Analog summing again

[arigy]

Dear friends! As I noticed, analog summers are all based on current summing, in one way or another. Traditionally, we use a resistors as a current source, then ground them all via load resistor (passive summing) or use inverting opamp (active aka virtual earth summing). If we are talking of passive summing, the source resistors with Rload are forming a divider, which leads to substantial level drop, so we have to use output amplifier with high gain, bringing up noise.

But what if we will use active current sources instead of resistors, such as Howland Current Source (http://www.national.com/an/AN/AN-1515.pdf)? As these current sources are independent of Rload, the output voltage will be independent of channel number, so we'll only need unity gain buffer on output.

Keeping in mind internal trimmed resistors, we can make Howland Current Source of any Balanced Receiver, like SSM2141 or INA134 with just two precision resistors.

What's the drawbacks of this approach? I know there should be

 

[Rob Flinn]

Had you thought about using VCA's, they output a current.

 

[JohnRoberts]

Dear friends! As I noticed, analog summers are all based on current summing, in one way or another. Traditionally, we use a resistors as a current source, then ground them all via load resistor (passive summing) or use inverting opamp (active aka virtual earth summing). If we are talking of passive summing, the source resistors with Rload are forming a divider, which leads to substantial level drop, so we have to use output amplifier with high gain, bringing up noise.

I ran into trouble explaining the distinction between combining voltages sources feeding through bus resistors into a virtual earth, and summing actual current sources (Howland or whatever) with a patent examiner back in the '70s when I tried to patent this. Most textbook explanations for how virtual earth summers work use similar language about "combining currents" because it it so much easier than explaining the complex interaction with loop gain margin et al.  I might have eventually educated the examiner by throwing a few more thousands of dollars at my patent lawyer, but I was spending my own dollars, and abandoned the effort.

But what if we will use active current sources instead of resistors, such as Howland Current Source (http://www.national.com/an/AN/AN-1515.pdf)? As these current sources are independent of Rload, the output voltage will be independent of channel number, so we'll only need unity gain buffer on output.

I actually used a variant on the Howland closest to their figure 5...

Keeping in mind internal trimmed resistors, we can make Howland Current Source of any Balanced Receiver, like SSM2141 or INA134 with just two precision resistors.

What's the drawbacks of this approach? I know there should be

I don't follow how to easily make a balanced receiver into a Howland but maybe you're smarter than me.

I sure wish I could have asked somebody how to do this 30+ years ago.  ;D

Long story short, yes it actually works, but to come up with a practical implementation you need to get familiar with how the Howland approach and positive feedback works, and what the noise important mechanisms are. There is little use replacing a 10k resistor with an active current source if that current source makes 10x the noise... 

I'd love to revisit this with modern parts, and wish the world really needed a better analog sum bus but unfortunately, if you're only combing 10 or 20 stems, KISS. Finally for any number of channels, digital combining should be better than analog, I still don't understand the attraction of OTB mixing, other than perhaps to use a bunch of old school analog effects.

FWIW I built one (split) console with >100 inputs to the L/R mix and the whole right side of the console (3x24 or 72 feeds) were through synthesized current sources.

There are many practical considerations, but you need to do a little homework before i give it all to your on a platter. 

JR

PS I even had a cute name for this "i-bus" (using i the symbol for current long before apple co-opted it)

 

[arigy]

Had you thought about using VCA's, they output a current.

Well, that was an original idea, actually. On second thought, if we do not want any input level/panning control, and we already have IC based balanced input, we can easily omit VCA, thus reducing noise and distortion.

I actually used a variant on the Howland closest to their figure 5...

I don't follow how to easily make a balanced receiver into a Howland but maybe you're smarter than me.

I was thinking on solution similar to figure 5, but using internal trimmed resistors and adding two more to set output current:

 

[JohnRoberts]

Had you thought about using VCA's, they output a current.

Yup, VCAs should work, but you do need to deal with their output capacitance if you throw a bunch of them onto a bus. I've pondered this at length on a another forum.  http://www.proaudiodesignforum.com/forum/php/viewtopic.php?f=6&t=211

I even generated some schematics, but never melted solder... (aka mental masturbation).  It seems VCA based consoles, could do this with little cost penalty, while they need to deal with the practical considerations.

Note: for uber bench performance the VCA running at unity gain could dominate the bus noise floor. 

JR

 

[JohnRoberts]

Had you thought about using VCA's, they output a current.

Well, that was an original idea, actually. On second thought, if we do not want any input level/panning control, and we already have IC based balanced input, we can easily omit VCA, thus reducing noise and distortion.

I actually used a variant on the Howland closest to their figure 5...

I don't follow how to easily make a balanced receiver into a Howland but maybe you're smarter than me.

I was thinking on solution similar to figure 5, but using internal trimmed resistors and adding two more to set output current:

Yup,, that'll work. Resistors obviously need to be precision, and both + and - input resistors need to be driven by low impedance sources or it will degrade the output compliance.

I used a slightly different arrangement of values, but you're in the ballpark. 

JR

 

[bezen4uk]

Hi,
I had  same idea but with penthodes with common plate resistor. Never tried to realize this.

 

[barclaycon]

Have a look at how old Neve desks do their summing. I think you might get a shock!

 

[PRR]

> analog summers are all based on current summing, in one way or another

???

 

[JohnRoberts]

Hi,
I had  same idea but with penthodes with common plate resistor. Never tried to realize this.

Summing together the high compliance current outputs from several common emitter/source (cathode?) stages,  not only has operating point issues, since the DC current from N input stages combines linearly, but the noise from combining several such stages is dominated by the relatively high class A operation of input stages, not the combining amp. 

I settled on the synthesized current sources (Howland or variant thereof) as a good way to get an AC current output with high source impedance, low noise, and good linearity, without breaking the bank. When doing this times a hundred inputs to L/R in one console the cost is not inconsequential.

The noise of these current sources matters, but the obvious benefit  of this topology is that the noise combines incoherently. A conventional 100 input virtual earth bus, would suffer a noise gain of 101x the sum amp's input noise.  Alternately combining 100 high compliance current sources would only combine to 10x the noise of a single current source.

Despite the noise benefit, and as I mentioned in my article on the subject back in 1980, the noise is already dominated by one microphone input at reasonable gain. The real benefit of this bus topology is the improved phase response and linearity from operating at lower noise gain.

INO to completely characterize a mix bus's performance we need to see not only output noise level, but phase response at 20 kHz, and distortion with all inputs assigned.

JR

PS: I used a variant on this for a different reason in a later Automatic Mixer design. When combining together multiple mixers to make them operate like one large Automatic Mixer, a problem is dealing with the insertion loss of simply adding together some internal buses. By using a synthesized current source feeding a separate termination resistor, that can be disconnected via a rear panel screw terminal, multiple current outputs can be summed into a single termination resistor with no insertion loss. I did a patent search for this application, and the patent examiner found an example in some (very old) art of a tube design that combined several such outputs. While the old prior art did this for some unrelated reason, it looked close enough to block mine. 

 

[JohnRoberts]

Have a look at how old Neve desks do their summing. I think you might get a shock!

Please expand on that... I am only aware of one other console designer using this technology, and he told me at a trade show that he got the idea from my 1980 article. I didn't say how, just that it could be done. He was clever enough to figure out how.

That said I haven't studied everybody's console schematics, so I'm always prepared for some shock and awe. 

JR

 

[ruffrecords]

Have a look at how old Neve desks do their summing. I think you might get a shock!

Please expand on that... I am only aware of one other console designer using this technology, and he told me at a trade show that he got the idea from my 1980 article. I didn't say how, just that it could be done. He was clever enough to figure out how.

That said I haven't studied everybody's console schematics, so I'm always prepared for some shock and awe. 

JR

There's no mystery. Old Neve consoles used passive summing.

Cheers

Ian

 

[JohnRoberts]

Have a look at how old Neve desks do their summing. I think you might get a shock!

Please expand on that... I am only aware of one other console designer using this technology, and he told me at a trade show that he got the idea from my 1980 article. I didn't say how, just that it could be done. He was clever enough to figure out how.

That said I haven't studied everybody's console schematics, so I'm always prepared for some shock and awe. 

JR

There's no mystery. Old Neve consoles used passive summing.

Cheers

Ian

Well perhaps a little mystery (to me) since this thread is about using current sources for loss-less combining.

Passive and virtual earth combining despite appearances involve almost identical amounts of insertion loss/make up gain, while the virtual earth summer's gain is concealed as as noise gain.

JR

 

[ruffrecords]

Have a look at how old Neve desks do their summing. I think you might get a shock!

Please expand on that... I am only aware of one other console designer using this technology, and he told me at a trade show that he got the idea from my 1980 article. I didn't say how, just that it could be done. He was clever enough to figure out how.

That said I haven't studied everybody's console schematics, so I'm always prepared for some shock and awe. 

JR

There's no mystery. Old Neve consoles used passive summing.

Cheers

Ian

Well perhaps a little mystery (to me) since this thread is about using current sources for loss-less combining.

Passive and virtual earth combining despite appearances involve almost identical amounts of insertion loss/make up gain, while the virtual earth summer's gain is concealed as as noise gain.

JR

Absolutely right. I have the Neve technical report on that very topic and it concludes just that.

Cheers

Ian

 

[barclaycon]

Not just 'passive' summing.
You are effectively mixing into a transformer.

 

[ruffrecords]

Not just 'passive' summing.
You are effectively mixing into a transformer.

Not really. There is a transformer that's true but the mixing itself is entirely due to the bus resistors and the transformer effectively 'bridges' the bus. It is no more than a gain make up amplifier with a transformer input. The mixing is entirely unbalanced; the transformer just gives a bit of 'noise free' gain.

Cheers

Ian

 

[abbey road d enfer]

Not just 'passive' summing.
You are effectively mixing into a transformer.

A transformer is passive...

 

[abbey road d enfer]

Have a look at how old Neve desks do their summing. I think you might get a shock!

Please expand on that... I am only aware of one other console designer using this technology, and he told me at a trade show that he got the idea from my 1980 article. I didn't say how, just that it could be done. He was clever enough to figure out how.

That said I haven't studied everybody's console schematics, so I'm always prepared for some shock and awe. 

JR

There's no mystery. Old Neve consoles used passive summing.

Cheers

Ian

Well perhaps a little mystery (to me) since this thread is about using current sources for loss-less combining.

Passive and virtual earth combining despite appearances involve almost identical amounts of insertion loss/make up gain, while the virtual earth summer's gain is concealed as as noise gain.

JR

Absolutely right. I have the Neve technical report on that very topic and it concludes just that.

Cheers

Ian

You're right. At the end of the day, there is no intrinsic difference in the performance of passive vs. virtual earth.

Each solution has its pros and cons, mainly:
Passive mixing needs an extremely stiff ground reference for the mixing stage, because the injection resistors actually inject current in it. If it's not stiff enough, there's a lot of x-talk. A serious concern in multi-track mixers.
OTOH, affixing an input transformer to a passive mix amp improves its noise performance drastically because its optimum noise impedance can be tuned to the low impedance of the mix bus (typically about 200 ohms for a 48 ch mixer).

Virtual earth mix requires a VLN mixer amp, precisely because of the VLZ nature of the mix bus. There's the added advantage of reduced x-talk because the voltage is zero on the bus (a consideration that matters only with shabby pan-pot arrangements).
The ground reference for the VE mixer does not need to be stiff, because it drives a high Z input, it just needs to be clean.

 

[JohnRoberts]

To perhaps UN-VEER a little...  The subject of this thread, substituting current sources for input resistors feeding a virtual earth sum bus, enjoys the benefits of virtual earth (like low crosstalk) while eliminating or significantly reducing the insertion loss or noise gain required. Another little inspected benefit of current source summing is that ground potential differences between channel and master are likewise reduced in influence. Since the noise contribution of the sum bus amp is no longer multiplied by a high noise gain term you can use a JFET opamp there for improved RF rejection.

As usual this isn't a free lunch, since you need to use practical (non-ideal) current sources that aren't noiseless nor do they exhibit infinite source impedance. IMO the approach has merit for very large analog consoles, but the sand is quickly running out of that hour glass....

In theory digital combining is lossless, noiseless, whatever less.... So current source summing is a good idea (IMO) whose time has lapsed.

JR

 

[arigy]

I've tried googling information on Howland Current Source noise, and found nothing but an article in International Journal of Electronics on Noise analysis of a Howland current source (http://testservice-eprints.gla.ac.uk/23847/). Unfortunately, this article is not available online, and I personally am not qualified enough to analyze it myself

Regarding practical implementation: as it is vital for Howland Current Source to have precisely matched resistors (order of 0.01%), it might be inconvenient for actual use. However, instrumentation amplifiers and balanced receivers are all already has matched set of resistors (0.05% generally) which we can easily use, and previously mentioned AN-1515 (Burr Brown also has an Application Bulletin on this, see fig.51 http://focus.ti.com/lit/an/sboa046/sboa046.pdf) also suggest alternative solution involving OpAmp voltage follower, which allows us to use only a single resistor for setting of current.

The problem is: different balanced receivers use different values of feedback resistors, and, if I understood correctly, it is better to have gain resistor value less then feedback resistors. If we stick to INA134/SSM2141, I suppose it is okay to have gain resistor of 10k (which will give us 1mA for 10V differential input). However, THAT1240, which considered a superior balanced receiver, has 9k resistors. Wouldn't 10k gain resistor too much for it? What bad will happen, if we change gain resistor, for example, to 4.7k (will give us 2.13mA for 10V diff.input)?