After reading on this reflector and elsewhere about avoiding out-of-band transmission I wonder if someone could clarify how the emissions are for CW and SSB on modern rigs. I’ve written this as statements I believe to be true (and to be challenged) rather than as a load of questions.
I think I understand the SSB case. Take transmitting in USB mode, for example, on the 3kHz-wide UK 60m sub-band, 5.4035 to 5.4065 MHz. Operators set the rig’s dial frequency to the lowest frequency in the sub-band (5.4035) so that their (suppressed-carrier) upper sideband occupies approx. 200 to 2600 Hz up from that frequency.
I assume most or all rigs transmit CW as double sideband with full carrier. My friend looked at the CW output from his TS590 on a spectrum analyzer and confirmed the symmetric distribution about the carrier. That suggests the dial frequency must be set at least half the total CW bandwidth up from the lower edge of the band or sub-band. For example, for SOTA CW ops I use the 5.3540 – 5.3580 sub-band (which overlaps with WRC-15) and I set my dial (carrier) frequency to 5.3545.
My rigs have CW and CW-Reverse. My understanding is switching between them affects only receiving (equivalent to LSB and USB), i.e. whether the sidetone is set above or below the received signal – but, when transmitting, it doesn’t affect the relationship between dial frequency and the transmitted signal (which is DSB anyway). Is that correct?
BTW: in case it matters, I have a FT857, FT817 and KX2.
As a DXer as well as SOTA Chaser on the 60m band and an FT8 MGM user, I am careful to avoid transmitting a data signal above 5357.95 KHz, which is the UK band limit for a 50 Hz wide FT8 signal. It is fairly common to see UK stations operating in USB DATA Mode more than 1000 Hz above 5357 KHz - a situation recently exacerbated by DXpeditions operating in Fox mode on 5357.300 KHz (approx) and selectively listening (due to the software design) 1000-2000 Hz above their TX freq.
The WSJT-X and JTDX MGM software in F/H Mode will only resolve Hounds calling 1000 Hz above their TX frequency (above 5358.3 KHz), before answering specific callsigns and then moving the Hound station to their working frequency of 5357.3 JHz (approx). This is tempting some UK DXers to call above 5358 KHz - out of band. Polite emails I sent to a few of them reminding them of our bandlet limit are usually answered and are well received.
Interesting subject! My FT991A has not only “USB-CW” and “LSB-CW” but also a menu selection to stop the dial readout changing when you go from SSB to CW. What this all means in terms of actual transmit frequency relative to readout is going to take a bit of work with a frequency meter to determine - it certainly doesn’t seem to be made clear in the operating manual!
On a technical point, I’d expect a normal CW transmission to be A1A - ie carrier only, no sidebands though there will be some close-in due to phase noise in the oscillators generating the signal and some dependent on the keying waveform and speed. General concensus seems to be that the bandwidth of a signal keyed at around 25wpm with a well designed and adjusted transmitter should be about 80 - 100Hz, though this may well be based on subjective assessments. When I get a minute I’ll hitch my radios up to the spectrum analyser and take some measurements while sending strings of dots (which should give the worst case scenario). At last I’ve found a use for the built in keyers
I don’t think it is technically or colloquially correct to say CW is transmitted as double sideband, but that’s a side issue.
On the FT8*7 series, the carrier for CW is injected after the balanced modulator. No audio ges near it. When you change from say, USB to CW, the displayed dial frequency goes up by an amount equal to your current sidetone setting, usually between 600 and 800 Hz. But the receiver “BFO” stays unchanged, so this allows you to commence a contact in USB and then switch to CW and the receiving station will hear a beat note equal to your CW offset frequency. You’ll still be able to hear the other station on either USB or CW, your receiver carrier frequency is the original, but the transmitted signal is on the dial frequency.
So with the FT8*7 series, set the dial to the channel frequency you quoted, then switch to CW and you’ll be safely 600-800 hz up from the dial frequency.
In a rig with CW-Reverse the whole thing happens in mirror image for LSB. But LSB is an abomination and we should drop it on 1/1/2020. (Seriously)
Switching from LSB to CW has you transmitting on a carrier freq that depends on the radio, and listening to either LSB or USB depending on the brand (Icom typically LSB, sadly, FT8*7 USB).
With the KX3 the dial frequency displayed on CW is the transmitted frequency, so the receiver “BFO” is offset from that. For CW-REV, that is USB. I have reluctantly continued to use CW on the LSB bands, but for the upper bands, I default my KX3 to CW-REV so that it tunes in the same direction on both modes. KX2 may be the same, probably is. I think the K3 works the same.
The rigs using DSP for the CW filtering have a big flexibility advantage over those that use intermediate frequency filters.
I also have noticed that when switching the mode from, say USB to CW, or back again on my FT857 and FT817 the rig changes the dial (carrier) frequency up or down by the sidetone frequency.
I already found this useful when helping a friend with his Morse practice on a mixed mode QSO (SSB and CW) on 60m. I set the dial frequency when using USB to the lowest frequency in our chosen 60m UK sub-band. Then, when I switch to CW mode it automatically adds 600Hz (my sidetone frequency) to the dial/carrier frequency, which ensures my CW emissions are not going ‘out of band’.
Incidentally, his TS590 (like my KX2) doesn’t change the dial frequency so he has to change it manually when switching between SSB and CW.
On the point about CW (A1A mode) being double sideband or not, have a look at the definitions for A, 1 and A at the Wikipedia definition …
If you concede a CW signal occupies spectrum bandwidth then it has one or two sidebands - unless this is a semantic argument about the use of the term ‘sideband’.
IMHO, I don’t think this is a side issue because knowing what the occupied CW bandwidth is w.r.t. to the dial frequency goes to the heart of the matter in avoiding going out of band when working at the band/sub-band edge.
Here’s what I think (feel free to disagree) …
If a (perfect) carrier is unmodulated, it has zero bandwidth (in theory) but it contains no information. As soon as you vary its amplitude you get a non-zero-bandwidth emission. For AM, (unless suppressed by the rig) that bandwidth comprises mirror-image upper and lower sidebands no matter if you change the amplitude to the complex variations of Chuck Berry (e.g. an AM broadcast) or the more-abrupt on/off AM with Morse-coded carrier switching. The shape and width of the sidebands will be different for each case but they are both DSB AM.
On a related point, many amateurs believe that the bandwidth of a CW signal is determined only by the character sending speed but, in fact, it’s only a secondary effect – the speed mainly affects the ‘ripples’ in the shape of the sidebands.
A Fourier Transform (i.e. intensity vs frequency graph) of a CW signal shows the highest frequency components of the modulation are in the on and off edges of the dots and dashes. And it’s the highest frequency component that determines the total bandwidth.
This appears to be confirmed by my friend’s [G8EJN] experiments viewing the bandwidths of his TS590 RF output on a spectrum analyzer when the rig is keyed with a continuous stream of dots with the e-keyer set at a range of slow to fast words-per-minute speeds.
Modern rigs carefully shape the carrier switching (making the rise and fall times of the edges 5 to 10ms) to create a trapezium-like RF output rather than a strict square wave. That avoids ‘key clicks’ and keeps the bandwidth to under 200Hz (BTW: some blogs I read get into heated arguments about the definition of bandwidth but let’s not go there).
Multiplying an RF carrier by your modulating signal corresponds to convolution in Fourier space (this is called the Convolution Theorem). Your pure carrier is a zero-bandwidth signal as you say - a delta function in the Fourier domain. Convolution with your Fourier transform (i.e. the spectrum) of your keying waveform is therefore just a shift (i.e. to your RF). Whatever is the spectrum of your keying waveform is simply being translated up to the radio frequency.
Simplifying it a little, consider a single square pulse. The Fourier transform of a square pulse is a sinc function (sin(x)/x)), which is symmetrical and spreads power out to infinity! A real-world keying function will introduce some rounding. For example, consider your square pulse convolved with a Gaussian. Invoking the convolution theorem again this means the spectrum gets multiplied by the Fourier transform of the Gaussian which is itself another Gaussian. Make the keying Gaussian wider in time (slower) and the Fourier-space counterpart gets narrower. This multiplies your nasty sinc function and pretty much kills off the high-frequencies (the clicks).
This stuff is really more familiar to us all than we might imagine. It’s the same maths that gives us side lobes. Consider a microwave feed to a dish. If it stops dead at the edge that’s the same as pure on-off keying. Nasty nasty sidelobes. Put in some edge taper and you have a clean beam.
Yes all true, but in practical terms, the question is whether assessing a signal as out of band will be done using a receiver and a pair of ears or using a spectrum analyser, fourier analysis and a super computer. Or at all?
In practical terms, cw signals are a big carrier and some very low sidebands of up to several hundred Hz, due to keying and synthesiser hash. (And receivers generally have the same synthesiser hash).
I suggest it’s the big carrier that really needs to be inside the band edge, otherwise where do you stop? Can anyone transmit in the 30m band at all? Or any band?
Going back to the original question, about how to ensure your carrier is not out of band, using the ft817/857/897 in the way described is a neat way to be reasonably confident of being “in band”, within the meaning of bandwidth and bands at the technical level of amateur examinations.
I don’t have a SA but my friend’s SA plots from his TS590S for SSB (LSB and USB), DSB AM and CW outputs gave me confidence about what I think I know in theory. But can his results be generalized to another modern rigs? I assume so.
I’ve done tests with my FT857 transmitting into a dummy load with it sending a series of dots using its e-keyer (at various wpm settings) whilst I’m tuning across the transmitted signal using my FT817 and noting the frequencies (above and below) at which I can no longer hear the signal. I’ve used wide and narrow filter settings on the receiver and assumed the narrow one represents the signal cut-off better.
I found this crude method to estimate my CW-signal bandwidth is consistent with my friend’s spectrum analyser plots of his TS590 CW-signal spectrum.
It’s 25 years since I did Fourier analysis on a DSP course for a MSc at Glasgow Uni and I’m more than a bit rusty on that [although I did go online yesterday to look at the maths for Convolution Theorem after Simon G4TJC replied].
Hah! I’m just popping down to PC World to pick up a super computer now. I wonder if they have any deals on quantum computers?
The upshot is, I concluded the CW bandwidth is under 200Hz (+/- 100Hz about the carrier frequency). If I double that figure to take account of the crudeness of my experimental data and differences in the definition of ‘bandwidth’ that should give me a safety margin. So, in practise setting the dial frequency at least 200Hz up from the sub-band edge should be okay for CW.
No-one (so far) has come back and said I’ve misunderstood how it works. That’s not to say some RF expert will challenge these ideas in the future.
Why not try a Web SDR Andy? You may want to run a (calibrated) frequency counter too, in case of offsets on the Web SDR. Then you’ll have a nice graphical indication. At least it’s all too often possible to see awful splatter from SSB stations with them. Key clicks might be a bit subtle to see above band noise and the other effects Andrew mentions unless well placed to get a nice big signal from your TX.
Brian, I don’t understand why that would be the case. Would you explain?
The Tx frequency is fixed. I’m moving the Rx dial frequency (slowly) across the Tx spectrum (in much the same way as you would on-air to find a clear frequency). The Rx filter has zero attenuation at the Rx dial frequency (and increasing attenuation at frequencies above and below that according to the filter shape).
So, when I get to a frequency where I can no longer hear the signal the emission at that frequency must be very low or zero. Now, it’s true that the filter has a certain ‘width’ (shape) so that I would be hearing a contribution of the signal from frequencies closer to the carrier frequency (more so for a wide filter than a narrow one).
But that only makes the perceived cut-off frequency a bit further away from the carrier so that the calculated bandwidth is wider than it really is. No bad thing.
I already said the method is crude but I don’t see why it would be invalid. Perhaps you would elaborate.
Hi Simon, I’ve not used a Web SDR before but, on your suggestion I just tried Hack Green on 60m http://hackgreensdr.org:8901/
I can hear my test transmission and see a trace on the waterfall at the correct frequency. But even if I zoom in the trace is too narrow to judge its width. Is there a way to get a better graphical display?
BTW: I don’t suspect any problems like key clicks with any of my rigs. I always get good to excellent reports about the quality of the CW signals.
I may be misreading you, but it appears to me that you are assuming that the cut-off frequency that you perceive defines the bandwidth of the signal, whereas I think that it defines the bandwidth of the filter at the sensitivity limit of the receiver. If you could generate a signal with an infinitely narrow bandwidth I think the result of your test would be the same.
That statement seems at odds with one’s everyday experience of tuning across the bands. If I tune off a SSB station I have to move frequency more before I get to a clear frequency than I would do with a CW signal [with the same filter setting].
Not really, because if this perhaps mythical infinitely narrow bandwidth signal were to be encountered its characteristics would be blurred by the imperfections of the receiver. You get phase noise with the receiver oscillator, too. Anyway it would seem that you would need to use different criteria for an SSB signal. Earlier you wrote:
By this argument you would have to keep an SSB signal 2200 Hz (or whatever the SSB filter width is) plus safety margin from the band edge, but this would make several commonly used 60m bandlets unuseable! The point is that if you tune up to a couple of kHz below 5.4035 MHz you will still hear an USB SSB signal with that as its carrier frequency, but that doesn’t mean that there is any (significant) RF being transmitted below 5.4035, it is just an artifact of the receiver bandwidth. I bracketed (significant) because suppression of the opposite sideband is very good nowadays, but not perfect. Yaesu quotes >50dB for the FT857.
We must bear in mind that it is out of band settings of the VFO that we must guard against, plus broadening of signals by bad practice. Artifacts in the receiver are outside the control of the transmitting station.
I’ve pondered the measurements you made and I think you are being misled by the apparent bandwidth of the signal shown when you tune across it with a (presumably 250 hz) filter in the receiver. As Brian suggested, you are seeing the bandwidth of the receiver filter.
In general you have to subtract the known bandwidth of the receiver from any measurement.
Another experiment you might want to try with the receiver. Instead of using a modern transmitter for the signal source, use a quality signal generator with a known phase noise performance. One of the good brands. When you measure the bandwidth of a carrier from one of those things, you know it should be close to zero.
I think your method will definitely say it is 250 hz wide on the narrow CW filter and it will be about 2700 hz wide on the standard Murata SSB filter.
I believe you are correct. It makes no difference at all on transmit. If you are listening to a CW signal which is the only one in the IF passband, it makes no difference on receive either.
Where it makes a difference is when there is an interfering signal close to the wanted signal. Suppose you are listening to a station on 7010.0 kHz and are set to hear it at a typical audio frequency of 700 Hz. If you switch between CW and CW-R it should sound exactly the same.
Now suppose an interfering station comes up on 7010.1 kHz, i.e. 100 Hz up. In CW mode, you will hear that at 800 Hz. In CW-R mode you will hear it at 600 Hz. You might find one easier to ignore than the other, so having the two modes gives you a choice.
Yes. CW is just amplitude modulation of a carrier with the keying envelope. The bandwidth is exactly double the highest frequency present in the spectrum of the keying waveform. That depends primarily on the rise and fall time of the keying envelope and its exact shape. Keying faster does not increase the bandwidth, but it does increase the energy present in the sidebands and may make components that were previously below the noise become visible, and thus give the impression of a wider signal.
If the rise and fall time of the keying envelope is 5ms and it has a decent shape, there will be a strong modulation component at 100 Hz, which will give a bandwidth of 200 Hz. This suggests that one should not operate CW with the dial frequency closer than 100Hz to a band edge. You might also need to make an additional allowance for the uncertainty in the calibration of your transmit frequency.