Want to Build Audio Summing Mixer

[Alcratin]

So a friend of mine that I work with just got a HD3 system and asked me if I would help him build an Audio Summing Mixer, now that he spent all of his money on the Digi system.

I'm familiar with the summing circuit using resistors and a non-inverting portion of an op-amp, but after looking at the Speck, SPL, and others, can anyone recommend or suggest the use of isolation transformers on input and output? Is this necessary to reduce noise floor?

Thanks in advance.

Alcratin

[Geoff_T]

Originally Posted by Alcratin

So a friend of mine that I work with just got a HD3 system and asked me if I would help him build an Audio Summing Mixer, now that he spent all of his money on the Digi system.

I'm familiar with the summing circuit using resistors and a non-inverting portion of an op-amp, but after looking at the Speck, SPL, and others, can anyone recommend or suggest the use of isolation transformers on input and output? Is this necessary to reduce noise floor?

Thanks in advance.

Alcratin

Hi

The reason folk use transformers in their designs is because the transformer floats your circuitry from the grounds of the external circuits you connect to. It also provides a balanced input with good common mode rejection so your wiring will be less susceptible to noise. Also, transformers do not care if they are connected to balanced or unbalanced circuitry though the common mode rejection will only work effectively with balanced signals.

This is not to say that transformers don't have their little foibles nor that you can't perform a balanced mix without transformers.

If I were you and you were doing this on a tight budget, I'd consider doing a balanced mix into two inverting amplifiers, then feed the outputs of those two amplifiers into a third amp's differential input....

There's several ways around this... if it were vintage Neve it would be transformers everywhere... but look at the price!

[Tim Farrant]

Originally Posted by Geoff_T

Hi
If I were you and you were doing this on a tight budget, I'd consider doing a balanced mix into two inverting amplifiers, then feed the outputs of those two amplifiers into a third amp's differential input....

Yeah this is a good idea, then you could use just 2 transformers to balance the output. I would recommend the Sowter 8403X, it has good performance without the need for much compensation. I think a transformer on every input would be excessive and unnecessary, might colour the sound too much, not to mention the price!

Tim

[Tim Farrant]

Below is a simple circuit that could be used for summing balanced outputs. There is however a minor error in my drawing - see if you can spot it!

The opamp can be another type, but good noise and output drive performance is critical. The OP275 will run on +/-20V rails for a bit more headroom. The 22k resistors should be 1%. The 100uF output caps should be bipolar bypassed by the 1uF polyestors. This circuit should be flat to 100kHz.

Cheers
Tim.

[LRRec]

Inverted polarity?

[Tim Farrant]

That was quick! Yes indeed, because of the inverting mix amps, the output polarity should be the opposite of what I have drawn.

But sorry LRRec, no prizes this time!

Tim.

[LRRec]

Tim,

I'm curious about the 100R resistor going into the summing node. I assume it is there to buffer any capacitance on the buss. I have also seen in schematics the use of a low value (~49ohms) resistor buffering the inverting input but also within the feedback loop. Is one more advantageous than the other?

[Tim Farrant]

mmmm, good question. I suppose I just add it because its good design practise. It is probably more important where the summing buss resistors are a long way away from the opamp, like in a console. The 100 ohm and the input capacitance of the opamp form a LPF which helps keep the circuit stable. I am not sure if putting inside the loop would be beneficial

Tim.

[ulysses]

That design is nice. Simple and elegant. I would run it at 18V rails, or even 20V, to increase headroom. In this application, where people are going to try and run their individual protools channels as hot as possible, headroom in the summing amps will be important. Increasing the supply voltage doesn't cost anything extra.
The other thing I would change would be to improve the simplicity by eliminating the output capacitors from the amplifiers. DC coupling the amplifiers to the transformer would be safe if you use good design practice to minimize and equalize the DC output of the two amplifiers. If they're equal, then no DC current will flow in the transformer and it'll be happy. To be safe, you can minimize the output offset by equalizing the input offsets on the inverting and non-inverting inputs of each amplifier. You do that by giving the non-inverting input a resistance to ground that matches the net effective resistance-to-ground of the inverting input. In the case of a summer with 8 inputs, this would be 2523 ohms. The catch is that the ideal non-inverting input resistance will change, depending on how many channels you had in use. There are ways around that.
Of course, you don't need the transformer. The output is already balanced without it. Shave a couple hundred bucks off the price tag. I would still DC-couple the outputs.

[tommyd]

I just had to chime in and say threads like this are one of the many reasons I visit this place for information, tips, help, etc. Here's someone looking for advice on building a summing box, and out come a bunch of designers who make this stuff to help him do it rather than sell one of their own.
Great forum Jules, and utmost respect to all here.
TommyD

[Tim Farrant]

Originally Posted by ulysses

...DC coupling the amplifiers to the transformer would be safe if you use good design practice to minimize and equalize the DC output of the two amplifiers. If they're equal, then no DC current will flow in the transformer and it'll be happy...

Hi Justin,

I agree it would be nice to DC couple the outputs, with the OP275 the offset would be minimal too boot. But, the mixer inputs are DC coupled too, and you never know what might be plugged into it, so IMO some form of protection for the opamp and transformer is required. Remember that at DC, the transformer is near short circuit to the opamp output.

If you use good quality bipolar caps (not electro's) and bypass them with polyestors (or polyprolene) then the sound quality should not suffer.

The use of the transformer ensures that any unbalanced inputs connected to the mixer will be converted to balanced output, plus, it will add a nice bit of colour.

Perhaps we could make a kit version of this circuit - how many would be interested?

Cheers
Tim.

[Phil Gorey]

Originally Posted by Tim Farrant

Perhaps we could make a kit version of this circuit - how many would be interested?

Cheers
Tim.

I would love to build this kit and would love for someone to make it...possibly provide everything except transformers and caps so that we could choose our own? great thread!

[Jim Williams]

Remove the 100 ohm resistors to improve crosstalk. Install a 100 uh inductor to trap rf. Replace the 10 pf feedback caps with 50 to 100 pf, stability will improve as the noise gain of the sum amp rises with the number of inputs summed. This also degrades the phase margin and can affect stability. If you use a low bias, low offset opamp like a Linear Tech LT1358, you can remove the coupling caps. Put the transformer in a box so you can choose whether you want the flavor or not.

Jim Williams
Audio Upgrades

[kevinc]

Originally Posted by Tim Farrant

Hi Justin,

Perhaps we could make a kit version of this circuit - how many would be interested?

Cheers
Tim.

I`d be interested as well.

[Tim Farrant]

Originally Posted by Jim Williams

Remove the 100 ohm resistors to improve crosstalk.

Crosstalk of what exactly?

Install a 100 uh inductor to trap rf.

Unnecessary, it's not a 12ft long console, it's an 8 input mixer and IME the choke will cause all sorts of ringing problems.

Replace the 10 pf feedback caps with 50 to 100 pf, stability will improve as the noise gain of the sum amp rises with the number of inputs summed. This also degrades the phase margin and can affect stability.

It's designed to be only 8 inputs, not 48. I would not make this any bigger than 22pF unless you want to turn it into a slug - and it will depend on the opamp used too. The values shown will provide excellent square wave response with the OP275.

If you use a low bias, low offset opamp like a Linear Tech LT1358, you can remove the coupling caps.

As I mentioned above, the caps are there as a saftey measure to protect the opamp/transformer from DC entering the inputs.

Put the transformer in a box so you can choose whether you want the flavor or not.

Could be made switchable IN/OUT if desired.

Tim

[oudplayer]

Count me in for the kit version with the switchable transformer...

[tINY]

Originally Posted by Tim Farrant

Crosstalk of what exactly?

You no longer have an ideal summer where the output is the sum of the inputs. Since the point where all of the 22k resistors connect is not a virtual ground any more, the current added by each resistor is modulated by the others. Worse, any noise between the ground reference of the driver and the reference used by the op-amp is only attenuated by 47dB.

Originally Posted by Tim Farrant

I would not make this any bigger than 22pF unless you want to turn it into a slug - and it will depend on the opamp used too. The values shown will provide excellent square wave response with the OP275.

I think you are asking for a lot of trouble with a bandwidth of 700kHz on a summing amp. This is worse if you don't use a small choke where you show the 100ohm resistor. A cap of 100pF is more sensible - it gives you a bandwidth of 70kHz, better than any speaker will reproduce. Then you can drop the inductor...

Originally Posted by Tim Farrant

As I mentioned above, the caps are there as a saftey measure to protect the opamp/transformer from DC entering the inputs.

You only need a cap on one side for that, no sense in using two of them that are twice as big as you need if you only use one.



-tINY

[Tim Farrant]

Well I don't really want to argue the in's and out's of it, but just because an SSL has a 100uH choke in it's mix amp does not mean it is required in every mix amp on the planet, and I do not see any technical reason for it in a compact 8 ch summing box and any such reactive element at the critical virtual earth should be avoided if possible IMHO.

I will say that I am into wide bandwidth audio electronics, so the wider you can get it the better! We love 700kHz! (of course it will not be that wide, more like 200kHz at most given the gain of the mix amp).

If you build it as drawn with good PCB layout techniques, put it into a steel or aluminium case (the best RF filter there is) it will work fine. If anybody would like to pursue the idea of this as a kit including power supply, maybe even the case and connectors, please send an email to stuff@buzzaudio.com

Cheers
Tim.

[tINY]

I'm starting to see why op-amps have a reputation of having "bad tone"....



-tINY

[Paul Frindle]

Originally Posted by Tim Farrant

Well I don't really want to argue the in's and out's of it, but just because an SSL has a 100uH choke in it's mix amp does not mean it is required in every mix amp on the planet, and I do not see any technical reason for it in a compact 8 ch summing box and any such reactive element at the critical virtual earth should be avoided if possible IMHO.

I will say that I am into wide bandwidth audio electronics, so the wider you can get it the better! We love 700kHz! (of course it will not be that wide, more like 200kHz at most given the gain of the mix amp).

If you build it as drawn with good PCB layout techniques, put it into a steel or aluminium case (the best RF filter there is) it will work fine. If anybody would like to pursue the idea of this as a kit including power supply, maybe even the case and connectors, please send an email to stuff@buzzaudio.com

Cheers
Tim.

BTW, the main reason that the SSL has an inductor at it's input is to maintain stability in the face of the considerable buss-GND capacitance in a large console (this provides better Xtalk than controlling it with a simple resistor). It also doubles as a rather gentle RF filter. But you have to watch it cos the inductor is a nasty reciever of magnetic noise :-( This can be minimised by mounting the inductors very close together so that induced noise is partially cancelled.

[Paul Frindle]

Originally Posted by Paul Frindle

BTW, the main reason that the SSL has an inductor at it's input is to maintain stability in the face of the considerable buss-GND capacitance in a large console (this provides better Xtalk than controlling it with a simple resistor). It also doubles as a rather gentle RF filter. But you have to watch it cos the inductor is a nasty reciever of magnetic noise :-( This can be minimised by mounting the inductors very close together so that induced noise is partially cancelled.

BTW - does anyone know why anything you do on this forum prompts you to install Macromedia? This message popping up all the time is a real bore :-(

[Tim Farrant]

Originally Posted by Paul Frindle

BTW, the main reason that the SSL has an inductor at it's input is to maintain stability in the face of the considerable buss-GND capacitance in a large console (this provides better Xtalk than controlling it with a simple resistor). It also doubles as a rather gentle RF filter. But you have to watch it cos the inductor is a nasty reciever of magnetic noise :-( This can be minimised by mounting the inductors very close together so that induced noise is partially cancelled.

Thank you Paul, that was my point, the choke is not necessary for a mix buss of this small physical size and an added complication at best. A large format console on the other hand is a completely different kettle of fish. In hindsight the 100R resistor might be a bit large, maybe 47R is better.

Oh and Paul, the chatroom uses Macromedia Flash but I can't see why you should get a prompt unless you tried to log into that. I'll let Jules know about this.

Tim.

[Jules]

Originally Posted by Paul Frindle

BTW - does anyone know why anything you do on this forum prompts you to install Macromedia? This message popping up all the time is a real bore :-(

Hi Paul the chat rooms & banner ads we run here (several with free give-away items) require macromedia flash software..

You might as well have it installed..

Regards

[Paul Frindle]

Originally Posted by Jules

Hi Paul the chat rooms & banner ads we run here (several with free give-away items) require macromedia flash software..

You might as well have it installed..

Regards

Ok Jules - thanks. Haven't been here in a while and didn't know :-)

[Paul Frindle]

Originally Posted by ulysses

To be safe, you can minimize the output offset by equalizing the input offsets on the inverting and non-inverting inputs of each amplifier. You do that by giving the non-inverting input a resistance to ground that matches the net effective resistance-to-ground of the inverting input. In the case of a summer with 8 inputs, this would be 2523 ohms. The catch is that the ideal non-inverting input resistance will change, depending on how many channels you had in use. There are ways around that.

BWT - putting resistors on the +ve opamp inputs will not help since the DC is differentially developed across the amps - so it's only the offset difference that counts. Putting resistors on the +ve inputs will only increase noise.

One thing that would increase performance somewhat (unless the 150R output Z is required) would be to do a pole splitting job around the 75R resistors to reduce the effective output Z. In this case the 22pF(?) caps would remain as they are from opamp out to -ve in, but the feedback Rs are taken from the other side of the 75Rs instead of the opamp out. This maintains HF stability but provides a low impedance output that is less affected by cable runs etc.. :-) The only downside is that DC offset will need to be more accurately balanced due to the loss of the 150R source Z.

[ulysses]

I agree that the circuit will perform better without the 100 ohm resistors feeding the amps. They degrade the virtual earth input such that channel input impedance will be affected (slightly) by the number of inputs in use, and to a lesser extent by the impedance of the sources. This will cause a shift in the mix balance when you add or subtract channels. These effects will be small, but this circuit isn't worth implementing unless you're paying attention to small effects. In an 8-channel mixer, the resistor won't be necessary anyway.
You're right though that capacitive cooupling on the outputs is a good idea since the inputs are DC coupled. But I see no reason besides coloration for the output transformer. I suppose if you fed all of the inputs from unbalanced sources and then fed the transformerless output of this circuit into a Symetrix 620 you'd lose 3dB of headroom, but that's a pretty narrow context for requiring differential signal output. The output is still technically "balanced" even when fed by all unbalanced sources. The transformers would be the single most expensive component (except perhaps the chassis) and can be easily omitted.
You're starting to talk about a kit, at some point you're going to have to consider that this circuit needs to be doubled to make it useful. I don't know of many people looking for mono outboard summing. That's easy enough, but it brings up the question of input connections and input switching. You need to be able to configure the inputs for a given mix situation. That is, some tracks will need to be fed to the Left buss, some to the Right buss, and some to both. That means at least a DP3T switch on each input (two DPDT switches seems to make the most sense here). Now you're talking about a control surface (i.e. faceplate) that is electrically close to (but usually physically far away from) the input connectors. It'll also increase the complexity and expense of the chassis.
One other thing - When you're calculating the size of the feedback capacitor, remember that the gain of the amps is dependent on the number of channels actually in use, not just the number of channels you build (unless you include switching that shorts unused inputs).

[Alcratin]

I haven't said thank you to all that responded to my question.

My friend asked me how were going to build this summing box and I said, let me check to see if anyone responded to my thread.

I'm really sorry I didn't thank you guys sooner.

Alcratin

[Alcratin]

Originally Posted by Tim Farrant

Crosstalk of what exactly?

Unnecessary, it's not a 12ft long console, it's an 8 input mixer and IME the choke will cause all sorts of ringing problems.

It's designed to be only 8 inputs, not 48. I would not make this any bigger than 22pF unless you want to turn it into a slug - and it will depend on the opamp used too. The values shown will provide excellent square wave response with the OP275.

As I mentioned above, the caps are there as a saftey measure to protect the opamp/transformer from DC entering the inputs.

Could be made switchable IN/OUT if desired.

Tim

What would change to make it 16 Channel vs 8 Channel, is going beyond 8 Channel a bad idea box for this design?

Alcratin

[Tim Farrant]

Originally Posted by Alcratin

What would change to make it 16 Channel vs 8 Channel, is going beyond 8 Channel a bad idea box for this design?

Alcratin

Hi Alcratin,

A 16 channel version might be possible, I think the OP275 is capable of the noise performance required. The gain would only be 24dB.

I have had a number of emails about the kit, sorry I have not replyed to everyone yet. If we were to do a kit, should it just be the PCB, parts, power supply and leave the rest (case, connectors) to the constructor?

Should we include the input switching Justin has suggested?

Cheers
Tim.

[Mr Thin]

This is George/Mr Thin

I was looking at your posts last night. It brought back a lot of memories. There is a lot of excellent brain power on this site. I used to go to studios and set everything up when they has tape recorders and record lathes.
I am in the middle of fingering out another website. I am going to post some pictures somewhere so people can see some of my work before I get a new site up.
A few thoughts.
A. Would it make sense to have the summing sends from the channels balanced as to remove the ground reference? I would use a differential receive of two inverting op amps into a differential op amp for balanced receive with summing. The massive ground floor in the console can be a real problem. When I built studios, we did a whole room prep-up before bringing the console in. This included ground planning the floor with copper, a large ground rod etc.. Op-amps are not unreferenced from ground. If you put a signal into the plus or minus input on a differential op-amp, it will have the same output as into both, balanced. In differential mode, you at least can have the CMRR to reject noise and take most of the ground plane effect away.
B. The problem with the bandwidth is that there are noise artifacts, harmonics and instant bursts that last milliseconds that ad harshness to a sound. The reason tube or older transistor or transformers add sweetness is because they will not pass very quick bursts of transit energies that harshly color the sound. So more bandwidth for this is not maybe a good thang...
C. I have a friend that got into Pro-Tools. There have been a few discussions about throwing out the mixer and just having a line summing system. I designed out a tube summer that would have +-6DB level adjust and pan per channel, pin 2+ drive I was going to make the pots with two photo-cells driven by lights so the whole thing could be automated.
D. The final thought and what I have been pestered about is doing it with 12au7's or 6sn7s at least in the output buffer so the line summer would be a Phat making deal. Using tubes to Phatting up the sound.

For some 5 or 7 band EQ's I built using 6DB/octave filters, I designed a 12au7 summer to add the bands together that works fine, no effect from other inputs.

Then a low impedance summer so the sum resistors are not to big and normal isolation between inputs.

I recently have been making mike preamps with these Yamaha PM-1000 boards that have two Phat little four-trans amp that are the building block amps for the whole console. They are single rail, output cap amps, but real phat with an open sound. They will drive down to 30 ohms at high level. A older style trans circuit could be used to get low impedance out without the trans.
E. You might consider having a Phat circuit before output that could be faded in/out for tone color. Even fading paper-oil or poly caps in/out for tone color.

I also have used simple resistor/capacitor networks that are slow, between stages to add color.
F. The last is the balanced out. When I worked at Sound Image, being the repair tech, I found there is no way to know for sure what any piece of gear is doing for the in/out wiring. Some gear is only pin2+ and 3 just hangs. Some have differential inputs and transistor buffers to pin2+ out. There was even some gear that was pin3+ and the cables to this gear had to be reversed..So the balanced out scheme is cool, but what is the input structure of the driven device?

I was building acoustic Git preamps for a frind for research and we found the input on his Sony mixer/HD recorder thang has a 1.5K resistor between 2-3 into a diff op-amp. We found a lot of the cheaper mixers do this. We were getting weird responses because of the load to the final tube stage in the preamp. Some mixer companies just assume a 1.5K load for line and mike.

Again though, what is the purpose of the mixer if not to add color???