This big cap's value suggestions go from 2nF (Scott) to 160nF (Wayne).
Any advice about the optimum sonic range of values?
Given my 3 DAC in parallel I suspect that I should at least triplicate the starting value.
Simulation doen't help in the choice...but shows that i would need a 5 ohm series resistance after the cap to damp a RF resonance if I used dozens of nanofarads..
If not I will go for listenig stepping tempts (for big part of this weekend...
)Hi Scott, why you ended at 2n and not further higher?
Stupid me, I could not get the AD797 Spice models to work for me.
So I copied Scott and used LT1364 instead.
Scott,
I want to build this as a reference IV to measure distortions of DACs.
No need to sound good, only for measurement purposes,
What else should I watch out for other than following the AD797 DS blindly ?
Should I not be worried about input bias current and input impedance of the AD797 ?
Or better to use a JFET input opamp (ADA4627) ?
Thx,
Patrick
.
So I copied Scott and used LT1364 instead.
Scott,
I want to build this as a reference IV to measure distortions of DACs.
No need to sound good, only for measurement purposes,
What else should I watch out for other than following the AD797 DS blindly ?
Should I not be worried about input bias current and input impedance of the AD797 ?
Or better to use a JFET input opamp (ADA4627) ?
Thx,
Patrick
.
Bigger input filter cap without a resistor between the cap and the I/V opamp will cause more opamp gain peaking at HF, IIRC. A resistor there helps a lot to isolate the cap from affecting the opamp gain so much. If no resistor, less trouble if pre-I/V cap is no bigger than I/V feedback cap (that is in parallel with feedback R). However, that may not work out as a practical matter due to other considerations. (All this talk reminds me I can across this stuff long ago for use with a multi-wire ion chamber used for neutron beam steering. Worked out the math for it at the time, but long ago.)
Don't know why not look into what abraxalito has been doing with multi-pole filters between dac and opamp. More attenuation of clock noise (or whatever other unwanted HF) without going to a huge cap. abraxalito?
Don't know why not look into what abraxalito has been doing with multi-pole filters between dac and opamp. More attenuation of clock noise (or whatever other unwanted HF) without going to a huge cap. abraxalito?
Ooohps, gotcha! Thanks for pointing to this issue.
So we might end up with a "best" topology with Abraxalito's damped CLC, with the damper split in complex conjugates, capacitive leg going to ground and inductive leg going to OpAmp -IN in an effort to isolate the second C from the noise gain equation...
Good to know that you will ALWAYS be around to help! Not sure I am able to make a similar promise
Would there be any advantages in using a degenerated op-amp (like the LME49860 mentioned by Markw4) instead of e.g. the AD797 ?
Or would a newer op-amp like ADA4898, which is claimed to have a "linear, low noise input stage" be useful?
Or doesn't it really matter as long as there is a capacitor across the DAC output?
The AD797 and ADA4898 have lower noise than the LME49860, but would the degenerated LME49860 (or similar) be able to achieve a lower distortion?
I can use LT now.
There are a lot of options today and I would try some of the newer low noise FET amps, the source impedance at low frequencies is the feedback resistor so ib and 4KTR noise get you on the 1nV bipolar parts.The component choices were a little arbitrary but the process follows from the photo-diode I to V problem. A PD is a current out device that can have a large source capacitance. There are some simple rules of thumb the max frequency response is roughly the geometric mean of the major RC (transresistance and diode capacitance) and the unity gain frequency of the op-amp. The small feedback cap tunes the response to a two pole damped one.
That is the signal response is flat but the noise gain still peaks. Here I simply arbitrarily picked values to push the peaking and phase issues a decade or more beyond 100kHz (IIRC the demo card ran at 88.2K sampling frequency) so the audio out was amplitude and phase perfect and no noise was added.
I have not explored the small input R that shows up in some of the examples and the 100 Ohm one I used is a quirk of the AD797. The input devices have an unusual layout in order to get low rbb in the process at the time and they had an unfortunate tendency to form a Colpitts oscillator (~60MHz) at low closed-loop gains.
So the 3K was what the AD1862 had on chip and I simply picked values to get ~1MHz as where all the action happens. If you model the op-amp as an ideal integrator with a given unity gain BW you can write the second order equation by inspection (IIRC it's on the AD711 DS). Yes it's still there.
Last edited:
I think Wayne do talk about gain peaking. He does not use TIA terminology, but does look at it like a resonance between the input cap and the created virtual inductance at the -IN node.
He talks clearly about the resonance peaking.
I think these are two equally valid approaches. And he does arrive to the same conclusion that a small series R after the cap damps the resonance.
Mark, i wanted to say that opamps do fail in that position, if there are no countermeasures applied.
Your example, the LME49860 is a perfect example for a completely inadequate part for high speed applications like this, with it's 20V/usec slew rate.
So it absolutely needs additional massaging like is discussed here.
Just for fun, the TIA that I use for reading out a high capacity detector is: < .6nV en; 14GHz GBW, 1600V/usec slew rate. So I know very well that there exist opamps quite ok...
Maybe so for 16-bit dacs with high levels of clock noise. For 24-bit ESS dacs the situation might be rather different with far less clock noise and a need for very low distortion out as far as 21-bits. A single cap should be possible, but extremely good sound quality has been shown without one. Could be adding a single cap of modest value could help to make a slight bit more improvement though.
I think I'll try it with LT1792 first.
Not my first choice, that's the one where they even copied the apps off of my datasheet including resistors that did nothing. Jim Williams called me when he saw it and asked if I did that on purpose.
Mark is right things have come a long way, this was almost from day one, multi-bit DAC's, first class trips to Japan, Kobe (real) beef every night.
I wanted us to do a multi-chip SD converter with super discrete bipolar op-amps up front and full 20V p-p output.
Last edited:
I would try the ADA4625-1 or both ADI/LT and TI/BB have an attempt at doing the classic OP627/37 (I forgot the numbers). Don't forget that due to the high noise gain you can get away with a decomped amp, this should help a lot.
Another thing not to forget at extreme low distortion levels it can matter if an amplifier is on an advanced complementary process. Clever designers squeezed a lot out of early processes where the pnp's were fairly marginal, but at the extremes you can only deal with so much imbalance.
Another thing not to forget at extreme low distortion levels it can matter if an amplifier is on an advanced complementary process. Clever designers squeezed a lot out of early processes where the pnp's were fairly marginal, but at the extremes you can only deal with so much imbalance.
> it can matter if an amplifier is on an advanced complementary process.
How would I know which one if not working for ADI / TI ??
What else, ask the designer.
There was a behind the scenes competition in the old days, I know the designer of that LT part (Bill G. now long retired) they put a stake in the ground with the LT1028 and we answered with the AD797 so I did the AD743/45 and they answered. I think it was not always the most financially smart use of all the talent, but it was the wild west more than folks know.Strange.. while I liked /like ADA4627(b) very much - in the applicaton where I tried (same as for 4627) the ADA4625, sonically, was a failure. And I expected great things.
The same place I like much the OPA828.
(As Scott mentioned, both ADA4627/37 and OPA827 /OPA828 are the evolution of the classic OPA627 )
The same place I like much the OPA828.
(As Scott mentioned, both ADA4627/37 and OPA827 /OPA828 are the evolution of the classic OPA627 )
Last edited:
Except the audio measurements do show that it is completely adequate for this application. It does not need any additional massaging, if you use it in the DAC datasheet circuits you will get excellent numbers. I am not sure you can achieve much of an improvement, honestly.
The LME49860 has good output drive, distortion, and mix of current/voltage noise. The ESS DACs high output current require a low feedback resistance.
Last edited:
I think I'll try it with LT1792 first.
Patrick
If you're only after distortion measurements, I think you'll find it hard to beat OPA1611 in this position.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.