Here we go again…
Having already spent months on this subject, I had promised myself I wouldn’t go down this damn rabbit hole. But here I am.
The BBC’s NICAM-728 engineering trial started in 1986 and ran for 5 years, with the public launch in 1991. A lot was learned in this period, quite understandably. It was a large and complicated change to the network. It was shut down in 2012. Today we’ll be looking at the kit which bought us BBC television sound during that 21 year period, where ever you were, whatever you were watching on.
Previously I have looked at off-the-shelf NICAM broadcast equipment. Solutions like that are what most smaller / non-UK broadcasters would have used. But not the BBC. In my previous post I looked at a NICAM-676 (FM radio backhaul) solution designed and used entirely in-house. For NICAM-728 (Television) the BBC had their own implementation too but the job of designing it was (sort-of) outsourced. It is stated that it was put out to tender and it seems barely a surprise that the successful bidder was in fact an ember of exact same company who had produced the previous generation of sound distribution equipment.
Enter a company you’ve never heard of
Above is a Varian NICAM-728 coder which belonged to the BBC, and now belongs to me. We can see four of these in this google street view of the BBC’s analogue bay in London. Varian has its roots in a company you might have heard of:
The units do not feature the PYE branding because Philips (its former owner) was frustrated that the company had become uncompetitive in an increasingly international market, so decided to call time on it. They sold that part of the business to Varian Associates in the USA, who fed it to the sharks shortly thereafter. The other parts of Pye didn’t fare much better.
And a technology you’ve never heard of
We cannot talk about NICAM-728 at the BBC without talking about Sound in Sync hereafter: SIS. It is also referred to as Sound in Syncs (with an ‘s’). This was the BBC’s preferred mechanism of audio distribution in their network. There was much grumble and grouse about the massive bills that had to be paid to BT for links between studios and transmitter sites, so the BBC devised a way to splice audio into the expensive high-bandwidth video links, eliminating the need for separate links. This was done by inserting digitised audio data into the blanking periods as a baseband signal, that would then be extracted and removed at the other end of the link. We’ll be looking at this in detail shortly.
The first version of the system dating back to the 1960s was mono, inefficiently encoding the data as raw PCM. The picture above shows one of these earlier units. Eventually a stereo version was developed. The data coded into the blanking periods by this system would be NICAM-728.
Let’s have a high level look at how the system was put together:
As we can now see the NICAM-728 at the BBC fell under the umbrella of the existing SIS technology, thus it was incorporated into its infrastructure as a core technology. This is why it took 5 years to get it right.
This was not the case in other adopting countries – for example New Zealand where Broadcasters did not face the same constraints, specifically they were free to use their microwave backbone to distribute additional audio channels at minimal additional cost. Thus it was treated as more of a “bolt-on” with a presence only at transmitter sites. This is why they managed to launch the technology two years before the UK (1989).
With all of that out of the way, let’s see what kind of treasures we can find in that big metal box. I did ask if any technical documentation could be released for these units. At the time of writing nothing has been released that I know of.
SIS (DCSIS) Coder
This piece of equipment lived at the studio, at the source of the analogue audio. The BBC internally refer to these as “DCSIS”. The DC means Dual Channel. As we’re not going to be looking at the earlier single channel SIS solution, that’s two letters too many for me.
While this unit appears to be missing it’s lid, this is actually a feature. They are designed this way to allow airflow through a vertical column as typically many are mounted in a rack. Occasionally they were loaned to third parties in a portable case. Like the Philips solution it has two thumbscrews to release the front panel, allowing easy access to the internal modules.
Audio input (filters / preamp)
The following two assemblies are the exact equivalent of Unit 3 of the Philips solution however they used a mezzanine arrangement to pack both on onto a single similarly sized card.
There’s not a great deal to see on here. Just a bit of pre-amplification, the two input isolation transformers (probably toroids) and the all-important input anti-alias filters. Unlike the Philips solution where we see this built from scratch using industry standard parts, this one is a pre-made module from RIFA. It’s actually a catalogue part but I bet it’s a bummer to find spares.
Of interest is the quarter inch jack. This allows a test headset to be plugged in to check the audio at the input of this unit.
Once again, not a lot to see. The ADC is an expensive, very nice part from Burr-Brown who were well regarded in this area.
The BBC’s original “NICAM I” (676) solution used a similar part. What we see here is actually a 16-bit ADC, overkill for NICAM which is only 14-bit. The two least significant bits aren’t connected.
Architecturally it is very similar to the Philips solution using a single ADC to digitise both left and right channels, however the analogue front end here is just a pre-made module from US company Analogic. Once again, probably a bugger to source replacements.
NICAM Coder (DSP)
This assembly is the equivalent of Unit 5 of the Philips solution, containing the digital part of NICAM coding process. Previously we have seen FPGAs, an LSI and a Motorola 68K doing the job. This time, it’s a TMS320 series DSP from Texas Instruments. Oh so many ways to skin a cat. The code for the DSP is stored in the two UV EPROMs.
Clock & I/O
This one generates all of the clocks used within the unit and handles many of the external I/O signals.
NICAM Decoder / Monitor
This unit has an empty slot for a NICAM Decoder card. It’s used only for monitoring purposes. Like the input card it has a quarter inch jack so that a headset can be plugged into it for testing. By doing so the whole digitisation and encoding process can be verified.
None of the pictures of these units I could find have this fitted. I suppose it’s pointless. How useful is it to know only half of the box is working? For proper monitoring you’d want to stack this box on an SIS decoder and plug your headphones into it instead.
The BBC Engineering archives do briefly mention it, and it is also found in the SIS decoder. It is stated to be built similarly to a consumer NICAM receiver. I would expect it is fitted with a Texas instruments CF70123 decoder (the only consumer decoder available when this was designed) and perhaps some kind of DAC from Burr-Brown. If I ever get my hands on one I will update this page.
CPU / Monitor
On here we have the BBC’s go-to – the Z80 CPU. It is not involved in any of the heavy lifting, instead just performing monitoring, control and housekeeping functions.
There are three additional cards plugged into this unit which encode the NICAM data into Sound-in-Sync format. At the time of writing I have no technical documentation for this unit. There is some information about the previous generation in this document. The SIS part of the previous version is likely very similar to this.
I’ve not studied this part in detail yet. For now let’s just look at the pictures:
Not a lot to say about this. It has thumping massive +5V linear regulator on the rear with a heatsink to match.
The rear is quite alien compared to what we’ve seen previously. It’s a funky completely BBC specific set of connections. It’s worth taking a quick look at the rear panel of the Philips PM5687 as this unit has many connections in common with it.
Specifically I have confirmed:
- Stereo analogue audio in – on a weird 5 pin XLR connector. Guess it’s a bit quicker and eliminates the possibility of L/R swapsies?
- Composite video input
- Composite video output (SIS)
- NICAM CLOCK/DATA IN (To SIS Coder)
- NICAM CLOCK/DATA OUT (From internal NICAM coder)
- Reference clock in
- Control inputs
- Status / alarm outputs
Aside from completely a completely different output the key differences are that all of the signals are TTL digital, whereas (by default) Philips use GPIB for control/status and bipolar signalling for the rest, causing me some troubles as we’ll see.
Some of the signals are doubled or quadrupled up too, for some reason, thus there aren’t as many as it appears.
This piece of equipment lived at the transmission site. Physically it nearly exactly resembles the SIS coder we’ve just looked at. The transmitter sites were sold off to Arqiva long ago complicating my efforts to obtain one. I wrote to Arqiva asking if they had one I could borrow. It won’t surprise you to hear that they didn’t respond. The BBC probably still have some somewhere but for now I’ll skip this pursuit.
Whether or not this would have been interesting to look at depends on your perspective. In terms of the end-to-end reproduction of the analogue audio signal, you have now seen everything involved. In terms of distribution of the digitised audio throughout the BBC’s network, this is a key component.
The SIS Decoder had two main jobs. Firstly it extracts the digitised NICAM data from the incoming video signal from the studio, this is then passed digitally to the DQPSK modulator for transmission. From a NICAM perspective it is (almost, see below) a pass-through, as is the modulator.
Secondly, it actually decodes that data and outputs an analogue audio signal, which feeds the FM mono (monaural) modulator. So even if you didn’t have a NICAM receiver, your mono sound was still courtesy of the standard.
The SIS decoder may have also contained the reframer which is described in the engineering archives. Let’s cover that briefly.
This part of the system was developed at the 11th hour and was said to have delayed the public launch. In the BBC’s setup there is not a single source of coded NICAM data. In fact everywhere there is an audio source, we would have found one of these coders, in regional studios, regional parliaments etc. Everywhere the BBC has a presence. This constant chopping the source lead to regular blips and drop-outs. Because the NICAM data stream was piped all the way to the public, each time this happened viewers would see a message on their TV that it had lost stereo sound. For older sets or the “stereo” LED would flash off and on. This is not acceptable because people do not expect this, and may even think their equipment is faulty.
Enter the reframer. It lives at the transmitter site (it’s probably a card in the SIS receiver) and solves this problem. The encoded audio is extracted from the source and inserted into a steady stream of newly generated frames. If the source is momentarily lost, no problem. The reframer just inserts silence, maintaining NICAM lock in the public’s receiving equipment.
Getting it working
Even though I don’t have the SIS Decoder we can still see the NICAM coder in operation as the two are entirely independent from each other. The 728kbit/s data stream from the NICAM coder is outputted on an auxiliary connector on the rear. This can be connected directly to a DQPSK modulator.
In the above setup I’ve used the venerable Arcam Delta 150 standalone receiver. The BBC liked these. The engineering archives state that they were used for some monitoring purposes. It’s largely the same test setup shown in this post.
Mindful of the unfortunate history between these two companies I was reluctant to give this tired old bit of kit the Philips treatment, but frankly I have no other way to assemble a test setup. The BBC engineering archives do talk a little bit about the modulator although stop short of saying what it actually was.
Whatever it was, it appears to have been a more rudimentary device than the Philips PM5686A as it is stated to require a separate external upconverter and amplifier to produce the 32.348 MHz 0dBm IF output. The Philips does this in a single unit.
The audio signal is fed into the SIS unit from an unbalanced to balanced converter box. The NICAM digital output is taken from the “Auxiliary Output 1” connector where it is then fed into the Philips PM5686A. These two units aren’t electrically compatible (TTL vs Bipolar). The DSub backshell contains a line driver which is why I added a sneaky power feed from the SIS box.
The NICAM intercarrier output is taken from the PM5686A and fed into the composite video input of a cheap modulator, which in turn feeds the receiver. It’s quite a quirky setup. I’ve already covered the technical details of it in this post.
Was the BBC’s implementation incompatible with the standard?
I’ve spotted multiple occurrences of internet forum conjecture generally centered on transcoding having to be done at transmitters because the BBC’s kit used non-standard coding. Being the first pleb with a rambling blog known to be in possession of the kit I think can clear this up.
As we just saw above I’ve connected a BBC coder more or less directly to a Philips modulator (as standard as it gets) which in turn is feeding a standard consumer receiver designed for the UK market. There is no transcoding here. Additionally there wouldn’t be any sense in the SIS decoder changing the format of the data. All it has to do is extract it from the syncs and pass it on. Easy. So even if I had the decoder, the picture isn’t going to be any different.
There might have been an earlier prototype solution which coded the audio as NICAM-676. This would require transcoding before transmission. We know they worked a lot on this standard because we saw it in my previous post.
There is a brief mention of development of NICAM-676 for television broadcast in the the archives but also that it was abandoned. This may be where this idea came from. Perhaps people thought the “Reframer” I mentioned earlier was a transcoder. It wasn’t. Beyond that I don’t know. There is no other evidence of it being true. As always, if you know something I don’t please leave a message.
The SIS bit
I cannot show an end-to-end but what we can do is look at the output on an oscilloscope. The SIS coder automatically comes out of standby once a composite video input is connected.
Sound-in-Sync was a interesting and innovative solution that solved a difficult problem. The SIS modules are very impressive looking and I hope we can learn more about them someday. They were probably mostly designed by the BBC. The NICAM specific parts bear absolutely no resemblance to the BBC’s internal design. Perhaps they re-did this part from scratch.
It was designed to do a job and for 21 years it did it, for tens of millions of viewers. That alone was a good enough reason to look at it.
All of the supporting information in this post (unless specifically linked) comes from: