Component choice for OPA Alice circuit

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

Thimras

New member
Joined
Mar 17, 2024
Messages
4
Location
Melbourne, Australia
Hey everyone,

I'm sorry if this is the wrong forum for my question.
I thought I'd put together my own Alice circuit to learn how it works, and using the BOM supplied in the instructables, I tried to order by uploading the list directly.
A few items were on backorder, so I checked if there was an alternative and lo and behold, of course, there are many. I can see why certain features, like "military" etc can make parts more expensive, but the parts listed by Jules were middle-of-the-road. There were some cheaper and some more expensive.

The only difference between the parts I could identify in the search results was the "series", which has cryptic initals like MF or MFS which I thought meant "metal film", but maybe means something in more detail? What would be the difference in the parts, and thus, what would be the difference for the resulting circuit?

Thank you
Eike
 
Ahh, I checked in the data sheets, and the series thing is referred there, it's got to do with how the components react to temperature per their rating and stuff like that that isn't in the listing. Sheesh, so many details :)
 
I think (for resistors) "S" = smaller. So the "MF" is the approx. 6mm long ver. and "MFS" is around 3mm length (or bit over). Smaller resistors (and parts) could be useful to pack more circuitry in a given space and to do things like making various loop areas ("electronic circuits" implies AC, DC (and both AC and DC) loops) smaller so they are less "influenced" by other loops, esp. if there is noise or say big signal or whatever that might produce a deleterious interaction (noise, instability). I suppose they could also help to shorten things like ground, power, traces (less voltage drop) (but not if you just "drop them in" to an existing layout. (Maybe not in everything, but) physical implementation (layout, grounding, thermal design, etc.) can make a difference in the outcome. (Obvious manisfestation = surface mount technology (despite what some people seem to think I don't believe SMT "sounds bad" compared to thru-hole--not in reference to anybody here BTW--just some sites offering things like replacement boards claiming such).
 
I think (for resistors) "S" = smaller. So the "MF" is the approx. 6mm long ver. and "MFS" is around 3mm length (or bit over).

That's a big "maybe" - depends entirely on how a particular manufacturer decided to "encode" their product ranges. Not all manufacturers will have a series called "MF" or "MFS" or whatever.

the series thing is referred there, it's got to do with how the components react to temperature per their rating and stuff like that

... Which will vary from manufacturer to manufacturer.

Much of "run of the mill"-electronics (which audio is, for the most part), is getting to know which specs are important and which aren't, at least for the application at hand.
 
I wouldn't worry too much about resistor types in this application.

Things that don't matter include:
  • Power rating. None of the resistors will get past a handful of milliwatts. 0.5W or higher ratings might be too physically big, so 0.25W is about optimal.
  • Tolerance: nothing in the OPA circuit depends on absolute accuracy or matching between resistors, particularly.
  • Temperature coefficient of resistance - as above.
  • 'Safety' or 'high voltage' features, relevant if they are operating at mains voltage/current but not an issue here.
You'll want 'low noise' but basically that means using a metal film type and avoiding carbon ones.

Hope that helps.
 
I wouldn't say that for the output stuff, i'd still go for matched 47ohm and 2k2 resistors.
I'm not sure even that is strictly necessary? .... 1% metal film for the 2k2 resistors would give less than 0.5dB gain variation between using resistors at each extreme values of the 1% tolerance for those values.
And the tolerances of the output electrolytics will be far greater than the tolerance of 1% 47R resistors......

Now whether it's worth matching output electrolytics or not?..... probably not worth it, in reality..
 
1% metal film for the 2k2 resistors would give less than 0.5dB gain variation between using resistors at each extreme values of the 1% tolerance for those values.

That's likely rather for the power tap-off from the XLR2/3, in order to achieve as high a CMRR as feasible (not the inverting opamp gain setting resistors).
 
That's likely rather for the power tap-off from the XLR2/3, in order to achieve as high a CMRR as feasible (not the inverting opamp gain setting resistors).
I would have thought the much greater tolerance of the output electrolytics - especially over time - would have a more significant influence on CMRR than the difference in the value of 1% resistors ?
 
That would likely affect CMRR only at the very low end of the frequency range. Mismatched resistors will affect CMRR at *ALL* frequencies. Would they not?
 
That's true of course -- although hum can often be a major cause of problems -- at the lower end of the frequency spectrum.
In reality, I don't think the kind of tolerances we are talking about will make a lot of difference?
If you need really serious CMRR then you probably need to think about galvanic isolation and transformers !
 
I'm not sure even that is strictly necessary? .... 1% metal film for the 2k2 resistors would give less than 0.5dB gain variation between using resistors at each extreme values of the 1% tolerance for those values.
And the tolerances of the output electrolytics will be far greater than the tolerance of 1% 47R resistors......

Now whether it's worth matching output electrolytics or not?..... probably not worth it, in reality..
That is interesting to think about - how do you arrive at those numbers? As in, how could I follow/predict what kind of tolerances where would/should result in what possible variance? Which values would affect linearity of the circuit vs noise from the lowered effectiveness of the common mode rejection?
 
You're right that the 47 ohm resistors should be matched, but I'd want to try to quantify this a bit...

Suppose our noise source has an impedance of 10K to each of pins 2 and 3 (e.g 1nF at 16KHz). If our effective output impedances are 46.5 and 47.5 ohms, we have attenuation of -46.7dB and -46.5dB respectively, but the difference between pins 2 and 3 (i.e. actual noise level imposed on the signal) is -80dB down.

(What's a real-world spread of resistor values, without taking special precautions? I measured some 47 1% ohm resistors and the first four of them next to each other on the bandolier all came out at 46.7 ohms, on my meter. I'd imagine you'd be seriously unlucky to get two outliers).
 
You're right that the 47 ohm resistors should be matched, but I'd want to try to quantify this a bit...

Suppose our noise source has an impedance of 10K to each of pins 2 and 3 (e.g 1nF at 16KHz). If our effective output impedances are 46.5 and 47.5 ohms, we have attenuation of -46.7dB and -46.5dB respectively, but the difference between pins 2 and 3 (i.e. actual noise level imposed on the signal) is -80dB down.

(What's a real-world spread of resistor values, without taking special precautions? I measured some 47 1% ohm resistors and the first four of them next to each other on the bandolier all came out at 46.7 ohms, on my meter. I'd imagine you'd be seriously unlucky to get two outliers).
I think i expressed my opinion a bit harshly, not that it would be a dealbreaker but i'd still go for matched ones, like why not? Doesn't cost a fortune, one can match them on their own, goes also for the output electrolytics if the OP has a meter, and some to choose from.
 
That is interesting to think about - how do you arrive at those numbers? As in, how could I follow/predict what kind of tolerances where would/should result in what possible variance? Which values would affect linearity of the circuit vs noise from the lowered effectiveness of the common mode rejection?
If you were to have a 1% 2k2 resistor which actually measured (2200 - 1%) it would measure 2178Ω
A second 1% resistor at the other end of the tolerance range (2200 +1%) would measure 2222Ω.
In an inverting op-amp configuration, imagine a resistor of the lower value as the input and the higher value as the feedback resistor, then the voltage gain of the circuit would be 1.0202. (2222÷2178) .... That would have a gain of 0.174dB
(Remember that this ciruit configuration also reverses the polarity of the signal - not an important parameter for this calculation)

(Use an online calculator like this: https://sengpielaudio.com/calculator-gainloss.htm for the conversion)

Similarly, with extreme 1% tolerance value connected the other way round the gain loss would be (2178÷2222) which is a loss of - 0.175dB

Those value differnces would mean a maximum gain difference of around 0.35dB using 1% resistors.
Which is of no significant consequence at all in this situation.

Remember that it is the impedance of both sides of the line being similar that affects the CMRR, not any differnces in the signal levels applied to each leg.

Which brings me onto the wisdom of using a differential audio path at all?..... Introducing a second inverting op-amp with it's associated resistors into the signal path to create a differential audio output will increase the system noise. The op-amp itself has a self noise figure, and each resistor in the signal path adds more (the higher the resistor value, the more the extra noise).
I prefer to lower noise single sided audio approach personally, but not everyone agrees.

What can affect the CMRR - especially as lower frequencies( as Khron has already mentioned) is the tolerance of the output electrolytics capacitors.
Not only will they have tolerance of 20% or so, but that can vary over time.
In reality, none of these things will make very much difference in most cases.

If you do have a situation that requires very long cable runs, then the much higher CMRR values obtainable from transformer coupling might be considered.
But that'a different conversation altogether!
 
That's true of course -- although hum can often be a major cause of problems -- at the lower end of the frequency spectrum.
In reality, I don't think the kind of tolerances we are talking about will make a lot of difference?
If you need really serious CMRR then you probably need to think about galvanic isolation and transformers !
...or think about using designs with well matched, symmetrical BJT outputs. My personal favourite: CFP Schoeps style outputs. No worries about degrading elcaps or selecting parts. Matched BJTs, if needed at all in a CFP output, are readily available at low cost from several manufacturers. Admittedly, they are all SMT and making a design more complex. As you strive for simplicity and want to use through-hole technology, this may not be something you will immediately embrace. I wouldn't mind a few extra (SMT) parts in my huilds if that would improve the design, but we all have our own preferences.

As you never know how much CMRR is actually required to prevent hum issues, I would personally select the caps for capacity and ESR. A 47uF 20% elcap could deteriorate CMRR to worse than 45 dB. Combined with other mismatches in the signal path, it could become even worse than that. The costs are neglible compared to the rest of the mic build.

I've been at several concerts where noise, hum, crackles, RF issues, drop-outs and you name it occurred, spoiling the performance and ruining the mood of the artist, so I would always try to optimize the design for best CMRR, RF immunity, moisture insensitivity and, to a certain extent, noise (hissss). When choosing elcaps, I would propose the Japanese brands, e.g. Nichicon or Rubycon, for extremely consistent quality and very low actual tolerance. I've measured Nichicon elcaps and compared them with same type Lelon. Of the new Nichicons, stored for several years or used ones, the ESRs measured within a few tens of Ohms. Lelons had higher ESR, values were all over the place, yet still within datasheet spec limits.

YMMV, of course...

I have Jules OPA PCBs lying around for a year or so, but never got around to actually assemble and test them. But I sure want to that and compare with other designs I'm working on. Curious to learn the differences, both audible and measured.

Jan
 
I would have thought the much greater tolerance of the output electrolytics - especially over time - would have a more significant influence on CMRR than the difference in the value of 1% resistors ?
Just to do a 'real world' check I grabbed a handful of 47uF50v Panasonic M series electrolytics (10 to be precise) just to check what the typical tolerance range was. They all measured within the range 45.6uF to 46.3uF. ...... So probably a tolerance of more like 2% rather than 20%.

Another point that kingkorg made earlier was matching the 2k2 DC feed resistors. As he said, it's not very time consuming to select resistors - although with most phantom power feed resistors only specificed as 6.81k 1% it's probably not strictly necessary?

I think I can probably expect acceptable CMRR results with 1% resistors and 2% electrolytics?....
 
AFAIK, the P48 phantom power specification prescribes 0.1% resistors for best CMRR. See sound-au.com/project96.htm for your reference. If your mixer brand chose to save $0.01 by using 1% resistors, then they are not following the standard. If your mixer does have 0.1% resistors, then to me it would make sense to build your gear up to the same level of quality, wouldn't it?

Your capacitance measurements confirm my statement that you can expect Japanese caps to havemuch tighter tolerances than specified. Just measured 10 pcs Lelon EAR series 22uF 25V caps: deviation ranged between +4.4% up to 11.5%. Still within the +/-20% spec limits, but not on par with the excellent actual tolerances you generally find with the Japanese elcaps.

Anyway, what CMRR is acceptable, is hard to say if you do not know the CMRR specs of your other equipment and the level of interference generated in your recording environment. And in the end, CMRR is a system property, so you cannot just judge the CMRR of the mic in isolation. But I would always try to get the best possible results, within reasonable costs and effort of course. Design for worst-case situations and your mic is not likely to become the weakest part of the chain.

Jan
 
Back
Top