summing large amounts of channels.

Help Support GroupDIY:

abbey road d enfer

Well-known member
Moderator
Joined
Jan 22, 2008
Messages
13,456
Location
Marcelland
JohnRoberts said:
I'm sure I have shared this before but back last century I used the improved howland current pump,
Litt seems to be a tad inconsistent as to what an "improved" Howland pump is; is it just the one with the current sense resistor, that reaquires compensating the PFB resistor for the current sense resistor, or is it the version with the voltage follower, that uses equal values for teh PFB and NFB paths? My inpression was the latter...

that I first saw in a national applications manual back in the 70s 
I discovered it in RTS intercoms, where it constituted the basis of their party-line system, allowing many beltpacks to be connected on a single line.

but with impractical values for audio use (they used 1M resistors trying to get highest possible output impedance).
Actually, the absolute value of the resistors is not as important as their matching. In fact, with 1k resistors it is still possible to achieve a dynamic impedance of about 200k. There may be other consequences, though.

For today's TMI, early this century I discussed with a senior engineer at THAT corp
Would that be Gary Hebert? I spoke to both him and Les Tyler about this.

*
** ICs are fabricated with subcomponents that get connected together by the final metallization layer, kind of like how traces on a PCB connect discrete components together...  ICs are like an upside down PCB but without the board.
It's the culmination of the PCB concept. "Printed" circuit boards are actually etched; metallization is a real positive print.
 

Winston OBoogie

Well-known member
Joined
Jun 3, 2004
Messages
2,685
Location
UK.
JohnRoberts said:
more significantly buffer the inputs to the improved Howland. The varying source impedance from the pan pot position imbalanced and degraded the current sources output impedance. The fully loaded bus noise floor was lowest with all pan pots hard panned left or hard right

The active pan also has the benefit of better left-right isolation or "offness" (90dB) and a closer to desired pan law. 
 

Attachments

  •  Active Pan.png
    Active Pan.png
    76.1 KB · Views: 98

ruffrecords

Well-known member
Joined
Nov 10, 2006
Messages
13,931
Location
Norfolk - UK
Seeing op amp schematics with the output connected somehow to the + input always gives me the jitters. I know it is common practice (JLM VU buffer for instance) but I just cannot get used to it. Seems to me to be contrary to nature.

Cheers

Ian
 

Winston OBoogie

Well-known member
Joined
Jun 3, 2004
Messages
2,685
Location
UK.
Dredging an old thread. Since we first talked here about it, there are now copies of the early EMI current summing schem other folks can look at. These were redrawn and made availble here:

Therefore I don't feel guilty about posting the relevant redraws for others to see. (Including a correction to Chris' redrawn Amp B schem.)

Amp B was the voltage to current amp which followed the channel's pan pot outputs:

And amp C1 was the virtual earth summing.

Sure, we can do better now, but this is a 1967 design using the then S.O.T.A. with all that entailed.
I don't think it was too shabby given the timeframe.
 

Attachments

  • TG Type B.pdf
    8.3 KB · Views: 36
  • TG Type C.pdf
    12.7 KB · Views: 31
  • Amp B correct Collector load R.png
    Amp B correct Collector load R.png
    58.2 KB · Views: 53
Last edited:

abbey road d enfer

Well-known member
Moderator
Joined
Jan 22, 2008
Messages
13,456
Location
Marcelland
Amp A was the voltage to current amp which followed the channel's pan pot outputs:
I guess JR and I could complain again about the ancients stealing our ideas (current summing).
However, I think there's a discrepancy between the concept and the realization.
Either a mistake in the schemo or the purpose of this circuit, because it has an output Z of ca. 2k, which IMO does not qualify as current gen. It has a "gain" of about -10dB (attenuation, actually).
As it is, notwithstanding polarity, a simple 6.8k resistor would do a better job.
And amp C1 was the virtual earth summing.
Yes. The input Z is about 100r, which is not too bad, except at LF where the input cap starts to dominate at 300Hz. that's for teh C1 version.
Sure, we can do better now, but this is a 1967 design using the then S.O.T.A. with all that entailed.
I don't think it was too shabby given the timeframe.
Indeed, I have a lot of respect for the EMI designers.
 

Winston OBoogie

Well-known member
Joined
Jun 3, 2004
Messages
2,685
Location
UK.
Yes I get that the actual realisation of the concept came short of ideal. (Edit: By the way - Yes, almost forgot - the redrawn schem that Chris posted in the TG Diagrams thread does indeed have an error: The collector load should be 27K rather than 2K7. I've added the attached hand sketch to the above post now as well as here).

But if I'm not mistaken, audio mixing up to this point had been voltage summing. Certainly the early professional transistor desks from that time that I've seen all were.

Given a few more years as I.C.'s became readily available, I have no doubt that the current summing idea Mike Bachelor had in '66/'67 would have been better realised by him with closer to ideal current sources.

But by the time I.C.'s were cheap as chips, EMI were purchasing desks from other manufacturers. The first being the big EMI/Neve's which, of course, utilised voltage summing so, possibly a step backwards in some regards?
 

Attachments

  • Amp B correct Collector load R.png
    Amp B correct Collector load R.png
    58.2 KB · Views: 48
  • Description of TG Amps B -a &  B-b .png
    Description of TG Amps B -a & B-b .png
    81 KB · Views: 38
Last edited:

abbey road d enfer

Well-known member
Moderator
Joined
Jan 22, 2008
Messages
13,456
Location
Marcelland
P.S. Yes, almost forgot: the redrawn schem that Chris posted in the TG Diagrams thread does indeed have an error: The collector load should be 27K rather than 2K7. I've added the attached hand sketch to the above post now as well.
Definitely much better now! Now for an assessment of the practical improvement, it depends on the operating level and designed headroom.
0.4mA rms (consistent with sim results) suggest a maximum operating level of about +5dBu, so nominal would be about -15.
Simply using 27k resistors would result in the same noise level (even a tad lower since no additional active elements) and would allow a much higher headroom, particularly considering the +/-20V rails.
I would think the channels could easly operate at -2dBu nominal, unless an enormous input headroom was required.
This would result in a much better noise performance of the summing amp.
JR would probably consider it not a significant concern, but I can't help wondering why EMI did that, since it seems that improving noise in the summing stage was the major incentive. Quite puzzling...
I guess someone has to start turning tables in seance...

EDIT: actually some of these amps (type E, F & G) run on a single negative rail, except for base bias. So the headroom was somewhat compromised there.
 
Last edited:

Winston OBoogie

Well-known member
Joined
Jun 3, 2004
Messages
2,685
Location
UK.
I can't help wondering why EMI did that, since it seems that improving noise in the summing stage was the major incentive. Quite puzzling...
I guess someone has to start turning tables in seance...
For sure.
Unlike for the earlier REDD series, I'm not lucky, nor been privy enough, to have copies of Mike Bachelors initial design notes regarding the TG stuff.
A lot of his notes and work regarding the valve designs from the years prior. I have covered but... ?

Unfortunately, by the time I thought to contact Mr Bachelor personally, it was decades after the fact by which time the man was suffering from Alzheimer's so was not in any great condition to, nor invested in, talk(ing) about this stuff.

👍
 

JohnRoberts

Well-known member
Staff member
Moderator
Joined
Nov 30, 2006
Messages
21,405
Location
Hickory, MS
Yes I get that the actual realisation of the concept came short of ideal. (Edit: By the way - Yes, almost forgot - the redrawn schem that Chris posted in the TG Diagrams thread does indeed have an error: The collector load should be 27K rather than 2K7. I've added the attached hand sketch to the above post now as well as here).

But if I'm not mistaken, audio mixing up to this point had been voltage summing. Certainly the early professional transistor desks from that time that I've seen all were.

Given a few more years as I.C.'s became readily available, I have no doubt that the current summing idea Mike Bachelor had in '66/'67 would have been better realised by him with closer to ideal current sources.

But by the time I.C.'s were cheap as chips, EMI were purchasing desks from other manufacturers. The first being the big EMI/Neve's which, of course, utilised voltage summing so, possibly a step backwards in some regards?
It seems we've been around this tree before (several times). We can call a common emitter transistor stage a voltage to current convertor wrt it's collector, but that does not make that collector output a high compliance current source. From your sketch each bus feed presents a source impedance of 27k in parallel with 150k which will load the bus and impact noise gain. This is in fact a better than 10 dB improvement to virtual earth bus noise gain. 22k effective source impedance vs. 6.8k with simple bus resistors feeding the bus.

A 10 dB sum bus noise improvement (estimate, reduced by noise of class A stage) is not nothing so congrats to those ancients. As I have shared multiple times, the benefits from reduced bus noise gain, is more than just lower noise. We also enjoy lower distortion and lower phase shift, while most people are fixated on noise, which is easily audible. I won't waste time casting shade on the apparent limits of that simple class A transistor stage.

====
I will surely be repeating myself so stop reading now if you already know my story.

I killed a lot of brain cells investigating sundry current output approaches. I am surely repeating myself but I settled on using an op amp synthesized current source where positive feedback (apologies to Ian who find that unnatural) elevates the current source output (source) impedance almost without limit. In practice that op amp synthesized current source output impedance is controlled by how well the feedback resistors are matched and the nominal starting impedance. This nominal impedance gets multiplied X times based on how closely the resistors are matched. I used 1% resistors in my production design.

The published examples of these synthesized op amp current sources routinely use 1M resistors to realize extremely high source impedance. My epiphany to make a practical current source for feeding VE bus was that I didn't need high megohm source impedance to realize benefits. Observe that the TG(?) design being touted above derives its benefit from a 22k source impedance vs 6.8k. IM ohm resistor current sources would contribute more noise than they save.

I realized that by substituting 10k and 20k resistors for the 1M resistors from the op amp current source example would allow me to deliver 10x to 100X that 20k nominal impedance with acceptable noise floor performance.

In bench measurements of a fully loaded prototype console I measured a real world source impedance from all the paralleled op amp current sources in the 200K ohm range... As I quickly realized when you reduce one noise term to insignificance, all the other noise sources remind you they are still there. ;)

If big analog buses were still a thing this might be worth pursuing, but nah....

JR
 

dogears

Well-known member
Joined
Nov 15, 2017
Messages
477
Location
Texas
I killed a lot of brain cells investigating sundry current output approaches. I am surely repeating myself but I settled on using an op amp synthesized current source where positive feedback (apologies to Ian who find that unnatural) elevates the current source output (source) impedance almost without limit. In practice that op amp synthesized current source output impedance is controlled by how well the feedback resistors are matched and the nominal starting impedance. This nominal impedance gets multiplied X times based on how closely the resistors are matched. I used 1% resistors in my production design.

The published examples of these synthesized op amp current sources routinely use 1M resistors to realize extremely high source impedance. My epiphany to make a practical current source for feeding VE bus was that I didn't need high megohm source impedance to realize benefits. Observe that the TG(?) design being touted above derives its benefit from a 22k source impedance vs 6.8k. IM ohm resistor current sources would contribute more noise than they save.

I realized that by substituting 10k and 20k resistors for the 1M resistors from the op amp current source example would allow me to deliver 10x to 100X that 20k nominal impedance with acceptable noise floor performance.

In bench measurements of a fully loaded prototype console I measured a real world source impedance from all the paralleled op amp current sources in the 200K ohm range... As I quickly realized when you reduce one noise term to insignificance, all the other noise sources remind you they are still there. ;)

If big analog buses were still a thing this might be worth pursuing, but nah....

JR
It would seem like a THAT1240 would be great for this application, no?
 

abbey road d enfer

Well-known member
Moderator
Joined
Jan 22, 2008
Messages
13,456
Location
Marcelland
It seems we've been around this tree before (several times). We can call a common emitter transistor stage a voltage to current convertor wrt it's collector, but that does not make that collector output a high compliance current source. From your sketch each bus feed presents a source impedance of 27k in parallel with 150k which will load the bus and impact noise gain. This is in fact a better than 10 dB improvement to virtual earth bus noise gain. 22k effective source impedance vs. 6.8k with simple bus resistors feeding the bus.
But with 6.8k bus resistors, the current would be 3 times higher, and the gain of the summing amp could be reduced as much. I don't see any practical improvement here.
 

abbey road d enfer

Well-known member
Moderator
Joined
Jan 22, 2008
Messages
13,456
Location
Marcelland
It would seem like a THAT1240 would be great for this application, no?
You mean as a V to I converter? I don't see how...
AFAIK, there are only two options: the Howland current pump and the "current mirror under control of an opamp", that works in class A but is very tricky to operate in class B, and probably not examplarily linear.
 

dogears

Well-known member
Joined
Nov 15, 2017
Messages
477
Location
Texas
Looking at the partial schem john showed on the previous page using TL072 with discrete resistors as an improved Howland current pump, you could do the same with a THAT124x.

Something like this, Design 3 on page six. Also talks about the buffer John mentioned.
 

JohnRoberts

Well-known member
Staff member
Moderator
Joined
Nov 30, 2006
Messages
21,405
Location
Hickory, MS
It would seem like a THAT1240 would be great for this application, no?
close but not quite there... 4 of the 5 precision resistors could be from the IC but the 5th added resistor could degrade the potential matching. Years ago I asked Gary Hebert (senior THAT corp engineer) about a custom metallization layer. The basic silicon for that family of parts has all the different resistors and they just connect up the ones they need for the given fixed gains. I could grab a 5th resistor from one of those unused Rs and get superior matching but the project stopper was that THAT corp refused to source ICs without full protection wells for every I/O pin. I needed one more protection well than available on the die and completely new silicon just wasn't practical. The cost of just a metallization layer was probably tens of $k.

====

The buffer I was talking about adding if I was to revisit that old production console design was after the pan pots feeding the current sources. To save op amps (remember there were 72 pairs of these current sources just in the monitor section), I drove the synthesized current sources directly from the pan pots so pan position slightly imbalanced the current source output resistance multiplier. There was a just barely audible difference in master bus noise floor when listening WFO. The lowest noise (highest current source output impedance) was with all monitors hard panned left or right.

JR
 

Winston OBoogie

Well-known member
Joined
Jun 3, 2004
Messages
2,685
Location
UK.
Posted this in the wrong thread:, meant it to be here:


Without doubt John, your elegant and upgraded I to V would be a superb way to sum many channels in a large desk.
If there was a market for these desks, the results you show would sway towards your system being de-facto.

The only reason I felt it was applicable to draw attention to the EMI, is that, if we stop being a little pedantic, I think it's fair to say that EMI's design drawing from 1966/67ish isn't a whole lot different from the basic idea you show in 1980.

No doubt in my mind that you believed this was an original idea, and why would you think otherwise since the prior art documents were filed away in EMI's "need to know basis" aisle. You thought of the concept on your own. Kudos. No way I could do that.

Also no doubt at all that you improved the schem beyond anything that came from the EMI.

I just like to see the folks whose shoulders we all stood also be acknowledged sometimes
 

JohnRoberts

Well-known member
Staff member
Moderator
Joined
Nov 30, 2006
Messages
21,405
Location
Hickory, MS
But with 6.8k bus resistors, the current would be 3 times higher, and the gain of the summing amp could be reduced as much. I don't see any practical improvement here.
Perhaps I am not being clear (or reading it wrong) but the output of that transistor circuit is roughly the same current as a 6.8K bus feed resistor, but with the higher source impedance loading the bus of around 22k for a measurable reduction in bus noise gain (22/6.8 or around 10dB benefit).

JR
 

abbey road d enfer

Well-known member
Moderator
Joined
Jan 22, 2008
Messages
13,456
Location
Marcelland
Perhaps I am not being clear (or reading it wrong) but the output of that transistor circuit is roughly the same current as a 6.8K bus feed resistor, but with the higher source impedance loading the bus of around 22k for a measurable reduction in bus noise gain (22/6.8 or around 10dB benefit).
No, you've been extremely clear, but I'm clear too about the fact that 400uA rms is equivalent to the current that a 27k resistor can provide.
The circuit clips at 4V p2p for 400uA output current. As I said earlier that means a nominal operating level of about 0.4V p2p, or -15 dBu.
If 27k resistors were used the nominal operating level could be increased by 10 dB, which would result in identical performance of the summing amp and probably better level management in the channels.
I understand that operating at -15dBu allows higher headroom, but this headroom does not extend to teh summing amp, which is probably where it matters more.
 

dogears

Well-known member
Joined
Nov 15, 2017
Messages
477
Location
Texas
The buffer I was talking about adding if I was to revisit that old production console design was after the pan pots feeding the current sources. To save op amps (remember there were 72 pairs of these current sources just in the monitor section), I drove the synthesized current sources directly from the pan pots so pan position slightly imbalanced the current source output resistance multiplier. There was a just barely audible difference in master bus noise floor when listening WFO. The lowest noise (highest current source output impedance) was with all monitors hard panned left or right.

JR
Ah gotcha. I thought you were talking about this. By buffering the positive feedback Rs no longer factors into the gain, right? Wouldn't this "fix" the issue of matching?
wHuwJxm.png
 

JohnRoberts

Well-known member
Staff member
Moderator
Joined
Nov 30, 2006
Messages
21,405
Location
Hickory, MS
Posted this in the wrong thread:, meant it to be here:


Without doubt John, your elegant and upgraded I to V would be a superb way to sum many channels in a large desk.
If there was a market for these desks, the results you show would sway towards your system being de-facto.

The only reason I felt it was applicable to draw attention to the EMI, is that, if we stop being a little pedantic, I think it's fair to say that EMI's design drawing from 1966/67ish isn't a whole lot different from the basic idea you show in 1980.
in my mind it is still quite different (been around this tree before too).

I dismissed using a class a common emitter stage as a voltage to current convertor for several reasons. To be generous I didn't count the ways. Abbey mentioned a few of them.


No doubt in my mind that you believed this was an original idea, and why would you think otherwise since the prior art documents were filed away in EMI's "need to know basis" aisle. You thought of the concept on your own. Kudos. No way I could do that.
I did an extensive patent search back in the late 70s and abandoned my patent application because I couldn't afford to educate the wet behind the ears patent examiner regarding the difference between a current and a current source at what my patent lawyer was charging me per hour.

From prior art it was common practice to sum the outputs from vacuum tubes together without bus resistors and loss requiring make up gain, BUT these tubes were effectively operating class A so always nominally at 50% to provide +/- AC signals.

If you want to make a bilateral current source using discrete transistors just look at a VCA topology without the final current to voltage op amp stage. These bilateral current outputs from multiple VCAs could feed a VE bus without the N+1 bus noise gain...

Note: I never bench tested this and suspect after enough of these VCAs get summed the combined output capacitance might need to be compensated for but this is probably manageable. Modern VCAs are almost cheap enough to do this, but it would be obsolete before begun.


Also no doubt at all that you improved the schem beyond anything that came from the EMI.
No, you've been extremely clear, but I'm clear too about the fact that 400uA rms is equivalent to the current that a 27k resistor can provide.
The circuit clips at 4V p2p for 400uA output current. As I said earlier that means a nominal operating level of about 0.4V p2p, or -15 dBu.
If 27k resistors were used the nominal operating level could be increased by 10 dB, which would result in identical performance of the summing amp and probably better level management in the channels.
I understand that operating at -15dBu allows higher headroom, but this headroom does not extend to teh summing amp, which is probably where it matters more.
I guess I was being overly generous by ignoring the the shortcomings of a class A voltage to current conversion. I dismissed using those early in my pursuit of current source summing.
I just like to see the folks whose shoulders we all stood also be acknowledged sometimes
I guess we will continue to disagree about this. I do not mean to be dismissive of EMIs work, only that it did not resemble my current source summing approach.

JR
 
Top