Op-amp requirements for a I/V converter?

[Audio Maniac]

Imput noise density, slew rate, gain bandwith, THD + N, settling time or open-loop gain, which of this parameters really matters when choosing op-amps for an I/V converter of a DAC?

 

[Osvaldo de Banfield]

All of them, and depend on how much you want to spend, but all things together is expensive.

 

[jcx]

if you put a C acrosss the feedback R giving a single pole low pass then there is little need for excessive slew rate - can be part of the anti-image/reconstruction filter

ideally higher frequency error V at the op amp input would then be reduced by ~ fc/GBW

with the feedback C there is a "feedthrough zero" that gives a small input Vdiff glitch at current steps due to the op amp output Z - decades old I/V app notes described the advantages of a low impedance output buffer inside the loop for op amp I/V

the op amp slew rate spec can still be a proxy for input diff pair linear operating range with degeneration, fet inputs improving both

but the low Vnoise parts can't have much degen

a compleltely new "highly linear" diff pair is used in AD8099/ADA4898 family parts - I think the ADA4898 is the best combination of specs for audio DAC I/V available today

you would need to look with a very good 'scope to see if ccs bias of the output to either rail helps - the ADA4898 has a 50 mA "linear current range" spec so you could bias at several times the fs current of most audio DAC chips

 

[Audio Maniac]

Let me place the question in a different way. Among the listed parameters, which are the most relevant?

I need to choose a pair of dual op-amps for I/V convertion but they must be electrically compatible with NE5532. I searched about this subject but all recommendations where based mostly on subjective tastes. I would like to get a technical explanation about the matter.

 

[wahab]

http://www.ti.com/lit/ds/symlink/pcm1794.pdf

 

[Audio Maniac]

The PCM1794's datasheet informs gain bandwith, settling time and slew rate affects the audio dynamic performance of I/V section.

So imput noise is less important for this application?

 

[jcx]

why trade off when you can get "very good" or "best" in all categories at once

you might want to read the Hawksford paper I refence in the 1st post: http://www.diyaudio.com/forums/digi...sford-iv-nested-loop-op-amps.html#post2218591

I show a sim of glitch at the input of the I/V op amp - Hawksford shows that you should want to minimize a function like the V^2 time integrated error V

GBW is the biggest help, low output Z as I mentioned makes the feedback C work better

the function to be integrated also depends on the linearity of the input stage - the ADA4898 uses a new linearized BJT input stage that is unmatched by previous monolithic op amp tech
if I've done the calc form the patent correctly the input stage gm is linear to -120 dB with |4 mV| input differential - the same number for a low noise bjt op amp is ~100 uV

so the ADA4898 has high GBW >50 MHz, high "linear output current" spec f = 100 kHz, SFDR = −70 dBc, RL = 150 Ω 40 mA, >50 V/uS slew rate, low 1 nV/rtHz noise - all together - and avaiable in duals


I/V S/N is limited by DAC I_fs and the larger of op amp Vnoise/Rfdbk or op amp Inoise - so op amp Vnoise can be reduced in importance by increasing R, increasing output V swing

 

[SoNic_real_one]

I just got today some THS4032 (TI). Noise: 1.6nV/sqrt Hz; Slew Rate: 100V/us; Settling Time: 60ns, THD: -90dB at 1MHz; BW: 100MHz!
Another contender that I have at hand is ADA4897-2 (AD). Noise: 1.0nV/sqrt Hz; Slew Rate: 85V/us; Settling Time: 96ns, THD: -115dB at 100kHz, -93dB at 1MHz; BW: 230MHz!

I am planning to use them as I/V stage after a PCM61P - it has a settling time of 350ns. I am planning to get rid of the SR limiting capacitors in the I/V stage.
I just need to wait to get the adaptors from SMD to DIL.

 

[jcx]

the ada4897 is new to me - the Vnoise looks very good but the chip has a max |Vsupply| of only 10 V so +/-5 V supply would be its limit

pkg is msop - 20 mil pin spacing is challengening to hand solder - but the size is probably required to keep lead/layout inductance low - I am suprised they didn't use a better pinout

I really wouldn't try to use a 230 MHz op amp on an adapter

 

[Audio Maniac]

The ADA4897-2 is also new to me. I'm amazed with its great specs, but unfortunately its power supply operating range is 3 to 10V. I need an op-amp which supports 15V.

 

[SoNic_real_one]

In my application I have +/- 5V so I could use the ada4897 (it's a "low power" device), but I was waiting for THS4032.
THS4032 has 33V max voltage (measurements for +/-5V and +/-15V are given in datasheet) and is almost identical for our I/V scope... And distortion figures are around -100dB all the way to 400kHz.

 

[Audio Maniac]

Sonic_real_one, do you think a LME49722 or a pair of LME49990 are good choices for I/V scope?

 

[abraxalito]

I am planning to use them as I/V stage after a PCM61P - it has a settling time of 350ns. I am planning to get rid of the SR limiting capacitors in the I/V stage.

Why get rid of those caps? Even with slew rate of 100V/uS you'll still be going into slew limiting directly from the DAC's output. The settling time (to 0.006% or 14bits for an allegedly 18bit part ) is going to be considerably longer than the rise time.

Just as an aside, why use PCM61P anyway when TDA1541A beats its specs?

 

[SoNic_real_one]

I am using now LM4562 (that is the same as LME49720) in most of my DAC's. IMO, it is one of the best sounding OpAmps, beats anything that is usually installed "stock" in players.
LME49722 is marginally better so I can assume it will be similar. LME49990 takes another small step in the "good" direction.
Basically , they have vanishing noise and distortions, no question. 20-22 V/us is good for an I/V stage that has the SR reducing capacitor and up to 8x OS of 44.1kHz...
Higher than that, I might see the argument for something faster. That's why I am looking to experiment with something elese.

Why get rid of those caps? Even with slew rate of 100V/uS you'll still be going into slew limiting directly from the DAC's output. The settling time (to 0.006% or 14bits for an allegedly 18bit part ) is going to be considerably longer than the rise time.

Just as an aside, why use PCM61P anyway when TDA1541A beats its specs?

As far as i know, there is no audio DAC that can output with 100V/us, they are limited too. PCM61P has 350ns settling time, so the 80nS of the THS4032 should be more than enough. Maybe I will just reduce that capacitor.
As for comparing 1541A, why don't make it fair and go to PCM61P-K ? I did play with my TDA1541 but I wasn't that excited about it.
Now, I just got this old Denon DCM-360 "toy" with Denon Alpha filter (SM5848), PCM61P and external I/V OpAmp that adds also 2 extra bits (19, 20 are done with resistors). After replacing the absolute crappy opAmps with LM4562 I am hearing things that I don't hear on my PCM1791A. I am almost... in denial about this.
I ordered two -K DAC's already, but I might just buy two more and piggy back them... And I want to hear what a even faster OpAmp can do (or not).

 

[abraxalito]

20-22 V/us is good for an I/V stage that has the SR reducing capacitor and up to 8x OS of 44.1kHz...

Have you checked incidentally that the output stage can deal with the current slew rate that's required? jcx's first post above indicates that relying on the SR reducing cap to prevent input stage overload may just move the problem to the output stage instead.

 

[SoNic_real_one]

The output stage? That's the limiting point of the SR in an OpAmp, that's why is only 20V/us!
So it's not "moving" the problem, it is the same and only problem...

I might try faster OpAmps with reduced capacitors (or none) when those chinese guys will stop partying and send me those SMD adaptors

 

[abraxalito]

As far as i know, there is no audio DAC that can output with 100V/us, they are limited too.

Well we're talking I-out DACs here, they don't have slew rates for voltage, rather they have current slew rates.

PCM61P has 350ns settling time, so the 80nS of the THS4032 should be more than enough. Maybe I will just reduce that capacitor.

I suggest just reducing the cap but not totally eliminating it. On my AD1955 using NE5532 as I/V (just for fun) I found elminating that cap totally gave rise to some very weird audible effects. Rather like birdies on an FM tuner. Even when I was using LM6172 I didn't much like the sound with zero SR limiting cap and there was an overshoot on the scope. I found I liked the sound of 5pF.

As for comparing 1541A, why don't make it fair and go to PCM61P-K ?

Because if we're playing off selected parts then the natural contender would be TDA1541AN1/S2. You'd not want to go there would you?

Now, I just got this old Denon DCM-360 "toy" with PCM61P and external I/V OpAmp that adds also 2 extra LSB bits (done with resistors).

Resistors are the easy part - what do they use for switches?

After replacing the absolute crappy opAmps with LM4562 I am hearing things that I don't hear on my PCM1791A. I am almost... in denial about this.

I've found that the S-D parts add things that multibit does not. So perhaps there's hope for you yet

I ordered two -K DAC's already, but I might just buy two more and piggy back them... And I want to hear what a even faster OpAmp can do (or not).

Been there, now I'm curious about what a faster DAC can do.... The slippery slope

 

[SoNic_real_one]

It's all in the filters... D-S noise is not a problem, it is cut below any relevance. What I hear I think are better transients/harmonics on high frequency sounds (fast cymbals/high-hat). I am a rock music type mostly, so those are important.

BTW, the internal OpAmp in PCM61 is capable of only 1000ns settling time (compared with 350ns of the straight current), I am glad Denon choose not to use it in this player (they used thou the internal laser trimmed, feed-back resistor).

Ah OK, if we compare, there is a PCM61P-"J" version in between, and there are the pin compatible AD1860 (two flavors there too). My "-K" versions costed 7.50US (so 15USD for two channels), I got two AD1860-J for 12USD... And there are no fakes (that I know) of in this field. Compare that with a S1/S2 - show me a real one outside a player that is at least 1000USD.

 

[abraxalito]

It's all in the filters... D-S noise is not a problem, it is cut below any relevance.

Only with averaging. But D-S is known to suffer from noise modulation - short term shifts in the noise floor - like what you'd get from using the wrong sort of dither.

What I hear I think are better transients/harmonics on high frequency sounds (fast cymbals/high-hat).

You're hearing absence of noise modulation I reckon. Its perceived as more 'surprise factor' on transients. Try listening to massed brass - S-D totally destroys that compared to multibit.

Ah OK, if we compare, there is a PCM61P-"J" version in between, and there are the pin compatible AD1860 (two flavors there too). My "-K" versions costed 7.50US (so 15USD for two channels), I got two AD1860-J for 12USD... And there are no fakes (that I know) of in this field. Compare that with a S1/S2 - show me a real one outside a player that is at lest 1000USD.

Ah if we start comparing costs then I agree, TDA1541AN1/S2 is totally out of the game. Pure unobtainium. My TDA1387s though do quite well on cost-benefit - I paid 5mao each (that's $0.08) - secondhand though

 

[SoNic_real_one]

I just did a quick A-B comparison.
Source: The Police - Message In A Bottle. CD, SACD, SHM-CD.
The CD layer from SACD sounds sensible the same on this Alpha-driven PCM61 as the SACD layer on PCM1791.
The CD layer sounds worse than SACD on PCM1791 (non-Alpha driven, I am too lazy to get to living room to the DVD-2030).
The big surprise is... SHM-CD. I hear all the brass clearer. I don't know if the japanese remaster has accentuated some dB the top of the band or they just compressed the same top, but it sounds... better. I don't belive in their statement that their polycarbonat is "better".