traderbam said:
I am not, i just don't see any difference between phase modulation and frequency modulation, isn't FM just one frequency modulating the phase of another ?
At least that's how i programmed FM-synthesis and doppler effects, worked perfect... (sorry to be a non studied idiot)
Mike
Frequency modulation is a kind of phase modulation. It is quite difficult to tell from spectra what kind of modulation it was. Anyway, it is not so important, as distortions caused by these modulations are subjectively very bad, ear is not tolerant to them. This is much worse than harmonic distortion. As the instrumentation is difficult, engineers better measure HD, which in fact need not say much about amplifier sound, especially when below certain level, which is easy to obtain now.
Well, I should then correct myself and call it PIM distortion.
Barrie Gilbert appears to call it "a peculiar kind of amplitude to phase distortion". Shall we call it that? Whatever you want.
I admit that I don't know the exact technical difference between phase modulation and frequency modulation. I will have to look it up.
For the rest out there, it IS some sort of distortion, that even concerns Barrie Gilbert, who is an op amp designer, and me also a little bit.
Barrie Gilbert appears to call it "a peculiar kind of amplitude to phase distortion". Shall we call it that? Whatever you want.
I admit that I don't know the exact technical difference between phase modulation and frequency modulation. I will have to look it up.
For the rest out there, it IS some sort of distortion, that even concerns Barrie Gilbert, who is an op amp designer, and me also a little bit.
I agree, PMA. I am just interested in making better amps and preamps. This is a new insight for me, and I think, very important, subjectively. But then, what do I know? My preamp is selling in Japan at a price that BMW's go for in this country. Just sold another one tonight at full retail. Happy days!
My point is this:
In my experience, good audio is about precision engineering. Precision is both in measurement and in precision of analysis. Being imprecise can lead to all sorts of dead ends and problems and red herrings. For both experts and DIYers alike.
There is a BIG difference between frequency modulation and phase modulation. The difference is important because designing a control system to deal with FM is whole different kettle of fish to designing one to deal with phase shift.
Gilbert's article is talking about the effect of variable phase shift. The term PIM (phase intermodlaton distortion) seems to me to be a fair way to refer to one of the impacts of this.
Both FETs and BJTs have this problem. So it is a matter of measuring/analysing the impact in the particular circuit. FETs tend to have the advantage, as JC said, of lower junction capacitances and lower order non-linearity of transconductance.
This effect needn't scare any of you who aren't maths graduates, at least in its concept. You already know that the gain and phase of a transistor vary with frequency. They also vary with current and voltage. This discussion is about the impact of these variations when occur in the input differential circuit. Gilbert's article simply puts these effects inside a negative feedback loop and explains, in a slightly simplified model, the type of distortion that results.
I think the relevance of this to this thread is whether JFETs or BJTs (with emitter resistors) are better or worse in their contribution to variable phase shift for a given input differential stage.
About feedback systems:
Things get much more challenging to analyse when you introduce these non-linearities into a control system and then try to analyse or measure it. Control theory is extremely mathmatically complex and the impact of these effects is very difficult to analyse and therefore difficult to deal with. This is why negative feedback can sometimes, if not commonly, make circuits sound worse in one or several regards, contrary to intuition and linear control systems theory. It doesn't have to, but it certainly can do.
I think it is quite understandable and justifiable, therefore, to design open loop systems and to depend on painstaking selection and matching of parts to achieve acceptable distortion as an alternative to the (perhaps more painful) challenge of resolving the control system issues.
In my experience, good audio is about precision engineering. Precision is both in measurement and in precision of analysis. Being imprecise can lead to all sorts of dead ends and problems and red herrings. For both experts and DIYers alike.
There is a BIG difference between frequency modulation and phase modulation. The difference is important because designing a control system to deal with FM is whole different kettle of fish to designing one to deal with phase shift.
Gilbert's article is talking about the effect of variable phase shift. The term PIM (phase intermodlaton distortion) seems to me to be a fair way to refer to one of the impacts of this.
Both FETs and BJTs have this problem. So it is a matter of measuring/analysing the impact in the particular circuit. FETs tend to have the advantage, as JC said, of lower junction capacitances and lower order non-linearity of transconductance.
This effect needn't scare any of you who aren't maths graduates, at least in its concept. You already know that the gain and phase of a transistor vary with frequency. They also vary with current and voltage. This discussion is about the impact of these variations when occur in the input differential circuit. Gilbert's article simply puts these effects inside a negative feedback loop and explains, in a slightly simplified model, the type of distortion that results.
I think the relevance of this to this thread is whether JFETs or BJTs (with emitter resistors) are better or worse in their contribution to variable phase shift for a given input differential stage.
About feedback systems:
Things get much more challenging to analyse when you introduce these non-linearities into a control system and then try to analyse or measure it. Control theory is extremely mathmatically complex and the impact of these effects is very difficult to analyse and therefore difficult to deal with. This is why negative feedback can sometimes, if not commonly, make circuits sound worse in one or several regards, contrary to intuition and linear control systems theory. It doesn't have to, but it certainly can do.
I think it is quite understandable and justifiable, therefore, to design open loop systems and to depend on painstaking selection and matching of parts to achieve acceptable distortion as an alternative to the (perhaps more painful) challenge of resolving the control system issues.
FM and PM are essentially the same, the only difference is that the modulation is passed through an integrator
in one case and not in the other.
If you have a carrier modulated with just one sine there is obviously no difference at all.
Doubling the frequency of a carrier makes the phase noise 6 dB stronger
because the absolute jitter stays the same while the period is halved.
Multiplying the frequency of a test signal several times makes it thus easier to measure the PM.
There is a lot of information on phase noise on the webserver of the Time&Frequency group at the NIST.
Goto www.boulder.nist.gov and search for the names
F.L.Walls
Ferre-Pikal
Craig Nelson
Aramburo
Some .pdf:
-Comparision of 1/f PM noise in commercial amplifiers
-Guidelines for Designing BJT Amplifiers with low 1/f AM and PM noise
-Reducing Errors, Complexity and Measurement time of PM noise measurements
The filenames have numbers only, search them via the publication lists.
I don't have the URLs handy, I printed the stuff a year ago.
This is serious RF / microwave work, probably way over the top for audio.
regards, Gerhard
in one case and not in the other.
If you have a carrier modulated with just one sine there is obviously no difference at all.
Doubling the frequency of a carrier makes the phase noise 6 dB stronger
because the absolute jitter stays the same while the period is halved.
Multiplying the frequency of a test signal several times makes it thus easier to measure the PM.
There is a lot of information on phase noise on the webserver of the Time&Frequency group at the NIST.
Goto www.boulder.nist.gov and search for the names
F.L.Walls
Ferre-Pikal
Craig Nelson
Aramburo
Some .pdf:
-Comparision of 1/f PM noise in commercial amplifiers
-Guidelines for Designing BJT Amplifiers with low 1/f AM and PM noise
-Reducing Errors, Complexity and Measurement time of PM noise measurements
The filenames have numbers only, search them via the publication lists.
I don't have the URLs handy, I printed the stuff a year ago.
This is serious RF / microwave work, probably way over the top for audio.
regards, Gerhard
john curl said:
It's great, congratulations!
If you had a japanese customer who could not afford the BMW price but only Nissan, please direct him to me
. I will try to satisfy him with my preamp. I have quite good knowledge of japanese customers, as I visit Japan once a year with quite a different technology than audio.Arigató gozai masu,
Pavel
, how about giving him some practical circuit examples?
PMA said:
Yes, jfets sound lovely, but i use bjts in inputstage because of their much higher Voltage rating and gm... And badly done cascoding can ruine the whole inputstage.
Sadly, my favourite jfet sk170 has become obsolete. (but still have a dozen)
In an earlier amp i compared bc546b against sk170 by swapping them and listen, sk170 was the definite winner. (smoother/nicer trebles)
Mike
John, do you mean DC-offset caused by base currents from the input-bjts ?
Well, it forces you to use DC-blocking caps (input cap,dc-cap in feedback), a big advantage of jfets...
But if using dc-blocking caps, and setting the correct resistors, dc-offsets perfectly cancel out through common mode.
For example, in my symasym the DC-offset (3mv) is only caused by a thermal inbalance in the current mirror of the vas.
If for some reasons these caps can't be used (DC-coupling needed) i would always go FET.
Mike
Well, it forces you to use DC-blocking caps (input cap,dc-cap in feedback), a big advantage of jfets...
But if using dc-blocking caps, and setting the correct resistors, dc-offsets perfectly cancel out through common mode.
For example, in my symasym the DC-offset (3mv) is only caused by a thermal inbalance in the current mirror of the vas.
If for some reasons these caps can't be used (DC-coupling needed) i would always go FET.
Mike
What about newer high beta, low noise transistors? The high betas would minimize offset and bias current. I've had little offset, i.e., <10mv using matched pairs of BJTs, that weren't particularly high beta. And they're quite easy to match.
As to PIM, I recall reading a paper that discredited this phenominon. I haven't read if for a while, and don't remember the author, but I seem to recall that any PIM in a properly designed op amp would be caused by non-linear capacitance in the VAS, not the input stage. The author might have been Cordell, but I'm not sure. I'll have to look for it.
I understand the noise issue with degeneration resistors, though.
However, what about the input capacitance of Jfets. Don't the lower noise and more linear ones have higher input capacitance?
As to PIM, I recall reading a paper that discredited this phenominon. I haven't read if for a while, and don't remember the author, but I seem to recall that any PIM in a properly designed op amp would be caused by non-linear capacitance in the VAS, not the input stage. The author might have been Cordell, but I'm not sure. I'll have to look for it.
I understand the noise issue with degeneration resistors, though.
However, what about the input capacitance of Jfets. Don't the lower noise and more linear ones have higher input capacitance?
No transistor will ever beat a jfet concerning these offsets, even the mpsa18 with its beta of ~1000 is no challenge to the leakage of picoAmperes from a jfet.
You are right, these very nice jfets like sk170 have horrible high capacitance, input capacitance = 30pF, output capacitance = 6pF.
Mpsa18: 5.6pF / 1.7pF...
I see the big advantage of jfets in their transfer curve, it's tube like, no e^x function, more like x^2. This results in very little high order harmonics.
Mike
You are right, these very nice jfets like sk170 have horrible high capacitance, input capacitance = 30pF, output capacitance = 6pF.
Mpsa18: 5.6pF / 1.7pF...
I see the big advantage of jfets in their transfer curve, it's tube like, no e^x function, more like x^2. This results in very little high order harmonics.
Mike
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.