Reducing gain in mic input stage to match line level

[gnd]

With above measurements i forgot to make note, that direct out level is some 10dB below master out level. In order to get it to proper level, channel is amplified, and noise with it. So all measured noise levels need to be increased by 10-12db.

So channels gains, amplified 12dB, would come to this:

Without termination:
- min ch. gain -96dBA
- working gain position -95dbA
- max gain -70dBa

gnd

 

[gnd]

John.

Perhaps a good experiment would be to measure the stock line input compared to DIY pad that terminates mic in with 200 ohm instead of 2k or so.

I'm a bit lost. :? How do i make such a pad? Is this 200ohm different from measurements I did with 150ohm resistor over pins 2 and 3 on XLR mic input?

I noticed, that when putting 150ohm resistor in, line level drops considerably, some 24dB. So putting such pad on mic input makes my line input too quiet. I cannot keep those pads inside anyway. Noise measurement may be better with pads, but it has nothing to do with real life....

When mixing, I will have to work without termination on mic inputs anyway. So why measure with them?

Am I missing somethig?

Should I replace 27k/1k resistors in line pad with lower, like 5k/200ohm?

gnd

 

[JohnRoberts]

[quote author="gnd"]John.

I'm a bit lost. :? How do i make such a pad? Is this 200ohm different from measurements I did with 150ohm resistor over pins 2 and 3 on XLR mic input?

I noticed, that when putting 150ohm resistor in, line level drops considerably, some 24dB. So putting such pad on mic input makes my line input too quiet. I cannot keep those pads inside anyway. Noise measurement may be better with pads, but it has nothing to do with real life....

When mixing, I will have to work without termination on mic inputs anyway. So why measure with them?

Am I missing somethig?

Should I replace 27k/1k resistors in line pad with lower, like 5k/200ohm?

gnd[/quote]

Your mixer is already using a pad formed by the 27k/1k so yes pads exist in the real world. Yes I'm suggesting scaling down the impedance in the existing pad to something lower that you still can drive from your preceding stages.

Rather than dropping the 1k to gnd lower I would add a small R from + to - which will help CMR too. It would be pretty easy to experiment with this coming directly into the mic jack.

This is all still an exercise in how little can we get away with changing. You may be better of to just blow off the mic pre and reconfigure the one opamp into a simple diff amp. That will be pretty easy to get quiet, but you give up the gain trim.

I would get a baseline noise (measured at direct out) than start experimenting. If you can get decent performance coming into mic inputs, you don't even have to butcher your console.

JR

 

[gnd]

OK, I did experiment with two pads. Existing line pad is 27k/1k. My pads were 1.2k/68R and 6.8k/270R.

I had input simulation created from before, so I picked R values to match gain of existing pad as much as possible. Real resistor values were a bit off marked values, but were pair matched within below 1%.

Pad with 1.2k/68R provided exact the same gain as existing pad, to within 0.01dB! The other one needed few dB of gain to provide same output level (1kHz signal at -30dBfs), so it may be a bit off because of gain boost.

Results:
- reference noise level = -106.4dBA (DAW line output going into console line input)
- 1.2k/68R pad gave noise at -107.1dBA (DAW line output going into pad on mic input)
- 6.8k/270R pad gave noise at -106.6dBa, but there was need for some gain boost to have signal in it on same level as reference.

So, there is tiny improvement of 0.7dB in channel noise by feeding it through low Z pad. I guess this 1.2k pad is on limit for DAW outputs. So this 0.7dB is probably as much as I can get from tweaking mic input. I thought it'd be more....

I guess I will have to blow off mic pres, and make new line-ins. Why would I have to give up pain trim with diff amps? I could make two stages, and put existing antilog inbetween. Or make three-op-amp-input (instrumentation amp), which has one resistor gain control.
How about these?
http://84.255.203.119/mixer/in1.jpg
http://84.255.203.119/mixer/in2.jpg

But what about summing bus?
Would I maybe gain more by rebuilding sum amps. I could make balanced summing, or at least quasi-balanced with (-) bus teminated over 1k to channel ground. Buses need rewiring anyway, now lines are to thin.
Or I could go for devolved balanced mixing. Right now my new DIY console power supply is distributed over several regulators and each 317/337 regulator pair is feeding separate group of 4 channels. I could also group outputs of four channels, and sub-mix them with lower gain mix amps. Thus 24 channels would give me 6 sub amps, which could have balanced outputs, and then they would feed main balanced sum amp. Just ideas... What do you think?

gnd

 

[JohnRoberts]

ok, your test demonstrates that termination impedance makes a difference.

It also suggests to tweak the pad for less cut (requiring less make up gain to boost noise) while still delivering low input Z termination will severely load earlier output stage.

There is still the possibility to rebias mic pre input devices for lower current density so pad values don't have to go so low (may make it a little happier at higher input Z) but much easier solution to get line input quiet is just use a unity gain stage.

You could reconfigure the one opamp in mic pre as diff amp or add two opamp balanced on a add-in pc board.

JR

 

[JohnRoberts]

yup, if you want to keep gain trim, and balance, 3 opamp instrumentation topology with varible R leg between two - inputs, works.

The pot law on your console trim will work OK in that configuration but you won't need that much range.

But now you need to add parts and will increase current draw on PS.

JR

 

[gnd]

John.

But now you need to add parts and will increase current draw on PS.

Power supply is not a problem. I upgraded it already, as I already mentioned.
Originally mixer was powered by single regulation by 317/7915, as you can see in schematics. I threw that away.

My new DIY PS can now feed 5amp+ into mixer. It is made of 8 separate 317/337 boards, 6 of them feeding 4 channels each, seventh is for both master channels, and eight is for monitoring and extra needs that may pop up. PS reg boards are mounted on heavy aluminium, and each will take 1A no problem, and thats regulated 8A/15V, possibly more. Trafo is external 200VA toroid, with really big capacitors.

So PS is no problem, ready for any modifications.

I will make one input stage, and post it here.

Any suggestions on my ideas regarding mix bus upgrade, going devolved and balanced? Is it overkill?

gnd

 

[JohnRoberts]

Devolved?

Balance sum bus structure is better but there is additional parts and another bus needed. If R values are kept low noise will not increase because of it.

While I don't care for bottom up analysis it is easier to get numbers for.

Your sum bus ein will be boosted 25x in a 24 input virtual earth (approx 28dB). Plus it looks like some 6-10 10dB of noise gain from pan insertion loss make up gain, so to keep degradation of final mix noise floor less than 1dB, ein needs to be 28dB+6dB+(6 to 10) dB < target output noise (40+ dB).

Noise from channels will sum incoherently, simple power law, so 24 equal noise channels will combine to raise the output noise floor some 14 dB.

So this simple "bottom up" analysis puts a lot of focus on the sum amp, but keep in mind the noise floor will never be lower than your one noisiest source. So one real world input source may trump many dollars of tweaked out summing amps. So try to keep it real.

I would invest in a good new generation opamp for summing stage, and see how everything looks after input stages are cleaned up and with real sources.

JR

 

[gnd]

John.
Devolved or distributed mixing. It means that instead of one sum mix amp for all 24channels, you submix channels in groups of 8 or 4, and then send outputs to main sum amp. I see this as advantage in case I'd go for balanced summing. Instead of having balanced output on every channel, grouping them by 4 means 4 times less balanced drivers. For 24 channel mix that meas just 6 balanced drivers.

Here is diagram from the book:
http://84.255.203.119/mixer/distributedmix.jpg

I'm working on rewiring busses. Instead of ribbon cable, I'm designing boards with pin headers, that will fit on channel PCBs, exactly in place of ribbon cable. There are several reasons for this:
1. My PS power is distributed, one regulator board is feeding 4 channels. Thus I could power each "bus board" with one power cable, and power cable would be neatly connected to board (not hanging from cut ribbon cable, like I have it now).
2. Each channel now has 4 AUX sends, but because I use channels as stereo pairs alone, I plan to make it so that each stereo pair of channels sends left channel AUX's to AUXout1-4, and right AUX's to AUXout 5-8. Thus I will get 8 aux sends per stereo pair. So I need 8 aux buses, and that will be easy to do with 'bus boards'.
3. I will need additional bus lines for balanced mixing, and it would be really bad making it on existing ribbon cable. With 'bus boards' it will be easy.
4. On 'bus boards' I can place sub-mix amps, with balanced outputs, which will drive main mix balanced bus.
5. With 'bus boards' i can implement star grounding for console, and connect center of each 'bus board' to common star point with equall length and thick ground cable. But LED grounds I would still daisy-chain through all 'bus boards', to keep switching away from audio ground. Same for solo switching.
6. With channels grouped by four, I could also take balanced signal from sub-mix amps and carry it to main sum amp through shielded cables. Maybe this would be better than daisy chaining balanced busses.

So there are several advantages to using 'bus boards' on channels groups of four.

I will go for quietest possible sum amp. Of course, single external noisy source will stick out over all, but I can deal with it in DAW, before it enters mix.

You mention good new generation opamp. Lets say, that I make sub-mix amps, for groups of 4 channels. Then main mix amp will be mixing just 6 channels. Less channels means higher input impedance, is it. Would it matter FET or bipolar opamp? Would you have any recomendations regarding main sum opamp choice?

gnd

 

[JohnRoberts]

:grin: Must be a Devo fan... :grin: ("we are Devo"). I described that summing structure in an article I wrote back in 1980 or so on console design.... but I didn't know it had a cute name.

As I recall from my analysis over 25 years ago, there was a modest benefit in the linearity and phase shift since the opamps weren't working as hard, but the noise improvement ran into quickly diminishing cost/benefit returns.

To wit: a single summing amp runs at 25x ein or 28 dB.

2x 12 input summers combined into a x2 master would be. 25 db

4x 6 input summers combined into a 4 input master sum would be 22 dB.

So you buy 3 dB or so for every doubling of sub summers, but beyond 4x 6, you need to consider sub-sub- summers for the 8x3 outputs. Say 8x3 by 2x4 by 2in... silly.

If you determine that your sum bus is your weakest link, this is a possible approach to buy 3 dB improvements at a cost that doubles for each 3db improvement.

You could also pick up 6-10 dB using a lossless pan circuit.

Not to be a broken record, but if this noise level is already well below your program noise floor, you're just throwing money on the ground. You could instead buy beer with that money and mail it to me at my PO box ...

Regarding a specific opamp recommendation I don't have any personal experience with the new silicon (or whatever it is made from) but some of the new stuff is silly quiet with linearity too low to easily measure. My only advice is to look for a models that is decompensated. A real term, that means it isn't fully compensated for unity gain stability, and because of that will deliver somewhat better performance.

JR

PS My conclusion 25 years ago was that my current source summing approach was superior. Some people still don't believe me.

 

[gnd]

Must be a Devo fan... ("we are Devo").

Yes, evolved Devo :grin: , not the one working in a coal mine, going down down..... :grin:

I described that summing structure in an article I wrote back in 1980 or so on console design.... but I didn't know it had a cute name.

Any chance to peek into that article? Is it online somewhere?

You could also pick up 6-10 dB using a lossless pan circuit.

Is there such a thing as lossless pan in practical life? I saw a thread on panning, but ... I don't know... Panning in my console is somewhat part of summing resistor, and I need those anyway, so why not just leave it as is.

Not to be a broken record, but if this noise level is already well below your program noise floor, you're just throwing money on the ground. You could instead buy beer with that money and mail it to me at my PO box ...

:grin: Yes, I know. I have checked noise levels of other consoles, and my toy mixer doesn't seem so bad in this regard after all. I've seen Sondcraft 3200 having 84dB S/N ratio with 16 channels up. And elsewhere 92dB SNR is considered very good. So I have almost 90dB SNR. I may just do the xlr pad trick at inputs after all, with a bit less gain reduction, and thus mic stage will work at lower gain, and lower noise. Maybe just a bit, few dB. I need to think about it for a while.

I may also change channel opamps from 072 to 5532. Noise would probably be the same, hopefully not worse.... I will try on one channel and see.

But bus rewiring with 'bus pcb boards' I will do surely, if nothing else, just to gain 8 aux sends per stereo pair, and to avoid cables hanging inside console. I may do balanced and devolved, or just one of those, or maybe just quasi balanced, lets see, need to let it settle a bit.

And for final summing bus I'm not so much into experimenting with new opamps after all. I'd rather go for SSL4000 mix bus, with transistor pairs and ne5534 - I already purchased SSM2210 transistor pairs for this purpose, months ago. And on post fader output stage I will put some lundahl transformers, which I cannot seem to use anywhere - would be shame to have them, but not use them.

gnd

 

[JohnRoberts]

Any chance to peek into that article? Is it online somewhere?

I have a scan of it. Maybe this weekend I'll try to put on my website somewhere. The magazine (Record /Engineer Producer ) is long gone so I doubt they'll mind.

Is there such a thing as lossless pan in practical life? I saw a thread on panning, but ... I don't know... Panning in my console is somewhat part of summing resistor, and I need those anyway, so why not just leave it as is.

I was not making a serious recommendation, but if you're willing to Devo your bus structure for 6 dB, why not get some serious dual section, pan law pots so you don't have any insertion loss to make up?

:grin: Yes, I know. I have checked noise levels of other consoles, and my toy mixer doesn't seem so bad in this regard after all. I've seen Sondcraft 3200 having 84dB S/N ratio with 16 channels up. And elsewhere 92dB SNR is considered very good. So I have almost 90dB SNR. I may just do the xlr pad trick at inputs after all, with a bit less gain reduction, and thus mic stage will work at lower gain, and lower noise. Maybe just a bit, few dB. I need to think about it for a while.

I may also change channel opamps from 072 to 5532. Noise would probably be the same, hopefully not worse.... I will try on one channel and see.

The 5534 is lower noise voltage than 072 but much higher noise current than 072 so depending on the circuit 5532's may not be an improvement, or cause other issues. I have used both in signal paths as appropriate,

But bus rewiring with 'bus pcb boards' I will do surely, if nothing else, just to gain 8 aux sends per stereo pair, and to avoid cables hanging inside console. I may do balanced and devolved, or just one of those, or maybe just quasi balanced, lets see, need to let it settle a bit.

And for final summing bus I'm not so much into experimenting with new opamps after all. I'd rather go for SSL4000 mix bus, with transistor pairs and ne5534 - I already purchased SSM2210 transistor pairs for this purpose, months ago. And on post fader output stage I will put some lundahl transformers, which I cannot seem to use anywhere - would be shame to have them, but not use them.

gnd

I made my own discrete opamps back in the '70s-'80s (using 394s) because you couldn't buy off the shelf stuff that kicks ass like today. I'm not sure I could match some of the new stuff discrete, certainly not without a huge effort.

Regarding transformers, I mean no disrespect to Lundahl, but IMO the only thing better than a great transformer is no transformer, unless you're looking for a "transformer sound" which a really good transformer shouldn't have that much of if it's really good... :roll:

JR

 

[JohnRoberts]

[quote author="gnd"]Any chance to peek into that article? Is it online somewhere? [/quote]

OK, it was a bit of work but I have assembled scans of my old console article. While it's probably older than some posters here (I wrote it 27 years ago) it also covers some general opamp design issues that may still be of interest.

The basic design theory and laws of physics haven't changed. Since the article is basically about performance limits in console design I do find some of the references to state of the art (circa 1980) a little amusing. Perhaps also interesting while recording media has advanced hugely in the almost 3 decades since, today's analog console is rather similar performance to what was possible back then.

I corrected a few typesetting/layout errors, and there is one unearned college degree in my bio that the editor spontaneously awarded me :wink: . I even left in my sundry misspellings of 'bus', so I guess I've learned at least that one thing in the last 30 years, but still no college degree.

Below are links to the individual pages.. Comments welcome, but please keep in mind I wrote this a while ago.

JR

http://circularscience.com/des_art_1.pdf
http://circularscience.com/des_art_2.pdf
http://circularscience.com/des_art_3.pdf
http://circularscience.com/des_art_4.pdf
http://circularscience.com/des_art_5_6.pdf
http://circularscience.com/des_art_7.pdf
http://circularscience.com/des_art_8.pdf

 

[JohnRoberts]

While I haven't changed my position that bus noise is not the biggest contributor to final output noise, sum bus noise gain can contribute other nonlinearities like phase shift and distortion.

This is a little counter intuitive but by adding some extra opamps as buffers after the pan circuit you can first reduce the amount of insertion loss (from 10 dB to less than 5 dB) and second, restore that gain at the channel, where it's noise contribution will contribute incoherently into the final result as compared to putting the make up gain in the master summing amp where it’s contribution is coherent.

Specifically changing the pan circuit from the 4 10k resistors with pan pot shunting at the junction of two resistor strings, to feed the same pan pot from a pair of 6.8k resistors (for 3 dB law). Then connect the pan pot end leads to the + inputs of non-inverting gain stages running at 1.68X. After these buffers the bus can now be run at unity gain instead of the former +10dB.

A simple "all things equal" analysis reveals a noise savings but more importantly (in my mind) the summing amp is now operating at a noise gain of 25x instead of 73x! For improved all around performance.

Another possible performance tweak if you are considering adding an active balanced output (with 6db of gain) to the L/R outputs is to pull that 6dB of gain out of the summing bus to line up the clipping points. This is academic again from a noise analysis as you would drop your post sum bus fader 6 dB to line up gains, but dropping the bus noise gain to 13x from 25x makes it’s work load that much easier.

Note: this running buses down 6dB cooler after adding active balanced outputs makes sense for aux buses also. When I designed consoles I routinely ran my internal 0VU down 6 dB for consistent headroom. From a noise POV this is still all moot compared to other dominant sources, but for overall sound quality it may be worth pursuing, if you're already thinking of cutting in some circuitry to add line level balanced inputs and or outputs.


JR

PS: The post pan make up gain stage, should have it's gain setting resistor from the - input, grounded close to the pan pot wiper for best pan kill.

 

[recnsci]

First, thank you very much for article John :thumb:

[quote author="JohnRoberts"]While I haven't changed my position that bus noise is not the biggest contributor to final output noise, sum bus noise gain can contribute other nonlinearities like phase shift and distortion.
[/quote]

This is one example where discrete could outperform even best ICs, since
summing amp doesnt need to be even close to stability at unity gain (shure
thing, bus capacitance and other parasitics will be a problem here). I think
that getting 20dB more in OL gain of summing amp is quite realistic.

cheerz
urosh

 

[JohnRoberts]

[quote author="mediatechnology"]JohnR: Thank you for that article particularly the current summing approach. As you know we've discussed that now in about three different threads: This one, the one I wrote on panning and matrixing and a third one that cited my panning post.

Perhaps we should start another thread specifically to the topic. I think we agree that it has merit.

Now I get the TransAmp summing approach and why this was also done in some of the SSM2017 apps.

As to VCA current summing I did check the datasheet recently and the output capacitance of the 2180 is approx. 15 pF. Sixteen summed (paralleled) 2180 current outputs would produce 240 pF at the mix amp summing node. Phase margin in the I-V converter would need to be dealt with seeing that the input capacitance would be relatively high. But it does seem doable.

If you don't mind I'd like to start that thread and cite your article.[/quote]

Reference away.. I published this over 25 years ago so it's public domain, while there there are obviously details about actually making them practical that may not be obvious. I only had one console designer tell me he used it (in a broadcast console). Most don't believe it works or find it too unconventional.

I never made one with VCAs as the current sources but I did make one console with a VCA for the summing amp, since the noise gain there was now trivial.

As I mentioned in a response to (I think) NYD, bus capacitance can also be dealt with by making the summing amp a 2 pole active LPF where the bus C can actually be paralleled with a real C that swamps out the VCA/bus' somewhat questionable for purity output C.

I still find this approach interesting, however staying in the digital domain, if already there should in theory deliver superior performance.

JR

 

[JohnRoberts]

[quote author="recnsci"]First, thank you very much for article John :thumb:

[quote author="JohnRoberts"]While I haven't changed my position that bus noise is not the biggest contributor to final output noise, sum bus noise gain can contribute other nonlinearities like phase shift and distortion.
[/quote]

This is one example where discrete could outperform even best ICs, since
summing amp doesnt need to be even close to stability at unity gain (shure
thing, bus capacitance and other parasitics will be a problem here). I think
that getting 20dB more in OL gain of summing amp is quite realistic.

cheerz
urosh[/quote]

Perhaps, but that was the beauty of the Transamp summing bus approach. The open loop gain would increase with the number of channels assigned. So you always had a high loop gain margin with good stability.

Indeed you can make a higher open loop gain DOA, but if you have to leave unused channels connected for stability, you throw away closed loop gain just for stability. I have made consoles with well over a hundred feeds to the L-R bus, but you don't use them all for every project. Note: I have played with a work around for that where the assignment switch instead of just disconnecting the resistor feeding the bus, alternately connects an appropriate RC to ground to maintain HF stability without the noise gain in the audio band.

This is one of the intriguing things about console design. On the surface very simple circuits but start combining a hundred or more channels and it can get interesting. :wink:

JR

 

[gnd]

John. Was away few days, but now I'm back again. Really thank you for your article and all other input. I have hard time understanding it all, but I'm trying my best.

...bus noise is not the biggest contributor to final output noise...

Untill now, to reduce channel noise a bit, I used the method with pads going into mic xlr method on few channels for test. Now I can pass by with much lower channel input gain setting, and noise is a bit lower. But I used higher values, 2.2k/390E, just to be safe for units with TL072 on outputs. So there is probably no improvement due to lower input load. But channel gain pot setting is lower.

Regarding panning, I did some simulating, because I cannot do calculations by hand. And with simulation it is easier for me to follow your points.
With 10k resistors loss is 10dB at middle position, and 6dB fully panned. Changing all 4 resistors to 6.8k gives 7.5dB in middle and 4.5dB panned. Thats 3dB law as you suggested.

I need to redraw channel schematics to include opamp buffers, otherwise I will get lost.

I also do not understand this:

... connect the pan pot end leads to the + inputs of non-inverting gain stages running at 1.68X...

Connection is clear, I will draw it. But why gain exactly 1.68x? Is it to compensate for loss on 6.8k pan resistors, 4.5dB gain?

Balanced output I will make too. That will reduce fader for 6dB, as you say. Will do it for AUX's too.

I will come up with edited schematics for channel output section, so it will be easier to follow.

thnx
gnd

 

[JohnRoberts]

The voltage gain of 1.68x immediately after the pan pot makes up for the insertion loss at unity gain (when pan pot centered L & R both -3dB).

With losses restored before the summing bus, the bus combining can be done at unity gain.

JR

 

[gnd]

[quote author="JohnRoberts"]The voltage gain of 1.68x immediately after the pan pot makes up for the insertion loss at unity gain (when pan pot centered L & R both -3dB).

With losses restored before the summing bus, the bus combining can be done at unity gain.

JR[/quote]

I see. Like you explained, mix bus would work cooler, at lower gain, meaning lower noise and distortion.

If you see schematics of my console, after master fader, there is another 3x gain. Would it make sense to make this additional 3x already at channel, with channel output post-pan gain buffer? Thus instead of gain 1.68x, it could have gain 5x. Thus sum amp would still combine with unity gain, but at higher voltage output.

How about summing resistors values? If I do usual summing bus (not devolved), there will be 24 channels connected to bus. Using NE5532 for channel buffers, I could go for lower summing resistor value, like 1k? But what about current then? 24 channels at high voltage output (possibly 10V with NE5532 at 5x gain) will create substential current for sum amp. If I feed 10V on 1k sum resistors, thats 10mA from each channel, together 250mA. I would have to go discrete then on sum amp? What is max practical source/sink current for existing NE5532? Any suggestion on this?

thnx
gnd

PS: Regarding panning resistors, now they are normal carbon type. Would it be better to use metal film type? Would it matter?