Instrumentation op-amp for DC-coupling to audio input?
in [Design]
|
From: Lostgallifreyan on 13 Jul 2010 11:57 John Devereux <john(a)devereux.me.uk> wrote in news:87tyo3o03q.fsf(a)devereux.me.uk: > Lostgallifreyan <no-one(a)nowhere.net> writes: > >> I'm considering an op-amp for making a DC coupling adapter to a >> soundcard to convert it to signal logging purposes while retaining its >> audio performance. It uses a passive adder and a gain of 2 to add a >> bias voltage to the signal before an ADC input. >> >> The sound card is one with external analog circuitry in a rack unit, it >> has 20 bit signal conversion, so this op-amp will have to be good to >> maintain that and the other specs this unit has. > > Audio ADCs usually have bad DC specifications - why wouldn't they. You > may want to verify this before you try too hard to find the perfect > opamp. > > [...] > > Possibly so, but I have already done somethign similar with a DAC and found it be be excellent at outputting any abitrary DC or slowly changing voltage and holding it where it's meant to be, and the ADC's and DAC's in this equipment were chosen from the same maker. I know I'm making an assumption that it will work, but I do think it will. Besides, I have to make some decent effort to get this op-amp decision right, partly to help me learn, and not least, to make sure than when I do test that assumption, it isn't my choice of ap-amp or adapter design method that is screwing it up. :)
From: Marte Schwarz on 13 Jul 2010 15:21 Hi, >>> 20 bit signal conversion, so this op-amp will have to be good to >>> maintain >>> that and the other specs this unit has. >> So did you think about noise? > Yes, several times. This amp has THD+N of 0.00005% and is going into a > device > specificied as 0.002%. Good enough? I think so. Oh I didn't talk about distortion, I mean noise SNR or ENOB in total system. Marte
From: Lostgallifreyan on 13 Jul 2010 15:50 "Marte Schwarz" <marte.schwarz(a)gmx.de> wrote in news:8a3sm2Fn40U1(a)mid.uni-berlin.de: > Hi, > >>>> 20 bit signal conversion, so this op-amp will have to be good to >>>> maintain >>>> that and the other specs this unit has. >>> So did you think about noise? >> Yes, several times. This amp has THD+N of 0.00005% and is going into a >> device >> specificied as 0.002%. Good enough? I think so. > > Oh I didn't talk about distortion, I mean noise SNR or ENOB in total > system. > Not my job. I'm adding an op-amp and four resistors per channel. Not much I can do about the rest of the system, except to note that it claims 20 bits while having a 98dB dynamic range that barely scrapes past 16 bits worth of resolution (though this does apply to the DAC too, the whole chain..). All I'm concerned with is that I do not add to the existing reduction significantly. George Herold already helped me with that, and I'd already suspected I could reduce noise to acceptable levels for real 20 bit (120 dB) dynamic range by lowering the value of the resistors in the passive adder (which George Herold confirms, but with useful detail I didn't yet have). Apart from choosing the op-amp there isn't anything else I can do. I'm not trying to improve the system, just to make effective DC coupling for it without harming it by adding complex circuitry where simple stuff will do it better. If I didn't have to handle the bias voltage as I do, I wouldn't be looking for an op-amp at all. Some systems are nice, they have no DC bias on either side of the DC blocking cap, but when they do, I can't think of a way to satisfy all conditions without an op-amp to mix the bias voltage into its output without it appearing on the output of the previous stage, so that op-amp must be there if the capacitor isn't. If there is a simpler and quieter way to mix two signals without output inversion than a passive adder based on an op-amp (gain of 2) and 4 resistors I'd like to know what it is, because in many years I have never found one.
From: George Herold on 13 Jul 2010 16:08 On Jul 13, 11:26 am, Lostgallifreyan <no-...(a)nowhere.net> wrote: > George Herold <gher...(a)teachspin.com> wrote in news:a02bdb85-1047-4ba3-aaec- > b24eee25f...(a)s9g2000yqd.googlegroups.com: > > > Your band width is maybe > > 100kHz? (Do you have any filtering before the A-D) > > Sorry, I didn't remember to directly answer that... They just have a > differential input stage that adds the two differential signals (doubles) > then quarters the output in the gain of the second of two amps they do this > with, so halving the total signal. After that it goes through a digitally > controlled analog variable resistor (CS3310-KS) followed by a DC blocking > capacitor and then an op-amp stage to make a differential signal for the ADC > (CS5335) with a 2.2nF cap with two 150R resistors to remove anything above a > few hundred KHz. 300 ohms inline with the signal and 2.2nF to ground? Something like 240kHz if I punched the numbers correctly. You could scale the above noise numbers up a factor of sqrt (2.4). If you want to be real fussy there is something called the equivalent noise bandwidth. (ENBW) (Your low pass filter is not a brick wall and higher frequencies add more noise.) For a simple RC the ENBW is 3.1415/2 times bigger than the 3dB corner frequency. > > Apart from using that variable resistor IC they kept it as simple as they > could. Total dynamic range through the system is 98 dB, so only a tad more > than 16 bits worth, but aiming for better in my DC coupler board means I can > adapt other devices reliably. OK that's much less of a constraint then. Sounds like you've got more than 100uV of 'noise' in the system already. George H.
From: Lostgallifreyan on 13 Jul 2010 16:53
George Herold <gherold(a)teachspin.com> wrote in news:734e6be6-17bd-465c-9aaf- ec3d7423fc08(a)j4g2000yqh.googlegroups.com: >> (CS5335) with a 2.2nF cap with two 150R resistors to remove anything abov > e a >> few hundred KHz. > > 300 ohms inline with the signal and 2.2nF to ground? Something like > 240kHz if I punched the numbers correctly. You could scale the above > noise numbers up a factor of sqrt (2.4). If you want to be real fussy > there is something called the equivalent noise bandwidth. (ENBW) > (Your low pass filter is not a brick wall and higher frequencies add > more noise.) For a simple RC the ENBW is 3.1415/2 times bigger than > the 3dB corner frequency. > Hmm, I didn't describe that filter fully, it's got one 150R in series with each new signal line (one original plus one from an inverting buffer amp that copied it at unity gain), and each line then feeds one of two differential inputs on the ADC, and the 2.2nF cap goes directly across those. By 1/(2*PI*R*C) I assumed that either 240 KHz or 480 KHz would be the relevant frequency, and as both easily leave 100 KHz of bandwidth I didn't inquire further. :) I doubt that any input I'll put into this system will have much signal above audio band. I'll rework any other calculations I did for power consumed, resistance for voltage reference etc, based on passive adders with 47K resistors. That should allow the extra scaling needed, or close enough. I don't want to go lower than that. If the OPA2277 will work ok it will also be good for power consumption, as it's more efficient that the LF412 I originally planned before I learned how limited it would be. I won't go any deeper than this now, I'll try it to see how it can illustrate what I've seen here. If my original question-and-answer post about slew rates is ok I've probably got the op-amp I'll stay with. BTW, I saw mention of OPA134 (and OPA2134 I think, a dual one) but I haven't looked into that in much detail yet. Might not have to, I hope to use what I have first. I drew a blank on OPA164 though, that one apparently doesn't exist. :) Like LT, they have so many amps that it can be very hard to choose, so I'll look more for contexts described by people who are working with similar situations, and see what they're using (is how I bumped into the OPA134 last night..). |