Group Delay Correction on System A (405-lines)

[SwanseaSteve]

This is my first post! Hope no-one minds me starting with a new thread and a question....

Basically: does anyone know what group-delay precorrecton was used by the BBC for the VSB 405-line transmitters? I believe this sort of information was in the CCIR official specification for System A. (Data for all the CCIR 'systems' used to be published in the back pages of the Proceedings of the CCIR Plenary Assemblies.) The university library used to have a copy 20 years ago, but when I went in there recently it seems thay don't have it any more.

I assume that Ally Pally (until decommissioned in 1957) didn't provide for any pre-correction, as TVs built for DSB reception wouldn't have had any relevant filters in the video path, but that all changed as soon as the first TVs built for VSB came out.

[Synchrodyne]

In case it might help, the attached BBC document, EID IS #2202-2, provides the transmitter delay specifications for the experimental 405-line NTSC colour transmissions, see §3.2.

Secondly, see BBC Research Report 1958/18, available at: http://downloads.bbc.co.uk/rd/pubs/reports/1958-18.pdf, which describes an experimental 405-line NTSC colour receiver. Appendix A, §2.8, on pge 16, includes the receiver delay specifications.

Cheers,

[Ray Cooper]

Originally Posted by SwanseaSteve

...does anyone know what group-delay precorrecton was used by the BBC for the VSB 405-line transmitters?...
I assume that Ally Pally (until decommissioned in 1957) didn't provide for any pre-correction, as TVs built for DSB reception wouldn't have had any relevant filters in the video path, but that all changed as soon as the first TVs built for VSB came out.

I can only speak for those specimens of 405-line gear that I came into contact with: but certainly, the HP TXs at Sutton Coldfield (and probably Holme Moss, too) were double-sideband beasts followed by a large VSB filter operating at high power. The TXs themselves certainly didn't have any formal group delay correction fitted: there was a specification on the group delay figure for the VSB filter which would have given acceptable results without any pre-correction.

Other HP TXs used by the BBC were largely low-level modulated, followed by a linear amplifier tuned to the required VSB response, and as such didn't need a VSB filter. Can't remember at this distance of time whether there was any formal GD correction on, say, the Marconi 5kW TXs used as reserve at SC, but I can't ever remember setting any up....

That EID report EID IS #2202-2 seems to indicate that group delay was only corrected for at carrier and subcarrier frequencies, and the allowable error of 220nS at frequencies just above vision was enormous by the standards of 625-line colour (about 40nS would have been considered large for this standard).

So the conclusion is that, for monochrome 405 line TV, group delay wan't something that they lost a lot of sleep over.

[SwanseaSteve]

Well, thank you both for those informative answers! However, like Ray Cooper seems to be implying, the group-delay numbers given in EID IS #2202-2 are so far off the numbers that I expected (which should have been more in the 20 - 40ns bracket) that I'm suspicious that the EID is answering a totally different question!! Something maybe to do with the characteristics of video amplifiers, but not what I had in mind.

Let me explain what I thought I was asking. It's all to do with the filters in the receivers that have to generate a full DC - 3MHz video signal from a signal transmitted in VSB style. I've got a Sobell T21 lined up for restoration, but I've not started it yet. I've been unwell, and though I'm on the mend, the T21 will likely remain in the stack for a bit longer yet. To keep my brain working, I thought I'd try and simulate the video I.F. response of the set and see how well it had been engineered to cope with VSB, since handling that signal is non-trivial.

The setup instructions for the T21 basically say to peak both the video I.F.Ts for 17.6MHz (video carrier is 16MHz on this set). The video I.F.Ts are double-tuned transformers, which gave me the challenge of how to model them, but I did all that and was actually quite pleased to see that assuming Sobell's designers got the coupling coefficent right in the transformers, the TV was capable of reconstructing the tricky area around the vision carrier very well.

However, it all does depend on the precorrection being present, and matching the response of the filter curve produced by the I.F transformers.

Consider a video signal of 1.6MHz. This is outside the VSB region, and the TV is working single-sideband to detect it. It also happens that in the I.F, this information will be at 17.6MHz which is the centre of the responses of the two video I.F.Ts. Let's call this point 0dB (the reference).

Consider a video signal of 250kHz (produced maybe by a backround in a picture consisting of nine or ten white doors in a dark wall). This part of the signal is carried by the double-sideband part of the system. In the I.F, two signals at 15.75MHz and 16.25MHz will be present, but because they are being passed by the skirt response of the video I.F.Ts, they will be a bit attenuated. My calculations show that the 15.75MHz component will be at -8.2dB and the 16.25MHz component will be at -4.1dB.

But this is deliberate! If you add those signals together (they both represent the same thing after all) they add up to almost exactly 0dB! (Which is what we need if we're going to get a good picture.)

Trouble is, because they got through the I.F. skirt response, they didn't just get attenuated a bit, they got delayed too. And they got delayed by different amounts. My calculations show that the 15.75MHz component takes a 16.26ns delay, and that the 16.25MHz component takes a 12.82ns delay. Now, this won't matter if the BBC pre-corrected those sidebands of the carrier by those amounts.

But if they didn't, then everything goes wrong. For everyone. Those sidebands only add up to 0dB if the signal components that they carry are delayed by equal amounts. Additionally, the edges (say, of the white doors in the dark wall) get damaged. On each door as shown on the Sobell TV, the left edge would appear, but be only 62% as bright as it should be, then 3.5ns later it will get brighter as the other sideband belatedly joins in (but not as bright as it should because the components' phases still don't add up).

This information was known when VSB was first used in the American 525-line system launched in 1941. Certainly the BBC can''t have failed to notice, and the BBC are usually known to be such technical sticklers that I just can't believe that they didn't correct for it!

Notice that you can't correct for group-delay like this at baseband. This has to be done post-modulation. Notice also that one of the components of a television signal most affected by bad group-delay issues would be the sync pulses. Again, not a place I can believe the BBC would ignore.

Or did they??

[Synchrodyne]

I think that I am getting out of my depth here, so this might not make much sense; apologies in advance if I am just confusing the issue. In terms of the BBC EID looking at something different, is it possible that on the one hand there is the transmitter correction required to offset the distorting effects of the vestigial sideband filter, and on the other hand pre-correction for the expected distorting effects of the receiver IF filters and traps.

The excerpt from Amos, attached, seems to address the former. On the other hand, the excerpt from Benson & Whitaker seems to provide information on the pre-correction curves for Systems B, G and M, although not any others. I have a vague notion that Systems B/G had a lot more receiver pre-correction than System I, but I cannot find now any confirmation.

Information on the Systems B/G correction numbers is given at: http://www.itelcast.com/_download/ITEL-RF-course.pdf, scroll down to page 5.

And for System M at: http://www.um.edu.ar/catedras/clarol...ue&cidReq=1051, scroll down to page 1107.

Additionally, Carnt & Townsend, in Volume 1, addressing the BBC 405-line NTSC colour system, provide the following comment in respect of receiver IF amplifiers: “The necessity of providing adequate rejection at sound and adjacent sound frequencies makes it difficult to keep the vision phase response linear at the edges of the pass-band. The situation at the vision carrier position is the same as in monochrome recep¬tion but it is more critical at the sub-carrier end of the band. The transmitted waveform is pre-distorted to correct for the phase response of the average receiver around the sub-carrier and this eases the receiver design.”

One might read into this that pre-correction was introduced only for the colour transmissions, and had not been used for monochrome transmissions.

Cheers,

[Synchrodyne]

I have now found the reference to System I pre-correction, in Hutson, as follows: “A cut-off between 1.25 MHz and 1.5 MHz is considered to be about the optimum. The shallow slope of the System I VSB characteristic, although allowing a considerable overspill into the adjacent channel, does not necessitate the use of additional group delay pre-correction at the transmitter for the errors introduced at the receiver by its VSB characteristic. The overall ideal transmission/reception group delay response for System I will therefore be flat and will have no receiver pre-correction. Other transmission standards which use steeper VSB slopes to better contain the signal within the channel allocation will often also pre-correct at the transmitter for the group delay error introduced by the steeper filter characteristic of a standard VSB receiver.”

Now I wonder if the same reasoning had been applied to System A? I do not know whether it is the absolute magnitude of the vestigial sideband or its size relative to the full sideband is the key determinant here. If the latter obtains, then the vestigial sideband for System A is a greater proportion of the full sideband than is the case for System I, and it might have been thought that receiver pre-correction was not necessary.

Some receiver SAW filters appear to have group delay curves that match their respective system pre-correction curves within a given tolerance range. And some more sophisticated receivers have had post-demodulation group delay correctors, such as the multistandard tuners in the Sony Profeel range, presumably because the SAWF characteristic is correct for one System only.

Cheers,

[Radio Wrangler]

On 405 line monochrome, did they specify anything in terms of group delay? I thought quality assessment was all in terms of waveform 'k' factor from pulse and bar signals viewed on a scope with a special graticule, plus high frequency limit from scanned test-cards.

David

[Ray Cooper]

Originally Posted by SwanseaSteve

...they got delayed by different amounts. My calculations show that the 15.75MHz component takes a 16.26ns delay, and that the 16.25MHz component takes a 12.82ns delay. Now, this won't matter if the BBC pre-corrected those sidebands of the carrier by those amounts.

But if they didn't, then everything goes wrong....On each door as shown on the Sobell TV, the left edge would appear, but be only 62% as bright as it should be, then 3.5ns later it will get brighter as the other sideband belatedly joins in...

Notice that you can't correct for group-delay like this at baseband. This has to be done post-modulation....

Fascinating stuff: but I feel that a healthy dose of pragmatism is called for here. We are, after all, talking System A. Group delay errors of this order have to be viewed against the fact that a picture element, which we'll be very kind and say is equal in length to a half-cycle at 3MHz, is going to be about 160nS long. So yes, there are errors, but no, they'll be quite invisible on the picture. (I've just dragged out my copy of the IEEE paper on Sutton Coldfield, dated 1950, and it gives absolute delay figures of 20nS at fv-2.75MHz, 95nS at fv, 120nS at fv+0.75MHz, and no attempt was made to correct for these delays. So your 16nS looks rather small beer...)

With regard to the statement 'can't be corrected at baseband', this needs qualifying by the addition of the word 'satisfactorily'. In fact, early marks of 625-line 'colour' TXs used correction at baseband, because in those days there was no other way of doing it. Until the later generations of I.F. modulation TXs came along, there was no prospect of being able to correct Group Delay at final carrier frequency - or at least, in an acceptably stable manner. The old baseband correctors didn't work well, it's true, but they could be codged into providing an acceptable average response. I recall that the GD correction provided on the original Marconi 625 transmitters took the form of a multi-position switch which enabled various pre-set corrections (as envisaged by the mind of the original designer) to be applied. Some of these positions incorporated correction for both RX and TX group-delay anomalies, though I don't believe that these were ever used in this country.

[Nymrod121]

I think you're right on the mark there, David. The only measurements I recall being done in respect of group delay were done on UHF plant only - use of a "Heucke" Group Delay Measurement Set was generally considered as being the most accurate way of doing it and the Senior Transmission Trainer at Wood Norton (Tony Larkham) had the procedure down to a fine art.

During a stint at Brookmans Park (BBC Transmitter Capital Projects Department), I remember reading a technical reference manual written by Nick Davies: "Group Delay, Ceefax and IF Mod. drives". Again, this covered the UHF network only. Nick's still around so I'll drop him a line and see if he can recall if there was any standard in respect of the VHF 405-line service.

I have to say that I'm inclined to agree with Ray's comment regarding the pragmatic approach ...

Regards
Guy

[SwanseaSteve]

Originally Posted by Ray Cooper

Fascinating stuff: but I feel that a healthy dose of pragmatism is called for here. We are, after all, talking System A. Group delay errors of this order have to be viewed against the fact that a picture element, which we'll be very kind and say is equal in length to a half-cycle at 3MHz, is going to be about 160nS long. So yes, there are errors, but no, they'll be quite invisible on the picture.

Hmm - good point, I am rather over-egging the timing errors aspect when really the only thing that would likely matter would be the phase errors which would prevent the components of the signal in the VSB area from adding up to 0dB correctly.

And my example was poor (I must have been half asleep when I wrote it, sorry). The edges of my putative "doors in a dark wall" would be carried by video components at much higher frequencies, probably up in the several MHz, and not in the VSB area anyway. The effect of the phase-errors in the VSB area would be to make the signal at (say) 250kHz several dB less than it should be, but the visual effect would be that the centres of the "doors" would display darker than the left and right edges of the same.

So, thank you very much, everyone who contributed here. I think it can (for now) safely be said that there's no evidence that the BBC did any GD correction to VSB transmissions other than possibly the correction mentioned in EID IS #2202-2 which looks to me to be related to something else! While we're still on the subject, has anyone got any idea what the BBC thought they were correcting for with those values in the EID??

I think it's something about the characteristics of video amplifiers. Many video amplifiers would employ "H.F. peaking" to try and compensate for driving a wideband signal into the capacitive load of the CRT. Could that be it? Or even part of the explanation? Was there ever a 3.5MHz sound-patterning notch filter in video amps of the day? Might that also be part of it??

[Synchrodyne]

Originally Posted by SwanseaSteve

While we're still on the subject, has anyone got any idea what the BBC thought they were correcting for with those values in the EID??

I think it's something about the characteristics of video amplifiers. Many video amplifiers would employ "H.F. peaking" to try and compensate for driving a wideband signal into the capacitive load of the CRT. Could that be it? Or even part of the explanation? Was there ever a 3.5MHz sound-patterning notch filter in video amps of the day? Might that also be part of it??

Possibly it was pre-correction for expected receiver characteristics? I haven’t done it, but perhaps one could compare the group delay curve for the colour receiver outlined in the Research Report 1958/15 with the transmitter curve described in EID 2202-2. I have extracted and attached the pertinent pages from each.

Presumably the overall group delay was more critical for colour transmissions anyway, and added to that, colour receivers would more likely have an IF bandpass shape that gave a vision response that was flatter out towards 3 MHz and then dived into a very deep 3.5 MHz sound notch, which I think would in turn be have greater adverse effect on the group delay curve, thus necessitating some pre-correction, whereas it was ostensibly not needed for monochrome transmissions.

Cheers,

[Ray Cooper]

Originally Posted by Radio Wrangler

... I thought quality assessment was all in terms of waveform 'k' factor from pulse and bar signals viewed on a scope with a special graticule...

That's the way 405-line tended to be done. But it's worth remembering, that on a historical note pulse & bar testing was not introduced until the early 1950s and didn't become widespread until the late 'fifties. Until that time, they had to make do with a rather entertaining signal called 'flagpole': this was simply a short rectangular pulse going to 100% peak white, sat in the middle of a black line waveform. On a picture monitor, it looked like, well, a flagpole...

To use it, you simply measured on a CRO any overshoot on the pulse edges, and if it wasn't too bad you felt yourself lucky. Peeping into the spec for the CG1 transmitter (a design that didn't appear until the mid-fifties) I see that pulse/bar performance info is totally lacking, and if any overshoots, as measured with 'flagpole', were less than 5% of peak amplitude, the thing was in spec. How times have moved on.

[SwanseaSteve]

Originally Posted by SwanseaSteve

I think it's something about the characteristics of video amplifiers. Many video amplifiers would employ "H.F. peaking" to try and compensate for driving a wideband signal into the capacitive load of the CRT. Could that be it? Or even part of the explanation? Was there ever a 3.5MHz sound-patterning notch filter in video amps of the day? Might that also be part of it??

FYI, my Sobell T21 does indeed have a 3.5MHz trap in the video amp and quite a few circuits that I would think contribute peaking of various sorts at various frequencies.