Just another summing mixer

[KMB-AUDIO]

I want to build a 32 channel mixer. I made up the following schematic (link below) on combining info from NYD's schemo, API notes, and JLM audio. I'm not a circuit designer, and I do not play one on TV!

David from Cinemag had the following to say:
1. The network's impedance is a bit high which can make it prone to rfi if you are very close to a radio station transmitter.
2. The S/N ratio is not idealized. However, it is simple and inexpensive to make.
3. Note that you will optimize power supply rejection ratio if the non-inverting inputs of the op-amps are teminated through 47K resistors for the values chosen in this circuit.

He recommeneded the CMLI-15/15B as a good choice for the inputs. I also want to try edcor X series 10K:10K in this as well to see what I like more.

My questions:
1. How do I, and should I lower the impedance?
2. How do I calculate S/N? I record rock music, so it may not matter?
3. What value should I use for the master volume control?
4. What other transformers might one recommend?

I can't think of anything else to ask at this point. If anyone has input to this design to help make it better, I would appreciate it!

http://laurenstephens.net/uploads/cd32619deb.pdf

 

[TomWaterman]

I think he means your 47k summing resistors are a bit high...but you'll have to wait for someone else to comment on that, I'm not sure. Sorry. You'd have to work out the load needed for your pan pot - I think Dave posted some info on that. Not sure about SNR, I suppose you need to consider the noise gain of the mix amp more than anything. Also, I remember reading something about the 47k on the non-inverting input to ground possibly NOT being a good idea, but can't remember what I read...or where.

For RFI you could add a small inductor in series with the summing node but I think that ruins some of the benefit of near zero-crosstalk virtual earth mixing...and its not altogether a cure for RFI.

However, the reason I'm being un-helpful and posting is this!!! LOL - you are using virtual earth mixing so the mix amp will compensate for the 'make-up' gain. The inverting mix amps should have a gain of roughly 30dB looking at your schematic.

Also bear in mind that your polarity will get inverted there too, so one option would be to run your fader booster amp (listed as 'make-up gain' on your schematic) as inverting too.

Or the most simple solution would be to reverse the polarity of the output transformer primary or wire its secondary outputs the wrong way around so pin 3 and pin 2 on an XLR).

As your inverting mix amp is doing all the hard work, you don't need the make-up gain amp after the fader set so high.

Typically a fader booster amp is used here, with a gain set to +6 or +10dB.

Right now you have it doing almost 40 dB. I would scale it back to 6 dB.

Use either a really nice stereo 10k fader for master level - or if you want perfect left to right matching and don't need to do analogue fade-outs - use a 12-position stepped attenuator, essentially a 10k pot divided up into useful steps.

I'd go:

Pos 1: -20 dB
Pos 2: -12 dB
Pos 2: -6 dB
Pos 4: -4 dB
Pos 5: -2 dB
Pos 6: -1 dB
Pos 7: 0 dB
Pos 8: +1 dB
Pos 9: +2 dB
Pos 10: +3 dB
Pos 11: +4 dB
Pos 12: +6dB

Or something like that.

Good luck!
Tom

 

[KMB-AUDIO]

Ok, thanks for the tips!

So if I change the input resistor on the "make-up gain" opamps from 470R to 47K, that will give voltage gain of 2, ~6db? or is my understanding wrong?

I'm going to have to do more reading to understand the relationship between the panpot and the bus impedance. I used values from a post that NYD posted some time ago.

I planned on inverting the ouput at the transformer, it seemed like the easiest way to do it.

You were not un-helpful at all. Thanks again!

 

[TomWaterman]

Yep non-inverting resistor to 47k will give you a voltage gain of 2 in the output amp.

Av = 1+(Rf/Rg) where Rf= 47k and Rg will now = 47k

20log 2 = 6.02dB

-T

 

[JohnRoberts]

[quote author="KMB-AUDIO"]

My questions:
1. How do I, and should I lower the impedance?[/quote]
You can scale valued proportionately down.

2. How do I calculate S/N? I record rock music, so it may not matter?

Early stage noise is amplified linearly by voltage gain of following stages.
Input fader and pan is lossy. Summing amp is slight loss from fader full up and hard pan L or R. unity, Final stage has 40dB+ of gain at full up.

So input noise is padded by fader-pan, Summing amp boosts it's input noise at roughly 33x (n+1 inputs). The output stage is boosting it's own noise by 40dB and the noise of the summing amp by whatever make up gain pot is set for.

These noises combine incoherently, or the square root of( (the sum of the summing amp noise times makeup gain squared)+( the noise of the makeup gain stage squared) ). Of course all the noise on the inputs sources also combine incoherently.

3. What value should I use for the master volume control?

Probably not very significant

4. What other transformers might one recommend?

No idea.

I can't think of anything else to ask at this point. If anyone has input to this design to help make it better, I would appreciate it!

I have more than few concerns.

#1 do you need 40 dB of make up gain? Putting it after the sum bus, just amplifies bus noise too. Not to mention it's running at 40 dB noise gain even when turned down. Consider a less noisy topology at reduced gain.

I suspect that much gain isn't needed at all, only perhaps if used after a passive sum bus (>30 dB loss pasive), not active like drawn which is only losing a few dB in input controls..

#2 bus topology is pretty basic but resistor in plus input is sized for DC, not AC balance. Preferably generate another similar bus of 47k resistors from every input/fader/pan ground to sum amp + input. This too can be cap coupled and keep the single 47k at plus input but reference it's ground forward to the make up gain amp grounds.

This differential bus will improve crosstalk but won't be perfectly balanced due to changing source impedance of fader pan circuit. The actual value of the differential ground bus doesn't matter, only that the ratio between input Rs, and R referencing forward matches the ratio on negative side. Perhaps something like 6ks for channel ground send resistors and 4.7K fed forward. To keep summing amp in dc balance make the 4.7K from a 47K connected directly to plus input, and say a 5.1k on the other side of the blocking cap.

Note these values are approximate since this bus will only be in balance for one setting of pan pots and input faders.

That sure looks like a lot of work..

Good luck.

JR

 

[ruffrecords]

Your problems seem to stem from the choice of pan pot value. I can see why you used 50K, so as not to unduly load the channel level pot but this means you have to have high value (47K) bus resistors which is not good from the noise or rfi point of view.

So, I would suggest you change the pan pot to 10K and scale the other resistors accordingly. The pan pot now loads the channel level pot too much so you need to put a buffer stage between the channel level pot and the mute switch. You can now reduce your bus feed resistors from 47K to 10K. Assuming you normally have 10dB in hand on the channel level pot then you should set the gain of the buffer amp to 10dB to make up this loss.

Assuming there is a 6dB loss in the pan pot when panned hard L or R then there is 6dB of loss to make up. You can do this in the summing amp by setting the feedback resistor to 20K. From the noise point of view it is better to do it here than in the make up amp you have driving the transformer. To keep bias currents balanced you need to change the 47K resistor feeding the non-inverting input of the summing amp to 20K. I am not familiar with this particular op amp but the noise from the 20K on the non-inverting input is amplified by an amount depending on the number of inputs connected so I would recommend it is bypassed say with a 10uF and 0.1uF in parallel.

Assuming you again have 10dB in hand on the master fader then you just need 10dB of gain in the gain make-up amp.

As far as noise is concerned, the buffer amp between channel fader and pan pot should achieve an output noise of better than -100dBu with 10dB gain and a 5K source resistance (half the channel fader pot value). You lose 6dB in the pan pot so channel signal and noise both go down by 6dB (to -106dBu). The effective source resistance of the channel is little more than the 10K bus resistor and this has a self noise of around -112dBu, some 6dB less than the channel noise itself. The noise resistance seen by the summing amp is 10K/n where n is the number of channels. For 32 channels this is about 300 ohms. From the summing amp point of view this appears in series with its own input noise which can itself be represented by an equivalent noise resistance. For a good summing amp this equivalent noise resistance will be 300 ohms or less (in a 20KHz bandwidth). The total noise resistance at the summing amp input is therefore about 600 ohms which generates a noise level of about -124dBu. This is amplified by n+1 times or in this case 2 x 33 times or about 36dB so the noise at the summing amp output due solely to the bus and the summing amp is -124dBu + 36dB or -88dBU. The noise the active channel feeds onto the bus is -106dBu and is amplified by 6dB to -100dBu. So the noise at the summing output is dominated by the noise generated by the summing resistors and the summing amp input noise.

Really good op amps will have a better (lower) equivalent input noise resistance than this which means the noise at the summing amp outputs becomes dominated by the mix bus resistors. To make this noise less you have to lower the value of the resistors still further.

Note that with your 47K resistors the bus equivalent noise resistance is 47K/32 = 1468 ohms. If we change the feedback resistor to 100K to get the same gain as before the summing amp output noise becomes about -84dBu.

An alternative strategy would be to keep the pan pots as is but add a buffer amp to both the L and R outputs before the buffer resistors. You would then need to move the mute switch to the output of this buffer (and use a double pole switch of course) so the buffer amps are no feeding noise to the bus when not selected.

HTH

Cheers

Ian

 

[TomWaterman]

Good post Ian.

I enjoyed reading that. Lots of useful info.
-T

 

[KMB-AUDIO]

Great info!

NYD PM'd me the following after I posted (he said it was ok to share)...and Ian suggested some of the same things that he did. However, Ian, you suggest splitting the gain duties between the summing stage and output stage. Do you think there will be much difference in regards to noise specs?

The simple high-impedance mixer circuit I posted (from which part of your circuit is derived) is not something I would recommend scaling up to a large format such as 32 inputs, at least not without some changes. The large impedances involved can be problematic on that scale.

First off, I would buffer the output of the fader with a noninverting amplifier with a gain of 10dB, since it’s normal to operate the fader at about 10dB of loss. Then I’d scale down the panpot and mixing resistors by a factor of five: 2.2K, 4.7K, 10K pot, 10K mixing resistors. Implement Mute by simply breaking the connection from mixing resistors to L/R buses. Change the feedback resistor in the mixing amp to 10K as well.

If you choose to try a series resistor on the + input to minimize offsets, the value of this resistor should equal the resistance seen by the inverting input, which would be approximately (10k/N+1), where N is number of inputs. (The “+1” in that equation accounts for the feedback resistor around the mixing amp). So, with 32 inputs, you’d use 300 ohms. You’ll probably find you don’t even need it.

For the output amplifier, I suggest an inverting amp with a gain of 16dB (10dB master fader loss + the 6dB insertion loss of the panpot circuit).

That’s all I have time for now, but I hope it helps. Best of luck!

I don't know if I want to use an IC or discrete opamp for the input stage. Any ideas for that? Maybe a JLM hybrid...

Here is the latest design: http://laurenstephens.net/uploads/22f8b1b99d.pdf comment?

Again, I really appreciate all the info that has been shared in this thread already.

 

[solder_city]

1.5 k input resistor on the output amp will load down the master pot big time.

either scale the resistors up (if opamp will allow it) or use non inverting amplifier and flip polarity at the output transformer. daves advice is good for getting correct output polarity but you have the luxury of fudging it afterwards since youre using an output transformer.

notice you have 400uf cap at the output and 47 uf at summing amp input. that seems backwards to me, one of the smarter guys can say if im right or not

other than that it looks ok to me.

 

[KMB-AUDIO]

Man, I need to take some EE classes...

I didn't even think of that...

So if I use a 10K input resistor (matching the 10K master control?), and a 63K feedback resistor, that should be ok? (providing the opamp will allow)...if not, should I lower the value of the master control and scale accordingly?

Another question, should I use a coupling capacitor in the input stage of the output amp?

 

[ruffrecords]

[quote author="KMB-AUDIO"]Great info!

NYD PM'd me the following after I posted (he said it was ok to share)...and Ian suggested some of the same things that he did. However, Ian, you suggest splitting the gain duties between the summing stage and output stage. Do you think there will be much difference in regards to noise specs?[/quote]

With the 6dB panpot loss gain make up in the output amp, the output noise will be 6dB higher even with the master pot turned right off. However, the noise from the output amp even then should be below 100dBu so it will likely be exceeded by noise from the mix bus. NYD and I have also had a private discussion on this topic. I said mixer design is always a balance between headroom and noise and as NYD rightly pointed out headroom is probably more of an issue in the mix amp so less gain here means more headroom. So I would advise you keep the mix amp a unity gain and make up the panpot and master fader loss in the output amp.

Looking at your new circuit I would suggest a couple of things. First the channel fader wiper feeds the buffer op amp directly. This is good from the frequency response viewpoint but unfortunately it means the dc bias to that input of the opamp flows through the fader - not a good idea as any unevenness of the fader can cause scratchy noise in the output as the fader is operated. Better to include a coupling capacitor and a bias resistor after it.

Next the mute switch. It will work the way you have it but you may well get some signal leakage across the switch to the bus due to the switch capacitance so in the mute position the 10K bus resistors should be grounded to prevent this. I am never sure whether it is best to leave bus resistors on the bus and grounded or open circuit as in your schematic. In theory your method should have lower noise but I would be a bit wary of possible bus pick up when only a few channels are unmuted. A lot depends on the physical implementation of the bus itself. Best thing is to try it and see.

Next the mix amp. I see you have a 300 ohm resistor on the non-inverting input. I have to disagree with others and say the value of this should be 10K in your design. Its purpose is to ensure both the + and - inputs of the op amp receive the same dc bias current. The - input gets it through the 10K feedback resistor alone because the bus mix resistors are not in the dc path due to the input coupling capacitor so the + input should also be fed through a 10K resistor. As I said earlier to stop this resistor introducing noise you should bypass it at audio frequencies.

Lastly the master fader is carrying op amp bias current jsut like the channel buffer one and I would suggest the same alterations are made here. Also, as others have mentioned, the 1k5 resistor loads the pot too much. I would suggest using a non-inverting topology here and get the necessary phase correction by changing the output transformer sense.

Cheers

Ian

 

[TomWaterman]

[quote author="ruffrecords"]I am never sure whether it is best to leave bus resistors on the bus and grounded or open circuit as in your schematic. In theory your method should have lower noise but I would be a bit wary of possible bus pick up when only a few channels are unmuted.[/quote]

Another good post Ian!

From a purist point of view, would it not make sense to ground all unused bus resistors to ensure the buss impedance remains the same?

In a large mixer (48ch+) - I'd say disconnect to possibly improve noise if less channels are used...but for a small summing system, the purist in me says ground them so the virtual earth mix amp gain remains the same no matter what is open or not...possibly keeping frequency response and distortion etc constant at all times?

I'm sure Fred Forssell talked about that...
T

 

[ruffrecords]

[quote author="TomWaterman"][quote author="ruffrecords"]I am never sure whether it is best to leave bus resistors on the bus and grounded or open circuit as in your schematic. In theory your method should have lower noise but I would be a bit wary of possible bus pick up when only a few channels are unmuted.[/quote]

Another good post Ian!

From a purist point of view, would it not make sense to ground all unused bus resistors to ensure the buss impedance remains the same?[/quote]

I have a Neve technical Report from 1976 which compares the now 'old fashioned' Neve passive mix bus with the then 'new fangled' virtual earth mixing system (remember the NE5532 had only just appeared then). It says that a virtual earth mix bus is theoretically quieter with the unused inputs open circuit but I don't like the idea of all those inputs sitting there in the air picking up who knows what, so I would be inclined to ground them for that reason alone.

As far as bus impedance is concerned, this is mainly determined by the op amp open loop gain and the value of the feedback resistor. So with a decent op amp with an open loop gain of at least 1000 (60dB) and a 4K7 feedback resistor the virtual earth impedance is effectively 4700/1000 or 4.7 ohms. Connecting a 100 4K7 bus feed resistors across that ain't gonna make much difference to the bus impedance. And anyway the bus impedance is only important from a crosstalk point of view. The virtual earth sums the current in the bus feed resistors so it does not matter how many are connected or not (except from the noise point of view).

In a large mixer (48ch+) - I'd say disconnect to possibly improve noise if less channels are used...but for a small summing system, the purist in me says ground them so the virtual earth mix amp gain remains the same no matter what is open or not...possibly keeping frequency response and distortion etc constant at all times?

As I mentioned above the virtual earth sums the currents in the bus feed resistors. How many you have connected or not makes no difference to the virtual earth amp's gain - it is always bus feed R/feedback R for each and every channel.

I'm sure Fred Forssell talked about that...
T

I confess I never heard of the guy until now. I looked him up on the web and I have to say his articles on summing were not impressive.

Cheers

Ian

 

[solder_city]

sorry if im not following all the ins and outs of this discussion, but am i to understand that maybe it would be best to mute by disconnecting the mix resistor from the bus and GROUNDING it- instead of letting it hang in the air, leaking capacitively to the mix bus through the mute switch? i hope thats correct because it def. makes sense to me.

i dig the logic of keeping the gain of the summing amp constant to avoid changes in 'sound' but it seems a waste noise-wise on a larger mixer because how often do you find yourself assigning ALL of the inputs to the mix?

 

[jeth]

[quote author="ruffrecords"]

In a large mixer (48ch+) - I'd say disconnect to possibly improve noise if less channels are used...but for a small summing system, the purist in me says ground them so the virtual earth mix amp gain remains the same no matter what is open or not...possibly keeping frequency response and distortion etc constant at all times?

As I mentioned above the virtual earth sums the currents in the bus feed resistors. How many you have connected or not makes no difference to the virtual earth amp's gain - it is always bus feed R/feedback R for each and every channel.

[/quote]

But is it not the case that the noise gain is dependant on the number of mix resistors on the bus?

 

[ruffrecords]

[quote author="solder_city"]sorry if im not following all the ins and outs of this discussion, but am i to understand that maybe it would be best to mute by disconnecting the mix resistor from the bus and GROUNDING it- instead of letting it hang in the air, leaking capacitively to the mix bus through the mute switch? i hope thats correct because it def. makes sense to me.[/quote]

What I meant was you disconnect the mix bus resistor at the pan pot end and ground it so the mix bus is just fed with a resistor connected to ground.

Cheers

Ian

 

[ruffrecords]

[quote author="jeth"]
But is it not the case that the noise gain is dependant on the number of mix resistors on the bus?[/quote]

It is but so is the noise resistance seen by the op amp. I believe I did mention elsewhere that the lowest noise is when the muted channel's mix resistors are left open circuit, but this does mean that unless you are very careful with your screening they are liable to pick up any airborne interference - think SMPSUs and PC monitors. On balance I think it is better to ground them. Neve built desks with over 48 inputs to a bus and used 'old fashioned' voltage mixing and still achieved respectable noise figures.

Cheers

Ian

Cheers

Ian

 

[KMB-AUDIO]

Ok...I've got a new schemo here...

http://i6.photobucket.com/albums/y206/ametth/updates/new_mixer_schemo.jpg

Comments?

I've decided to use JH 990 (kit) as the fader buffer amp.

This is going to be alot of work, but I think I'm up for it.

All help is truly appreciated!!

 

[ruffrecords]

[quote author="KMB-AUDIO"]Ok...I've got a new schemo here...

http://i6.photobucket.com/albums/y206/ametth/updates/new_mixer_schemo.jpg

Comments?
All help is truly appreciated!![/quote]

OK - first the channel fader amp. I think the lower circuit is better because the upper one tries to set the gain in the feedback loop to zero when the fader is completely down and I am not sure this would work.

The lower circuit I would modify as follows:

1. The feed back resistor I would multiply by 10 to 100K and 45K. On the + input you need a dc path to ground to bias the + input so where you have put 10K here? I would put a 100K from the + input to ground - I chose 100K so as not to load the channel fader pot. and that's why I said increase the feedback resistors 10 fold so the - input gets its bias through approximately the same resistance.

2. I would include a coupling cap between the output of the channel opamp and the 2K2 resistor that is the input to the pan circuit otherwise all the op amps outputs are connected together at dc - probably not a problem but if one goes wrong and hits a rail who knows.

MIX AMP

Mix amp is OK except you need to bypass the 10K from the + input to ground - I think I have mentioned this a couple of times before ;-)

OUPUT AMP

I would suggest using the lower version of the output amp for the same reasons I recommended it for the channel amp - and you need to make the same mods to it.

Cheers

Ian