functional enhancements to my console

[Rob Flinn]

Hello Folks

I've been thinking about some functional enhancements to my MCI 428b console.    One thing I like to do is use stereo compressors on subgroups.  The problem is that the MCI 428b doesn't have subgroups.

My idea was to make a stereo return module that would occupy the space below where the patchbays are.  That way I could create a stereo sub mix through the tape routing pairs (my desk has a mod that enables panning to the tape routing pairs), come out of the line outs/tape sends on the patchbay through a compressor & into the added stereo return module.

Since the MCI has quad outputs that are summed to make the stereo output I was considering disconnecting the back bus from the back summing amps & using them to sum my stereo returns which then get summed with the front bus to make my stereo bus. Is this a good idea or should I make a dedicated summing amp for the stereo returns.  I don't use the back bus for anything else so I just figured that this would use it for something.

I was thinking I could hook the balanced inputs to the stereo returns parallel to the tape sends on the patchbay.  That way I would have say the 1st 10 pairs of tape routing also normalised to the stereo returns.  This gives me an insert point to them. Since when tracking I would not use the st subgroups & when mixing I wouldn't be sending to tape I can't see a problem.     Question - would this give me an impedance issue paralleling them to the tape/PT inputs.  

I would like to have a solo function on the subgroups.  I was thinking that I could have a standard solo function for the stereo pairs using relays for each pair & another relay to switch off the mono channels by disconnecting the front bus.  The only problem with this is that I think if I solo a st pair & a mono channel I think the mono channel won't be heard. So I'm trying to find a way to make it part of the whole solo system on the console, needs a bit of thinking about. .

Anyway above is a preliminary diagram for one half of a st pair.  The resistor after the 220u & before the mute switch is the 10k summing resistor.

Things I'm not too sure about are the values & polarities of the caps, & if I actually need the 2u2 at all considering that the dc is blocked by the 100u preceding the fader.

I would add that I have taken the basic building blocks from Doug Selfs Book, small signal Audio Design.   Some of the things like the muting switch arrangement look a bit radical, but according to Doug this is a lower noise than decking the 10k with it still connected to the bus.    If anybody has any arguements for other topologies, critisms or enhancements I would be interested to hear them.

 

[abbey road d enfer]

I was thinking I could hook the balanced inputs to the stereo returns parallel to the tape sends on the patchbay.  That way I would have say the 1st 10 pairs of tape routing also normalised to the stereo returns.  This gives me an insert point to them. Since when tracking I would not use the st subgroups & when mixing I wouldn't be sending to tape I can't see a problem.     Question - would this give me an impedance issue paralleling them to the tape/PT inputs.

None at all, in particular considering the type of diff amp you've chosen, which has a very high input Z.

I would like to have a solo function on the subgroups.  I was thinking that I could have a standard solo function for the stereo pairs using relays for each pair & another relay to switch off the mono channels by disconnecting the front bus.  The only problem with this is that I think if I solo a st pair & a mono channel I think the mono channel won't be heard. So I'm trying to find a way to make it part of the whole solo system on the console, needs a bit of thinking about.

I must admit I don't get your description; can you sketch something?

Things I'm not too sure about are the values & polarities of the caps, & if I actually need the 2u2 at all considering that the dc is blocked by the 100u preceding the fader.

Values are correct, polarities are indifferent, the voltage they will see is just the offset of the opamp, a few millivolts... The 2u2 is not indispensable, but is recommended since the offset and noise current of the opamps can be modulated by the fader, resulting in some noise.

I would add that I have taken the basic building blocks from Doug Selfs Book, small signal Audio Design.   Some of the things like the muting switch arrangement look a bit radical, but according to Doug this is a lower noise than decking the 10k with it still connected to the bus.    If anybody has any arguements for other topologies, critisms or enhancements I would be interested to hear them.

Nothing wrong with this topology, it's a well-proven design. But you need a DC leak resistor at junction of 220u and 10k (about 100k to ground).
The 820R resistors are a tad low. I would use 2-5k. And put 100pF caps on the FB resistor and foot resistor.
Post-fader amp is wired for isolating output from any reactive load. Not necessary here.
You've made it unity-gain. There is usually about 10-12dB gain there. A marginal advantage of having gain is somewhat better high-level transient response.

 

[JohnRoberts]

Life is too short to discuss all the different possible sum bus variants in a brief post, so I won't.

Re: your schematic, I didn't really closely follow for dialog, but if those inputs are interfacing with the outside world, you might want to cap couple, add some C to ground (maybe with a small series input R to catch RF) and even add some simple diode clamps (to protect against static or whatever. In general we ASSume that outside boxes could exhibit DC errors and the outside world is always noisy and full of threats.

820 ohms seems low impedance, while within the drive capability of the 553x . You could afford to scale up a few times without compromising your noise floor. Note: those mAs of signal current being dumped into your audio ground will need to go somewhere, so layout and ground management will be important to keep crosstalk low.

For a really tweaky design I might be inclined to send to the meter through a different cap, or cap and resistor, especially if you are talking about a real VU meter. that nonlinear load will also be in parallel with second stage feedback network so coming out of the 553x drive current available. If the meter is a high linear input impedance, never mind, while I'd rather not corrupt the audio path.

The unity gain NI 553x stage with 47k feedback resistor will cancel out DC offset, which is not really a problem when output is cap coupled (without it you actually can predict a DC output voltage to bias a polar cap). Further the 47k Johnson (thermal) noise is in series with the input signal, and the 47k forms a pole with the 553x input capacitance reducing stability margin. IMO you could probably lose that 47k entirely.

The technique of taking the negative feedback from after a small build-out resistor is useful to decouple from large capacitive loads (which you should not have there) and improve stability in those high capacitance applications. I am not familiar with the extra RxC pole loading the output, (perhaps a Self thing?). It looks like a more common compensation technique used in a power amp than on an IC.

Console designers, generally put 10 dB of gain in most fader stages, which would also improve the stability of that stage (even allow use of a decompensated 5534 for slightly better performance).  

You might be able to substitute in more precision modern ICs than the 553x and lose some of the DC blocking caps (the best cap is no cap), but the 553x when properly applied is a good soldier, relatively inexpensive, and will respect your audio signal.

    JR

 

[Joechris]

If you use tielines as input for your subgroup mixer, you could go for a simple "passive" style mixer, with unball inputs. Just a stereo fader and a 10k sum resistors. You could sum this with your front(1,3) or back(2,4) sum. I always just used output 1 and 3 as main out to avoid the extra stereo suming stage. I see that MCI carefully loads the input transformers with a 600ohm load, and since they use the same transformers on the output, mabye you should try to load the stereo 10k fader with a 1k in parallel to see if it keeps transformer ringing low. Unball is not so bad, API has done it with great results... Stereo solo...not so simple. Mabye just sum subgroups with with buss 2 and 4 and kill buss 1 and 3 with a relay, and use a solo inplace relay system on your submixer.. late nigth thougts
j

 

[Rochey]

Dumb question - more for my knowledge than helping you:

Why create your own instrumentation amplifier with the 3x 553x's?

Wouldn't it be easier to use an ina134 or heaven forbid, a THAT equivalent?

Cheers

/r

 

[PRR]

> You've made it unity-gain.

I read it as gain of 47K/2K2 or 22, 27dB.

> usually about 10-12dB gain there.

Agree. Unity after the fader is too low, 27dB is rather high, especially since levels are already "known", no large boost should be needed, but we want a little reserve gain in hand.

> Wouldn't it be easier to use an ina134 or...?

He still needs the post-fader booster. So it's still two chips. A cup of 5532 is cheaper than a half-cup 5532 and a half-cup of INA/THAT chips.

It is possible the input stage can overload. If driven from a push-pull balanced driver it could have +13V on "IN+" and -13V on "IN-". The third opamp must swing to -26V, which it can't. Perhaps the source won't swing to its rails, probably the case here. In general, if the source is not under your control, the diff to single-end stage should have gain of 0.5.

I would prefer another diffamp, 2-opamp, but it would still be two chips. Actually you could put the fader in the NFB loop and get it down to one dual-opamp chip per channel.... but a wiper-skip goes MAXX! gain instead of zero signal, so it is frownable.

> I am not familiar with the extra RxC pole loading the output

Odd, mis-copied, or not applicable.

Right away, 100pFd against 47K is -3dB at 35KHz, -1dB at 17KHz, which matters to some people.

Below 30KHz, the output impedance is near-zero for good line drive. (But then we pad it up 10K?) Above 40KHz, the opamp is isolated  by 68 ohms from the line capacitance (but we only have a 10K resistor?).  

 

[JohnRoberts]

I didn't even consider the possibility of 27 dB gain. :-[ 0 dB on the fader would sure end up pretty low..  

+1 to PRRs suggestion for -6dB in the diff amp,,,  I usually did that too. I like everything to clip at the same time. You usually pick up +6 db swing in active balanced outputs so running everything around inside at -2 dBu nominal 0VU worked for me.

JR

 

[abbey road d enfer]

Note to Rob. If your intention is 47k, 2k2, 100p and neg input of opamp to be connected together, you should put a large blob there.

 

[Rob Flinn]

Thanks you very much for all your responses, they have really got me thinking.

Addressing a few comments :

re the fader buffer gain I made an error in the value of one of the resisitors.  Should be 4k7 & 2k2  which should give 10dB.    I will also up the 820's on the differential amp, possibly also giving it 0.5 gain as in PRR's suggestion.

Re: your schematic, I didn't really closely follow for dialog, but if those inputs are interfacing with the outside world, you might want to cap couple, add some C to ground (maybe with a small series input R to catch RF) and even add some simple diode clamps (to protect against static or whatever. In general we ASSume that outside boxes could exhibit DC errors and the outside world is always noisy and full of threats.

The unity gain NI 553x stage with 47k feedback resistor will cancel out DC offset, which is not really a problem when output is cap coupled (without it you actually can predict a DC output voltage to bias a polar cap). Further the 47k Johnson (thermal) noise is in series with the input signal, and the 47k forms a pole with the 553x input capacitance reducing stability margin. IMO you could probably lose that 47k entirely.

The technique of taking the negative feedback from after a small build-out resistor is useful to decouple from large capacitive loads (which you should not have there) and improve stability in those high capacitance applications. I am not familiar with the extra RxC pole loading the output, (perhaps a Self thing?). It looks like a more common compensation technique used in a power amp than on an IC.

JR

Hi John

What sort of value should the input caps ?  If I use a 2u2 for the input cap that gives me a cut off F of 1.5Hz or should I make it larger ?

Also the cap to ground on the input ?  100pF ?

I have changed the fader buffer for the next revision of my circuit, because having read your reply & refering to Doug Selfs text again, like you said it was for driving high capacitive loads "in some layouts the panpot, routing, & post fade sends can be physically spread out so that stray capacitance can be seen ..............".  Obviously not the case here.

Where you say I could lose the 47k around the unity gain stage are you talking about the fader buffer op amp +i/p to ground 47k ?  In Dougs book he says that because the 553x has a significant bias current means that yopu need blocking cap (2u2) & because of thsi the 47k to ground is needed to bias the op amp.  I'm don't really understand what he means by this, but that is the reason I included it.

It is possible the input stage can overload. If driven from a push-pull balanced driver it could have +13V on "IN+" and -13V on "IN-". The third opamp must swing to -26V, which it can't. Perhaps the source won't swing to its rails, probably the case here. In general, if the source is not under your control, the diff to single-end stage should have gain of 0.5.

Point taken, I'll have a think about this.  I remember somewhere in the MCI 500 there is an op amp configuration that allows a +&-18v op amp to swing between higher voltage rails in symapthy with the i/p signal to extend it's output capability.  Maybe that's a possibility because the console has +&-24v power rails.



Rochey

The point of using 2 x 5532 rather than a dedicated balanced line receiver is three fold.  Firstly aa PRR states the whole channel rather elegantly uses 2 x 5532, secondly it's a lot cheaper than using a dedicated bal line receiver, & thirdly according to mr Self's book it's actually less noisy or around the same noise floor depending on which dedicated bal line receiver is used.  Obviously the CMRR depends on making sure the differential amp resistors are tightly matched.  I'd love to use transformers because most of the i/o on my console is transformer balanced, but doing these stereo subgroups like this, I will still have all the colour I need from the channel strips, so I figured I might as well make this part clean(ish) sounding.



JoeChris

The reason I was thinking of using the back bus summing amps rather than adding it to 1-3 summing chain was that I thought I could take advantage of some distributed summing.  I know the console isn't exactly huge, but these channels will be in a different part of the console to the main channels & I was suspsecting doing it this way would give me less problems. 
I understand your suggestion to strap a fader across the tape send, slugged with a 1k.  However, since the tape send is transformer fed & could be paralleled with a variety of recording inputs tape deck, Protools, or something like an La2a if I inserted a compressor I think I'd rather avoid this topology to "try" to avoid issues.

The solo capability I was thinking on the same lines as you, but I reckon it would be possible to use the same solo sense line as the normal channel strips, so if I use bus 2-4 for the solo path of this subgroup mixer it should be possible to get this to work.


Below is my Mk II diagram.  I have added input caps & RF filtering caps to the voltage followers & the differential amp.  Dropped the gain of the differential amp to -6dB.  Changed the fader buffer to a simpler design with +16dB of gain to give me a 10db boost on the fader & offest the drop from the differential amp.  I also added the 100k resistor that Abbeyrdenfer suggested before the summing resistor.

Things I'm not sure about

The value of the i/p caps & rf filter caps on the 47k.
The exact values of the new gain resistors on the fader buffer.  I just plucked them out of thin air to give me +16dB whilst trying to keep them reasonable low in value.
Any other comments welcome ?

 

[JohnRoberts]

Thanks you very much for all your responses, they have really got me thinking.

Hi John

What sort of value should the input caps ?  If I use a 2u2 for the input cap that gives me a cut off F of 1.5Hz or should I make it larger ?

In general I target 1 decade below passband. A real pole at 1.5Hz will be 1/10th dB at 15 Hz so quite respectable. That said in a console with many such HPFs the effects are cumulative so you need to look at the combined effect of your entire path.  Another consideration I should have mentioned, when you add poles here, the matching of these caps (can) become a CMRR issue at hum frequencies. The pole being well below mains frequency helps reduce the sensitivity to mismatch, but these caps are typically only 20% tolerance. If you are comfortable that you won't see significant DC at these inputs, you can blow off the DC blocking.

Also the cap to ground on the input ?  100pF ?

Again there are complexities. Making this cap large can destabilize poorly designed gear with the cap to ground. Also the pole frequency will be unpredictable depending on the source impedance of sundry external boxes.  In my experience I have seen source impedances as low as zero (well not really zero), and as high as 2k for some consumer gear. In practice you should see in the range of 50-600 ohms, so 100pF will be suitably high. If worried about AM radio, you might want to add some series input resistance, to make sure the pole is low enough. Of course these networks and values need to be matched for best CMRR at HF. Series R can affect all CMRR. 

I have changed the fader buffer for the next revision of my circuit, because having read your reply & refering to Doug Selfs text again, like you said it was for driving high capacitive loads "in some layouts the panpot, routing, & post fade sends can be physically spread out so that stray capacitance can be seen ..............".  Obviously not the case here.

I have mixed feelings about that "decoupling inside the feedback loop" topology. It's been decades since I looked at it closely but IIRC there was no free lunch. Yes it arguably delivers flatter amplitude response, but the gyrations the opamp is going through gets reflected into the input referred error voltage. This is a subtle and esoteric point so don't lose any sleep over it. My personal preference is just to use a simple series resistor if appropriate, but apparently opinions vary. . 

Where you say I could lose the 47k around the unity gain stage are you talking about the fader buffer op amp +i/p to ground 47k ?  In Dougs book he says that because the 553x has a significant bias current means that yopu need blocking cap (2u2) & because of thsi the 47k to ground is needed to bias the op amp.   I'm don't really understand what he means by this, but that is the reason I included it.

This comment was based on my apparent misreading of your schematic.

Looking at the revised version your gain is closer to normal for that position. The choice of how much gain to put here, depends on what the output stages of you console looks like. If you use active balanced outputs with +6dB voltage gain, I would stick with only 10 dB here, if your output stages are unity voltage gain, then +16dB is reasonable so you can hit the bus hard from weak sources.

re: input bias current, yes the 553x is an old school bipolar opamp, without first order bias current compensation, so there will be a roughly equal DC current flowing into both input pins due to the NPN LTP transistors. Modern bipolar opamps typically cancel out this bias current. To cancel the DC error caused by this current you need the DC resistance to ground (and the opamp output) seen by both input pins to be equal. 

In your schematic the DCR is 47K at the + input, but 10k in parallel with 1.8k at the - input so clearly mismatched. Note: in your first schematic, if the 2.2k resistor was cap coupled, then both inputs would have seen 47k and balanced out- or if the 47k was the only feedback R to minus input, like I apparently misread your schematic, but the first schematic was mismatched too with - input seeing 47k in parallel with 2.2k.

But not to worry... since you are using a DC blocking cap, this predicted DC offset will give you a working DC bias voltage on the output capacitor. That opamp output will be sitting a modest negative DC voltage so orient the output cap with it's + terminal away from the opamp. 

Things I'm not sure about

The value of the i/p caps & rf filter caps on the 47k.
The exact values of the new gain resistors on the fader buffer.  I just plucked them out of thin air to give me +16dB whilst trying to keep them reasonable low in value.
Any other comments welcome ?

Those values look reasonable +/- my other comments above.

Some of this could be considered over thinking this. Many of my suggestion, are to deal with more extreme inputs than you may have to deal with.

JR

 

[Rob Flinn]

John

Thanks again for your response.

The reaso I made the fader buffer +16dB was because I changed the differential amp gain to lose 6dB to avoid the possibility of clipping the output as in PRR's comment.

I'm still not really getting the DCR thing in the fader buffer.  are you saying that the combination of the 47k (i/p to ground) & the 1k8 in my mk II diagram should be equal to the 10k feedback resistor ??
Or if I used a cap in series with the 1k8, the 47k should be an equal value to the 10k ?
I'm guessing the 47k shouldn't be a particularly low value ?  But I think that the gain resistors for the fader buffer are better to be a lowish value to reduce the Johnson noise ?

regards

Rob

 

[JohnRoberts]

John

Thanks again for your response.

The reaso I made the fader buffer +16dB was because I changed the differential amp gain to lose 6dB to avoid the possibility of clipping the output as in PRR's comment.

I'm still not really getting the DCR thing in the fader buffer.  are you saying that the combination of the 47k (i/p to ground) & the 1k8 in my mk II diagram should be equal to the 10k feedback resistor ??
Or if I used a cap in series with the 1k8, the 47k should be an equal value to the 10k ?
I'm guessing the 47k shouldn't be a particularly low value ?  But I think that the gain resistors for the fader buffer are better to be a lowish value to reduce the Johnson noise ?

regards

Rob

No the DC resistance to ground and/or the opamp output must be the same in both the + and - inputs.

With 47k in + input, you would need the feedback and R to ground in parallel to equal 47k for DC balance. This is impractical. Putting a cap in series with the leg to ground, takes it out of the equation, so 47k feedback with any R to ground would balance, but like I said, with DC blocking cap in the output, a little DC offset that you can predict is not a bad a thing so you can orient a polar cap.

JR

 

[Rob Flinn]

So d.c meter the quiescent output of fader buffer to measure offset & orientate the 220u accordingly ?

 

[JohnRoberts]

So d.c meter the quiescent output of fader buffer to measure offset & orientate the 220u accordingly ?

As I already said, we can predict which way the output will go.

Typical bias current is -200 nA (into the input pins)

So, using Ohm's law, -200nA X 47k = -9.4 mV  at the + pin.

Since the opamp has active negative feedback we know the - pin also be forced to follow the + pin down to -9.4 mV. 

Again using Ohm's law we can calculate how much current will be flowing in the 1.8k resistor between ground (0v) and that -9.4mV - onput, or I=-9.4mV/1.8k = 5.22 uA.

Applying Kirkoff's current Law, we know that the sum of currents entering and leaving that node net out to zero. So we have 5.22 uA, flowing down the 1.8k and "into" the node, with 200 nA being sucked "out of" the node by the - input. This means that the balance of 5.02 uA must be pulled out of the node through the 10k resistor.  Ohm's law tells us this will drop 50.2 mV.  Using superposition, we subtract this 50.2 mV from the -9.4mV that both + and - pins are sitting at, giving us a nominal -59.6mV at the opamp output.

It is a little silly to calculate this to too much precision since the bias current, worst case could be 4-5x higher, but i feel safe in predicting which direction to point your capacitor. Of course feel free to measure with your VOM if you don't trust me.

JR 

PS: It was fun using all those fancy terms, but this was routine design math back in the day when using these opamps. Nowadays modern opamps tweak out the bias current completely, leaving only tiny error terms. 

 

[Rob Flinn]

John,

I trust you, believe me, I'm just finding it hard to follow the theory, you're taking it further than the level I studied to !  I should probably look at the theory more, so I can work all this stuff out for myself, & stop having to bother you & all the other professionals on this forum.

Thank you again for your help !

 

[JohnRoberts]

If I was bothered I wouldn't post answers.

I am mostly self taught, and despite throwing in some $2 words, the concepts are pretty simple, ohms law, etc. If my explanations are confusing ask more specific questions, but don't give up.. if I could figure this stuff out, you can. 

JR

 

[PRR]

> DC resistance to ground and/or the opamp output must be the same in both the + and - inputs.

For zero output DC error.

But we don't care about small DC on the +10dB stage. There's a 220u cap(??) blocking it from the mix resistor.

The key problem is: the 5532's input bias current can be a large part of a microAmp. If this flows through the wiper of the fader-pot, it will scratch.

It's not fatal. If the fader is rarely moved, and never "live", the scratch is tolerable. I've used PA mixers with worse gain-pots and nobody fired me. Since his input is probably nearly DC-free, and he cap-blocks the output, a volt or so of stray DC is an insignificant reduction of headroom.

The 5532 could have 5(?)mV offset voltage, times 7, is 35mV output offset (either way). Enough to block from the next box's input, not enuff to hurt headroom.

5532's input bias current is under 1uA. Try a 1Meg input resistor. 1V input offset! Times 7, 7V output offset! Half your headroom shot.

Try 50K input bias resistor. 20 times better, 0.3V output offset, quarter-dB loss of headroom, bah, no problem. We may look at 5532's input bias current direction, I *think* the input will rise positive, which suggests which way the output cap should go, but at 0.3V it isn't real important.

The 10K||1K8= 1K5 equivalent resistance on the other input causes a smaller error, 1.5mV, in the same direction. Yes, if we made the NFB network 33 times larger 330K+60K, the bias current error cancels.... but those are large noise resistances. 50K+9K+cap also cancels, but needs another cap. And why try to cancel a negligible DC error?

Being cheap and lazy, for a small loop-back in a small studio, I'd perf-board one of these and see how bad it sucked----

 

[Rob Flinn]

Thanks PRR.

I understand your point about the simpler circuit.  One thing I've tried to keep in mind is to build something that is in sympathy with the 1978 console rather than using an "Upgrade" approach just for the sake of it.

That said, my circuit in total (disregarding cost of fader) is significantly cheaper than a single ssm or that's balanced line receiver chip.

What I did like with your circuit is the simplicity in making the input trimmable, which is a useful feature......

I need to ponder a bit more I think ..... 

 

[JohnRoberts]

While this is a pretty esoteric concern, so take it with a grain of salt, I actually prefer the symmetry of your simpler input, especially if you don't heavily LPF the inputs. The inverting stage effectively in series with one of the inputs means that there will be some lag in that one path, not in the other. This should be insignificant for audio frequency signals, but don't expect an unfiltered common mode square wave or RFI to make it through the inverter perfectly unscathed to cancel out (also any roll off across the inverter feedback R will degrade HF CMRR).

Of course good design practice is to LPF at the very inputs which will also reduce sensitivity to this different path lag in only one leg.

JR

 

[PRR]

> there will be some lag in that one path, not in the other

If I understand the requirement: the whole loop is a few feet inside a working audio studio. Therefore there "should not" be any heavy RF to lag and fail to CMR.

Of course stuff happens, two chips is hardly expensive, my fader may scratch and will AAACKKK! if the wiper loses contact.... things the designer/user must balance for the situation.