Friday, 15 November 2019

FT8 and 100hz ripple, a strange one for me

The general idea behind sound-card based data modes is that you transmit some tones from your computer using a transceiver, and then use the computer to decode the tones created by others which you receive in reply.

This makes things pretty easy technically. You need steady tones, stable radios and well adjusted sound levels. Modern equipment available to radio amateurs make the whole process fairly simple and reliable. Having said that, rather than amateur radio gear, the tones come from your computer sound card, and who knows about the purity or fidelity of that?

Neil, G4DBN, gave a useful talk on FT8 at the 2018 RSGB Convention. This was recorded by the RSGB and you can find it on YouTube here. If you use data modes I really do encourage you to take the time to watch Neil's presentation.

Neil suggests that it is a good idea to monitor your signal to make sure there is nothing odd going on.

Now, as it all (i.e. all those modes including FT8) relies on everybody being in the same receive and transmit filter bandwidths, the potential for an unwanted signal getting through are quite high. If you are transmitting a 500hz tone, you might generate a 1500hz harmonic, and as that is within a typical 2.7kHz or so filter, you will transmit that and it will be heard by everyone.

Luckily, that type of harmonic is not so common and you can solve it by raising your transmit frequency to 1500hz, whereupon that harmonic becomes 4500hz and is way outside the filter. It would still be there but it would never get transmitted and never get heard. If we ever go beyond the standard SSB-type filter to using really wideband signals then we will have to find a better solution than that, but then we will probably have such superb modes then that we won't care.

Returning to the real world, I thought it best to keep monitoring my digital signals and I still check from time to time. I was thinking that I might find a harmonic of the type I describe above, maybe capable of being resolved by altering the drive at some stage or other.

Oh dear. This latest test was not what I expected to see...
GM4FVM monitoring his own FT8 signal producing multiple decodes (click to enlarge).
In this instance you can see two parallel signals on the wide graph, each 100hz away from the main signal. Both of them are decoding and the worst is showing up as -30dB below the main signal. You can just about make out another pair 100hz further out.

The set-up I was using was transmitting into a dummy load on an IC-7100, receiving on an IC-9700. I could juggle it about, tx on the IC-9700, rx on the IC-7100, or between two IC-7100s, on 70cm, or 2m, and then using an IC-7300 and the 7100s, on 70MHz or 50MHz bands the results were broadly the same.

These are not harmonics of the wanted signals. They have nothing to do with over-modulating using the wanted signal. These are extra modulation products introduced either in the computer sound card or the transmitter.

Something was allowing 100hz into my system. The FT8 signal was then effectively being modulated at 100hz in what I imagine to be an AM-style (unwanted) method. The harmonics of the unwanted 100hz signal are producing mirror images 100hz away. All of this is at far higher levels of RF output than I consider to be acceptable. -30dB is FAR too strong for an in-band unwanted signal for me. In this case it is not just in-band, it is in-filter (if such a compound word exists). I was generating something that others could hear, and those others might be listening for weak signals and not liking what I was doing.

I consulted various people about how important this is and the consensus of opinion was that "it depends". "It isn't good but did anyone complain?" "Is it somehow appearing in the receiver and you am not transmitting it at all?" And several who have heard me say that there is nothing untoward visible or audible on my signal. This last comment is very comforting, but of course it only adds to my knowledge of my signal if they are hearing me 30dB above their threshold - if I am weaker than that they would not hear or see the unwanted signals. However, say during a Sporadic E opening when signals are strong, someone else might hear them loud and (not) clear.

Before going too far I checked to see that this was not an artifact turning up in the receiver. Using four radios, both ways, it was still there. And it was not on most signals received from other amateurs. The theory that this was in my transmitted signal was tested later and it certainly does not seem to be the receivers at fault.

Nevertheless, some received signals do have that effect on them and it was pretty easy to see that it was not introduced at my end. I have been on the receiving end of these hums being transmitted by others, but equally strong signals received at the same time are clean. Here is one from a station in Hungary ...
A certain HA station as decoded on 70MHz by GM4FVM in 2019
Note that as this guy faded out as shown by the trace at 13:15:30, the two unwanted signals faded not long before the main one, meaning that they were strong in relation to it. -18dB according to WSJT that day. This evidence made me more certain that I needed to eliminate my unwanted signals. Although some advice was that "if no locals could hear it then it didn't matter", I know that I can see such things on DX stations who are loud. And anyway, I do not want to radiate unwanted signals.

And these things can get worse. The evidence I produced suggested it changed depending on the radio in use and maybe just over time. Here is the worst I saw me manage, with 6 of my traces which were all decoded..
GM4FVM and his unwated FT8 friends captured at their worst by ... GM4FVM
In this case the strengths recorded for the unwanted signals relative to the wanted one by WSJT were -21, -29, -34, -42 and -43. This translates to the strongest unwanted one showing up as a signal of +01dB. As the wanted signal was +19, that means that during a Sporadic E opening similar to the one I saw the HA station, I would decode 3 of his traces and he would decode 6 of mine.

It just isn't good enough. Something must be done.

It does not really matter to me that other stations are doing it too, nor that no locals are complaining (there are no locals to complain anyway). I just had to clean up my signal.

Lacking any calibrated way to measure this, and recognising that some oddities might remain and pass through my filters, the target for fixing it was no decodable traces here apart from the wanted one. That suggests better than -43dB. Further, I wanted to get to no multiple traces visible at all, decodable or not, which looks like -50dB or more.

"ALARP" they called it during my time working with the railway industry. Any unwanted thing not capable of full control should be "as low as reasonably possible". Maybe I could not reach perfection, but I should try.

Turning back to G4DBN's RSGB lecture he quite correctly states that this type of effect is usually caused by 100hz ripple from switch-mode power supplies (SM-PSUs). It is well known that SM-PSUs can produce ripple on the DC output at twice the input AC frequency. Thus in parts of the world with 60hz mains supply this presumably causes traces with 120hz spacings.

My first line of thinking was that the station SM-PSUs must be generating 100hz ripple and that is getting into the radios. I changed all of the main SM-PSUs for analogue ones. There are two PSUs for the main rigs, one for the ancillaries (SWR meter, network radio,  Hy-Gain rotator controller ...) and one for the 70cms linear amplifier. No change. So they are in the clear. Despite this finding, I changed these around several times because I felt sure the fault lay there. No, it didn't.

Second were SM-PSUs which are not directly connected to the radios and are not what you would call "amateur grade" kit. Shop bought things. An early target was the computer, where any ripple reaching the sound card would explain the multiple traces. I changed the computer, but this made no difference. Then there were the computer displays, the supply for a multi-socket USB 3.0 outboard hub and finally one that produces 20 volts for the SPID rotator and 5V for the GPS standard. Nope, nothing.

By the end of this second phase I was beginning to think that the ripple must be generated somewhere else and be getting back down the mains into the equipment. This seemed highly unlikely. The whole idea of an SM-PSU is that it chops up the mains, generates higher frequency AC, and processes that. How 100hz hum would pass through  that I have no idea. For sure, SM-PSUs can generate their own 100hz hum, but can they allow 100hz hum coming down the mains to pass through?

However unlikely something is, if you have eliminated all the likely options, the unlikely ones have to be checked out. Sherlock would have loved me.

By this stage I had checked all the likely and all the unlikely options and nothing was the culprit. I still thought this must be something wrong with the FT8 signal going into the radio.

I should have realised sooner. The audio input graphical display on the IC-9700 and IC-7300 both showed clean audio coming out of the computer into the radios. This ripple had to be getting into the radios by some other means, but the penny had not dropped with me yet. That was a result I dismissed, and thus overlooked a clue.

Phase three had to be more radical. Even while I ignored a clue that proved it, I was beginning to think that this was getting into the radios by some odd means. If it was possible that this effect was not generated or passing through the computer then just monitoring a plain carrier would show no multiple traces. How could it? A plain carrier is unlikely to be modulated by 100hz hum coming from a computer which is turned off and disconnected.

So stage three saw one radio transmit a single carrier (CW key down or RTTY with no modulation), into a dummy load which was monitored by an SDR receiver with the trace observed on the waterfall. Both radios were powered from separate linear PSUs and entirely separate on both the DC and RF sides. Everything else was turned off and unplugged at the mains. Sure enough no multiple images. But then as I gradually connected various parts of the shack and powered them at the mains (there are three separate mains circuits), look what I saw ...
Tah-dah! The effect as seen on an otherwise unmodulated carrier.
When I connected the third mains circuit multiple traces appeared. In fact, I could toggle them on and off.

It took a while for this to make sense. Somehow a radio sending plain carrier and a receiver receiving it were producing or detecting 100hz modulation from some device not connected to them. The only obvious point of contact was the mains and that did not seem like a likely route (though possible). This had nothing to do with sound cards or FT8, though that was how I first noticed it. My computer was disconnected the whole time during phase three testing.

On then to determine which item on that circuit could be doing this. Obviously it must be ... nope, nothing obvious was doing it. The only thing left is the shack light, a fluorescent anglepoise which appears in my QRZ shack photo. As early mains powered LED lights were (radio) noisy I used a U-shaped fluorescent when the shack was built, and later replaced it was an identical one when the base broke. Could this be the cause? Surely not ...
Effect of light on and off as seen on the IC-9700 waterfall (click to enlarge if necessary).
Yes, the shack light was doing the damage. I will spare you the rest of the investigation. I tested with a reasonably close station on FT8 showed that with the light turned off they were seeing no sign of multiple traces. No amount of testing on my part showed any effect on received signals, so I am pretty sure (but not absolutely certain) that these multiple traces were generated on transmit  only. On testing on FT8, with any combination of radio (IC-7300, IC-9700 and two IC-7100s), turning the light on and off produced and removed the multiple traces. With the light off there were NO extra traces here, decoded or otherwise, visible or even partly visible, other than the main one.

The culprit
Gradually turning on the other SM-PSUs there were some slight bumps at 100hz intervals visible on the SDR waterfalls. They seemed to be -50dB or lower as I reached full operation. None of these were visible on the FT8 waterfall. I suppose these are the residual products of the switch mode process and I shall probably have to accept that. Some things, like computers and their display screens, would be hard to convert to analogue PSUs. I have to accept these unwanted outputs as ALARP, as low as I can make them. They have not gone away though, and will still be heard if my signal is very strong. I will return to the task of making them weaker later.

So the obvious thing to do is to turn off the shack anglepoise lamp. OK. I bought a cheap (£6) LED light with a flexible mount. This runs on 5 volts which is supplied by my (now tested) auxiliary PSU for the SPID rotator and the GPS receiver. So that is it sorted.

Or is it?

I am a radio amateur. I cannot just turn off the light without some investigation of how four separate radios managed to produce unwanted products without some explanation. How did it happen?

Also, as a radio amateur I am used to chasing round RF interference, clamping ferrites all over the place. And ferrites really do work very well against RF interference. But I am no expert on 100hz. I have read up that "electronic ballasts" on fluorescent lights, which have largely replaced the "electric" ballasts and starters, can indeed generate 100hz ripple. But how does this get into my radios?
The ballast for the lamp was held in a plastic box which broke up when opened to reveal this

Searching the Internet for the ballast using the reference number finds the manufacturer and the product name but anything else comes up as "404 error, page not found". I suspect that the BS EN standard referred to relates to double insulated (non-grounded) power supplies. Anyway I cannot get into the ballast itself and I only have generic circuit diagrams to work out what it is. Whilst it certainly might produce 100hz ripple, the means of transmission remains unclear to me.

The only suggestion I can put forward is that somehow this hum is being transmitted inductively. My set-up has lots of potential for hum loops. The radios are all grounded in multiple ways which are hard to disentangle - via the linears, at the power supplies, via the computer, via the SWR meter. And then the computer and the SWR meter are grounded together via the computer cables anyway ... and so it goes on. There is a rat's nest of DC wiring to provide lots of scope to pick up some low frequency energy and share it around via various routes with different impedances. Then again it might simply be passing through the mains (though I cannot really see how).

Even if that is the case, how can it be getting into the rigs? All four rigs in fact? I cannot test removing the DC supply cable, but I did remove the USB lead and the GPS frequency standard for the IC-9700. It cannot be getting down the coax, surely.

I am left with the impression that the lamp is leaking 100hz ripple down its mains lead (not grounded of course), and then inducing a similar ripple in the radio DC leads, or in the other DC connections and thus linking indirectly to the radio DC leads. In then gets inside the radios where it must be doing some mischief. How that modulation occurs I have no idea.

The inevitable consequence of all this is that there must a be a lot of energy coming out of the lamp down the mains cable, which also seems intuitively wrong. But then, touching the lamp holders of my bedside lamp in the days of coiled fluorsecents allowed me to feel the 100hz ripple via my skin. Those lamp ballasts must be a big inductive load.

Anyway, it has gone now. Banished from the house into the garage where it can do no harm. I did try it on a different circuit in the house but the effect was the same. As I cannot be in the garage and the shack the same time I doubt if it will cause any harm out there.

As for the future, there are lots of "mains conditioner" type multi-socket boards designed to cope with this type of thing. They are mostly aimed at musicians using amplifiers during stage shows. The drawback is that they generally have no technical specifications worth the title. They claim to stop mains hum, but do they? Even then, they will not stop anything generated in the PSU and passed on down the DC line into the radio, though they would stop anything radiating through the mains (if they work at all, that is).

I am left feeling uncomfortable. After 10 days of heaving about power supplies and digging under the bench I have an answer to "what" but no answer to "why". The lamp and the immediate effects are gone, but my equipment has been shown to be susceptible to a form of interference against which I have no real defence.

The inquiring mind is never satisfied.




No comments:

Post a Comment