Nonlinearities manifest as THD
Mikeks asked "Have you encountered any form of non-linearity that does not generate spurious harmonics, and is not, therefore, manifest as an increase in THD+N?"
First of all, you need to allow for the possibility that it may manifest itself in THD+N at some frequency and at some level. For example, it is easy to build an amplifier that has good 1 kHz THD but which still has poor high frequency nonlinearity. But that same amplifier will not likely escape without measurable 20 kHz THD. Similarly, the nonlinearity of an electrolytic capacitor may only show up in THD at very low frequencies.
Given those caveats, I'm not sure that I have seen a nonlinearity that did not generate some THD, although I don't foreclose on the possibility. That is why I always believe in doing different types of distortion measurements. But it tends to be generally true that a nonlinearity will virtually always generate added spectral components that were not originally present in the stimulous. There are some cases where one type of distortion measurement will be more sensitive than another. A simple example is that 20 kHz THD may not register in a sharply band-limited system, while an IM measurement of the distortion that results from the same nonlinearity will register because it is in-band.
20 kHz THD is a much better power amplifier test than some have given it credit for, but I still don't believe in relying solely on it. I'm a big believer in spectral analysis of the distortion products out to, say, 9th order or more. That is not always practical with THD-20. For that reason, I lean toward full spectral analysis of 19 kHz + 20 kHz twin-tone IM. The tradeoff is that not everyone has a spectrum analyzer for that, and in some cases the measurement floor for twin tone IM is not as low as it is for THD-20. It really depends a lot on what kind of instrumentation you have available.
Bob
Mikeks asked "Have you encountered any form of non-linearity that does not generate spurious harmonics, and is not, therefore, manifest as an increase in THD+N?"
First of all, you need to allow for the possibility that it may manifest itself in THD+N at some frequency and at some level. For example, it is easy to build an amplifier that has good 1 kHz THD but which still has poor high frequency nonlinearity. But that same amplifier will not likely escape without measurable 20 kHz THD. Similarly, the nonlinearity of an electrolytic capacitor may only show up in THD at very low frequencies.
Given those caveats, I'm not sure that I have seen a nonlinearity that did not generate some THD, although I don't foreclose on the possibility. That is why I always believe in doing different types of distortion measurements. But it tends to be generally true that a nonlinearity will virtually always generate added spectral components that were not originally present in the stimulous. There are some cases where one type of distortion measurement will be more sensitive than another. A simple example is that 20 kHz THD may not register in a sharply band-limited system, while an IM measurement of the distortion that results from the same nonlinearity will register because it is in-band.
20 kHz THD is a much better power amplifier test than some have given it credit for, but I still don't believe in relying solely on it. I'm a big believer in spectral analysis of the distortion products out to, say, 9th order or more. That is not always practical with THD-20. For that reason, I lean toward full spectral analysis of 19 kHz + 20 kHz twin-tone IM. The tradeoff is that not everyone has a spectrum analyzer for that, and in some cases the measurement floor for twin tone IM is not as low as it is for THD-20. It really depends a lot on what kind of instrumentation you have available.
Bob
If you have a small SMPS (like laptop battery charger), it is usually portable, about the size that hand can handle.
Put this SMPS in duty (chargering battery), then move this SMPS near the pcb, near input section, near output section, near zobel network, near speaker cables, near speakers, etc. SMPS radiation REALLY gets in to the amplifier's loop, making it's way to distortion artifact. You will not see this on output trace (too small too see if compared to output signal), but it will be seen with distortion meter (after nulling the fundamental first).
Maybe some will consider this as "cheating", but in year 2006 almost everything in the house is powered by SMPS. Printer, computer, TV (has flyback), laptop, even neon lamps have SMPS. The environment of home audio is full of SMPS radiation.
We haven't count for WiFi radiation, handphone signals yet
Maybe power amps designed in 2006 upwards should consider these ?
Put this SMPS in duty (chargering battery), then move this SMPS near the pcb, near input section, near output section, near zobel network, near speaker cables, near speakers, etc. SMPS radiation REALLY gets in to the amplifier's loop, making it's way to distortion artifact. You will not see this on output trace (too small too see if compared to output signal), but it will be seen with distortion meter (after nulling the fundamental first).
Maybe some will consider this as "cheating", but in year 2006 almost everything in the house is powered by SMPS. Printer, computer, TV (has flyback), laptop, even neon lamps have SMPS. The environment of home audio is full of SMPS radiation.
We haven't count for WiFi radiation, handphone signals yet
Maybe power amps designed in 2006 upwards should consider these ?
Hi Bob,
i am not sure if this is appropriate but since you are here i might as well ask you.
i have read your article in AUDIO magazine about the THD analyser, in three parts, i only got the first and the third , and the second part only recently.
now my question is, do you have this analyser offered as a kit anywhere?
i am sure many DIY'ers out here would like to have one.
thanks.
i am not sure if this is appropriate but since you are here i might as well ask you.
i have read your article in AUDIO magazine about the THD analyser, in three parts, i only got the first and the third , and the second part only recently.
now my question is, do you have this analyser offered as a kit anywhere?
i am sure many DIY'ers out here would like to have one.
thanks.
Homework...
Perhaps folks would like to figure out whether the first pair of drivers in this output stage operates in Class A if the output devices are in Class B?
I am knackered.
Perhaps folks would like to figure out whether the first pair of drivers in this output stage operates in Class A if the output devices are in Class B?
I am knackered.
Re: Homework...
Because everything depends on value of resistor between emitters of driver stage, coefficient of current amplification of output stage, and load resistance.
If current through this resistor is more than peak current consumed by output starge from driver the driver works in A class, otherwise it works in AB class.
mikeks said:Perhaps folks would like to figure out whether the first pair of drivers in this output stage operates in Class A if the output devices are in Class B?
I am knackered.
Because everything depends on value of resistor between emitters of driver stage, coefficient of current amplification of output stage, and load resistance.
If current through this resistor is more than peak current consumed by output starge from driver the driver works in A class, otherwise it works in AB class.
Re: Hello Bob..
That's a very good question. I don't think that it's possible to have distortions that don't manifest as spurious harmonics. And even if there are, would the ear be able to hear them ?
What looks quite sure is that there are distortion types that won't show up with a single sinewave.
Nice example: IM-test with a 50hz+10khz. As openloop distortion will vary the openloop gain (feedback factor) on a 50hz period, the feedback amount for the 10khz signal will modulate with 50hz, varying the compensation for the openloop phaseshift with this 50hz. Voila, phasemodulation (fm). If openloop distortion and phaseshift is high enough, you will even get a "doppler" effect.
Mike
mikeks said:
That's a very good question. I don't think that it's possible to have distortions that don't manifest as spurious harmonics. And even if there are, would the ear be able to hear them ?
What looks quite sure is that there are distortion types that won't show up with a single sinewave.
Nice example: IM-test with a 50hz+10khz. As openloop distortion will vary the openloop gain (feedback factor) on a 50hz period, the feedback amount for the 10khz signal will modulate with 50hz, varying the compensation for the openloop phaseshift with this 50hz. Voila, phasemodulation (fm). If openloop distortion and phaseshift is high enough, you will even get a "doppler" effect.
Mike
HF MIM for EMI ingress
Milan,
You make some good points, but keep in mind that with the test I described in the earlier post, we don't really care about the level of the original test signal that gets through - in fact, the less the better. All we care about is the amount of in-band IM products that are generated by the amplifier in response to this high-frequency stimulous.
In another post, the question was asked as to why three tones were needed in the HF MIM test rather than just two. If you use only two tones, only the even order IM products will fold down into the audio passband. By using thee tones, you also get the odd-order IM products to fold down into the audio passband.
You are certainly right that many other types of HF stimulous could be used for an EMI ingress susceptibility test for amplifiers. The HF MIM one I described just seemed to be a fairly simple one to generate with fairly predictable IM product frequencies to look for.
Thanks for your interest.
Bob Cordell
Milan,
You make some good points, but keep in mind that with the test I described in the earlier post, we don't really care about the level of the original test signal that gets through - in fact, the less the better. All we care about is the amount of in-band IM products that are generated by the amplifier in response to this high-frequency stimulous.
In another post, the question was asked as to why three tones were needed in the HF MIM test rather than just two. If you use only two tones, only the even order IM products will fold down into the audio passband. By using thee tones, you also get the odd-order IM products to fold down into the audio passband.
You are certainly right that many other types of HF stimulous could be used for an EMI ingress susceptibility test for amplifiers. The HF MIM one I described just seemed to be a fairly simple one to generate with fairly predictable IM product frequencies to look for.
Thanks for your interest.
Bob Cordell
Re: Superbly intuitive analysis!
Actually, i have now found this not to be entirely true:
Whether Q7/Q8 operate in Class A is also a function of the size of the cross-coupling resistor, the magnitude of the load and is dependent on the entire output stage not being under-biased.
This, to a much lesser extent, is also true of the first pair of cross-coupled drivers in Locanthi's arrangement.
Be that as it may, with the later circuit the first pair of drivers can be guaranteed to operate in class A provided the output devices are biased at least for class B, and the cross-coupling resistor is made sufficiently small (typically =<1K).
mikeks said:
Actually, i have now found this not to be entirely true:
Whether Q7/Q8 operate in Class A is also a function of the size of the cross-coupling resistor, the magnitude of the load and is dependent on the entire output stage not being under-biased.
This, to a much lesser extent, is also true of the first pair of cross-coupled drivers in Locanthi's arrangement.
Be that as it may, with the later circuit the first pair of drivers can be guaranteed to operate in class A provided the output devices are biased at least for class B, and the cross-coupling resistor is made sufficiently small (typically =<1K).