What's the component-selection and tweaking process for using the ubiquitious feedback cap + R seen in opamp-based I/Vs (see schematic below -- taken from PCM1794 datasheet). So, for given schematic topology, how did TI/BB come up with the values for C1/C2 and R1/R2?
Attachments
The R's in the feedback path converts the current (see datasheet for value, approx 4 to 10mA) to voltage
-> U=IxR.
Keep in mind, that you have a swing around a single value,
e.g. voltage=0 -> +4mA,
voltage=+max -> +8mA,
voltage=-max -> 0mA.
The resulting offset is removed by the difference amplifier (3rd OPAMP).
The C can be calculated by fg=1/(2xPIxRxC) -> C=1/(2xPIxRxfg)
-> U=IxR.
Keep in mind, that you have a swing around a single value,
e.g. voltage=0 -> +4mA,
voltage=+max -> +8mA,
voltage=-max -> 0mA.
The resulting offset is removed by the difference amplifier (3rd OPAMP).
The C can be calculated by fg=1/(2xPIxRxC) -> C=1/(2xPIxRxfg)
Voegelchen: Thx for the tips + formula. As far as the NE5534...In myriad CDPs and/or outboard D/A processors that use this opamp, I can pretty much roll-in opamps** (w/o any other component change, e.g., C1 or R1) with results that other opamp rollers note for specific opamp brands/models. A further question is just how sensitive opamp performance i to the values for C and R are in the feedback locations. E.g., for R, I've seen this value vary from 500 - 1200 ohms.
**Note: I realize a lot of so-called "experts" DIYers shun opamp rolling. But if you are looking for opamp suggestions, and don't have a lot of time to 'scope each and every opamp, simply look at the sheer number of opamps rolled (e.g., at Tangentsoft.net and head-fi.org -- look for topical threads/articles there). The statistical significance of all these rollers is not insignificant.
**Note: I realize a lot of so-called "experts" DIYers shun opamp rolling. But if you are looking for opamp suggestions, and don't have a lot of time to 'scope each and every opamp, simply look at the sheer number of opamps rolled (e.g., at Tangentsoft.net and head-fi.org -- look for topical threads/articles there). The statistical significance of all these rollers is not insignificant.
Last edited:
Have you looked at the output waveforms of DACs like these? I have, although with ADI's equivalent part the AD1955, not the TI/BB one. The edges are very fast - a few nS - so fast that the output of an NE5534 (which looks inductive due to the falling OL gain) cannot really control them. So I think the transients get more 'smeared out' than controlled. I conclude that TI chose this large capacitor because it gives a low enough measured noise for the datasheet.
<edit> notice also that the textbook decouplers are shown in this application, about the worst possible thing for the sound. That's because the fast edges from the DAC go through C1, via the output stage of the 5534, straight through C11/12 and thereby into pin3. duh!
Last edited:
My Asus sound card uses this DAC (or one of the same family) and I found a marked improvement in the sound by decreasing the C1/C2 values. But the NE5534 whilst being an excellent audio opamp, is way out of its league in this particular application where the bandwidth of the incoming signal is huge. This particular application does demand a much wider bandwidth opamp - I'm at present using LM6172 (dual) with my AD1955 and am happy as a pig in clover with the sound when compared to a 5532.
In substituting a wider bandwidth opamp, the C1 and C2 values do need to be reduced considerably - I found that a wider bandwidth opamp when used with the stock values sounds worse than a 5532. Only with much lower valued caps do the video opamps come into their own.
I think cascading single poles is poor practice - I'd try for Butterworth maximally flat - even the phase/group delay isn't too bad at 1/2 fc - which would be 20KHz if you set the filter corner at 40KHz - and as long as phase is reasonably smooth, monotonic it really does't mater much as long as it matches between channels
few commercial recordings will have been done with higher fc mics
some other DAC I/V app circuits use 2nd order multiple feedback differential-to-single ended summer/filter which allows for designing a 3rd order response with the I/V pole
NP0/C0G ceramic smt caps will have good characteristics including very low esl to pass the fast edges - I'd still prefer polystyrene for audio frequency - it is possible to use both where they excel but it requires extra parts to "crossover" from the poly to the NP0 as frequency rises
several of this decade's better op amps likely could improve on the 5534 - I like the specs of the ADA4898 for DAC I/V
few commercial recordings will have been done with higher fc mics
some other DAC I/V app circuits use 2nd order multiple feedback differential-to-single ended summer/filter which allows for designing a 3rd order response with the I/V pole
NP0/C0G ceramic smt caps will have good characteristics including very low esl to pass the fast edges - I'd still prefer polystyrene for audio frequency - it is possible to use both where they excel but it requires extra parts to "crossover" from the poly to the NP0 as frequency rises
several of this decade's better op amps likely could improve on the 5534 - I like the specs of the ADA4898 for DAC I/V
Okay, cool. What values of C (and R for that matter ) are you using with your AD1955/LM6172 combo?
At the moment, 1k and 5pF (two 10pF in series). I also have a series resistor for the 5pF, 47R as damping. It works with no feedback cap at all but the scope shows a little overshoot - the sound then is quite interesting to listen to but I'm not sure how realistic. The 5pF was chosen to give the fastest settling time with no overshoot on the scope. I'm reconsidering the 1k to give it a bit more gain, but the jury's still out...
Have a lok at page two here,
http://www.resolutionmag.com/pdfs/SWEETS~1/DIGITA~1.PDF
on HF noise being a possible cause of nonlinearity in the 5532/4.
I know from my experience that I "rubbished" the 5534 because I heard the sonic improvements obtained from swapping it for other (sometimes much lesser devices), but this was when they were used as I/V convertors. Used as pure "gain blocks" I am much less sure.
It was fashionable to criticise the 5532, something I was guilty of too, and yes I swapped it and got better results... so it had to be true. Perhaps it was also true they were not used correctly.
The above link could partly explain that.
I/V opamp feedback cap + R, tweaking AD825
Any suggestions for a PCM1738/NE5532 combo (stock in Musical Fidelity A324 D/A processor) -- see schematic below. I simply rolled in AD825 per GeorgeHiFi's mod, but given the feedback from folks in this thread, I'm wondering whether the MFA324 ckt cannot be further tweaked for specific opamps. The NE5532 is 12V/uSec whilst the AD825 is 125V/usec. So perhaps lowering the value of the stock 3300pF feedback cap would be a start (any suggestions?). How about that 620-ohm R?
Ok thx.
Any suggestions for a PCM1738/NE5532 combo (stock in Musical Fidelity A324 D/A processor) -- see schematic below. I simply rolled in AD825 per GeorgeHiFi's mod, but given the feedback from folks in this thread, I'm wondering whether the MFA324 ckt cannot be further tweaked for specific opamps. The NE5532 is 12V/uSec whilst the AD825 is 125V/usec. So perhaps lowering the value of the stock 3300pF feedback cap would be a start (any suggestions?). How about that 620-ohm R?
I chose the feedback cap with my NE5532 to try to ensure that it did not enter slew rate limiting - with the AD1955 this was somewhere around 100-200pF. The PCM1738 may be a little different so try starting off with 470pF and then have a listen. If you're convinced it sounds better than the 3300pF then experiment downwards from there. I found when the feedback cap is too low, the NE5532 produces birdies with no signal present. For the time being I'd be inclined to keep the 620R as is, but definitely I'd not want to go lower in value with this opamp.
For the AD825 I'd start with 47pF and go lower based on listening.
The first stage differential filtering (2k2/1nF) looks too intrusive btw - I'd reduce the 1000pF to 220pF.
Very interesting, yes. Thanks for the heads up. I've been learning up about the effects of RF on opamps quite recently - partly because National have now in their catalog some 'EMI hardened' CMOS opamps I'd like to have a play with. They don't show much if any offset change when fed RF - I can't quite see yet how a change in the DC offset results in a change in the sound. I'm more inclined to believe that RF results in intermodulation which is what we hear, but its early days yet and more evaluation with CMOS opamps is needed. I did try an LMC662 as I/V but it sounded poor - perhaps taking into account its extremely low bandwidth though it wasn't that bad. I don't have any other CMOS opamps to hand to try but I'm keen to get some. MAX4489 is one that looks particularly interesting...
Didn't know about the potential compromise in diff stage filtering! The full ckt schematic of the MF A324's DAC and output stage is here, so if you see other possibilities for improvement, please let us know! Overall great tips -- thx.
FYI, I'm not going back to 5532/4's (well, not unless enough experienced DIYers convince me that there's untapped potential in them). My previous queries were referring to "better" op-amps like the AD825. The MF A324 schematic I posted is the stock ckt. Nevertheless, it's informative to have your input on the 5532/4’s for the sake of comparisons and reference.
BTW, have you used the AD825? Any other noteworthy opamps worth trying in the I/V? Also, which opamps are best for I/V vs. output (e.g., in the earlier posted PCM1794 datasheet ckt, TI/BB suggested LT1028 for output, so is this an “ideal” output opamp?).
Last edited:
I'd love to but can see no schematic at that link. Perhaps because I'm not considered bona fide enough to be a member at diyhifi - so images are blocked for me.
No, never ever played with an AD825. It looks a little underpowered in the bandwidth department - when I was doing my search for the best I/V amps, I think I ruled out anything under 50MHz as a matter of course, irrespective of the slew rate. LT1028 is a fine opamp but its a bit expensive for just doing the differencing function where its phenomenal low noise isn't needed. I'd also avoid it here because its optimised for low noise - higher bias currents and more likelihood of being affected by RF due to no degen on the LTP input. Currently I'm using NE5532 for this function, sounds fine and its cheap. Did try an old HA5222 in this position, which was my I/V of choice before the LM's fell into my hands. It sounds no better than the 5532.
The other opamp I'm dying to try in I/V is the (not very well known) OPA890. Its very similar to the LM6171 but a lower voltage process so a bit more bandwidth. And still very cheap and low power to boot. But no DIL so more difficult to experiment with.
Oh, about the LT1028 - some time ago I think TI had a 2nd sourcing agreement with Linear on some of their opamps, perhaps they like LT1028 because of that history. Noticed another funny thing - in the datasheet for the PCM1738 there's an 'OPA5534' - how weird is that?
Thx for your reply (i'll respond to the rest of it soon). Here's the MFA324 output schematic...
Attachments
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.