Virtual Earth Summing Questions

Dear friends, I found this schematic for virtual earth summing, and I'm a bit puzzled.
Hoping someone can help answering my questions

1. Is the signal before the mix resistor an unbalanced signal or a balanced non polarized signal?
2. If the signal is indeed an unbalanced signal, can I use the GSSL's input circuit (all the way to 15K/27K resistor but before the VCA) to convert my balanced input to unbalanced?

Thank you!!!

Yes, unbalanced.

Yes, any bal to unbal convertor would work.

The difficulty is in the details... namely doing X channels of whatever and keeping signal integrity good.

JR

Thank you!!!! ;D

Just to help you untangle the mistery 

Right! Thanks abbey for that info

In the first drawing, the 10K gain control is in the wrong place.  It should be between the two op-amps.

Speedskater said:

In the first drawing, the 10K gain control is in the wrong place.  It should be between the two op-amps.

Click to expand...

It depends...putting between the two stages will make the output impedance constant (FWIW), but the noise of the second stage will be constant too, even when the level is attenuated to suit low-level inputs. And there is no advantage in terms of dynamic range (if the inverter had gain, it could be pertinent to put the gain control mid-stage). The way it is on the proposed schemo, there is no real flaw, except due to the rather high impedance, the output level is load-impedance-dependant; as a result, when a load is connected to the second output, the level of the first is affected.

Drawn a schematic based on the distributed balanced summing amplifier theory written in the book "Small Signal Design"
Can anyone help me verify this? BTW, I left out the power pins connection on purpose.

Thank you!! ;D

i don't have a copy of the book... so i can't verify that it is the same...  what are you trying to do?

JR

John, what I'm trying to do is, learning how to make a 16x2 active summing box, hehe ;D

In the first schematic, the input is for an unbalanced signal, and therefore requires a converter for a balanced signal. However, in this book I mentioned, I found out that there's such thing as a balanced summing system. The author also explained a bit about a distributed summing system, which I'd like to apply for my summing box as well.

This is the schematic taken from the book.

In the schematic I've drawn, I omitted the pan pot, and the switching. I omitted the buffer amplifiers (A1 & A2) because I assumed (please correct me if I'm wrong) they are part of an unbalanced to balanced converter circuit, and omitted the subtraction amplifier (A5) because, again I assumed it's a balanced to unbalanced converter circuit.

Distributed summing system for 24 inputs, taken from the book.

Yes  I understand the concept.. I wrote about the generally in a console article back in 1980.

Are you doing 8x2x1 ?

Decompensated opamps could buy you a little more loop gain margin.

Yes, that will give fully balanced input to output.

JR

I'm thinking of splitting 8 inputs (left or right channel) into two 4x1. Then the outputs of these 4x1 will be summed together as a single channel output.

From what I understand, theoretically this approach has a better signal integrity than summing 8 inputs at once. It also seems to correct the signal phase that got inverted by the first summing stage.

But I'm still learning, so there's a big chance I might be totally wrong in this matter, hehe

Thank you everyone for the tremendous help! Really appreciates it!

John, can you please explain what did you mean by "decompensated op amps could buy you a little more loop gain margin"?

According to TI's notes:

"A decompensated op amp has internal frequency compensation designed to work with external gain-setting resistors such that the resultant closed loop gain is restricted to a number that is greater than a specified minimum. This
minimum gain is specified on the decompensated op amp’s datasheet. Compensated op amps, or simply op amps, are traditionally designed to be stable for gains down to and including unity gain. Decompensated, or less compensated op amps, exhibit higher bandwidth and slew rate than op amps compensated for unity gain."

But the summing amplifier is actually a unity gain amplifier, isn't?

If yes, then why the need to use op amps with a higher bandwidth and a higher slew rate, say, compared to NE5534 (10 MHz bandwidth, 13V/uS slew rate and 4 nV/Hz typical noise figure)?

Thanks again for the help.

metalb00b00 said:

I'm thinking of splitting 8 inputs (left or right channel) into two 4x1. Then the outputs of these 4x1 will be summed together as a single channel output.

From what I understand, theoretically this approach has a better signal integrity than summing 8 inputs at once.

Click to expand...

Theoretically, doubling the number of "layers" would give 3dB less noise; in practice, the difference would be less than that.
Noise in a single-layer 5534-based 8:1 mixer should measure at about -100dBu. Do you really want the added complexity for a tentative -103?  IMO, it's not worth the it. No source, except maybe some fully electronically generated signals are capable of -100dBu noise. Multilayer mixing is really valuable in very large mixers (100+ channels). Before multilayer mixing, there are several ways to optimize the noise figure, particularly by using lower noise-voltage stages and optimizing the level diagram.

It also seems to correct the signal phase that got inverted by the first summing stage.

Click to expand...

This is true. The second mixer, being also of the inverting type, dispenses with the need for a dedicated polarity inverter.

abbey road d enfer said:

Speedskater said:

In the first drawing, the 10K gain control is in the wrong place.  It should be between the two op-amps.

Click to expand...

It depends...putting between the two stages will make the output impedance constant (FWIW), but the noise of the second stage will be constant too, even when the level is attenuated to suit low-level inputs. And there is no advantage in terms of dynamic range (if the inverter had gain, it could be pertinent to put the gain control mid-stage). The way it is on the proposed schemo, there is no real flaw, except due to the rather high impedance, the output level is load-impedance-dependant; as a result, when a load is connected to the second output, the level of the first is affected.

Click to expand...

I'll take low (and fixed) output impedance over possible higher noise any day.

metalb00b00 said:

John, can you please explain what did you mean by "decompensated op amps could buy you a little more loop gain margin"?

According to TI's notes:

"A decompensated op amp has internal frequency compensation designed to work with external gain-setting resistors such that the resultant closed loop gain is restricted to a number that is greater than a specified minimum. This
minimum gain is specified on the decompensated op amp’s datasheet. Compensated op amps, or simply op amps, are traditionally designed to be stable for gains down to and including unity gain. Decompensated, or less compensated op amps, exhibit higher bandwidth and slew rate than op amps compensated for unity gain."

But the summing amplifier is actually a unity gain amplifier, isn't?

Click to expand...

No.  For a typical virtual earth sum section, the forward gain may be unity, but the gain from the perspective of the opamp is N+1, with N the number of stems being summed. This is also called "Noise Gain" and is the closed loop gain wrt the + input pin.

I adressed this in that old article. While people seem overly focussed on noise in a sum amp, the higher "noise gain" of such sections also erodes all the benefits of loop gain margin via NF to linearize the transfer function, i.e. increased phase shift and increased distortion result. The noise is generally swamped by other larger input noise sources, but phase shift and distortion are not so easily dismissed. 

If yes, then why the need to use op amps with a higher bandwidth and a higher slew rate, say, compared to NE5534 (10 MHz bandwidth, 13V/uS slew rate and 4 nV/Hz typical noise figure)?

Thanks again for the help.

Click to expand...

The 5534 is actually a good candidate, since it is decompensated and does not need added compensation caps when operated at 10dB or more closed loop gain (or noise gain). Your 8 input sum amp has a noise gain of 19 dB so comfortably compatible with decompensated application.

Note: Look at the 5534 data sheet there should be a slower slew rate spec and reduced open loop gain bandwidth when used with added unity gain stable compensation cap.

Note2: when operating decompensated, you can't use feedback cap across feedback resistor since that looks like unity gain or less (100% NF)  to the opamp at very high frequency so could become unstable without added compensation.

Here is a Link to my very old article about console design. Published in 1980 so only 3+ decades old, and before I learned how to spell bus.  :-[  some general opamp info in there.
http://www.circularscience.com/des_art.pdf

There will be a test.....  8)

JR

Speedskater said:

I'll take low (and fixed) output impedance over possible higher noise any day.

Click to expand...

It's a matter of circumstances. If the mixer was connected to a mic input, I would be happy to turn off the level and noise at the same time. I like low impedance, I like very low noise too, but given the little resources, one has to do some trade-offs...

abbey road d enfer said:

Theoretically, doubling the number of "layers" would give 3dB less noise; in practice, the difference would be less than that.
Noise in a single-layer 5534-based 8:1 mixer should measure at about -100dBu. Do you really want the added complexity for a tentative -103?  IMO, it's not worth the it.

Click to expand...

Agree with ya

JohnRoberts said:

The 5534 is actually a good candidate, since it is decompensated and does not need added compensation caps when operated at 10dB or more closed loop gain (or noise gain). Your 8 input sum amp has a noise gain of 19 dB so comfortably compatible with decompensated application.

Note: Look at the 5534 data sheet there should be a slower slew rate spec and reduced open loop gain bandwidth when used with added unity gain stable compensation cap.

Click to expand...

The usual 22pF compensation cap?

Note2: when operating decompensated, you can't use feedback cap across feedback resistor since that looks like unity gain or less (100% NF)  to the opamp at very high frequency so could become unstable without added compensation.

Click to expand...

Most of the virtual earth summing schema I have seen, used the NE5532/LM833 op amp with a feedback cap of 100pF paired with a 10K or 22K feedback resistor.