For the non-inverting circuit, it is a good practice to minimize resistance in series with non-inverting input. This greatly reduces effect of the non-linear input capacitance. The 10k resistor is too high. Decreasing it in a magnitude will decrease resulting delta f. Resulting FB network resistance and resistor in series with non-inverting input should be matched in case they are greater than some 2k.
BTW-how about Doppler effect at the speaker, in case we speak about 3.7Hz/10000Hz deviation?
BTW-how about Doppler effect at the speaker, in case we speak about 3.7Hz/10000Hz deviation?
jarek said:
This is a hyphotetical question, as it does not correspond to any real circuit. One does not design input stages with 10k resistors in series.
OK. Let’s use a 10k ohm pot. The worst case of resistance will be about 2.5k ohm. Level of fault will decrease to 0.01% or so. Coloration or not?🙂
The best is 50 ohm " binary " attenuator, which is on output of preamp 😉 . Coloration in this case = zero 😎 .
Re: Phase distortion
I have seen this argument aired somewhere here before.....i confess i have never understood its rationale...
Just why would you consider the input stage to be outside the global feedback loop for non-inverting applications?
smoking-amp said:
I have seen this argument aired somewhere here before.....i confess i have never understood its rationale...
Just why would you consider the input stage to be outside the global feedback loop for non-inverting applications?
jarek said:
The question might be a little bit different. Can we say that possible frequency modulation is the same as distortion?
If you say distortion of 0.01% what you mean? Overall THD number? That is quite useless! 0.01% of the 2nd harmonic is no problem but 0.01% at 10th harmonic is a great problem.
And finally I would like to mention again that matching equivalent resistors for + and - input greatly reduces effect of non-linear input capacitance.
PMA said:
I always wonder about this. 0.01% at 2nd harmonic is essentially a base sin wave plus a sin signal at 2x the frequency and 0.01% of the magnitude. and the same holds true for 0.01% at 10th harmoic.
Has anyone done actual listening tests, trying the signals on real people to see if they could tell the difference?
or better yet, you can generate a random signal and manually introduce distortion. Do people really show preference for 2nd harmonics in such a test?
This depends on distortion in human ear and masking - you will find publications about the phenomenon
The question is how these 0.01% are important according to driver's distortions? And what the spectrum of drivers distortion is.
jarek said:
Yes it is. The driver's distortion is mostly of low harmonics order. 0.01% of high order harmonics "electronics" distortion is quite different, harmful and audible. Crossover distortion, even when compensated by feedbach, is a good example. I have tested an amp having 0.05% THD, but spread at similar level up to very high harmonics. By sinus signal 1kHz. If you put your ear onto tweeter, you can easily hear the higher harmonics from the tweeter. Repeating the same test with the class A amp with very low distortion the tweeter remains silent.
PMA said:
but that's a test for the existence of high order harmonics: the fact you can hear anything from the tweeter tells you that there is high frequency content from that amp. That's a given in this particular case.
The question still remains that if such low - level of high order harmonics is more disliked than the same low-level of low order harmonics.
I am sure there are studies out there with empirical and statistically significant data but I haven't seen one.
Yes there are studies like that. Unfortunatelly don't have enough time now, will try to post link several hours later.
millwood said:
Some resources on audibility of higher harmonics and human ear self distortion may be find here:
http://w3.mit.edu/cheever/www/cheever_thesis.pdf
I do not speak about amplifier stuff at this link.
revised phase distortion calculation
After sleeping on the problem overnight, I realized I made a mistake in calculating the phase distortion due to a bipolar current source for the diff. amp. in post #99. The varying collector capacitance reactance must also be transformed thru the diff. amp. transistor beta (lets say 200) before doing the RC calculations. This effectively reduces the effect by 200 times, so is not a realistic problem. In addition, as pointed out, a lower input impedance would diminish this further. So, I have been barking up the wrong tree here I think.
However, the phase distortion effects of varying gm of the diff. amp. transistors (with differential current variation) still remains (just as in the Op Amp. papers) and is not corrected in the non-inverting feedback case (although high open loop gain obviously still minimizes the problem by reducing the differential voltage swing).
Since the question was raised as to why the non-inverting configuration does not correct errors within the diff. amp. stage, I will explain my reasoning. In the inverting configuration two resistors are depended on to perform a linear subtraction, and are normally pretty darn linear. In the non-inverting configuration the diff. amp. performs the subtraction. If say one of the diff. amp. transistors introduces a distortion of the signal it is the same as if the reference signal were being corrupted. If both diff. amp. transistors introduce the same distortion effects, then they would obviously cancel out. But, since the currents in the two diff. amp. transistors are complementary, not tracking, many distortion effects will not cancel out. Just to illustrate, suppose we put a diode between the two emitters, anyone still think the feedback will correct for this! Since gm is roughly Ie/26mV, the two diff. transistors have non tracking gm variation. And this leads to the phase distortion problem treated in the Op. Amp. articles.
I will try and gin up some crude estimate of the phase distortion from the varying gm after I get a better handle on how it happens (ie. crack the books), but I think I can see already that this will a much bigger effect since one trans. gm will dip to near zero when the input stage reaches 26mV unbalance.
Don
After sleeping on the problem overnight, I realized I made a mistake in calculating the phase distortion due to a bipolar current source for the diff. amp. in post #99. The varying collector capacitance reactance must also be transformed thru the diff. amp. transistor beta (lets say 200) before doing the RC calculations. This effectively reduces the effect by 200 times, so is not a realistic problem. In addition, as pointed out, a lower input impedance would diminish this further. So, I have been barking up the wrong tree here I think.
However, the phase distortion effects of varying gm of the diff. amp. transistors (with differential current variation) still remains (just as in the Op Amp. papers) and is not corrected in the non-inverting feedback case (although high open loop gain obviously still minimizes the problem by reducing the differential voltage swing).
Since the question was raised as to why the non-inverting configuration does not correct errors within the diff. amp. stage, I will explain my reasoning. In the inverting configuration two resistors are depended on to perform a linear subtraction, and are normally pretty darn linear. In the non-inverting configuration the diff. amp. performs the subtraction. If say one of the diff. amp. transistors introduces a distortion of the signal it is the same as if the reference signal were being corrupted. If both diff. amp. transistors introduce the same distortion effects, then they would obviously cancel out. But, since the currents in the two diff. amp. transistors are complementary, not tracking, many distortion effects will not cancel out. Just to illustrate, suppose we put a diode between the two emitters, anyone still think the feedback will correct for this! Since gm is roughly Ie/26mV, the two diff. transistors have non tracking gm variation. And this leads to the phase distortion problem treated in the Op. Amp. articles.
I will try and gin up some crude estimate of the phase distortion from the varying gm after I get a better handle on how it happens (ie. crack the books), but I think I can see already that this will a much bigger effect since one trans. gm will dip to near zero when the input stage reaches 26mV unbalance.
Don
- Status
- Not open for further replies.