Balanced Inputs

[mediatechnology]

I still remain apprehensive about tying a shield (antenna) to a virtual earth.

Yeah, that bothers me too. Been thinking about that one. Might be a good place for an inductor...

Though it was in the BW of the amplifer, did you see how good it did with 820 kHz?

[JR.]

This figure 3 is an odd bird, with my beloved 5 resistor bilateral current sources (using 6 resistors) as input loads. For normal (differential) inputs, those current sources look like high impedance, but for common mode signals they look like an impedance to ground (<47k?). Since those bilateral current sources provide the input DC path, bias currents and dc offsets in those opamps probably matter.

JR

[mediatechnology]

Since those bilateral current sources provide the input DC path, bias currents and dc offsets in those opamps probably matter.

Yep, in our audio world I suppose we could AC-couple the current sources.

Though I've seen the balanced version I'm still more intruiged by the SE input. It might be able to do more for us. There's a lot of silicon in figure 3.

[JR.]

I would certainly cap couple somewhere in the path with a 200x gain stage.

Taking a wild guess that was probably not a production schematic.

JR

edit- understood for HRT monitor they want vLF response, but they can still do that with cap coupling./edit

[mediatechnology]

It's got a DC gain of 1 (or near 1). Checkout in the text the description of R2, R3, C3. He uses a very clever trick. Looks like a feed-forward servo?

Dobrev Link (page 2.)

I haven't fully wrapped my head around the current pump at the input much less his AC-coupling trick. I think there's something else he did we can learn from.

with my beloved 5 resistor bilateral current sources (using 6 resistors) as input loads.

Oh, that's how you did the current outputs on the channels...

[JR.]

OK the DC path is even more convoluted...kind of like following an Escher drawing. I'd need to look up the opamp's input bias currents and DC offsets,, but yes i was incorrect the 200x is AC gain.

Since the DC path to ground for the input is through that current source, if there is an error current in or out, that error will work against that current source's output impedance, so the better matched the R values they worse a small error current could be. In other words not a trivial calculation.

========

I don't think my 5 resistor current sources have been much of a secret since Peavey published schematics of the AMR consoles back in the '80s-90s.

While i still think ti would be slick to have uber-matched resistors on the chip itself. The mechanism for noise is resistor values, while the mechanism for output impedance is resistor values and quality of matching. So better matching could deliver the same output impedance with lower resistance for lower noise. but yawn... they want ice water in hell.

JR

[mediatechnology]

OK the DC path is even more convoluted...kind of like following an Escher drawing.

Yes indeed. Some of the servo drawings I've seen elsewhere with the servo op-amp pointing backwards look like Escher's "Hands Drawing Hands."

Dobrev's DC compensation is more like one of the Escher stair drawings. It's a very clever trick - one we might use somewhere. Fair to call it a feed-forward servo? (It's got that big passive RC network integrating at the input of the follower.)

I don't think my 5 resistor current sources have been much of a secret since Peavey published schematics of the AMR consoles back in the '80s-90s.

Well they were well-kept from me. I never saw them but I finally figured out what you'd done. Did you also use the VCA sum amp in that one?

I hope to be distracted less today and return to the bench to work on the SE version of Dobrev. I also want a clean drawing of it re-configured as a coaxial input.

One thing that occurred to me about Dobrev's method of bootstrapping the input (the original one) is that it could allow the use of bipolar op amps for an instrument input. The noise current develops across a 10K but the input impedance is 10K ~ 1M depending on what you want it to be via a second termination resistance. There's nothing new about bootstrapping an input, but the use of an INA134 or THAT1240 as the bootstrap amp provides more accurate bootstrapping.

[JR.]

No the Peavey/AMR was a mixed current source, with resistor feeds to L/R mix. As many as 72 current sources on the monitor side of the 24 bus console, with 40 odd resistor feeds from input side of the split.

The VCA in the master section with all current source feeds was done in a earlier Loft console back in the late '70s early '80s

I had an improved second generation version of the AMR console working in the lab, but the program was dropped when the Mackie 8bus for $4k retail sucked all the air out of the market for mine. Developing a big dog console working for a company like Peavey was a struggle. One of my first prototypes was sold into Canada, and I had to go up there and tweak it. arghhh.

JR

[mediatechnology]

OK. Had a chance to play with the SE Dobrev circuit with imbalanced source impedances and the common mode rejection is poor. With that inverting input summing node hanging out there in the wild world there's a lot to not like about it.

I think there's merit to the current pump at the input and the fully-balanced version - which has some other neat tricks in it - might make a good INA. I don't think I want to go much further with it.

One common theme linking the Super Bal, Super-Duper Bal and the Dobrev circuits are the use of active correction signals at the input. In the cases of the "Super" inputs it's non-inverting input which gets lowered.

The bottom line is that the best performer is also the simplest - The "Cohen" cross-coupled impedance and level-balanced "Super-Duper-Bal." No external parts, just a THAT1286 and bypass caps.


Super Duper Bal showing the THAT1286 internal resistor networks.

I wanted to visit one point I missed earlier in the thread where we were discussing switching the output and DC-coupling. Any DC appearing at the input in common mode (within its operating range) will be rejected by the CMR of the line receiver. Only differential DC will be of concern if the output is switched.

With active line drivers such as the THAT1646, DRV134 and SSM2142 that do not use sense capacitors, the majority of the offset is common mode. The typical differential offset is quite low, IIRC for the 1646 6 mV max. That figure is again reduced by half due to the 6 dB attenuation of the 1286.

[JR.]

Looking at the 6 resistor version of 5 resistor bilateral current source, I am once again curious about wether THAT has some spare resisostors hanging around on the substrates that could be repurposed. I precision low noise current source may be a limited market, but cooler than the batmobile.

JR