pucho812
Well-known member
Was asked about this and I never heard the term current summing. What is the low down?
Voltage summing vs current summing
- Thread starter pucho812
- Start date
Help Support GroupDIY Audio Forum:
hmmmmm ;D
Back in the 70's i came up with a way to sum current sources, completely different from the widely held belief that virtual earth summing amps sum "currents". Even the wet behind the ears patent examiner didn't understand the difference. :
In a VE sum amp there will be a N+1 noise gain where N is the number of stems being summed. With ideal current sources being summed there is no noise gain (In practice I could reduce the noise gain more than 20dB). .
I wrote about this in my 1980 console technology article, but I didn't reveal exactly "how" I did it.
JR
Back in the 70's i came up with a way to sum current sources, completely different from the widely held belief that virtual earth summing amps sum "currents". Even the wet behind the ears patent examiner didn't understand the difference. :
In a VE sum amp there will be a N+1 noise gain where N is the number of stems being summed. With ideal current sources being summed there is no noise gain (In practice I could reduce the noise gain more than 20dB). .
I wrote about this in my 1980 console technology article, but I didn't reveal exactly "how" I did it.
JR
Bo Deadly
Well-known member
In a standard inverting op amp the output will source or sink current as necessary to hold the voltage at the negative input at the same voltage as the positive input (which is held at ground so 0V). This means the voltage changes very little. So your mixing currents and not voltages.pucho812 said:
I don't know the advantages / disadvantages of current vs. voltage summing but virtually all of summing mixer circuits that I have ever seen use the inverting op amp stage so it would be quite the revelation if it turned out not to be the best way.
Yup that's what the patent examiner thought too, and even pulled a reference from a text book that said that too. :'( :'(squarewave said:
Converting the input signals to currents in a VE sum amp is a convenient way to calculate gains and output results but not literally what is going on. I'm not sure I really want to go off into the weeds about what is really going on, but I have given this much thought over the decades.
Op amp output (voltage) is a function of the difference between + and - op amp input (voltages). Only when we connect negative feedback from output to the - input does that force it to follow the + input (how closely it follows depends on the open loop gain, there will be a tiny AC error voltage between the inputs ).
I realize I may be pissing into the wind since even the common nomenclature used to describe this is "Virtual Earth" and most people make the same ASSumption you have.
I'm not sure I even know what the difference is between what you call current or voltage summing.
http://groupdiy.com/index.php?topic=62460.msg791493#msg791493
http://groupdiy.com/index.php?topic=50034.msg632932#msg632932
http://groupdiy.com/index.php?topic=24788.msg294630#msg294630
I know I have written about this a lot more than just these few threads I found with a search, perhaps I'm thinking of more extensive discussions about current source summing over on Wayne's forum.
JR
PS: BTW contrary to popular opinion there isn't squat difference between passive and active summing, since the passive sum circuit needs make-up gain that coincidentally is almost exactly the same amount as the noise gain of a VE summer. Only my current source summing approach reduces the noise gain consequence, but yawn, modern digital combining is arbitrarily perfect so why bother?
ruffrecords
Well-known member
JohnRoberts said:
I am sure there is some truth in both if expressed appropriately. Taking your description as read, the - input is forced by NFB to closely follow the + input which for VE summing is connected to 0V. Under these conditions, the sum of the currents in the input resistors will be equal and opposite to the current in the NFB resistor, which is probably why some people call it current summing.
Cheers
ian
Yes, as I said the math works out neatly "thinking" about it that way.ruffrecords said:
I am probably over sensitive to this because I was arm wrestling with a patent examiner over the difference between summing current sources and summing currents, and I ended up abandoning my patent application back in the '70s because I didn't have enough money laying around to educate the patent examiner about what a current source is (hint, it's different from a resistor connected between two voltage nodes).
JR
I believe VE summing should not be called current summing, although part of the math applies.
Even voltage-summing is not correct; real voltage-summing can only be accomplished by connecting floating sources (such as xfmr secondaries) in series; to my knowledge it's never been used in audio mixers.
All the common summing topologies should really be called Thevenin/Norton/Kirchoff summing because both voltage and current are involved. VE simplifies the math, but is in the same category.
Only a system where signals are converted to current and then summed should be called current-summing.
I have already mentioned that I had the same basic idea as JR, at about the same time, but didn't try to patent it; I submitted it to Midas, who were interested, and conducted some experiments, but concluded that the improvement was not spectacular enough to justify the added complexity. That would have worked only if someone was ready to invest in developping a specific current-gen. As JR wrote, it is moot since digital summing, but, just for the same reasons there are fans of "passive" summers, there could be "true-current" summers; purists could skip the VE I-to-V converter and use their favorite triode...
More seriously, true current-summing would have been beneficial only on large-frame mixers (64+ channels)
Even voltage-summing is not correct; real voltage-summing can only be accomplished by connecting floating sources (such as xfmr secondaries) in series; to my knowledge it's never been used in audio mixers.
All the common summing topologies should really be called Thevenin/Norton/Kirchoff summing because both voltage and current are involved. VE simplifies the math, but is in the same category.
Only a system where signals are converted to current and then summed should be called current-summing.
I have already mentioned that I had the same basic idea as JR, at about the same time, but didn't try to patent it; I submitted it to Midas, who were interested, and conducted some experiments, but concluded that the improvement was not spectacular enough to justify the added complexity. That would have worked only if someone was ready to invest in developping a specific current-gen. As JR wrote, it is moot since digital summing, but, just for the same reasons there are fans of "passive" summers, there could be "true-current" summers; purists could skip the VE I-to-V converter and use their favorite triode...
More seriously, true current-summing would have been beneficial only on large-frame mixers (64+ channels)
moamps
Well-known member
What's wrong with already used name ACN (active combining network) for VE summing?abbey road d enfer said:
P.S.
A true current summing needs at the and some I-V converter so it is V-I-V network in reality, IMO.
or just call them summers.abbey road d enfer said:
Even I called it current source summing to differentiate between the all too popular terminology.
Steve Dove confided to me that after reading my 1980 article (where I didn't say how I did it), he used current source summing in a broadcast mixer (Alice?) and said the lower noise gain helped RF rejection at the bus. Funny half the people in the industry though I was full of bull, and Steve Dove figured out how to do it... 8)
Actually not all that hard... A decent op amp with 5 precision resistors can get you most of the benefit for modest cost. Even without trimming a one op amp synthesized current source delivering the equivalent current of a 20k resistor, can express an output impedance >>200k so easily 20 dB less noise gain vs. resistors, with acceptable channel noise floor (and this was using 1990's parts. )
A while back I talked with an IC design engineer at THAT corp about repurposing the precision resistors inside one of their differential input chips to make a very accurate and high compliance synthesized current source. We could have accomplished it with just a new top metallization layer, but the stopper for them was that they didn't have enough static/over-voltage protection clamps to protect every external pin in that configuration. IC makers these days put great effort into designing reliable chips, and an unprotected external pin was a no-go.
I used it in a console with over 100 stems to the L/R bus while I only used current sources on 72 of the stems. The other 40 or so used conventional resistors.As JR wrote, it is moot since digital summing, but, just for the same reasons there are fans of "passive" summers, there could be "true-current" summers; purists could skip the VE I-to-V converter and use their favorite triode...
More seriously, true current-summing would have been beneficial only on large-frame mixers (64+ channels)
Yes, I could make a better*** analog sum circuit than pretty much any other analog summing topology, but yawn why bother?
JR
**** As I discovered from developing this technology, reduced noise is not the only or even the largest benefit. Reduced noise gain results in less phase shift and less distortion especially in numerically large (100+) bus configurations.
John, you've mentioned your current summing patent application before. Have you got a note and schematic or link to show how its done in detail?
The 'standard' 1980s Calrec summer for big desks was VCA current sources but this fed the usual VE. We did it for the automation and considered the better noise a bonus.
The 'standard' 1980s Calrec summer for big desks was VCA current sources but this fed the usual VE. We did it for the automation and considered the better noise a bonus.
When cost is an issue, I believe the Howland pump is probably the best bet; just takes one, or preferrably two, opamps and a few resistors and caps.ricardo said:
ruffrecords
Well-known member
I was going to add that it is important to distinguish between what you are trying to achieve and how you achieve it - but I thought it would make the post to long. The point is, what you want to achieve is the summation of the signal voltages. If you had balanced floating outputs from each channel you could achieve this simply by wiring them in series but there are good reasons why you would not do it this way in practice. The hard bit is finding a good way to achieve what you want that has a minimum of compromises - like added noise etc. It is an unfortunate fact that patent attorneys are only skilled in the art of writing patents, not in understanding the(genuine novelty of) underlying concepts.JohnRoberts said:
Cheers
ian
While I have little opinion about that name to call this, the circuit i settled on resembles the "improved Howland current pump" fig 5 http://www.ti.com/lit/an/snoa474a/snoa474a.pdfabbey road d enfer said:
JR
Schematics for the AMR production series recording consoles are published and around. There are 72 current sources in the 36x24 so shouldn't be that hard to find, 3 in each monitor strip.ricardo said:
I'm sure I've posted schematics before, including discussing what I would do different, next time around.
I recall starting a thread over at Waynes forum about a hypothetical VCA based summing system... at some point the bus capacitance might be an issue, but I expect that is manageable.
I did one old Loft console with current source summing and a dbx VCA in place of the sum amp. Since noise gain was no longer a concern the VCA as a L/R bus master amp added some interesting headroom and other features.
JR
pucho812 said:
I guessed you asked a question that nobody answer and they used your topic to talk about boring current sources.
You were talking about summing a bunch of signals using a bunch of resistor and a gain stage. What's the difference between using the non inverting input (don't say passive) against using the inverting input (here we can say active but it shouldn't describe much)? If that was your question I will answer with a non trivial or empty answer, after a lot of thoughts no to while ago I found one difference, one inverts the polarity and the other doesn't.
Nah, talking seriously, but don't let anybody know :-X :-X :-X
The difference is:
X-talk!
In practical terms, when building a few bus mixer with as few gain stages as you can, you could avoid a buffer before the summing resistors and live with a bit of source impedance (pan pot, fader, both, whatever) If you accept some gain error in either case. When using the inverting input the crosstalk is much lower than if you where using the other one.
Let's say you have a source impedance of 2k5, 10k fader half way. Two channels. 10k summing resistors. 2 bus, one channel going both bus and the other going to just one. Calling X-talk the signal of the channel going to just one bus in the bus which only has the other channel. (I should make a drawing but I'm feeling lazy today)
If using the inverting inputs the X-talk should be really low depending on the parasitic resistance of the traces, and the OL gain of the opamp and other variables. With ideal components zero.
If using the non inverting input you have a wired network of resistors going back and forth, let's passivate the voltage at the channel going to both bus and put one volt at the other. It only goes to one place so it has 1V with 12k5 source meeting the other 10k in the non inverting node with infinite impedance, so 22k5 impedance meeting the 2k5 of the fader of the passivated channel. We have 0.1V there, across a 10k resistor going to a bus which should have 0V or nothing at all, but it has 100mV or 100000µV, that ain't nothing, it's something, 10% leaking from somewhere or 20dB X-talk I'd said.
Note that I ignored the gain errors, I did nothing with the unused resistor which should be connected somewhere, changing the numbers, I guess I made my point here. The moment you place buffers right before the summing resistors this effect disappears and the only difference you end is the polarity I mentioned but as you have buffers now you could swap it if you want there.
JS
PS: Remember :-X :-X :-X
Talk about hijacking! ;D
ruffrecords
Well-known member
joaquins said:
For a detailed treatment for tube mixers see this:
https://drive.google.com/file/d/0B_n67A1hN3qtaFdhNW8zcmpkZDg/view?usp=sharing
Cheers
Ian
