What's the best dual I/V opamp these days???

[diyralf]

If I take your 2 matched transistors and change the room temperature 3°C, then they are no longer matched.

[diyiggy]

comon heatsink ? and close ?

[EUVL]

The only critical matching in that circuit are the NN & PP pairs at the input.
They are in ONE SOT23.5 package.
So they also track each other thermally.

But you do not have to take my word.
Use opamps if it pleases you.
That is afterall the only thing that matters.


Cheers,
Patrick

[Jon Lord]

Originally Posted by diyralf

If I take your 2 matched transistors and change the room temperature 3°C, then they are no longer matched.

This could be cured by a very slow acting servo.

These pseudo-symmetrical discrete circuits are not as symmetrical as one might think, so matching NN and PP is not the dominating problem, its the NP matching of dynamic parameters.

IMO, such CFB circuits make more sense on an IC with very careful design like AD844 or AD846-(RIP) or very interresting:

The LT6275, which shares the advantages of CFB (Virtually no slew limiting) and VFB (high input impedances)

[Jon Lord]

In case you add a PI-Filter (with large PP or C0G caps up to 100nF) to the input node, you can easily come away with a good, but cheap audio OPA like OPA1642 or the like, because this filter unloads the I/V Converter from the speedy current transients, overloading the input stage, which can cause momentary slew limiting in traditional VFB designs.

The added noise gain (leading to a rise of noise outside the audio band), the succeeding aliasing filter stage can cope with. A second order filter may be enough, since you (overall) get a 5th order response.

I successfully adopted this trick as an upgrade for the SONY SCD-1 SACD-player, whose "bitstream" current DACs are very demanding regarding the I/V-Converter. The original LM6172 was even not fast enough in the original design, but after the PI filter change a LME49720 was more than sufficient regarding speed and provided a nice boost in sonic transparency.

You can easily monitor your success of this trick by measuring the inverting input at 0dB fs D/A-converted 20kHz sine wave with a scope. If the voltage spikes remain below lets say 20mV, you are likely safe.

NOS-Multibit and true one bit DACs are more demanding here.

[danville]

You have the inverting and non inverting inputs reversed.

I like your idea though. A little passive low passing should definitely help keep any amplifier in a "friendlier" operating situation.

[Markw4]

The trick may work with Hi-Z output dacs, but not with ESS Sabre or AK4499 where dac output impedance is very low. In the latter case the dac outputs need to be held at a precisely fixed output voltage by the I/V opamp or distortion rises unacceptably. Passive filtering before the I/V opamp allows the dac output voltage to be less tightly regulated and or causes frequency peaking issues in the I/V opamp. The differential summing opamp is a different matter of course. Passive filtering before it can be useful.

[diyiggy]

Thanks Jon Lord,


A noob question if I may, please.


Using multibits old dac chips, lowering the impedance output is very important.


Is this shematic is what you suggest after a oap I/V stage and also act as a buffer ?


For instance : a 1612 as I/V conversion than your shematic and then a buffer oap, or a low pass to attack for illustration a 25k input pre amp ??
As I have trimming so no serie caps nore traffo output in my dac (I/V opa861 made), I drive the pre with a serie 150 ohms after the buffer (reversed opa861)... is you shematic could be an improvment after the I/V stage please ?

[danville]

Originally Posted by Markw4

The trick may work with Hi-Z output dacs, but not with ESS Sabre or AK4499 where dac output impedance is very low. In the latter case the dac outputs need to be held at a precisely fixed output voltage by the I/V opamp or distortion rises unacceptably. Passive filtering before the I/V opamp allows the dac output voltage to be less tightly regulated and or causes frequency peaking issues in the I/V opamp. The differential summing opamp is a different matter of course. Passive filtering before it can be useful.

I see your point with the AK4499 etc. It seems to me the passive filter idea might be useful with a PCM1792A where the output Z is much higher. This may be a pretty good tradeoff given the fast edges.