Date
1 - 20 of 21
Locked FST4 modes with TS-890S - slow RF envelope rise problem #txaudio
|
Hi Bill,
Many thanks for that - I was about 75% of the way there yesterday but it was getting late and the neural fade was pronounced :) Anyway, I am happy to report that, having set the NOSHAPING environment variable, my Class D amplifier is happy. No hiccuping at the FST4(W) transmission start, even with the hardware signal level gate disconnected. I'll give it a more thorough waveform checkup later but it looks good at this point. I guess that with time and experience the switch may find its way to the WSJT-X GUI but it works well enough in the present form. Thanks again to you and the team, and 73, Peter. |
|
|
|
On 05/02/2021 15:53, Peter Hall, VK6HP
wrote:
Thank you to the WSJT-X team for incorporating a "no shaping" option into v2.3.0. It will be greatly appreciated by many FST4 and FST4W users. Reading the release notes, the comment is: Hi Peter, there are several ways you can do it. Environment variables can be set in MS WIndows at the system level, at the user level, and in each process. You probably want the user level option which will set it for all processes started by your user. You set them in the "Control panel->System->Advanced system settings->Environment variables" panel, just add a new user variable called FST4_NOSHAPING and give it a value of 1. 73 |
|
|
|
Thank you to the WSJT-X team for incorporating a "no shaping" option into v2.3.0. It will be greatly appreciated by many FST4 and FST4W users. Reading the release notes, the comment is:
FST4/W: Disable envelope shaping at start and end of transmission when environment variable FST4_NOSHAPING=1. Works for fst4sim too. Not being a Windows power user, I wonder if someone might tell me the easiest way of setting the required environment variable? My WSJT-X 64-bit installation is pretty standard, with default paths etc. The v2.3.0 manual gave some clues about the use of environment variables for other applications but I must admit that a quick search hasn't turned up the variable list I was expecting. 73, and thanks, Peter. |
|
|
|
N1BUG
Well said, Peter.
toggle quoted message
Show quoted text
Karl, I was not suggesting it be changed for one person. There are already several affected by this and the FST4(W) mode is new, not even out in a GA release yet. There will be more who run into issues with this. Neither is it just one amplifier design that is or will be affected, but many. I find it ironic that waveform shaping seems to have appeared for the first time in WSJT-X (I am not aware of older modes that use it) in modes designed for LF and MF where the majority of amplifiers cannot reproduce the shaping and may have other problems as a result of it. This kind of waveform shaping might make sense on bands where most people use linear amplifiers. That's just not the case on the bands this mode was targeted at. 73, Paul On 1/6/2021 10:52 PM, Peter Hall, VK6HP wrote:
Karl, |
|
|
|
Karl,
If you'll permit me to comment, it's not a matter of revising the software to work with one design. LF/MF users, where the FST4 modes are targeted, very often use highly efficient, non-linear amplifiers (such as Class D or E) with the prevailing digital and CW modes. Within these common designs the proper place to control RF envelope shaping is in the amplifier, not the exciter and, indeed, the usual topologies cannot reflect any exciter waveform shaping in their outputs. Furthermore, not only is exciter waveform shaping wasted, it's counterproductive in a spectral containment sense if it leads to some amplifiers emitting spurious transients. While I strongly suspect the present heavy envelope shaping can be reduced to keep everyone happy (as they are with WSPR, JT9, ...), another simple solution is to have the option to allow LF/MF users to retain or disable shaping, catering to both linear PA users and the large body of non-linear amplifier users. In the end, there's no point making life difficult for a large fraction of the target user base, especially if the ambition is to have the FST4 modes replace the earlier modes. 73, Peter. |
|
|
|
PAUL -
I hope you are not suggesting that the widely accepted WSJT-X software should be revised to work with one specific homebrew amplifier circuit design. In the long run that would set an unpleasant precedent for future software development efforts. Why not put some emphasis on the amateur radio station operator to learn how to interface to the RF deck he/she elects to use, or on those who design the RF transverter chain? -- Karl WA8NVW OH WA8NVW@... in WSJTX@groups.io |
|
|
|
Hi everyone,
First, should say that I have now submitted a WSJT-X bug report to the developers' list and perhaps when the team has had a chance to consider it we may hear back from them. Paul, on the matter of the eBay simple noise gate circuit, I can say that it does work, although it is not as sophisticated as Rik's digital implementation. It's the same general concept but uses a JFET as an attenuator element in response to a control signal from an envelope detector. There is no hysteresis but setting the threshold high enough works out well in practice. I bought the blank PCB from the Polish seller simply because I was curious as to whether I could make it work at 475 kHz. The actual circuit (MXR Noise Gate) has been around in one form or another for ages and is well-known in the music community. You can find details of the PCB and a link to the schematic etc by searching for eBay item 254048537744. I made two changes to allow good 475 kHz operation: the dual op-amp U1 is now an NE5532 (from my parts drawer) and the gain of the envelope detector is reduced (and its bandwidth increased) by changing R7 from 4k7 to 33 k ohm. For RF applications use a 51 ohm input termination and make R20 (the output resistor following the unity gain buffer) 51 ohm. I outline the option as a starting point, not as a fully-tested solution. It does drive my PA fine, and you can see the general action on the snapshot of the beginning of an FST4-60 transmission below, 73, Peter.  |
|
|
|
Rik Strobbe
Hi Paul, we are getting off topic but in short:
The circuitry around IC4 is an inverting comparator with hysteresis (see https://www.ti.com/lit/an/snoa997a/snoa997a.pdf?ts=1609852900466&ref_url=https%253A%252F%252Fwww.google.com%252F). Via IC5 it starts passing the signal to the MOSFET drivers (IC6, IC7) as soon a the amplitude at the input (rectified and amplified by IC3) exceeds a "high value" (eg 300mV) and keeps it passing until the amplitude drops below a "low value" (eg 100mV). That way there are not instabilities during a slow rising of falling signal. |
|
|
|
N1BUG
Hi all,
Rik, thanks for sharing your driver circuit. I can't really say I understand it yet, but I will save it for reference. Peter, I would be interested to see your first circuit and also the one you are now working on. My FET driver has an enable pin. It will not provide gate drive to the FETs until that pin is pulled low. I have just been leaving it hard wired low, as that works for all modes except FST4(W). I also 'works' for these modes but I have alarming waveform glitches, especially at the start and end of transmissions. I was thinking about sampling the RF drive, rectifying it, and feeding a comparator with reference level set to pull the driver enable pin low when the input waveform has reached sufficient amplitude to be 'safe'. This may not be the best approach but for now it is what I know how to do. Let's hope it's not too late for the developers to consider a change to the software, as that would be the best solution. 73, Paul N1BUG |
|
|
|
Rik Strobbe
Hi Paul, all,
toggle quoted message
Show quoted text
as could be expected the signal has the same (or better said mirror) shape at the end of a transmission. I just checked that for FST4-60 and I assume that applies also to FST4W and other cycle durations.  Best solution would be that the rise time / fall time can be set by the users, either a user selected value or at least an "on/off" option. In case you want to go for a hardware solution, here the circuit diagram of my driver:
 The circuitry around IC4 takes care of the hysteresis, levels are set by R11 and R12. In case the inserted images got lost, they are attached as well.
73, Rik ON7YD - OR7T
________________________________________ Van: main@WSJTX.groups.io <main@WSJTX.groups.io> namens N1BUG <paul@...>
Verzonden: maandag 4 januari 2021 20:01 Aan: main@WSJTX.groups.io Onderwerp: Re: [WSJTX] FST4 modes with TS-890S - slow RF envelope rise problem Hello Rik, Chris, Peter, The different rise times depending on T/R period might explain why some appear to be OK for me while others do not. Very strange things can happen when you start putting some FET drivers into undefined states. I say appear because surely all of them are transitioning through an undefined state for some period of time, it's just that I don't see the symptoms with all T/R periods. Has anyone checked to see if the envelope tapers slowly at the end of a transmission? I would guess it does. I have lost many FETs in the past due to slow envelope decay putting the driver into undefined states. It has not been happening much since I upgraded to higher voltage and current rated FETs but it is still bad for them and perhaps other components also. We should be protecting our PAs from slow envelope rise and slow decay times. I guess a properly designed circuit should do both easily enough. I don't know how to design such a circuit but I it seems I better start working on it. 73, Paul N1BUG On 1/4/21 9:54 AM, Rik Strobbe wrote: > Hello Chris, Paul, Peter, > > > I'm afraid I made a mistake in my earlier measurement: instead of > waiting for the next half hour for FST4-1800 to start I just pressed > "tune". Here I measured 0 rise time. > > Today I had a look at the rise times for the real transmission start > (rather than waiting 30 minutes I changed to PC clock) : > FST4(W)-1800 = 2.5s > FST4(W)-900= 1.2s > FST4(W)-300= 0.4s > FST4(W)-120= 0.15s > FST4(W)-60= 0.075s > Rise times seems to be identical for FST4 and FST4W. > I assume that these slow rise times are meant to avoid excessive > bandwidth at the start (and end?) of a transmission. But these can > indeed cause problems with class D amplifiers as used by the majority on > 630m and 2200m. > Since I did blow the final power MOSFETS a few times in the past due to > the slow rise times in other software and therefore added a electronic > "hysteresis switch" between TX and PA to avoid these kind of problems. > So I didn't notice the slow rise time when using FST4. > But the better solution would be an option in wsjt-x to set the rise > time to 0 (as it is with JT9). > > > 73, Rik ON7YD - OR7T > > > ------------------------------------------------------------------------ > *Van:* main@WSJTX.groups.io <main@WSJTX.groups.io> namens Peter Hall, > VK6HP <p.hall@...> > *Verzonden:* maandag 4 januari 2021 15:30 > *Aan:* main@WSJTX.groups.io > *Onderwerp:* Re: [WSJTX] FST4 modes with TS-890S - slow RF envelope rise > problem > Chris, Paul > > Thank you for your comments, which I think indicate there is an issue to > be at least understood. Chris, your waveforms look much like mine from > the TS-890S. I don't know if you had a chance to check out the URL to > the data provided by Rik, ON7YD, in the earlier RSGB-LF group discussion > which I initiated. Rik seems to have a simpler soundcard (rather than > an internal transceiver interface) and does not see the effect I > described. However, with several transceiver-based systems showing the > effect, it would be useful to have a discussion with the WSJT-X > developers. Being new to this group, I'm not entirely sure whether > issues get picked up as a matter of course from these forums or whether > a more direct report is required. Perhaps Paul might know the answer to > that question. > > My "noise-gate" is a hardware solution and, in retrospect, it's not such > a bad thing to have a circuit which only clocks the amplifier when the > drive is above a certain level. However, it'd obviously be useful to > correct any software issue, if only to allow users of very simple > non-linear PAs to use them reliably with the new modes. > > Paul, I've never had any issues with WSPR or JT9 either, and it sounds > as though your PA is also unhappy about a very slow rising clock > amplitude ambling through a transition point at the beginning of the > transmission. I can't explain your difference between the long and > short duration transitions, though; all my transmissions in the new > modes hiccup at the start but I've had good contacts using transmit > durations from 15 sec to 900 sec. and local test decodes at up to 1800 sec. > > 73, Peter. > > > > 
|
|
|
|
Hello everyone,
Thanks for the continuing interest. I've consolidated a few responses and comments below. First, it might be useful to point out that my Class D PA is built around the G0MRF module, with details of my implementation described in a short document at http://users.tpg.com.au/adslif89/ph/vk6hp/MF_amp/VK6HP_630m_Amplifier.pdf Rik, many thanks for the correction and update - very useful. If I extrapolate the rise times to the FST4-15 mode, we might expect a rise time of about 20 ms, which seems excessive for spectral containment. Perhaps we can verify that number but, in any case, it's clear that the envelope rise on the longer transmissions is very slow indeed. Most of my measurements were done with FST4W-120, for a direct comparison with WSPR. With that mode set, I see rise time numbers very similar to the ones mentioned by you and Chris. Paul, the tail of the transmission is also extended to a similar extent. I suspect that, in fact, my T/R sequencing (which effectively removes drive to the PA) may be activating prior to the trailing envelope causing any issues. Regarding the hardware "noise-gate" circuit, I am using a JFET as a level activated switch. Actually, I have a second (similar) circuit just about to be tested now. It's based on a guitar noise gate (!) PCB available on eBay and, if you're interested, I can post details and results when I've fired it up and done some tests. Chris, I suspect the fault condition in your friend's G0MRF PA is "over-current" rather than "reverse power". The two conditions get "OR-d" to produce a timed shutdown of the PA drive. In practice, and depending on the load, one can also trigger the other. Reino, it'd certainly be interesting to have a look at the tone-spaced carrier. My experience of those modes is in a digital exciter where, for 475 kHz operation (for example), the carrier is at 950 kHz and the tones spaced x2, producing the correct 475 kHz WSPR drive for a push-pull PA after dividing by 2. However, the transceiver systems most commonly used (Kenwood, Flex, ....) produce a 475 kHz signal at low level, normally requiring an external doubler prior to the PA divider. Such transceivers don't generally allow the generation of a 950 kHz carrier. In fact, I am testing a PA input circuit based on an LT1016 fast comparator which produces very symmetrical square waves, over a range of levels, from a 475 kHz signal. This removes the need for the doubler in the push-pull PA but it's largely irrelevant in the present discussion. On balance, I think Rik's take on the fix is correct: allow the WSJT-X user to select whether envelope shaping is used for the FST4 modes. It's probably true that the majority of LF/MF power amplifiers are non-linear types and the proper place to control the envelope of such systems is in the PA, not the exciter. For example, see the CW conditioning arrangements in the G0MRF amplifier shown in my earlier URL. 73, Peter. |
|
|
|
Reino Talarmo
Hi All, Is there any rise time information available, when Tone spacing is set to “x 2” or “x 4”? To my understanding those are intended for TX implementations, where signal frequency is divided by those values. Most probably in those cases digital processing is simpler without any signal slow ramp up/down. 73, Reino OH3mA
From: main@WSJTX.groups.io [mailto:main@WSJTX.groups.io] On Behalf Of Rik Strobbe
Sent: 4. tammikuuta 2021 16:54 To: main@WSJTX.groups.io Subject: Re: [WSJTX] FST4 modes with TS-890S - slow RF envelope rise problem
Hello Chris, Paul, Peter,
I'm afraid I made a mistake in my earlier measurement: instead of waiting for the next half hour for FST4-1800 to start I just pressed "tune". Here I measured 0 rise time. Today I had a look at the rise times for the real transmission start (rather than waiting 30 minutes I changed to PC clock) :
73, Rik ON7YD - OR7T
Van: main@WSJTX.groups.io <main@WSJTX.groups.io> namens Peter Hall, VK6HP <p.hall@...>
Chris, Paul |
|
|
|
N1BUG
Hello Rik, Chris, Peter,
toggle quoted message
Show quoted text
The different rise times depending on T/R period might explain why some appear to be OK for me while others do not. Very strange things can happen when you start putting some FET drivers into undefined states. I say appear because surely all of them are transitioning through an undefined state for some period of time, it's just that I don't see the symptoms with all T/R periods. Has anyone checked to see if the envelope tapers slowly at the end of a transmission? I would guess it does. I have lost many FETs in the past due to slow envelope decay putting the driver into undefined states. It has not been happening much since I upgraded to higher voltage and current rated FETs but it is still bad for them and perhaps other components also. We should be protecting our PAs from slow envelope rise and slow decay times. I guess a properly designed circuit should do both easily enough. I don't know how to design such a circuit but I it seems I better start working on it. 73, Paul N1BUG On 1/4/21 9:54 AM, Rik Strobbe wrote:
Hello Chris, Paul, Peter, |
|
|
|
G4SDG
Thanks Rik. I was just about to post asking someone to double check as I'd seen today that the delay was related to the mode transmission length and I was beginning to doubt my measurements!
I can understand the desire to reduce interference but surely that solution would be the same regardless of the transmission length? I received an email from someone else who has also had issues with a G0MRF based FET amplifier with FST4 modes, in his case triggering SWR protection (and blowing one FET) which hadn't happened in JT9 mode. 73 Chris G4SDG |
|
|
|
Rik Strobbe
Hello Chris, Paul, Peter,
I'm afraid I made a mistake in my earlier measurement: instead of waiting for the next half hour for FST4-1800 to start I just pressed "tune". Here I measured 0 rise time. Today I had a look at the rise times for the real transmission start (rather
than waiting 30 minutes I changed to PC clock) :
73, Rik ON7YD - OR7T
Van: main@WSJTX.groups.io <main@WSJTX.groups.io> namens Peter Hall, VK6HP <p.hall@...>
Verzonden: maandag 4 januari 2021 15:30 Aan: main@WSJTX.groups.io Onderwerp: Re: [WSJTX] FST4 modes with TS-890S - slow RF envelope rise problem Chris, Paul
Thank you for your comments, which I think indicate there is an issue to be at least understood. Chris, your waveforms look much like mine from the TS-890S. I don't know if you had a chance to check out the URL to the data provided by Rik, ON7YD, in the earlier RSGB-LF group discussion which I initiated. Rik seems to have a simpler soundcard (rather than an internal transceiver interface) and does not see the effect I described. However, with several transceiver-based systems showing the effect, it would be useful to have a discussion with the WSJT-X developers. Being new to this group, I'm not entirely sure whether issues get picked up as a matter of course from these forums or whether a more direct report is required. Perhaps Paul might know the answer to that question. My "noise-gate" is a hardware solution and, in retrospect, it's not such a bad thing to have a circuit which only clocks the amplifier when the drive is above a certain level. However, it'd obviously be useful to correct any software issue, if only to allow users of very simple non-linear PAs to use them reliably with the new modes. Paul, I've never had any issues with WSPR or JT9 either, and it sounds as though your PA is also unhappy about a very slow rising clock amplitude ambling through a transition point at the beginning of the transmission. I can't explain your difference between the long and short duration transitions, though; all my transmissions in the new modes hiccup at the start but I've had good contacts using transmit durations from 15 sec to 900 sec. and local test decodes at up to 1800 sec. 73, Peter. |
|
|
|
Chris, Paul
Thank you for your comments, which I think indicate there is an issue to be at least understood. Chris, your waveforms look much like mine from the TS-890S. I don't know if you had a chance to check out the URL to the data provided by Rik, ON7YD, in the earlier RSGB-LF group discussion which I initiated. Rik seems to have a simpler soundcard (rather than an internal transceiver interface) and does not see the effect I described. However, with several transceiver-based systems showing the effect, it would be useful to have a discussion with the WSJT-X developers. Being new to this group, I'm not entirely sure whether issues get picked up as a matter of course from these forums or whether a more direct report is required. Perhaps Paul might know the answer to that question. My "noise-gate" is a hardware solution and, in retrospect, it's not such a bad thing to have a circuit which only clocks the amplifier when the drive is above a certain level. However, it'd obviously be useful to correct any software issue, if only to allow users of very simple non-linear PAs to use them reliably with the new modes. Paul, I've never had any issues with WSPR or JT9 either, and it sounds as though your PA is also unhappy about a very slow rising clock amplitude ambling through a transition point at the beginning of the transmission. I can't explain your difference between the long and short duration transitions, though; all my transmissions in the new modes hiccup at the start but I've had good contacts using transmit durations from 15 sec to 900 sec. and local test decodes at up to 1800 sec. 73, Peter. |
|
|
|
N1BUG
On 1/3/2021 4:32 PM, G4SDG wrote:
I've also successfully used the FST4 modes but am distinctly uneasy about the effect the trips might have on my FETS and/or power supply! It's odd that nobody else from the LF/MF fraternity has mentioned it though.This is interesting. I was part of the early FST4/W test group but I never realized this ramped amplitude was happening. In general I would say it is not good news with class D and E amplifiers which were never designed for amplitude varying waveforms. I've been having problems with FST4/W transmissions with T/R period of 120s or less. Sometimes, but not always, my class D amplifier gets into an abnormal state with significant oscillation in output amplitude immediately at the start of a transmission and never recovers unless I ramp the drain voltage down drastically and then bring it back up. I never mentioned it because others didn't seem to be having issues and I thought it was something unique to my equipment. Now I don't know what to think. I have never had it happen with WSPR or JT9. What seems really strange is that FST4/W seems OK at T/R periods of 300s and up. So I'm not sure if my problem is related to this or not. 73, Paul N1BUG |
|
|
|
Hello Peter,
I've checked the audio waveform out of the Intel NUC PC and the FST4W-120 start is indeed much slower to rise (~140mS) in comparison with WSPR. In the shots below I've altered the timebase to get the full rise of the FST4W onto the screen. It's the only difference that I can see between the two modes at start-up. My amplifier is based on the K3MF design here https://njdtechnologies.net/k3mf-2200-meter-amplifier/ with the FT-817 running at 10.136MHz and followed by a mixer with a GPS locked 10MHz LO all of which works fine on WSPR but trips almost every time with FST4W. I guess VOX might be a solution but at the moment I'm using the FT-817 in data mode with the audio connected to the rear which is my preferred set-up. I've also successfully used the FST4 modes but am distinctly uneasy about the effect the trips might have on my FETS and/or power supply! It's odd that nobody else from the LF/MF fraternity has mentioned it though. 73 Chris G4SDG |
|
|
|
Hi Chris
Thank you for the information. It's very useful to know of a similar problem, especially since you're using an entirely different exciter. I condensed the story a bit for a concise description of the issue but I actually use a doubler after the transceiver output, followed by a divide-by-two flip-flop to give very symmetrical drive waveforms to the MCP1404 FET drivers and push-pull HEXFETs. With the FST4 modes, the (Schmitt) FF clocking becomes intermittent as the RF envelope spends a long time rising through the switching point, activating the protection circuit until things have settled down. However, using the segmented memory triggering options on my scope to account for delays, I can quite clearly see the problem starts with the envelope shaping of the waveform emerging from the TS-890S. I did wonder if the rise time issue may be a product of an over-zealous attempt to contain "key clicks" with fast FST4 but I come back to the demonstration by Rick, ON7YD, with his bare bones sound-card system: https://groups.io/g/rsgb-lf-group/topic/77212765#11958 However, perhaps if there's a capable transceiver selected, there's an attempt made to slow things down. Just my speculation, of course. Anyway, with the FST4 modes aimed at LF/MF users, it'd obviously be good to have those modes compatible with typical hardware. 73, Peter. |
|
|
|
G4SDG
Hello Peter,
I'm using a Class E amplifier and am having similar issues to you. I've not investigated deeply so far and have no problems with WSPR but on FST4 modes the power supply often momentarily trips at the beginning of a sequence. When seen on an analyser the carrier rise from the FT-817 driver rig seems to be stepped in level which perhaps upsets the TC4427 FET driver chip. 73 Chris G4SDG |
|
|