just be sure to put the (-) end at ground if you use a polarized capacitor. 73, Wayne N6KR On Dec 4, 2011, at 5:50 PM, Wayne Burdick wrote: The K3's hardware AGC time constant is a compromise between recovery time and IMD due to modulation of the loop. C238, on the bottom of the RF board near the front, sets this time constant. We did a little experiment today __ Elecraft mailing list Home: http://mailman.qth.net/mailman/listinfo/elecraft Help: http://mailman.qth.net/mmfaq.htm Post: mailto:Elecraft@mailman.qth.net This list hosted by: http://www.qsl.net Please help support this email list: http://www.qsl.net/donate.html __ Elecraft mailing list Home: http://mailman.qth.net/mailman/listinfo/elecraft Help: http://mailman.qth.net/mmfaq.htm Post: mailto:Elecraft@mailman.qth.net This list hosted by: http://www.qsl.net Please help support this email list: http://www.qsl.net/donate.html
Re: [Elecraft] K3 hardware AGC experiment shows promise ( KE7X config)
It's always a good idea to present the lowest-level of RF energy to the receiver input. If you can hear the band noise, any more gain at the RF is wasted and can result in various issues with strong signals near or in the passband. On the lower HF bands, the band noise is usually perfectly audible with the ATTN on. I find that the narrowest filter seldom improves copy on a weak CW signal and sometimes makes it worse. The narrower the filter, the softer the attack and decay of each CW element becomes. I only use tight filters to kill nearby QRM that threatens to swamp the weak signal or when the signal is so weak that I must suppress the background band noise to the greatest degree possible. In those situations the other station may need to QRS for good copy. Any other time I find it easiest to copy with a bandwidth of 500 or even 1000 Hz and, if only one station is nearby, use the notch filter to remove that signal. 73, Ron AC7AC -Original Message- Fred's configuration struck me as quite surprising. With DSP at 5-700, his K3 would be using the 1000 kHz roofing filter, so there would be a substantial frequency range blocked only by the DSP. And ATT on -- why would this be attractive? I'd think to hear weak signals close together, you'd turn off ATT, turn up RF gain, and squeeze down to a narrow roofing filter. Wouldn't that make the signals sharper and more legible? I wonder if this choice of operating settings is the reason not everyone hears the mushy effect. Peter W0LLN __ Elecraft mailing list Home: http://mailman.qth.net/mailman/listinfo/elecraft Help: http://mailman.qth.net/mmfaq.htm Post: mailto:Elecraft@mailman.qth.net This list hosted by: http://www.qsl.net Please help support this email list: http://www.qsl.net/donate.html
Re: [Elecraft] K3 hardware AGC experiment shows promise
No. We don't have a large stock at the factory, as we keep them at our contract manufacturing site south of here. I'll post here when we get some in here. 73, Eric www.elecraft.com On 12/5/2011 11:50 AM, Cady, Fred wrote: Should we call ordering to have a cap or two sent out? Fred Cady fcady at ieee dot org -Original Message- From: elecraft-boun...@mailman.qth.net [mailto:elecraft- boun...@mailman.qth.net] On Behalf Of Wayne Burdick Sent: Sunday, December 04, 2011 6:59 PM To: Elecraft Reflector Subject: Re: [Elecraft] K3 hardware AGC experiment shows promise By the way, we'll be happy to send out these surface-mount 1 uF capacitors (or whatever we determine is the best value) to anyone who would like to try the experimental mod, immediately and at no charge. These are huge as SMD parts go. Anyone with a fine-tip iron would have no trouble putting one in. There's no need to take the original, smaller one off; you can just stack the new one on top. If you're in a hurry, you could use any type of capacitor that will fit, including an electrolytic. Just be sure to put the (-) end at ground if you use a polarized capacitor. 73, Wayne N6KR On Dec 4, 2011, at 5:50 PM, Wayne Burdick wrote: The K3's hardware AGC time constant is a compromise between recovery time and IMD due to modulation of the loop. C238, on the bottom of the RF board near the front, sets this time constant. We did a little experiment today __ Elecraft mailing list Home: http://mailman.qth.net/mailman/listinfo/elecraft Help: http://mailman.qth.net/mmfaq.htm Post: mailto:Elecraft@mailman.qth.net This list hosted by: http://www.qsl.net Please help support this email list: http://www.qsl.net/donate.html __ Elecraft mailing list Home: http://mailman.qth.net/mailman/listinfo/elecraft Help: http://mailman.qth.net/mmfaq.htm Post: mailto:Elecraft@mailman.qth.net This list hosted by: http://www.qsl.net Please help support this email list: http://www.qsl.net/donate.html __ Elecraft mailing list Home: http://mailman.qth.net/mailman/listinfo/elecraft Help: http://mailman.qth.net/mmfaq.htm Post: mailto:Elecraft@mailman.qth.net This list hosted by: http://www.qsl.net Please help support this email list: http://www.qsl.net/donate.html
Re: [Elecraft] K3 hardware AGC experiment shows promise
Wayne, For those of us ordering the cap through Mouser or DigiKey, can you reply with the SMD/SMT cap dimension code (e.g., 0603, 1608, etc.) Tnx! Paul, W9AC - Original Message - From: Wayne Burdick n...@elecraft.com To: Elecraft Reflector elecraft@mailman.qth.net Sent: Sunday, December 04, 2011 8:50 PM Subject: [Elecraft] K3 hardware AGC experiment shows promise Hi all, The K3's hardware AGC time constant is a compromise between recovery time and IMD due to modulation of the loop. C238, on the bottom of the RF board near the front, sets this time constant. We did a little experiment today (thanks to Tree, N6TR) that suggests increasing the size of C238 substantially might be a worthy change. In the case of signals just large enough to tickle the hardware AGC, the first IMD products were reduced by something like 18 dB. This will also increase the recovery time for very strong signals, so the jury is out on whether this is OK for the average user. For the experimentally inclined: C238 is easy to get to; just remove bottom cover A (the front half). C238 is a large-ish surface mount capacitor nestled between two 20-pin connectors. The present value is 0.1 uF. Tree tacked a 1-uF cap on top of it. Then he tacked another one on, which improved things by another few dB. Some ops have mentioned problems with signals much lower than this, which has always baffled me. But I got to thinking: Suppose you're listening to a bunch of S4-S5 signals in your DSP passband. You could have larger signals outside the DSP, but inside the crystal filter. Or you could have clicks from strong signals that get inside the crystal filter but you can't hear because you're using a narrow DSP filter. Or you could have noise spikes. Any of these could ping the hardware AGC just enough to cause IMD between all of the signals in the passband. My point is that increasing the loop time constant could have a more general benefit when a band is busy and/or noisy. Let me know if you try this and whether the results are of interest. (I live in an RF-free zone, it seems, so I can never recreate the problem here. Frustrating!) 73, Wayne N6KR __ Elecraft mailing list Home: http://mailman.qth.net/mailman/listinfo/elecraft Help: http://mailman.qth.net/mmfaq.htm Post: mailto:Elecraft@mailman.qth.net This list hosted by: http://www.qsl.net Please help support this email list: http://www.qsl.net/donate.html __ Elecraft mailing list Home: http://mailman.qth.net/mailman/listinfo/elecraft Help: http://mailman.qth.net/mmfaq.htm Post: mailto:Elecraft@mailman.qth.net This list hosted by: http://www.qsl.net Please help support this email list: http://www.qsl.net/donate.html
Re: [Elecraft] K3 hardware AGC experiment shows promise
0805. tnx Wayne On Dec 5, 2011, at 4:31 PM, Paul Christensen wrote: Wayne, For those of us ordering the cap through Mouser or DigiKey, can you reply with the SMD/SMT cap dimension code (e.g., 0603, 1608, etc.) Tnx! Paul, W9AC - Original Message - From: Wayne Burdick n...@elecraft.com To: Elecraft Reflector elecraft@mailman.qth.net Sent: Sunday, December 04, 2011 8:50 PM Subject: [Elecraft] K3 hardware AGC experiment shows promise Hi all, The K3's hardware AGC time constant is a compromise between recovery time and IMD due to modulation of the loop. C238, on the bottom of the RF board near the front, sets this time constant. We did a little experiment today (thanks to Tree, N6TR) that suggests increasing the size of C238 substantially might be a worthy change. In the case of signals just large enough to tickle the hardware AGC, the first IMD products were reduced by something like 18 dB. This will also increase the recovery time for very strong signals, so the jury is out on whether this is OK for the average user. For the experimentally inclined: C238 is easy to get to; just remove bottom cover A (the front half). C238 is a large-ish surface mount capacitor nestled between two 20-pin connectors. The present value is 0.1 uF. Tree tacked a 1-uF cap on top of it. Then he tacked another one on, which improved things by another few dB. Some ops have mentioned problems with signals much lower than this, which has always baffled me. But I got to thinking: Suppose you're listening to a bunch of S4-S5 signals in your DSP passband. You could have larger signals outside the DSP, but inside the crystal filter. Or you could have clicks from strong signals that get inside the crystal filter but you can't hear because you're using a narrow DSP filter. Or you could have noise spikes. Any of these could ping the hardware AGC just enough to cause IMD between all of the signals in the passband. My point is that increasing the loop time constant could have a more general benefit when a band is busy and/or noisy. Let me know if you try this and whether the results are of interest. (I live in an RF-free zone, it seems, so I can never recreate the problem here. Frustrating!) 73, Wayne N6KR __ Elecraft mailing list Home: http://mailman.qth.net/mailman/listinfo/elecraft Help: http://mailman.qth.net/mmfaq.htm Post: mailto:Elecraft@mailman.qth.net This list hosted by: http://www.qsl.net Please help support this email list: http://www.qsl.net/donate.html __ Elecraft mailing list Home: http://mailman.qth.net/mailman/listinfo/elecraft Help: http://mailman.qth.net/mmfaq.htm Post: mailto:Elecraft@mailman.qth.net This list hosted by: http://www.qsl.net Please help support this email list: http://www.qsl.net/donate.html __ Elecraft mailing list Home: http://mailman.qth.net/mailman/listinfo/elecraft Help: http://mailman.qth.net/mmfaq.htm Post: mailto:Elecraft@mailman.qth.net This list hosted by: http://www.qsl.net Please help support this email list: http://www.qsl.net/donate.html
[Elecraft] K3 hardware AGC experiment shows promise
Hi all, The K3's hardware AGC time constant is a compromise between recovery time and IMD due to modulation of the loop. C238, on the bottom of the RF board near the front, sets this time constant. We did a little experiment today (thanks to Tree, N6TR) that suggests increasing the size of C238 substantially might be a worthy change. In the case of signals just large enough to tickle the hardware AGC, the first IMD products were reduced by something like 18 dB. This will also increase the recovery time for very strong signals, so the jury is out on whether this is OK for the average user. For the experimentally inclined: C238 is easy to get to; just remove bottom cover A (the front half). C238 is a large-ish surface mount capacitor nestled between two 20-pin connectors. The present value is 0.1 uF. Tree tacked a 1-uF cap on top of it. Then he tacked another one on, which improved things by another few dB. Some ops have mentioned problems with signals much lower than this, which has always baffled me. But I got to thinking: Suppose you're listening to a bunch of S4-S5 signals in your DSP passband. You could have larger signals outside the DSP, but inside the crystal filter. Or you could have clicks from strong signals that get inside the crystal filter but you can't hear because you're using a narrow DSP filter. Or you could have noise spikes. Any of these could ping the hardware AGC just enough to cause IMD between all of the signals in the passband. My point is that increasing the loop time constant could have a more general benefit when a band is busy and/or noisy. Let me know if you try this and whether the results are of interest. (I live in an RF-free zone, it seems, so I can never recreate the problem here. Frustrating!) 73, Wayne N6KR __ Elecraft mailing list Home: http://mailman.qth.net/mailman/listinfo/elecraft Help: http://mailman.qth.net/mmfaq.htm Post: mailto:Elecraft@mailman.qth.net This list hosted by: http://www.qsl.net Please help support this email list: http://www.qsl.net/donate.html
Re: [Elecraft] K3 hardware AGC experiment shows promise
By the way, we'll be happy to send out these surface-mount 1 uF capacitors (or whatever we determine is the best value) to anyone who would like to try the experimental mod, immediately and at no charge. These are huge as SMD parts go. Anyone with a fine-tip iron would have no trouble putting one in. There's no need to take the original, smaller one off; you can just stack the new one on top. If you're in a hurry, you could use any type of capacitor that will fit, including an electrolytic. Just be sure to put the (-) end at ground if you use a polarized capacitor. 73, Wayne N6KR On Dec 4, 2011, at 5:50 PM, Wayne Burdick wrote: The K3's hardware AGC time constant is a compromise between recovery time and IMD due to modulation of the loop. C238, on the bottom of the RF board near the front, sets this time constant. We did a little experiment today __ Elecraft mailing list Home: http://mailman.qth.net/mailman/listinfo/elecraft Help: http://mailman.qth.net/mmfaq.htm Post: mailto:Elecraft@mailman.qth.net This list hosted by: http://www.qsl.net Please help support this email list: http://www.qsl.net/donate.html
Re: [Elecraft] K3 hardware AGC experiment shows promise
This is the kinda stuff that makes me so glad that I purchased an Elecraft product. With which other Ham radio gear manufacturer do you get this kind of support huh?. Well done Wayne, Eric and the whole Elecraft crew. 73 de Jeff Cochrane - VK4BOF East Innisfail QLD, Australia K3 #4257, P3#1629, KPA-500 #161 - Original Message - From: Wayne Burdick n...@elecraft.com To: Elecraft Reflector elecraft@mailman.qth.net Sent: Monday, December 05, 2011 11:50 AM Subject: [Elecraft] K3 hardware AGC experiment shows promise Hi all, The K3's hardware AGC time constant is a compromise between recovery time and IMD due to modulation of the loop. C238, on the bottom of the RF board near the front, sets this time constant. We did a little experiment today (thanks to Tree, N6TR) that suggests increasing the size of C238 substantially might be a worthy change. In the case of signals just large enough to tickle the hardware AGC, the first IMD products were reduced by something like 18 dB. This will also increase the recovery time for very strong signals, so the jury is out on whether this is OK for the average user. For the experimentally inclined: C238 is easy to get to; just remove bottom cover A (the front half). C238 is a large-ish surface mount capacitor nestled between two 20-pin connectors. The present value is 0.1 uF. Tree tacked a 1-uF cap on top of it. Then he tacked another one on, which improved things by another few dB. Some ops have mentioned problems with signals much lower than this, which has always baffled me. But I got to thinking: Suppose you're listening to a bunch of S4-S5 signals in your DSP passband. You could have larger signals outside the DSP, but inside the crystal filter. Or you could have clicks from strong signals that get inside the crystal filter but you can't hear because you're using a narrow DSP filter. Or you could have noise spikes. Any of these could ping the hardware AGC just enough to cause IMD between all of the signals in the passband. My point is that increasing the loop time constant could have a more general benefit when a band is busy and/or noisy. Let me know if you try this and whether the results are of interest. (I live in an RF-free zone, it seems, so I can never recreate the problem here. Frustrating!) 73, Wayne N6KR __ Elecraft mailing list Home: http://mailman.qth.net/mailman/listinfo/elecraft Help: http://mailman.qth.net/mmfaq.htm Post: mailto:Elecraft@mailman.qth.net This list hosted by: http://www.qsl.net Please help support this email list: http://www.qsl.net/donate.html __ Elecraft mailing list Home: http://mailman.qth.net/mailman/listinfo/elecraft Help: http://mailman.qth.net/mmfaq.htm Post: mailto:Elecraft@mailman.qth.net This list hosted by: http://www.qsl.net Please help support this email list: http://www.qsl.net/donate.html
Re: [Elecraft] K3 hardware AGC experiment shows promise
Some ops have mentioned problems with signals much lower than this, which has always baffled me. Remember, IMD is driven by total *peak* signal level. The peak signal level is not simply the scalar sum of the levels ... it is a vector sum as the individual signals add in phase. W8FN explained some of the constraints in terms of peak to average power of multiple CDMA transmit signals and IMD ... the same applies to receive signals and the peak voltage handling capability in the IF amplifier/ADC. Six to 10 S4-S5 signals can create a greater peak voltage (if just for the shortest time) than a single S9+40 signal. It's been 25 years since I was exposed to the math behind this (as part of work in dealing with multiplexed power amplifiers for analog TV) but it would be instructive to convert the average power levels of a bunch of S4-S5 signals to P-P voltage, sum those P-P values and convert the peak voltage back to power levels to see the true impact. Solid state IF amplifiers and particularly DACs don't have the soft compression characteristics we were used to with remote cut-off pentodes or even dual gate MOSFETs in the analog receivers of previous generations. 73, ... Joe, W4TV On 12/4/2011 8:50 PM, Wayne Burdick wrote: Hi all, The K3's hardware AGC time constant is a compromise between recovery time and IMD due to modulation of the loop. C238, on the bottom of the RF board near the front, sets this time constant. We did a little experiment today (thanks to Tree, N6TR) that suggests increasing the size of C238 substantially might be a worthy change. In the case of signals just large enough to tickle the hardware AGC, the first IMD products were reduced by something like 18 dB. This will also increase the recovery time for very strong signals, so the jury is out on whether this is OK for the average user. For the experimentally inclined: C238 is easy to get to; just remove bottom cover A (the front half). C238 is a large-ish surface mount capacitor nestled between two 20-pin connectors. The present value is 0.1 uF. Tree tacked a 1-uF cap on top of it. Then he tacked another one on, which improved things by another few dB. Some ops have mentioned problems with signals much lower than this, which has always baffled me. But I got to thinking: Suppose you're listening to a bunch of S4-S5 signals in your DSP passband. You could have larger signals outside the DSP, but inside the crystal filter. Or you could have clicks from strong signals that get inside the crystal filter but you can't hear because you're using a narrow DSP filter. Or you could have noise spikes. Any of these could ping the hardware AGC just enough to cause IMD between all of the signals in the passband. My point is that increasing the loop time constant could have a more general benefit when a band is busy and/or noisy. Let me know if you try this and whether the results are of interest. (I live in an RF-free zone, it seems, so I can never recreate the problem here. Frustrating!) 73, Wayne N6KR __ Elecraft mailing list Home: http://mailman.qth.net/mailman/listinfo/elecraft Help: http://mailman.qth.net/mmfaq.htm Post: mailto:Elecraft@mailman.qth.net This list hosted by: http://www.qsl.net Please help support this email list: http://www.qsl.net/donate.html __ Elecraft mailing list Home: http://mailman.qth.net/mailman/listinfo/elecraft Help: http://mailman.qth.net/mmfaq.htm Post: mailto:Elecraft@mailman.qth.net This list hosted by: http://www.qsl.net Please help support this email list: http://www.qsl.net/donate.html
Re: [Elecraft] K3 hardware AGC experiment shows promise
Joe Subich, W4TV wrote: Remember, IMD is driven by total *peak* signal level. The peak signal level is not simply the scalar sum of the levels ... it is a vector sum as the individual signals add in phase. And those peaks could be slightly modulating the AGC loop. Wayne __ Elecraft mailing list Home: http://mailman.qth.net/mailman/listinfo/elecraft Help: http://mailman.qth.net/mmfaq.htm Post: mailto:Elecraft@mailman.qth.net This list hosted by: http://www.qsl.net Please help support this email list: http://www.qsl.net/donate.html
Re: [Elecraft] K3 hardware AGC experiment shows promise
This is why I bought an Elecraft. Good luck getting this type of support with IKY. I would try the mod but since I haven't noticed the issue, I don't think it makes sense for me to try. We were not digging signals out of the mud in J6 except for 160 but nobody mentioned any issues. On 12/4/11 6:50 PM, Wayne Burdick wrote: Hi all, The K3's hardware AGC time constant is a compromise between recovery time and IMD due to modulation of the loop. C238, on the bottom of the RF board near the front, sets this time constant. We did a little experiment today (thanks to Tree, N6TR) that suggests increasing the size of C238 substantially might be a worthy change. In the case of signals just large enough to tickle the hardware AGC, the first IMD products were reduced by something like 18 dB. This will also increase the recovery time for very strong signals, so the jury is out on whether this is OK for the average user. For the experimentally inclined: C238 is easy to get to; just remove bottom cover A (the front half). C238 is a large-ish surface mount capacitor nestled between two 20-pin connectors. The present value is 0.1 uF. Tree tacked a 1-uF cap on top of it. Then he tacked another one on, which improved things by another few dB. Some ops have mentioned problems with signals much lower than this, which has always baffled me. But I got to thinking: Suppose you're listening to a bunch of S4-S5 signals in your DSP passband. You could have larger signals outside the DSP, but inside the crystal filter. Or you could have clicks from strong signals that get inside the crystal filter but you can't hear because you're using a narrow DSP filter. Or you could have noise spikes. Any of these could ping the hardware AGC just enough to cause IMD between all of the signals in the passband. My point is that increasing the loop time constant could have a more general benefit when a band is busy and/or noisy. Let me know if you try this and whether the results are of interest. (I live in an RF-free zone, it seems, so I can never recreate the problem here. Frustrating!) 73, Wayne N6KR __ Elecraft mailing list Home: http://mailman.qth.net/mailman/listinfo/elecraft Help: http://mailman.qth.net/mmfaq.htm Post: mailto:Elecraft@mailman.qth.net This list hosted by: http://www.qsl.net Please help support this email list: http://www.qsl.net/donate.html __ Elecraft mailing list Home: http://mailman.qth.net/mailman/listinfo/elecraft Help: http://mailman.qth.net/mmfaq.htm Post: mailto:Elecraft@mailman.qth.net This list hosted by: http://www.qsl.net Please help support this email list: http://www.qsl.net/donate.html
Re: [Elecraft] K3 hardware AGC experiment shows promise
I should point out that the very common habit of not using the ATT or even turning on the PRE on lower bands would be commonly putting this circuit in this condition. Perhaps that is why there is a contingent of us that never seems to have the problems reported. Constantly running the RF end at max gain is bad for other reasons besides. Since the DSP AGC treats the outcome of the hardware AGC and the RF string switch settings as propagation, why would the hang time matter? If it was me, I would be working on something in firmware that did NOT leave it to the user to remember to turn off PRE and turn on ATT based on the signal levels being slammed into the hardware AGC. 73, Guy. On Sun, Dec 4, 2011 at 8:50 PM, Wayne Burdick n...@elecraft.com wrote: Hi all, The K3's hardware AGC time constant is a compromise between recovery time and IMD due to modulation of the loop. C238, on the bottom of the RF board near the front, sets this time constant. We did a little experiment today (thanks to Tree, N6TR) that suggests increasing the size of C238 substantially might be a worthy change. In the case of signals just large enough to tickle the hardware AGC, the first IMD products were reduced by something like 18 dB. This will also increase the recovery time for very strong signals, so the jury is out on whether this is OK for the average user. For the experimentally inclined: C238 is easy to get to; just remove bottom cover A (the front half). C238 is a large-ish surface mount capacitor nestled between two 20-pin connectors. The present value is 0.1 uF. Tree tacked a 1-uF cap on top of it. Then he tacked another one on, which improved things by another few dB. Some ops have mentioned problems with signals much lower than this, which has always baffled me. But I got to thinking: Suppose you're listening to a bunch of S4-S5 signals in your DSP passband. You could have larger signals outside the DSP, but inside the crystal filter. Or you could have clicks from strong signals that get inside the crystal filter but you can't hear because you're using a narrow DSP filter. Or you could have noise spikes. Any of these could ping the hardware AGC just enough to cause IMD between all of the signals in the passband. My point is that increasing the loop time constant could have a more general benefit when a band is busy and/or noisy. Let me know if you try this and whether the results are of interest. (I live in an RF-free zone, it seems, so I can never recreate the problem here. Frustrating!) 73, Wayne N6KR __ Elecraft mailing list Home: http://mailman.qth.net/mailman/listinfo/elecraft Help: http://mailman.qth.net/mmfaq.htm Post: mailto:Elecraft@mailman.qth.net This list hosted by: http://www.qsl.net Please help support this email list: http://www.qsl.net/donate.html __ Elecraft mailing list Home: http://mailman.qth.net/mailman/listinfo/elecraft Help: http://mailman.qth.net/mmfaq.htm Post: mailto:Elecraft@mailman.qth.net This list hosted by: http://www.qsl.net Please help support this email list: http://www.qsl.net/donate.html
Re: [Elecraft] K3 hardware AGC experiment shows promise
I find those comments exceptionally interesting, Joe. a. it might help explain why keyed CW signals seem (in my experience, at least) to be a bigger issue .. more yanking around of the AGC. b. it might help explain why at least some of us feel that the issue is worse for weaker signals. The AGC is definitely less linear at the low end (I use K8ZOA's web site for reference), but if signals were uniformly weak that wouldn't be such an issue since weak signals wouldn't have much level excursion anyway. But if the P - P effect you describe is repeatedly dragging the AGC across that really non-linear portion of the curve I would think it could cause a lot of mixing. 73, Dave AB7E On 12/4/2011 7:36 PM, Joe Subich, W4TV wrote: Some ops have mentioned problems with signals much lower than this, which has always baffled me. Remember, IMD is driven by total *peak* signal level. The peak signal level is not simply the scalar sum of the levels ... it is a vector sum as the individual signals add in phase. W8FN explained some of the constraints in terms of peak to average power of multiple CDMA transmit signals and IMD ... the same applies to receive signals and the peak voltage handling capability in the IF amplifier/ADC. Six to 10 S4-S5 signals can create a greater peak voltage (if just for the shortest time) than a single S9+40 signal. It's been 25 years since I was exposed to the math behind this (as part of work in dealing with multiplexed power amplifiers for analog TV) but it would be instructive to convert the average power levels of a bunch of S4-S5 signals to P-P voltage, sum those P-P values and convert the peak voltage back to power levels to see the true impact. Solid state IF amplifiers and particularly DACs don't have the soft compression characteristics we were used to with remote cut-off pentodes or even dual gate MOSFETs in the analog receivers of previous generations. 73, ... Joe, W4TV On 12/4/2011 8:50 PM, Wayne Burdick wrote: Hi all, The K3's hardware AGC time constant is a compromise between recovery time and IMD due to modulation of the loop. C238, on the bottom of the RF board near the front, sets this time constant. We did a little experiment today (thanks to Tree, N6TR) that suggests increasing the size of C238 substantially might be a worthy change. In the case of signals just large enough to tickle the hardware AGC, the first IMD products were reduced by something like 18 dB. This will also increase the recovery time for very strong signals, so the jury is out on whether this is OK for the average user. For the experimentally inclined: C238 is easy to get to; just remove bottom cover A (the front half). C238 is a large-ish surface mount capacitor nestled between two 20-pin connectors. The present value is 0.1 uF. Tree tacked a 1-uF cap on top of it. Then he tacked another one on, which improved things by another few dB. Some ops have mentioned problems with signals much lower than this, which has always baffled me. But I got to thinking: Suppose you're listening to a bunch of S4-S5 signals in your DSP passband. You could have larger signals outside the DSP, but inside the crystal filter. Or you could have clicks from strong signals that get inside the crystal filter but you can't hear because you're using a narrow DSP filter. Or you could have noise spikes. Any of these could ping the hardware AGC just enough to cause IMD between all of the signals in the passband. My point is that increasing the loop time constant could have a more general benefit when a band is busy and/or noisy. Let me know if you try this and whether the results are of interest. (I live in an RF-free zone, it seems, so I can never recreate the problem here. Frustrating!) 73, Wayne N6KR __ Elecraft mailing list Home: http://mailman.qth.net/mailman/listinfo/elecraft Help: http://mailman.qth.net/mmfaq.htm Post: mailto:Elecraft@mailman.qth.net This list hosted by: http://www.qsl.net Please help support this email list: http://www.qsl.net/donate.html __ Elecraft mailing list Home: http://mailman.qth.net/mailman/listinfo/elecraft Help: http://mailman.qth.net/mmfaq.htm Post: mailto:Elecraft@mailman.qth.net This list hosted by: http://www.qsl.net Please help support this email list: http://www.qsl.net/donate.html __ Elecraft mailing list Home: http://mailman.qth.net/mailman/listinfo/elecraft Help: http://mailman.qth.net/mmfaq.htm Post: mailto:Elecraft@mailman.qth.net This list hosted by: http://www.qsl.net Please help support this email list: http://www.qsl.net/donate.html
