Clever circuit or?

Attached is a small schematic of a balanced output driver copied from a Jensen web page. The top op amp is a non inverting buffer with an odd configuration. The obvious way to do it is to throw away the two resistors and just connect pin 2 to pin 3 to make a classic unity gain buffer. As drawn, if the 2K connected to pin one was connected to 0V you would get minus 2X the input at the output. I can sort of see that connecting it instead to the input signal add 1X so you get -1X at the output but why would you want to do it this way. Is it very clever and subtle? The + and - inputs must be at the same level and the + is the signal so the - must be signal which means no current can flow in the input 2K. By simialr reasoning, none can flow in the output resistor.

I am confused.

Cheers

Ian

Or maybe it's just trying to balance impedance as seen by the - inputs for the purpose of matching offsets or such?

At least it is interesting that the impedance seen by pin 6 is the source through 2K in parallel with OA output through 2K and the "odd configuration" of the upper amp matches that exactly.

ruffrecords said:

Attached is a small schematic of a balanced output driver copied from a Jensen web page. The top op amp is a non inverting buffer with an odd configuration. The obvious way to do it is to throw away the two resistors and just connect pin 2 to pin 3 to make a classic unity gain buffer.

Click to expand...

2 and 3 are op amp inputs, perhaps you mean connect pin 2 to pin 1?

As drawn, if the 2K connected to pin one was connected to 0V you would get minus 2X the input at the output. I can sort of see that connecting it instead to the input signal add 1X so you get -1X at the output but why would you want to do it this way.

Click to expand...

A1 is simple unity gain follower, but not that simple. A2 is unity inverting so transformer gets 2x input voltage opposite polarity on the two transformer legs.

Is it very clever and subtle? The + and - inputs must be at the same level and the + is the signal so the - must be signal which means no current can flow in the input 2K. By simialr reasoning, none can flow in the output resistor.

I am confused.

Click to expand...

Deane was picky about not letting DC into his transformers. Speculating that pin numbering scheme looks like a dual op amp, and the 5532 was the most likely candidate to drive 600 ohms, circa when that was published.

I'll take the easy one first. The unity inverting stage ASSuming a bipolar op amp will have to suck (or supply) the input bias current through the feedback resistor resulting in a small DC error voltage at that output. Ideally to keep DC out of the transformer we want the non-inverting op amp to exhibit the same voltage error at that op amp output. It appears we would already get that without the added R1.

The copy says drive should be 50 ohms with no DC, so I do not see an obvious DC balance contribution from R1. 

For subtle, adding R1 between + and - op amp inputs attenuates the NF feedback making the op amp behave like it is running at a noise gain of 2x, same as the bottom inverting stage.  In theory this would make the transfer function of the top and bottom op amps more similar for stability, phase shift, distortion, etc.

This looks like over-thinking a possible problem, but might be the kind of thing Deane or his crew would do. 

JR 

Cheers

Ian

Click to expand...

Actually there are other ways of balancing offsets, but as John says, this one has the potential advantage of balancing the dynamic response of the two halves, but who cares about it when there's a transformer in the signal path?
IMO there are much better uses for 6 resistors and two opamps, particularly driving the primary with negative impedance, which would be a significant improvement over the 48 ohms of this circuit.

Yeah, I don't get it it either. Even a 5532 would have an offset below 1mV with a 1K impedance which is less than a uA of DC through the transformer. And these days there are plenty of OAs that have lower offset and better drive (RC4580 comes to mind).

But the circuit is kinda clever. There might be some scenario where you would want to cancel offsets somehow. I will file this away in the back of my brain somewhere.

squarewave said:

Yeah, I don't get it it either. Even a 5532 would have an offset below 1mV with a 1K impedance which is less than a uA of DC through the transformer. And these days there are plenty of OAs that have lower offset and better drive (RC4580 comes to mind).

But the circuit is kinda clever. There might be some scenario where you would want to cancel offsets somehow. I will file this away in the back of my brain somewhere.

Click to expand...

IMO This was all about normalizing the loop gain margin (noise gain) of both op amps.

I don't think it was all that useful of a strategy but i can see their motivation. (i have killed my share of brain cells after beer o'clock, over-engineering designs.)

JR

Thanks guys. Nice to know you are as puzzled as I am.

Cheers

Ian

80hinhiding said:

Abbey, 24R9 + 24R9 is 49.8R, so it's closer to 50ohms than 48 no?

Click to expand...

ther is no value in driving a xfmr with precisely 50 ohms; whatever the real value , it's not as good as driving with almost-zero ohms, but driving with negative impedance beats it.

In cases where I've seen transformer datasheets where they recommend the tx be driven by xR (i.e. 600ohms) are they really suggesting that there needs to be a series resistance of that amount, split between both inputs of the tx?

Click to expand...

There is no benefit in driving a xfmr with additional resistors. I think they put resistors to minimize possible effects of offset voltages (doubt it since they went a tad anal with their circuit) and to protect/isolate the oamps.

This negative impedance you speak of, that sounds like something very difficult to achieve.

Click to expand...

No it's not. https://www.lundahltransformers.com/wp-content/uploads/datasheets/feedbck.pdf

  I think I remember you mentioning something about that a few years ago when I was more active here, in the form of feedback to the secondary? 

Click to expand...

It's one of the possibilities. The only disadvantage is that it takes an additional winding.
Compared to the negative impedance drive, the circuit is very stable and does not need to be tuned specifically to the xfmr resistance.

abbey road d enfer said:

ther is no value in driving a xfmr with precisely 50 ohms; whatever the real value , it's not as good as driving with almost-zero ohms, but driving with negative impedance beats it.

Click to expand...

I vaguely recall a phone call with Deane Jensen where I asked him about driving transformers. As I recall he suggested the importance of driving transformers from very low impedances to damp non-ideal artifacts from the magnetic components, the discussion was somewhat over my head but Deane suggested that driving transformers from higher source impedances would exhibit distortion from magnetic artifacts.

Deane was also experimenting with different transformer core metallurgy. IIRC he shared that his sister was a chemist with experience in material science, so he would pick her brain about core metal choices. 

There is no benefit in driving a xfmr with additional resistors. I think they put resistors to minimize possible effects of offset voltages (doubt it since they went a tad anal with their circuit) and to protect/isolate the oamps.
No it's not. https://www.lundahltransformers.com/wp-content/uploads/datasheets/feedbck.pdf
It's one of the possibilities. The only disadvantage is that it takes an additional winding.
Compared to the negative impedance drive, the circuit is very stable and does not need to be tuned specifically to the xfmr resistance.

Click to expand...

I suspect we are over analyzing this and doubt that circuit was designed by Deane. I stand by my speculation that it was an attempt to normalize the noise gain of the two drivers for matched stability margin, et al.

I also doubt it was worth doing or copying but YMMV.

JR

ruffrecords said:

Thanks guys. Nice to know you are as puzzled as I am.

Cheers

Ian

Click to expand...

Actually I'm not puzzled at all. I think John and I understand the reasons for this topology, and I maintain that the premisses are wrong. The effect of adding these two resistors is much worse than whatever could happen if the noise gains, offsets and dynamic responses were not über-balanced.
I believe the guy who designed this circuit knew much less about xfmrs than opamps.

I have not done any tests and therefore I answer by feeling: it may be that in this way the I / O impedances of the two sections are the same so that the two op-amps take exactly the same input signal level and, at the same time, do not unbalance the output transformer? In the end a way not to use servos