This is the article:
https://www.edn.com/adjust-op-amp-gain-from-30-db-to-60-db-with-one-linear-pot/
Very nice, as usual, but something bothers me though: Stephen recommends the use of low excess noise resistors in the feedback divider, but I don't see why: these resistors do not see any static bias, except that caused by the input current and offset of the OP37, which are normally minuscule.
The signal itself, if large enough, could cause excess noise on high-amplitude excursions, but it should remain moderate and drowned in a much larger signal.
It would result in a noise component having its enveloppe modulated at twice the signal frequency, thus excess noise from a frequency-domain perspective since the modulation should change nothing there, and maybe a kind of "breathing effect" for large, VLF signals in the time-domain.
To summarize, poor resistors would not affect the noise floor of the system (it would be quiet), but it could degrade the THD+noise distortion figure, but I don't think it would be relevant because of the low amplitude, even for carbon-film resistors, and because the noise would be random, uncorrelated to the signals processed.
Did I miss something? Can a poor resistor have a hotter virtual temperature than its physical temperature?
https://www.edn.com/adjust-op-amp-gain-from-30-db-to-60-db-with-one-linear-pot/
Very nice, as usual, but something bothers me though: Stephen recommends the use of low excess noise resistors in the feedback divider, but I don't see why: these resistors do not see any static bias, except that caused by the input current and offset of the OP37, which are normally minuscule.
The signal itself, if large enough, could cause excess noise on high-amplitude excursions, but it should remain moderate and drowned in a much larger signal.
It would result in a noise component having its enveloppe modulated at twice the signal frequency, thus excess noise from a frequency-domain perspective since the modulation should change nothing there, and maybe a kind of "breathing effect" for large, VLF signals in the time-domain.
To summarize, poor resistors would not affect the noise floor of the system (it would be quiet), but it could degrade the THD+noise distortion figure, but I don't think it would be relevant because of the low amplitude, even for carbon-film resistors, and because the noise would be random, uncorrelated to the signals processed.
Did I miss something? Can a poor resistor have a hotter virtual temperature than its physical temperature?
Excess noise specifically isn't thermal by definition.
You don't need DC to see excess noise, just current.. Think of it as time fluctuations of the resistance. Pass any signal current through and those variations create voltage fluctuations. Yes the noise rises with signal amplitude when excess noise is present, which is why its a good idea to measure with a full-scale signal present, not just at quiescent. Excess noise often has a 1/f spectrum.
Note a simple model suggests the size of the signal doesn't affect the ratio between excess noise and signal, twice the signal makes twice the noise, and the noise spectrum is modulated around the signal frequencies (looking like phase noise?). So the shape of the noise spectrum follows the signal (the close in noise to a spectral peak may be hidden if using a Hann window - a good high dynamic-range window is required for FFT analysis for this sort of effect).
But yes, adding a DC bias is going significantly worsen the noise situation...
The imperfections that lead to excess noise often produce popcorn noise too (usually much more of a problem). And of course real component noise is more complex than any simple model of it... Sometimes there are gross noise effects due to poor contact with the end-caps, which can be thermally sensitive, current sensitive, or even microphonic - another reason to stick to metal film resistors if you can.
You don't need DC to see excess noise, just current.. Think of it as time fluctuations of the resistance. Pass any signal current through and those variations create voltage fluctuations. Yes the noise rises with signal amplitude when excess noise is present, which is why its a good idea to measure with a full-scale signal present, not just at quiescent. Excess noise often has a 1/f spectrum.
Note a simple model suggests the size of the signal doesn't affect the ratio between excess noise and signal, twice the signal makes twice the noise, and the noise spectrum is modulated around the signal frequencies (looking like phase noise?). So the shape of the noise spectrum follows the signal (the close in noise to a spectral peak may be hidden if using a Hann window - a good high dynamic-range window is required for FFT analysis for this sort of effect).
But yes, adding a DC bias is going significantly worsen the noise situation...
The imperfections that lead to excess noise often produce popcorn noise too (usually much more of a problem). And of course real component noise is more complex than any simple model of it... Sometimes there are gross noise effects due to poor contact with the end-caps, which can be thermally sensitive, current sensitive, or even microphonic - another reason to stick to metal film resistors if you can.
Ok, that's more ore less what I remarked, but it doesn't answer the fundamental question: are low excess noise resistors required in this circuit, and if it is the case, why?
When reading the article, I thought the author simply meant "These resistors are already uncomfortably large values, so they'll generate undesirable Johnson-Nyquist noise anyway, therefore don't make your life even worse by choosing resistor types with excess noise."
The values don't look particularly large (for me anyway), and the worst offender, although not explicitly visible is the wiper to track parasitic resistance. Each part of the track is used as a resistor, with a common access via the wiper.
The optimum use of a potentiometer is in a potentiometric circuit, with negligible loading on the wiper
The optimum use of a potentiometer is in a potentiometric circuit, with negligible loading on the wiper
I think there is a problem with this circuit unless you add a minimum resistance to the top of the pot so that the feedback cannot be reduced to nothing. A series cap would also improve DC offset. But, I don't care enough to peruse it father. We have seen a number of relay etc. volume controls in DIYA. Many years ago, I was enthralled with R/nR ladders for that purpose, but today it all seems futile vs a DSP ~preamp.
Is excess noise still a thing in resistors? I haven't looked in detail, but it was my impression that modern metal film resistors don't suffer from excess noise. The resistor values in question should be easy to find in metal film.
You do have signal current flowing through the resistors in the circuit. Keep that in mind when you choose the potentiometer for the circuit as the signal current flows through it.
I'd be more worried about the 1 pF feedback capacitor. That's small enough that the feedback capacitance will be affected by the neighbour's cat walking by outside.
Tom
You do have signal current flowing through the resistors in the circuit. Keep that in mind when you choose the potentiometer for the circuit as the signal current flows through it.
I'd be more worried about the 1 pF feedback capacitor. That's small enough that the feedback capacitance will be affected by the neighbour's cat walking by outside.
Tom
Most surface mount resistors (which lets face it means most resistors) are thick-film, which have excess noise (not as bad as carbon, they use ruthenium oxide powder fused into glass so moisture doesn't affect them, but its basically still a composite of conductive granules like carbon resistors).
Even metal film have a tiny amount of excess noise, metal foil and wire are the gold-standard, but you need to be rich to own the equipment that can detect the difference between metal-film and metal-foil.
Even metal film have a tiny amount of excess noise, metal foil and wire are the gold-standard, but you need to be rich to own the equipment that can detect the difference between metal-film and metal-foil.
Indeed - its an indication that slack layout shouldn't be tolerated, 1pF of stray capacitace between non-inverting input and feedback would null that out.. Just make it 10--22pF like everyone else?
Excess noise is often signal-correlated, and signal-correlated noise has been claimed to be more audible than fixed level noise.
In my own experience with a discrete resistor DSD FIR dac, excess noise effects in the resistors could be quite audible. For example, Susumu RG, which are often considered to be quite good resistors, were some of the worst offenders in the dac experiments.
If feedback resistors do produce any signal correlated noise, then of course the feedback loop should produce the corresponding inverse of that noise at the amplifier output.
Doesn't seem to be much published research on the issue but possibly these taken together might lend a little more insight:
https://www.mdpi.com/1424-8220/23/3/1107
https://www.sciencedirect.com/science/article/abs/pii/S0378437109004476
Last edited:
The smaller a metal film resistor is, the less ldeal it becomes, for for instance an 0603 thin-film chip resistor is outperformed by 1210 thin-film which is out-performed by MELF metal-film (all else being equal). "thin-film" is just SMT speak for metal-film and "thick-film" means metal-oxide.
When I've done a feedback resistor network in SMT I tend to go for larger resistor sizes like 1210 and thin-film.
When I've done a feedback resistor network in SMT I tend to go for larger resistor sizes like 1210 and thin-film.
There are tons of metal film SMD resistors out there. The Susumu thin film ones are very good. From what I understand they're made from sputter deposition of NiCr onto a ceramic base. They've available with tolerances down to ±0.05 % and temperature coefficients down in the single digit ppm/ºC. I use the ±0.5 % and ±0.1 % ones quite a bit and they're only a few cents each if you buy them by the reel.
As @Mark Tillotson points out, there's also the MELF resistors, which are plain metal film resistors in an SMD package. I hate MELF because of their form factor. They tend to fly off the tweezers and roll off the PCB. But the performance is hard to beat, though Susumu RG gets you very, very close.
I agree that the thick film resistors are to be avoided. I mean... They're fine as a series resistor for an LED or a pull-up resistor for a logic gate. But I wouldn't rely on them for audio except maybe for a power resistor where thick film or MOX are often the only options.
Do watch the power dissipation in SMD resistors in particular in applications where the temperature variation can introduce unwanted signal components, such as in the feedback network of an audio amp.
Tom
Is this based on listening tests or measurements? If the former, how many participants?
Tom
Listening tests. IMHO and IME the number of participants is far less relevant the innate and acquired skills of the participants. For example, JBL uses trained listeners for evaluation of speakers rather than naive subjects for good reasons. BTW, some of the resistor effects were far from subtle. OTOH, certain brands and models of metal thin film resistors are pretty good. Some time after the initial experiments, someone funded a trial of SMD metal foil resistors in a different DSD RTZ FIRDAC. The metal foil resistors were judged only slightly better than the original metal thin film resistors. We also felt that the much higher cost for metal foil resistors was not warranted by only very slight improvement in SQ. Bottom line for me, there are some Yageo and Pansonic resistors I would use for a FIRDAC, but not Susumu RG.
Also it might be of interest to note that Susumu has a line SMD resistors for audio that are claimed to be of even lower noise than RG, but their other specifications are worse. It suggests to me that most resistor designs may involve tradeoffs of characteristics not all of which may result in optimal resistors for audio.
Last edited:
Interesting discussion. I use thin film 1206 resistors in a lot of my designs and favor the Panasonic thin films, followed by Vishay and Yageo. I have avoided Susumu resistors for no rational reason.
In terms of excess noise and over all resistive linearity what your recommendations be? I have not really given a lot of thought to testing them. Years ago there was an article apparently at least partially aimed at ATE and precision test equipment hardware designers that listed the best performing resistor families. I never saw the entire article so am curious if anyone else has run across it?
Someone mentioned to me recently that bulk foil resistors in some cases have significant issues with pop corn noise. is there any basis in fact to that allegation?
In terms of excess noise and over all resistive linearity what your recommendations be? I have not really given a lot of thought to testing them. Years ago there was an article apparently at least partially aimed at ATE and precision test equipment hardware designers that listed the best performing resistor families. I never saw the entire article so am curious if anyone else has run across it?
Someone mentioned to me recently that bulk foil resistors in some cases have significant issues with pop corn noise. is there any basis in fact to that allegation?
Last edited:
So no measurements and N = 1?
I'm not looking to argue. I'm just trying to evaluate how much to weight to put on your results relative to my measurements.
Tom
Last edited:
N = 3, not all listening at the same time, and not communicating results with each other before giving their listening impressions. Each one is experienced and skilled.
EDIT: In the spirit of avoiding argumentative content, this post has been edited. Also if it matters at all, the speakers here are large panel ESLs from Sound Lab. They are about as different from cone speakers as condenser mics are from dynamic mics. IOW, unsurprising if we can hear a lot of little details here.
EDIT: In the spirit of avoiding argumentative content, this post has been edited. Also if it matters at all, the speakers here are large panel ESLs from Sound Lab. They are about as different from cone speakers as condenser mics are from dynamic mics. IOW, unsurprising if we can hear a lot of little details here.
Last edited:
No, not without an external power source, or it would violate the second law of thermodynamics.
Suppose you had a resistor with more than thermal noise, without needing any form of DC or AC bias. You could then connect it to a resistor that has only thermal noise. The noise power that the first resistor would supply to the second resistor would then be greater than the noise power that goes from the second resistor to the first. All in all, there would be a net power going from the first to the second resistor. Thermally insulate them and you have a free fridge.