SSL 9k - background

Found this little nugget
The poster only posted once... but a nice ittle story
http://recforums.prosoundweb.com/index.php/m/67805/0#msg_67805

I know that the designer of the SSL 9000 has also known to lurk -and even post- on this forum. There are probably a few people who can claim to have done part of the 9000, Andy may very well have done the layout or something similar, but I think that Dave is the guy with the golden ears who actually designed and engineered the 9k preamp.

Keith

so what did he find so bad about the FET switching, I thought that was supposed to be one of the purer things (sonically)?

Well, at high signal levels, 'off' FETS can start to bleed 'on' again at the waveform peaks, since the gate-source voltage difference is lowered as the signal swings towards the rail... bleedthrough is a problem. SSL solved that later with a low current, higher-than signal rail supply for the gate switching only...

Anyhow, Rupert Neve recently said that he can get better sounding and better measuring results from a FET than a relay, but he can't get people to treat them without deep suspicion so he still uses relays...

Keith

Hi, just found this thread by accident.

If anyone's still wondering, the problem with single FETs is the distortion, as Keith says E series desks did also have a bleedthrough problem but we sorted that on the G series with a mod to the logic card. On the stillborn ARC we used relays and it sounded WONDERFUL, but we couldn't afford them for the 9k. The second best switch we found after relays is the SSM2412, particularly if used balanced.

Yes, you can make FETs work, but we found that you need so much rubbish around them to do it that it almost wasn't worth it.

I'm sure Dave would be highly amused to hear himself described as having golden ears! We just designed the mic amp, and all the other amps, to be as flat and as low distortion as possible - the secret is going for a flat response from 2Hz to 200kHz (which ain't easy). Anything different that's said about the design is marketing b******t. Just to clear up the confusion, Dave designed the mic pre for the ARC, and I cut and pasted the circuit into the 9k strip. (Dave being older than me knows how to use these three legged transistor thingys.) Otherwise I designed the strip, Dave designed the centre section - yes, by end 1992 / early '93 SSL only had two analogue designers left!

awesome!

I once asked John Stephens (of Stephens Electronics fame) why, being such a purist/minimalist, he used fets for audio switching instead of relays. He told me that they were fine if you use them only in series, never shunting.

Ummmm....errrrrrr...yes if they're going into a high enough impedance, but they will still leak and you'll get crosstalk.

But I guess some perspective is needed for this, certainly for a live sound desk fets would be fine, and plenty of recordings were done on E series desks which people were happy with.

A desk with mechanical switches will always sound better than one with fets, but whether it's enough to bother about is down to the ears of the user. It's a question of where do you stop. The ARC I mentioned above was the best sounding desk SSL ever made, but it was cancelled because nobody would ever have been able to afford to buy it.

However, the nice thing about diy is that often you make better equipment than the professionals - the cost of the components in a piece of outboard kit is minute compared to everybody's mark-ups (manufacturer's, distributer's, retailer's).

By the way, let's be honest, I can't hear the FET distortion, but I've conducted many listening tests which show that many people can. Everybody picks up different problems with the sound.

[quote author="Andy Millar"]A desk with mechanical switches will always sound better than one with fets,[/quote]
Rupert claims that he can get FETs to sound better than relays, -even good relays- but that people expect FETs to be worse, and a good relay is so good that he has no problem with it's limitations.

I can't say I see that any problems with a good relay are ANYTHING to concern myself with as being 'limitations'!

BTW, here I am, working on repairing an E/G series logic card as we speak!

Keith

[quote author="Andy Millar"]Ummmm....errrrrrr...yes if they're going into a high enough impedance, but they will still leak and you'll get crosstalk.

[/quote]

Yep, but I don't remember any problems with crosstalk on my old 24 track. I going to see if I still have any schematics around, maybe there was more to it...

I've always used the little Aromat DF series relays. Probably over a thousand by now, with only one failure. I've had plenty of FET switches go bad though. And not always "all the way bad" either, especially power-up mutes, for some reason. They would start sounding crappy enough to where I'd have to replace them. And of course, sometimes a relay is just easier to implement, especially when you need a DPDT.
I've just started using Axicom relays, pretty inexpensive (and Swiss made!).

Perhaps I also like the re-assuring "click" of the relay too...

Usually on the MCIs (like the speed switching on JH-series machines) they used series FETs to a common point. One was always live, because the machine was always set to one speed or another. As a result, the low output-impedance of the op-amp feeding it was a low-impedance ground shunt for any leakage from any 'off'-state FETs.

...-Works well enough for the MCI (Munchy, Crunchy & Intermittent) stuff.

Also, 'crosstalk' in this instance is going to be between the same channel's different EQ settings, and -40dB of crosstalk is insignificant. Crosstalk between channels is NOT going to be affected either way, and I'd guess that's what you're thinking of.

Keith

Yes, I was thinking of crosstalk as on, say, the '11/'294 card of a 4k desk so between record and monitor paths.

But absolutely, as everyone's said, if it sounds ok to the 'customer' (whether that's you or someone else) it's probably ok!

Also, the 4k suffers from FET distortion because the signal has to pass through so many of them, one or two probably wouldn't matter.

But nowadays I'm really lazy about my building - and cost isn't a huge problem since I only design audio stuff for myself - so it's much easier to solder a DIL relay onto a board than a three legged FET!

Any thoughts on analog switches such as the 4066??

Very good question. Again, you have to be careful how you use them. They are (relatively) high impedance at 80ohms, so need to feed into a few tens of kilohms to not distort. And they can't take very high voltages - +/-7.5V max. But for something like feeding onto a mix bus in a semi-pro system they're ok, and they are very easy to use.

Also, unlike JFETS you don't get pops and bangs when they switch. And of course they are much lower current and cheaper than relays.

I would seriously consider them for a demo studio or live sound type system, but for highest quality it would have to be relays first, then JFETS.

this brings up an interesting idea. My console uses JFETS as mutes which I suppose is great for the simple mute automation it has. I will try some hard bypassing to see if I can hear the FET.. i doubt I will though.. :shock:

with all the other mods I have done, I should try this one too!

Andy, maybe you could answer a question for me:

Who designed the compressors in the 4000 series SSL (i.e. the ones that are cloned here)?

People seem to attribute it to Alan Smart, but he has never claimed to AFAIK, and never mentions it in his bio. It seems from the bio that he worked for SSL after the board came out, and wasn't directly involved in any design work while there.

Thanks,

Paul

The compressor design dates back to the late 70's, certainly the earliest E series desk if not the B series. I'd always assumed that Colin Sanders designed them himself, but since I didn't join SSL until 1984 I may be wrong and Alan may well have been involved.

Sorry I can't help more!

[quote author="Svart"]this brings up an interesting idea. My console uses JFETS as mutes which I suppose is great for the simple mute automation it has. I will try some hard bypassing to see if I can hear the FET.. i doubt I will though.. :shock:

with all the other mods I have done, I should try this one too![/quote]

This is turning into a 'spot the FET' challenge! Great stuff! Although I don't think you'll hear it either (even I would admit that they're very good for mutes :wink: ).

FETs have one distinct advantage over relays and switches; If they're in an automation system and you may be cutting or re-making the signal during (for example) a high-spot in a LF waveform, relays or switches WILL put a click in the signal. It's just a fact of life. the sudden interruption or resumption of signal introduces a large, vertical stair-step.

With FETs, you can slug the gate slightly and get a fery fast fade-in or fade-out. There might be some 3rd harmonic distortion for a millisecond, but it's too brief to identify, and the shift in the level of the signal also helps mask it. -It's essentially indetectable to the ear.

If you wanted to go all audiophile, you could use time-delayed relays to bypass the FETS in "full-short-only" mode, (Q. -What's shorter than a short circuit??? A. -Two short circuits in parallel!) but why bother, really, if it's the only FET in the system?

The E and G-series problem was that you had FETs at the output of every input preamplifier (for logic and status-driven input selection and input muting during solo), then FETs at the input to the VCA, and FETs at the output of the VCA (for master status and logic-controlled fader swap) then FETs at the monitor path feed (for logic-controlled group/tape/group-&-tape-at-minus-3dB "supercue" monitoring) and FETs into the monitor fader and out of the monitor fader (for the rest of the fader reverse) then FETs in the master path, then FETs in the monitor source selection...

That's just off the top of my head... I'm sure that's much less than half of them!

In the E series they ran into (positive? -I forget...) waveform excursion part-turn-on distortion issues, even the disable for the buss compressor was a FET, which used to -at high signal levels- kick in slightly even when switched out, until we did a modification to hard-kill the detector output when the buss comp. was switched off.

With such a long cascaded chain of Fet switches, even subtle FET-related issues begin to become rather more significant. In addition, the E and G both had literally hundreds of mother-to-daughter card edge connector transisitions, which made servicing the module an absolute dream for techs (like me... when I had my "fixing hat" on) but -specially with age- made the sonic performance just a little less then stellar for engineers (like me... when I had my "mixing hat" on).

The card design of the 9k very sensibly reduced audio signal edge transistions to a tiny fraction of what they were on the E/G, (thanks Andy! :wink: :thumb and used a seperate, elevated rail (as retrofitted to the G series on the later logic card design) which moved the FET gate-switching rail further away from the maximum audio signal excursion level. This made the remaining FET switching vastly better and basically eliminated the punch-through.

Incidentally, Neve V-series boards used the same FET switching (though not quite so much) so it can't be all that bad!

The 9000 series used SSM2142s for the fader reverse switching, if I recall correctly. There was also a VERY cunning use of a 5534 that managed to pass signal despite both of its + and - inputs being tied to the negative power rail... -Anyone here smart enough to figure out how?

Andy M, -I think that might have been a Dave Mate-ism... though I don't wish to wrongly attribute the design. -Do you remember it? -It's the large fader (non-VCA path) wiper follower/buffer circuit...

Keith

SUPERB description of the FET issues!

It's interesting that Neve were always really worried about slugging their FETS - for the reasons you give - whereas SSL were never too bothered, hence the huge thumps you can get when switching from 'Record' to 'Mix'.

If the strange 5534 circuit was on the channel strip then it was probably mine - I don't think Dave did anything other than the mike amp - but to be honest I don't remember it at all, and of course it might have been changed after I left. I'd love to see a circuit!

And as most of the 9k was copied from the ARC which was designed by Tony Doy and Bernard (Reeve?), as well as Dave and myself, there are all sorts of hands on the design. The only bits which you can really nail down are the mic amp (Dave), the dynamics (me, but see below) and the EQ (me, but of course heavily based on Paul's and mine G series design).

I was also trying to remember - now this forum's started taking me back in time - what the 9k dynamics section looked like. Although it was my design I don't remember it at all! Was it like the 4k but using an MDAC instead of a VCA?