Quoll - a balanced, discrete IV stage

[PigletsDad]

Attached below is a schematic for a discrete I/V stage, intended to be used with the Twisted Pear Buffalo or COD DACs. It uses Nelson Pass' supersymmetry concepts, with a couple of twists of my own. The gain block effectively has 5 terminals - ground sense, Input A and B, Output A and B. In use feedback loops are closed by putting resistors from Output A to Input B and vice versa.

The symmetry of the output is maintained by a separate path that compares the outpoint mid point with ground. In principle, if an application needed the complementary outputs to move centred on some other voltage, this could be fed into the base of Q1 instead.

The schematics have been checked by simulating the extracted Spice file, but not built.

The most important idea is in the next post, which will show how the basic block is used.

By the way, a quoll is an Australian marsupial carnivore, about the size of a cat. I saw them on holiday in Tasmania, and they most attractive. See
Wikipedia for more details.

[PigletsDad]

The file attached here shows how the bock would be used.

Note that the dominant pole compensation components are placed before any active devices, at the "virtual earth" points. This removes high frequency rubbish before it encounters any devices that could possibly overload, fixing the major objection to feedback based IV circuits.

For a Buffalo, about 180 Ohms for RF and RF1 should give roughly Red Book levels.

Just a reminder, this has only been simulated, not built!

[Russ White]

PD,

Clever.

That's very similar to the "original" counterpoint design I presented in the "Haiku" thread where I took feedback at the bases of the input pairs. I used a folded cascode where you are using a resistor load. I also used a diamond buffer type output stage.

I am sure the cct would work fine. Mine worked very well indeed, but in the end I decided I did not want to use Nelsons IP, and would try to create some of my own.

You will need a way to trim the output offset unless you can match things very very well.

Good work!

Cheers!
Russ

[PigletsDad]

Good point about the offset.

Because the DC gain is low (about 1) and the impedances very low, the offsets will be dominated by the Vbe mismatches of the differential pairs. If they come from the same batch, I think you would hope for less than 20mV, which should be acceptable for many applications.

If lower offsets than that are needed, the pairs of 47 ohm emitter resistors can be swapped for a trim pot.

I went for resistor loading, rather than folded cascode, as it gives some extra CMRR in the second stage.

[Russ White]

Yes, like any opamp used for I/V, this one has a high open loop impedance, and low closed loop impedance.

I tried resistor loading too, but found it did not work as well for me.

Also you would likely need to augment the output devices to retain decent THD into lower impedance loads.

I also would look closely at how the circuit handles the fast slewing of the DAC as well as capacitive loads. This is the one major weakness of direct I/V using massive feedback.

How will the cct react to a short at the output?

Build one and give it a shot.

My discrete SUSY opamp I/V sounds excellent but is essentially the same as a THS4131.

Cheers!
Russ

[PigletsDad]

Originally posted by Russ White
Yes, like any opamp used for I/V, this one has a high open loop impedance, and low closed loop impedance.

I tried resistor loading too, but found it did not work as well for me.

Also you would likely need to augment the output devices to retain decent THD into lower impedance loads.

Sure - I think Diamante was designed to drive headphones? I wanted to keep things simpler, and save at least one stage by targetting line level loads (although it drives say 600 Ohms very nicely). This is a deliberate trade-off - I didn't want to redo you work from Diamante!

I also would look closely at how the circuit handles the fast slewing of the DAC as well as capacitive loads. This is the one major weakness of direct I/V using massive feedback.

I have studied this extensively, and this is why the compensation caps are in front of the input stages; you literally cannot make this slew rate limit with any signal in the DAC output range, no matter how fast; the dominant pole lies before the first gain stage.

The downside of doing this is that noise floor rises above the pole frequency, but I view this as a viable tradeoff at these signal levels.

Capacitive loads up to several nF are OK. Obviously a big enough load can make it unhappy at high frequencies - but it can drive tens of meters of cable without trouble.

How will the cct react to a short at the output?

As shown, with loss of the magic smoke! In practice, I would put say 50 Ohm stoppers in the output lines to isolate it from cable capacitance, and these will just about save the output devices. This also gives more or less matched impedance drive to STP.

[Russ White]

You will have to explain how you accomplish the pre-input filtering without raising the input impedance.

In the scheme you show in you second pic both the negative feedback and the input will be equally shunted to ground.

I just simulated, and this does not produce an effective filter.

I also have tried this with fully symmetrical opamps, it does not work.

To make it work, you need an input resistor after the filter cap. this will raise input impedance.

Also I was not referring to Diamante, to my opamp version of counterpoint.

I will try to dig up a link.

Cheers!
Russ

[Russ White]

http://www.diyaudio.com/forums/attac...amp=1204571940

[PigletsDad]

The caps are not for filtering - they are for stability!

If you want filtering, you add a cap across RF; I have simulated this a bit but need to do more.

For simplicity, lets start by considering a single ended case, with an inverting amplifier.

Consider the parallel combination of feedback resistor and input resistor (if any, say 200 Ohms for Sabre) as Thevenin equivalent source, with the FB loop open, but the far end of the feedback earthed (or otherwise driven from a low impedance). This interacts with the capacitor, to give HF rolloff. If we think of the input terminal of the circuit as lying before the equivalent source resistance of the Thevenin source, we find that open loop response of the amp rolls off, with a dominant pole set before the active device, giving overall stability without needing to put a roll-off at some interior point.

Closing the loop flattens the response (of course).

I have to go home now (wife stuck in traffic, dogs to walk, horses to muck out), but will put some simulations together tomorrow to show what I am getting at.

[Russ White]

Originally posted by PigletsDad
The caps are not for filtering - they are for stability!

I have to go home now (wife stuck in traffic, dogs to walk, horses to muck out), but will put some simulations together tomorrow to show what I am getting at.

I see now what you are saying.

I will give this a shot. I should work even for THS4131.