E/G channel overload indicator

it's fairly common practice to overload the channels for drums and stuff on the older SSL desks as a creative effect to give them a bit of grit - as indicated by the overload LED. i'm aware that this is merely indicating that the signal is within 4db of clipping at either post channel fader, post insert point or ch front end (pre any processing) but i'd love to know more about what is ACTUALLY happening. i'm in the market for a drum sub-mixer and i was hoping i could somehow recreate this without the need for a fully blown channel strip.

anybody got any ideas?

M

... by the way i should probably mention that i was planning on buying some old ssl channel strip sub-cards or something and cobbling them together but i was hoping somebody knew which parts of the strip are the main culprit for the overload effect.

I was aware of overdriving magnetic tape to get some free compression, but clipping solid state paths is more characterized like a raspy guitar fuzz tone, than anything sounding good for drums.

It is worth note that drum overtones are not even harmonics so any added harmonic distortion even or odd order, is not natural.

An old studio trick to add some snap to weak drum tracks is a noise gate set with high threshold and slow attack to get some intentional click when it opens.

JR

yeah... sure, i see your point. there are certain mix engineers i've spoken with who complain about the 9000s overloading a really harsh way but that like mixing on Gs and Es because they overload more gradually. I've never understood how having looked at the schematics. i guess any soft clipper would create this effect though. maybe i'm barking up the wrong tree trying to emulate this.

Hey Mark,

I'm aware of this too - Spike Stent is a big example of this style of E/G console abuse. Have you had a word with Alex? I'm sure he could shed some light as to where they really push the channel.

As far as I was aware it was driving the line input gain...hence the VHD on Duality to add some of the bite back to the superanalogue circuitry. At least that was what SSL told me.

I suppose its the clipping characteristics of the opamps as they get close to the power rails? What are they 5532/42 and TL072?

I'd be very careful about sticking some soft clipping circuits in a full channel path...maybe MosFet...but diodes won't have enough headroom and LEDs clip pretty abruptly. You may as well run your mix through a guitar drive pedal at that point.

I know this isn't the answer but I use PSP Nitro set to use just its waveshaping process as a plug-in substitution to something similar to VHD...it works pretty well IMO and sounds good.

Personally thinking about demo'ing an X-Rack for the summing part. Maybe add some E-Sig channels if you want 'that' sound. Spendy though.

Or maybe VHD modules in an X-Rack, but not the same thing.

Hmmmm not really sure. I'd love to know. DIY SSL SL-611V Channel strip? LOL

Where's Keef?

-T

Well the overload indicator is a red herring.

If you're looking for SONIC performance it's not affecting ANYTHING.

The big thing with older SSLs overloading ever-so-slightly softly is to do with the general topology. The SSL9k pre overloads unpleasantly because of it's DESIGN. If you add EQ gain and fader gain, you never hit the input amplifier overload, or else -if you do- you scissor it off later anyhow...

I'm suspicious about this because if you ask five people who THINK that they agree, when you dig a little deeper, you find that they're all making different assumptions: Some think it's the input amps (hence the 9k being unloved for this purpose) others think it's the EQ and dynamics, others will say that it's the mix summing amp...

All that I can say is that the console has a cohesive 'sound' which is largely the sum if its parts. I've racked outboard EQs before now, built and racked outboard mic pres, and racked outboard dynamics sections... None of these items singly seem to have the desired effect... I believe that it's the totality and the typical way in which the console 'steers' the user which results in the 'SSL sound'... if you're trying to bottle it in an outboard device, I don't really think that you can...

But the overload indicator is a red herring.

Keith

Hahaha why did I know you'd say that?! LOl

Nice one matey.

So is it generally slightly poor headroom that allows the waveform to softclip towards an edgy sound?

And this happens in very small amounts through out the channel?
T

I have a never-confirmed suspicion that the accumulative effect of a signal path of a hundred or so 5534s all with 10Ω 'fusible' resistors to the power rails may 'soften' the hard edge of the clipping ever so slightly, but I'd bow to more scolarly people if they call it hogwash...

The effect -if I'm not on strong drugs- would only become significant after a LARGE number of op-amp stages... so unless you want to try DIY-ing a load of cascaded op-amp gain stages to test the theory, we may never really know...

Keith

That is an awesome test tho...I may just do that!!!

Better the find out once and for all than not.

Maybe I'll order 25 5534s and veroboard a thing with probe points after each stage...go through and take a shot of the waveform, measure THD at each with various input levels.

25 unity gain stages? or with some amplification and voltage dividers in between?

-T

BTW when you say fusible resistors what do you mean?
Oh I guess I'd have to make them jumperable to you could compared clipping knee with and without...

-T

All of the various SSL stages have 10Ω resistors in the power rails, to act as fuses and burn up in the event of an op-amp power stage failure or similar disaster...

Keith

Ah thanks I'd always referred to them as 'build out resistors' - I thought it was a stability thing as opposed to fuses.

Cool - thanks for the info.

I'm really thinking about trying this out.
-T

Keith> All of the various SSL stages have 10 ohm resistors in the power rails

Tom> I'd always referred to them as 'build out resistors'

Different things.

Build-out is on the audio output. It is semi-required when driving strange loads and long wires.

Small simple systems don't "need" resistors in the power rails. Large systems have so much crap happening on the system power rail that it is wise to R-C filter it at each card. Large systems have a non-zero risk that a card will short-out mid-take... it would be better to smoke a resistor and lose one channel than to take-down the whole power system, lose the take, and set the mixer on fire.

I'm not in love with the theory "accumulative effect of a signal path of a hundred or so 5534s all with 10 ohm 'fusible' resistors", I'd expect to to have little effect; but I've never been blessed enough to abuse machinery like this.

Loading on 5532 chips may affect the roundedness of the clipping. 100K load may shear sharp, 500 ohm load may go so far and then strain roundly as nodes run short of drive current.

Overload inside EQ circuits is complicated.

Indeed, the general design of the complete board "steers" the user in good (or bad) ways. Each new board I meet needs some mind-shifting to get happy results. (Less today than in past decades.)

[quote author="SSLtech"]...I'd bow to more scholarly people if they call it hogwash...[/quote]
[quote author="PRR"](paraphrasing) hogwash (/paraphrasing)[/quote]

...Now, where's that animated 'bow' GIF?

:green:

Seriously though, the point about 'steering' is awfully important and frequently overlooked. The automatic make-up gain on the SSL (more makeup gain is automatically and progressively applied with increasingly aggressive threshold and ratio settings) tends to 'place' the signal in a certain range no matter what you do, and habit and good housekeeping tells us that the fader should be somewhere near zero.

Loading on 5534s (SSL's E and G-series op-amp of choice for audio porpoises) is generally kept light, but still.... who has really studied the detailed cumulative effects of three-digit serial chains of these things?

Keith

I used to use 10 ohm/100nF decoupling about everywhere as a precaution, in designs that weren't absolutely minimal cost. It's a bit like one of those Madman Muntz situations though: you could probably go in and remove a number of them with no evident effect on performance. But none at all, and things would go awry in systems with lots of gain here and there.

The subject of a cumulative effect from a long chain of amplifiers came up recently with a high-end friend. He said that a designer at Audio Note didn't like to use the same tube in too many places in an amplifier, lest the "sound" of that device became too evident, and referred (the friend, not the designer) to this effect as "sonic signature buildup". I couldn't resist a bit of a tease and said that phrase sounded like a detergent ad :razz:

Since mere measureable distortions, particularly as measured with sinusoids, often get fairly short shrift in audiophile circles, as a guide to how things sound, I didn't expect my argument that cascading stages can actually cancel at least second harmonic would carry much weight.

In general opamps are 5 terminal devices and careful inspection of data sheets will reveal a relationship between PS terminal voltage and output (usually referenced to the inputs). A PS terminal RC will LPF voltage generated there by supply current. In an ideal world if load in linear and both PSRRs symmetrical this is pretty innocuous. (Note: PSRR is not usually identical so a very small distortion term is expected.) This phenomenon is probably hard to measure in the presence of other larger effects. Of course hard clipping reduces PSRR to nil as the output follows it's PS rail voltage.

One potential cumulative effect from running through a long series of opamps is phase shift. While rarely discussed the open loop response of opamps is typically lagging 90' due to the stability compensation capacitor across an internal gain stage. Negative feedback pulls this back toward in phase with the input but not completely as limited by the amount of loop gain margin.

For argument, if one opamp stage corrected the 90' to only 1' of lag at say 20 kHz, and you run through 100 stages in series, you are now up to 100' of phase shift! The reality depends on average loop gain margin across all those stages. (EDIT: The reality depends on average times N)

I have never given much credence to the voodoo audio products that add some phase lead to make up for imagined lags elsewhere, but maybe in a path with that many opamps in series? Who knows? Not me... :?:

JR

For a good example of time/frequency domain cumulative shifting effects, listen to Steve Reich's ancient tape piece Come Out, which takes the short tape fragment and repeats it many many times through the same system.

The designer of the Neotek consoles used to brag that their gain stage design was totally transparent, and to "prove"it, they claimed to have cascaded 100 of them together and compared that to the original signal. They claimed the only difference was a slight increase in the noise floor due to the 100 opamps. A point of interest, most of the opamps in the Neotek were TL0 types, and used a transistor feeding the non-inverting input. A little buffer stage, I guess.
Neoteks were, and still are great sounding boards, and very different from Neve and SSL. A very "American" type of sound, if there is such a thing.

[quote author="Jim Zuehsow"]The designer of the Neotek consoles used to brag that their gain stage design was totally transparent, and to "prove"it, they claimed to have cascaded 100 of them together and compared that to the original signal. They claimed the only difference was a slight increase in the noise floor due to the 100 opamps. A point of interest, most of the opamps in the Neotek were TL0 types, and used a transistor feeding the non-inverting input. A little buffer stage, I guess.
Neoteks were, and still are great sounding boards, and very different from Neve and SSL. A very "American" type of sound, if there is such a thing.[/quote]

There are hybrid topologies involving a discrete device in front of a TL0 that deliver constant (high) loop gain margin for excellent phase shift and linearity performance. I first saw the topology used by Buff, but there were other variants around before him, and discrete designs can dial in as much open loop gain as a known closed loop gain might require. Console designers typically use such approaches only at the "heavy lifting" areas like mic preamps and bus summing amps, due to cost and real estate involved.

I am reluctant to start discussing "sound character" of the sundry desks, but yes they did have them.

I am not a student of the sundry schematics but if the SSL had that many off the shelf opamps, even decent ones for the day, in series there could be a measurable sonic impact.

Another perhaps important point, my phase shift comment (caused by inadequate or cumulative loop gain margin) doesn't fully characterize the error. It is the nature of NF related errors to be the inverse of the forward transfer function, in this case the stability compensation integrator. The error for that is the electrical derivative of the signal. For a simple sine wave this is a phase shifted sine wave but for complex waveforms this derivative is less benign, while likely down a few tens of dB below the signal and subtle.

Appropriate LPF of the audio path early on will keep some nasty stuff out of this cumulative error term.

JR

My specified three-digit op-amp count should perhaps be clarified. I was guessing the number of audio-path op-amps in a round trip from mic pre to tape, then back from tape to the mix bus...

I've just pulled a late 1980's module onto my knee, to make an accurate count:

Twelve op-amps on the input card: 9 of which are signal-path 5534s

Two 5534s on the dynamics card (and twenty non-audio path op-amps in the DC and metering sections)

Thirteen 5534s on the EQ card.

Two 5534s on the fader card. (and a few other non-audio op-amps)

Fifteen 5534s on the channel amplifier card.

That's a total of forty-one 5534s per channel (and about thirty non-signal path op-amps).

Two trips through that lot is about eighty 5534s, then the center section of course adds a good chunk more. -In the case of a 6000 (E or G-series), the A/B/C mix bus summing setction (the SL651) and then the master program section, compressor/master fader, signal output buffers and distribution etc.

I'd say that probably takes it to well over a hundred on both the 4000 and 6000 series, but if you're just going 'once through' from input direct to the stereo bus, that drops to sixty or so.

If you switch out the EQ and filters, you lose plenty of op-amps, and obviously bypass the record bus summing stages etc.

In fact, the I/O module (when used with just a mic input, with no EQ, no dynamics, no insert, going straight to the mix bus) only has about four 5534s in the signal path... but there's really no reducing what happens at the center section! -Does the module sound significantly different if the EQ (set flat) gets switched out?

I'd really like to take a Neve or an API module, power it and resistively feed the stereo outputs onto the mix bus, then see if it has some of that classic SSL 'softness... -It wouldn't surprise me if it did...

So did I just undermine my own suppositions? -I suspect that I did.

But until we know, we may never know... :wink:

Keith