Virtual earth summing with transformer

[abbey road d enfer]

The acid test is the interaction of inputs, If it is acting as a VE summer then there will be no difference in gain on a channel between the case where all the other input channels are shorted and the case where they are all open. If there is a difference then it is more likely regular voltage summing.

If your definition was correct, NO summing amp would qualify for VE, because the impedance of the node is never a true null, due to parasitics; in particular, at HF, the decrease in open-loop gain of the opamp has a measurable action on inter-channel feedthrough.
Current-to-voltage conversion and creation of a VLZ node by NFB are what characterize VE.

 

[L´Andratté]

There are advantages, such as elimination of one source of longitudinal noise (between reference bus and summing-amp refrence point) and possible noise optimization of the source impedance seen by the opamp.

For my better comprehension (sorry):
1. Do you mean a voltages developing along the ground path?
2. Then the 1:2 transformer in the OP´s example should be in the ballpark for a bjt-input opamp?

 

[ruffrecords]

If your definition was correct, NO summing amp would qualify for VE, because the impedance of the node is never a true null, due to parasitics; in particular, at HF, the decrease in open-loop gain of the opamp has a measurable action on inter-channel feedthrough.
Current-to-voltage conversion and creation of a VLZ node by NFB are what characterize VE.

What you say is correct if a tad pedantic. Perhaps if I had said :

If it is acting as a VE summer then there will be VERY LITTLE difference in gain on a channel between the case where all the other input channels are shorted and the case where they are all open. If there is a SIGNIFICANT difference then it is more likely regular voltage summing.

And if you query what I mean by very little and significant I shall ignore you.  ;D

Any while we are being pedantic, it was not a definition it was a method of determining the type of mixing.

Cheers

Ian

 

[abbey road d enfer]

What you say is correct if a tad pedantic.

It's my duty here, as a mod...  8)

 

[abbey road d enfer]

For my better comprehension (sorry):
1. Do you mean a voltages developing along the ground path?

Yes, particularly whatever voltage can develop between the reference bus and the summing amp reference point (that is also eliminated by the use of balanced bus).

2. Then the 1:2 transformer in the OP´s example should be in the ballpark for a bjt-input opamp?

A xfmr with 800 ohms reflected impedance is not adequate. The bus impedance (assuming 10k stem resistors and 32 channels) is about 300 ohms, so in order to qualify as proper VE, the resulting input Z should be significantly lower, I would say below 30 ohms. The inductance does not need to be big; in fact, the highest the nominal inductance, the higher the leakage inductance, which is a big factor in degradation of HF response.
The ratio will depend on the OSI (optimum source impedance) of the BJT stage, which depends essentially on the quiescent collector current. By varying the collector current from 50uA to 1mA, the OSI can be tuned from about 10k to 500 ohms.
With a 1:2 xfmr, the impedance presented to the BJT stage would be about 1-1.5k, which would suggest a current of about 200-500uA.
All in all, this suggests strongly a custom-wound xfmr, but a mic input xfmr could be used satisfactorily.

 

[fripholm]

Thanks for the replys so far! That's why I love this place, a vast store of knowledge! 

Ian, there are eight inputs from the DB25 connector, and four more on TRS, of which two are mono (using 20k bus resistors to each channel), the other two go to either left or right with 10k. If I'm calculating this correctly, the bus source impedance should be about 830 ohms.

The output level changes only slightly, when the additional inputs are shorted together - about 0.3dB for each shorted input vs. open. If this bench test circuit actually makes it into a working unit for use at my studio, the inputs will always be connected so that shouldn't be a problem.

Good tip with the mic transformer and its lower winding resistances and inductance. This lowers the noise even more. I had a couple of OEP J30A11C here, which - configured to 1:2.5 - also give a flat frequency response and ultra-low LF distortion.

At the moment there's no way to short all the channels at once to measure the output noise correctly. But considering the fact, that just one channel on the DB25 already has -111dBFS of output noise (about -98dBu), the summing circuit's noise of -105.5dBFS with all channels connected and a noise gain of about 10 (or 20dB) is pretty good, right!?

 

[warpie]

Very interesting topic!

 

[madswitcher]

I don'tnot sure what amplifier type B meansI described current source summing in my console design article published in 1980... At least one console designer shared with me that he used the technique after reading about it, (in a broadcast console he designed).

Most people just didn't believe it could work.  ;D ;D

JR

Well you were right John and I don't know how long this will stay up but  if it works there should be a pdf of a E/I (type B and VE summer (type C).  The E/I outputs 1/6.8 mA per input volt from a source impedance of 27K.

I tried to PM this to you John, but I couldn't insert an attachment.



Cheers

Mike

 

[ruffrecords]

Ian, there are eight inputs from the DB25 connector, and four more on TRS, of which two are mono (using 20k bus resistors to each channel), the other two go to either left or right with 10k. If I'm calculating this correctly, the bus source impedance should be about 830 ohms.

So would I be right in saying that each bus has four inputs from the db25 and another four from the TRS connectors?

If so, each bus is fed from four pairs of 10K resistors and four pairs of 20K which according to my calculator is a bus impedance of 3.3K

The output level changes only slightly, when the additional inputs are shorted together - about 0.3dB for each shorted input vs. open. If this bench test circuit actually makes it into a working unit for use at my studio, the inputs will always be connected so that shouldn't be a problem.

Just to be clear, are you saying that it changes by 0.3dB when just one input is changed?

Cheers

Ian

 

[fripholm]

So would I be right in saying that each bus has four inputs from the db25 and another four from the TRS connectors?

If so, each bus is fed from four pairs of 10K resistors and four pairs of 20K which according to my calculator is a bus impedance of 3.3K

Not quite. Each bus has five pairs of 10k (four from the DB25 and one from TRS) and two pairs of 20k coming from two mono inputs.

Just to be clear, are you saying that it changes by 0.3dB when just one input is changed?

That is exactly what I'm seeing.

 

[nielsk]

I have seen + & - inputs connected together on unused inputs in a balanced passive summing mix buss instead of connecting to ground, how do the 2 approaches compare?

 

[JohnRoberts]

Well you were right John and I don't know how long this will stay up but  if it works there should be a pdf of a E/I (type B and VE summer (type C).  The E/I outputs 1/6.8 mA per input volt from a source impedance of 27K.

I tried to PM this to you John, but I couldn't insert an attachment.



Cheers

Mike

OK not exactly a current source. Operating class A means not very quiet either.

My apologies I don't want to feed this veer away from the original topic. I've discussed current source summing multiple times since 1980.

JR

 

[abbey road d enfer]

The output level changes only slightly, when the additional inputs are shorted together - about 0.3dB for each shorted input vs. open.

That is consistent with the input impedance of about 800 ohms (essentially due to the DC resistance of windings).

At the moment there's no way to short all the channels at once to measure the output noise correctly. But considering the fact, that just one channel on the DB25 already has -111dBFS of output noise (about -98dBu), the summing circuit's noise of -105.5dBFS with all channels connected and a noise gain of about 10 (or 20dB) is pretty good, right!?

  The actual gain of this stage is -6dB, so that suggests an Equivalent Noise Voltage referred to input of -92 dBu, which is a rather poor performance (a pedestrian TL08x would be about 16dB quieter). This suggests your measurement is flawed, maybe the measurement BW is not correct, or hum/buzz/EMI pollute the connection. The transformer is actually a good candidate for culprit #1.

 

[PRR]

The 30dB crosstalk with an idle mono connection seems worse than a 0.3dB change when muting inputs. (Though muting a bunch of inputs for a solo might be a level jump.)

The mono crosstalk is cured if there is a ZERO impedance somewhere. This coil-input mix amp is non-Zero. But putting a zero impedance source buffer at the mono point cures this. For real-world 10 Ohm buffer and 10K mix resistors, crosstalk becomes -60dB, a decent figure. A half a 19-cent TL072 will do this.

 

[ruffrecords]

I have seen + & - inputs connected together on unused inputs in a balanced passive summing mix buss instead of connecting to ground, how do the 2 approaches compare?

They should never be connected to ground. In a balanced system the signal exists only between hot and cold.  In many systems they a referred to a common 0V but the do not need to be (just think transformers) You should always short hot to cold. That is all that is necessary.

Cheers

Ian

 

[ruffrecords]

Not quite. Each bus has five pairs of 10k (four from the DB25 and one from TRS) and two pairs of 20k coming from two mono inputs.

OK, that make the bus impedance about the same 3k3. As Abbey says, the 0.3dB drop is consistent with a bus impedance of 800 ohms  so the transformer impedance clearly dominates. However, the 0.3dB drop when just one input is disconnected does inpmy it is voltage summing rather than VE.

Cheers

Ian

 

[abbey road d enfer]

However, the 0.3dB drop when just one input is disconnected does inpmy it is voltage summing rather than VE.

When does the bus loading susceptibility implies that the circuit becomes voltage summing instead of VE? 0.1dB, 0.01dB?
VE is not a matter of performance, it's a matter of topology. In voltage summing, the actual invariance of gain vs. bus loading is based on a shunt resistor of sufficiently low value that the varaiation is minimized.
In VE, the invariance is based on current NFB.
Both concepts have actual limitations that tend to blur the line if only this aspect of performance is considered. VE summing with 100 stems taxes the open-loop gain of the opamp, resulting in non-zero input impedance and some sensitivity to bus-loading, as do any parasitics, particularly the inductance of the bus.
OTOH, a "voltage" summing amp with a 200 ohms shunt resistor would exhibit a variation of less than 0.1dB in the same conditions. Does it qualify it for VE?
It seems the Studer engineers agree with me (See attached). Are they pedantic?

 

[ruffrecords]

When does the bus loading susceptibility implies that the circuit becomes voltage summing instead of VE? 0.1dB, 0.01dB?
VE is not a matter of performance, it's a matter of topology. In voltage summing, the actual invariance of gain vs. bus loading is based on a shunt resistor of sufficiently low value that the varaiation is minimized.
In VE, the invariance is based on current NFB.

I agree, but at the risk of repeating myself, I was trying to help the OP determine if the specific topology he has is voltage or VE summing.  I asked the OP how many inputs he has, Based on that small number, VE mixing should have no detectable change when you remove one input. If there is a clearly detectable change in such a small number it is not certain, but it is more likely to be acting as a voltage summer.

Both concepts have actual limitations that tend to blur the line if only this aspect of performance is considered. VE summing with 100 stems taxes the open-loop gain of the opamp, resulting in non-zero input impedance and some sensitivity to bus-loading, as do any parasitics, particularly the inductance of the bus.
OTOH, a "voltage" summing amp with a 200 ohms shunt resistor would exhibit a variation of less than 0.1dB in the same conditions. Does it qualify it for VE?

I agree that, properly implemented, both schemes are  virtually indistinguishable from a performance point of view. Only their topologies differ.  However,  what we are trying to help the OP determine is if his circuit is VE or voltage summing. Which do you think it is because it is not clear to me?

Cheers

Ian

 

[abbey road d enfer]

However,  what we are trying to help the OP determine is if his circuit is VE or voltage summing. Which do you think it is because it is not clear to me?

With the risk of repeating myself, the circuit is undoubtedly VE, although the poor implementation (inadequate DCR of the windings) makes the gain sensitive to bus loading. The low impedance of the summing amp is not due to the shunt resistor, but to current NFB.

 

[Newmarket]

Theoretically, the two techniques are almost identical. In practice, both have limitations. Passive summing requires unused sends are connected to 0V to maintain bus impedance (and hence gain and crosstalk figures). Large numbers of channels connected to a virtual earth can cause serious instability. In noise terms there is nothing to choose between them.

Cheeers

Ian

Fair points. I think I still prefer Virtual Earth approach - to my way of thinking it offers more flexibilty wrt impedances etc.
How you implement analogue summing in a large system / console. That's a whole other question...