I only get tin stars these days.I published this in 1982, it was an EDN cover article on the AD524. I could make a perfect pink noise generator with a bridge of 3 MF and one carbon, but now I can't find a carbon comp bad enough to repeat the circuit.
A couple of other things, you don't listen to resistors i.e. there are no amplifier circuits that would support seperability of -147dB resistor distortion from the -110 to -120 dB of the best amplifiers.
Actually very few resistors in a typical circuit see full (or any) signal swing, usually only 2 or 3.
I don't see any concensus on resistors to support a "mesurement = sound quality" stance, there are plenty of people who won't give up their naked Vishays.
There is little evidence that anything more than the thermal properties of the resistors is being measured. In that case normal thermal management would handle the problem and the square law of power vs I or V would reduce the distortions dramatically. Already it is normal to make the feedback resistor in a PA 2-5W.
There may be no correlation between present measurements and sound quality, however different resistors in amps have different sound. May be the major question is what and how to measure.
I only get tin stars these days.
A couple of other things, you don't listen to resistors i.e. there are no amplifier circuits that would support seperability of -147dB resistor distortion from the -110 to -120 dB of the best amplifiers.
There is little evidence that anything more than the thermal properties of the resistors is being measured. In that case normal thermal management would handle the problem and the square law of power vs I or V would reduce the distortions dramatically. Already it is normal to make the feedback resistor in a PA 2-5W.
Scott,
Tin stars, my goodness you are lucky, you can roll out the tin and make audio capacitors!
Afraid we need to disagree. I have changed the feedback resistor in real amplifiers to see what happened. I could hear differences and thought it would be interesting to see what measurements changed. The delta C was unmeasurable as was the delta L, so I was left with delta R!
Virtually all resistors show thermal distortion at levels well below their rated power. So 2 volts across 150 ohms is as bad as 15 volts at 1000 ohms. I think the results make it clear that the power ratings and thermal coefficients need to be at least given a cursory glance in any design.
The interesting part about the distortion is some of the samples actually showed some distortion that was not thermal! This to me hints that some folks might want to look at the alloy they use to make the metal film.
There are bad resistors still being made. I see no reason to show the really bad examples as they are being sold as cheap or miniatures.
The claim that Naked Vishays "sound better" to me means the heatsinking change may actually be distinguished by some or it could just be more fertilizer. But they do not measure the same.
As to feedback resistors already being 2-5 watts, I got in half a dozen amplifiers for review, the biggest feedback resistor is a .33 Watt Vishay. The rest use 1/4 watt! High end stuff... no, but very common products.
But the real question is did this project find a really cheap resistor that works as well as the really expensive ones. Well that depends on your point of view is $25 per 1000 pieces a good price for one of the best?
Since the AP is an FFT I presume that would be 2.5nV rms in each bin. I have no idea what the bin width is in this case, but note a bridge of 1K resistors has 4nV/rt-Hz noise so further explanation is due.
BTW as 1audio pointed out and I mentioned a few weeks ago, you would be amazed at how good a sound card and in-amp can do this on your computer. Play out of phase excitation in the left and right channels so the common mode is very low and sense the bridge differentially with a good in-amp. Remember also for purely resistive measurements the residual distortion of the source is also mostly nulled. You could probably find a trick to get the thirds from L/R mis-match to null >170dB.
OK, let's see if I got this.
FFTlength is 65536 points
FFTfreq is 96kHz
This would give 96000/65536 = 1.46484375 Hz/bin
I want to find the rms voltage over, let's say, 10 bins, starting at around 1500Hz, so that would be the rms voltage from 1500Hz to 1514.6484375Hz. This is where I'd like to calculate the voltage which Simon says it is about 2nV. From the plot we could say that there are values anywhere around 2nV, could be 5nV, 1.5nV, 3nV, etc.
startbin: 1500 Hz / 1.46484375 Hz/bin = 1024
endbin: 1034 which will be at 1514.6484375Hz
V(rms,noise) = K_wc * sqrt( sum_n=startbin^endbin V^2
)where K_wc is the window correction factor, which for no windows it will be 1, for Blackman-Harris will be 0.761, for Hanning will be 0.8165.
Let's assume Simon used a Blackman-Harris window, then the approximate rms voltage over ten bins would look like this:
V(rms,noise) = 0.761 * sqrt ((2e-9)^2+(1.5e-9)^2+(5e-9)^2+...+(2.5e-9)^2) = X nV
Where X would be somewhere between 1 and 7 nV. I'm only trying to illustrate, approximately, what my understanding of what you said is.
I'm still missing the real parameters of Simon's test, i.e. the FFT freq and bin size. I did ask for the experiment setup, didn't I?
Scott,
The AP doesn't quite trust the user, The only setting I get is to tell it 65K samples or "Optimum" So I do 65 K samples and 4K averages.
I assume a single good 1K resistor at room temperature is 4nv/rt-hz. Each side of the bridge is indeed 500 ohms or 2.8 nv/rt-hz and both sides add back up to 4nz/rt-hz, on this we agree. As I have push pull distortion generation the distortion signals may just add, so I can look a bit deeper
As to the 2.5 number that was just a calculation based on the numbers Iko took off of the plots.
As to just using a sound card, that is the whole idea. Wheatstone was a very smart guy. A DIY resistor test anyone can do.
The other bit of disagreement is that I find the measurements do seem to show what are considered good resistors do measure better, those that are thought worse of do not measure as well.
In my carbon comp example the distortion can rise to -85db re rated power. I expect that this can be heard.
If it can be heard it, can be measured. If the measurements do not agree with what is being heard then the measurement may be of the wrong thing.
simon7000 said:
I definitely would like to try some version of this experiment with my 0404 sound card. Mostly out of curiosity.
scott wurcer said:
Not sure that I could find the trick, baby steps here. By all means any suggestion would be appreciated.
OK, so it may seem like I'm just trying to get spoon-fed here, but what I'm really trying to say is that I'm aware of my current limitations. Will try to work thing out by myself.
Probably so, at least it sounds logical.
The question is what measurements predict, or indicate, the sound quality of an amp and its' various components.
I would like to address a couple of points:
First, this type of test depends on some sort of 'nulling' of the fundamental.
It doesn't actually matter if it is by phase shift cancellation like the HP339, the ST1700, or the AP, in normal mode, or by bridge, as used by Ed Simon. Either way will work.
The REASON that we need to null the fundamental, is the limited dynamic range of the following test instrumentation that is often less than 100dB, or so.
Second, this is ONLY A TEST, of the resistance nonlinearity at a relatively high AC voltage. Another interesting TEST would be the same sort of measurement at 15mV or 1000 times lower voltage, in my dreams, at 15uV as well. This would bring out different distortion mechanisms from the resistors, more like crossover distortion. This is just as important, BUT even harder to do.
A serious scientific establishment might dedicate man-months and days of computer time for a serious measurement, to solve the most difficult questions. That is why we often rely on our ears, and then try to show the 'doubting' that there is something to what we actually hear, often over months or years, in listening to different parts.
Third, what is really important here, is the RELATIVE distortion between the different resistors, not the ABSOLUTE amount. We must remember that we don't listen to a single test tone, and call it music. Multiple tones, especially in music, can be a more efficient way for the human ear to discern differences. If we could use music directly for this test, AND with the same sensitivity, we would see even bigger numbers referred to as distortion, but our test equipment doesn't quite do it, yet.
First, this type of test depends on some sort of 'nulling' of the fundamental.
It doesn't actually matter if it is by phase shift cancellation like the HP339, the ST1700, or the AP, in normal mode, or by bridge, as used by Ed Simon. Either way will work.
The REASON that we need to null the fundamental, is the limited dynamic range of the following test instrumentation that is often less than 100dB, or so.
Second, this is ONLY A TEST, of the resistance nonlinearity at a relatively high AC voltage. Another interesting TEST would be the same sort of measurement at 15mV or 1000 times lower voltage, in my dreams, at 15uV as well. This would bring out different distortion mechanisms from the resistors, more like crossover distortion. This is just as important, BUT even harder to do.
A serious scientific establishment might dedicate man-months and days of computer time for a serious measurement, to solve the most difficult questions. That is why we often rely on our ears, and then try to show the 'doubting' that there is something to what we actually hear, often over months or years, in listening to different parts.
Third, what is really important here, is the RELATIVE distortion between the different resistors, not the ABSOLUTE amount. We must remember that we don't listen to a single test tone, and call it music. Multiple tones, especially in music, can be a more efficient way for the human ear to discern differences. If we could use music directly for this test, AND with the same sensitivity, we would see even bigger numbers referred to as distortion, but our test equipment doesn't quite do it, yet.
Before reading about the resistor test here I was going to get/build a notch filter to extend the dynamic range of my device. This article has a nice illustration of how this works.
Notch Filter Extends Spectrum-Analyzer Range - 2000-01-01 07:00:00 | Test & Measurement World
High Q and low distortion seem to be the two features one would like in a good notch filter. While we're at it, any suggestions? Passive, active? Anything at all?
Notch Filter Extends Spectrum-Analyzer Range - 2000-01-01 07:00:00 | Test & Measurement World
High Q and low distortion seem to be the two features one would like in a good notch filter. While we're at it, any suggestions? Passive, active? Anything at all?
Last edited:
If the resistors are mostly linear we should be able to infer impulse response from the frequency/phase response. The first thing to look at is the frequency response at the bridge. If the resistance shifts with frequency differently than the reference resistors any harmonics in the source will get amplified by the difference.
A simple test would be to quantify the difference in resistance vs. frequency by the change in the null settings. With the bridge network it should be possible to change the drive level and see if the null point changes as well. While distortion is interesting I would not be surprised if the resistance changes with frequency are as interesting.
A simple test would be to quantify the difference in resistance vs. frequency by the change in the null settings. With the bridge network it should be possible to change the drive level and see if the null point changes as well. While distortion is interesting I would not be surprised if the resistance changes with frequency are as interesting.
I rather mean behaviour during impact (surge stress). Measured by single surge wave, and evaluated in time domain.
I do not mean calculation of transfer function from impulse response, as suggested by Demian.
Regards,
- Status
- Not open for further replies.