Hartono,
you wrote me an e-mail with several good points regarding possible effect of output coils. My answer is like this - we always have to take into account real values when evaluating phenomenae. Though I do not use output coils, I believe that a coil of 2uH in parallel with damping resistor of few ohms will not be audible, with regard to speaker cable inductance and coil values used in passive crossover. A phase shift added by 2uH coil is IMHO negligible in audio band. You made some remarks about passive crossovers - yes, IMHO they are an evil and I do prefer active crossovers.
Regards,
Pavel
you wrote me an e-mail with several good points regarding possible effect of output coils. My answer is like this - we always have to take into account real values when evaluating phenomenae. Though I do not use output coils, I believe that a coil of 2uH in parallel with damping resistor of few ohms will not be audible, with regard to speaker cable inductance and coil values used in passive crossover. A phase shift added by 2uH coil is IMHO negligible in audio band. You made some remarks about passive crossovers - yes, IMHO they are an evil and I do prefer active crossovers.
Regards,
Pavel
Hi Pavel,
Maybe you misunderstood my point, the back EMF like to see 0 ohm ,if using output coil will degrade this at high frequency, hence my suggestion to tap the speaker and see what happen at the amp output. The back EMF going to the amp will be delayed, and produce error voltage because of non zero output impedance of the amp, and become worse because of the output coil + resistor, and the amp will try to fix it, when the amp try to fix it, the output from the amp must pass delay from the output coil,but it would already be too late, this causing what Graham Maynard said as first cycle distortion.
Best regards
Hartono
Maybe you misunderstood my point, the back EMF like to see 0 ohm ,if using output coil will degrade this at high frequency, hence my suggestion to tap the speaker and see what happen at the amp output. The back EMF going to the amp will be delayed, and produce error voltage because of non zero output impedance of the amp, and become worse because of the output coil + resistor, and the amp will try to fix it, when the amp try to fix it, the output from the amp must pass delay from the output coil,but it would already be too late, this causing what Graham Maynard said as first cycle distortion.
Best regards
Hartono
In case you say delayed, you have to say how much. Dealyed of 0.5us probably. This is IMHO negligible. We have already discussed this in another threads. "Back EMF" can be easily simulated by speaker model and we should not mix linear and non-linear issues. In case you come with calculation and assessment of the coil influence, we can discuss it.
On listener fatigue
This is more than 30 years ago, but I first saw the term "listener fatigue" in connection with Peter Walker's tests to find why his very good first transistor amplifier was rejected in preference to others with higher thd. The experiments seem to indicate that the brain can build up resistance to stimulation ("sounds"?) that was at or even below the threshold of audibility. (One must remember that the threshold of audibility is just that; it is not the threshold of stimulation recognisable by the brain.)
It was then explained (much as someone explained here some posts ago) that high order harmonic generation can at any moment in an orchestral performance, cause many hundreds of different frequency "impulses", of which a considerable combination can be so strident that resistance is created in the brain. This either by the pitches themselves or additional intermodulation products. This will also be impossible to quantify as the content changes every moment.
That to me appeared to be a reasonable postulation - and thus, as quantification is so difficult, simply avoid high order harmonic generation like the plague. I have often cited acoustical tests here mainly in the Scandenavian countries, but I have not come across research in this connection. That would be interesting to read.
This is more than 30 years ago, but I first saw the term "listener fatigue" in connection with Peter Walker's tests to find why his very good first transistor amplifier was rejected in preference to others with higher thd. The experiments seem to indicate that the brain can build up resistance to stimulation ("sounds"?) that was at or even below the threshold of audibility. (One must remember that the threshold of audibility is just that; it is not the threshold of stimulation recognisable by the brain.)
It was then explained (much as someone explained here some posts ago) that high order harmonic generation can at any moment in an orchestral performance, cause many hundreds of different frequency "impulses", of which a considerable combination can be so strident that resistance is created in the brain. This either by the pitches themselves or additional intermodulation products. This will also be impossible to quantify as the content changes every moment.
That to me appeared to be a reasonable postulation - and thus, as quantification is so difficult, simply avoid high order harmonic generation like the plague. I have often cited acoustical tests here mainly in the Scandenavian countries, but I have not come across research in this connection. That would be interesting to read.
I am still puzzled by some of the "coil audibility" arguments. As an example, the square wave response showed by PMA (post #82) can serve. Fine, there is some ringing. But as well as I could scale, the rise time of the 10 KHz square wave is about 4 uS. The oscillations are of the order of 300 KHz. Pray how can that have any audible effect at all? Also, what was the rise time of the input signal? I would guess about 1 uS or better. Those overshoots would not be there with a siganl closer to audio. Why does it seem that we want exemplary performance at r.f.? What are such frequencies doing in the input in the first place?
What can happen is that those oscillations can cause the amplifier to overload at a slightly lower output - but that will be a few % and inconsequential.
Then once more, let us please stop considering the damping factor in these discussions. It appears to be forgotten that the real damping factor depends on the total resistance (impedance) in the loudspeaker circuit. The main part of this is the 5 - 6 ohm odd resistance of the voice coil. I wonder if the series coil resistance (or impedance at audio frequencies) will be even 1% of this.
I also wondered why 2 uF across the load is sometimes quoted in stability measurements. How close is that to any practical circumstance? I have never come across loudspeaker cable capacitances > about 15 nF in domestic circumstances. (Excluding electrostatics. I understand one uses a series resistance there.) Perhaps someone can broaden my horizons.
Regards
What can happen is that those oscillations can cause the amplifier to overload at a slightly lower output - but that will be a few % and inconsequential.
Then once more, let us please stop considering the damping factor in these discussions. It appears to be forgotten that the real damping factor depends on the total resistance (impedance) in the loudspeaker circuit. The main part of this is the 5 - 6 ohm odd resistance of the voice coil. I wonder if the series coil resistance (or impedance at audio frequencies) will be even 1% of this.
I also wondered why 2 uF across the load is sometimes quoted in stability measurements. How close is that to any practical circumstance? I have never come across loudspeaker cable capacitances > about 15 nF in domestic circumstances. (Excluding electrostatics. I understand one uses a series resistance there.) Perhaps someone can broaden my horizons.
Regards
Johan Potgieter said:
These are all very good questions. It is important to separate the possible linear and nonlinear effects of the coil, and to understand under what conditions certain input stimuli are relevant.
It is difficult to get excited about the linear effects of frequency response and phase response of a uH or two, when it is realized that such effects of, for example attenuation at 20 kHz, can be just as easily caused by the amplifier's input RFI filtering or its natural rolloff due to its finite bandwidth. Who can get excited about a tenth of a dB at 20 kHz when there are so many ways for such a rolloff to happen?
Obviously, it is important that any such output inductor not be nonlinear. If it is nonlinear, it can cause HF IM distortion whose products may lie in the audible range. But these should be readily measurable on a distortion test, at least I would think.
Non-band-limited square waves as input are entirely appropriate for stability analysis, but they may not be appropriate for drawing conclusions about the audibility of an output coil, especially if the rate of change of the signal causes slew rate limiting or excessive edge currents. Keep in mind that if you confine your listening to CDs, then the fastest possible signal that can be produced is a full-amplitude sinewave at something a bit less than 20 kHz. This, due to the antialias/reconstruction filter in the CD player. CD's are certainly not the only source, but they are probably in the great majority (sadly so, due to the lack of software for alternative formats like SACD).
Cheers,
Bob
I've yet to see the CD with full amplitude 20kHz signals, unless it's one of my test CDs. The situation with real music is far less severe. I can't hear 20kHz anyway, nor can hardly anybody else on this forum who's made it beyond their teenage years. IMO, the output inductor is the least of my worries. Though I like the usual bench tests, and simulation as well, my preference is for comparing the input with an attenuated version of the output. This will show up anything that isn't buried in the noise floor, but has the flaw (if it is a flaw) of being very sensitive to phase shifts. Thus, I like to see zero phase shift in the audio band, just for my own convenience. No idea if it's audible. My bench tests to date have shown me that an output inductor of a few uH, is of no consequence, but the input network and the LF rolloff from the cap in the feedback network, are significant. The phase shift from that stupid cap goes way up into the audio band. Sometimes I test with a passive preamp, and rolloff from cable capacitance can get involved too. IMO, the output inductor threads have been interesting because the issue needs to be looked at, but in the end the direct effects of the thing are more theoretical than practical. The stability issues are not, so I'll keep using 'em until I prove a given design can't possibly go unstable without it.
Johan Potgieter said:
As I remember, I did not speak about audibility, but stability in that post.
My words on audibility were those: " Though I do not use output coils, I believe that a coil of 2uH in parallel with damping resistor of few ohms will not be audible, with regard to speaker cable inductance and coil values used in passive crossover. A phase shift added by 2uH coil is IMHO negligible in audio band. " (post no. 106)
Conrad Hoffman said:
I agree on all points.
For comparing the input with an attenuated version of the output, one can easily get a pretty good phase match by putting a single high-frequency adjustable pole in the input side to the comparison. See the reference to the Distortion Magnifier on my web site at www.cordellaudio.com. That description will soon be updated with more detail.
Cheers,
Bob
PMA said:
My apology, PMA.
I simply used your handy square wave graph as a picture to illustrate what I had in mind about super-audio oscillations in such a display in general. I did not imply any earlier comment by you, and none of my post had you specifically in mind. As far as I can recall I was always in agreement with your views.
Regards
Hi Bob,
Thanks for reply (your #112). Yes, I had linear coils in mind (I use air-core, mounted such that I hope they act linear!). Also regarding testing with fast rise-time square waves. Mmmmm .... I suppose you are correct; even if no such signals are contained in the intended input, something can come along to upset the system. That, naturally, must not happen, stability-wise, slew-wise or whatever-wise. Only I sometimes read that reviewers et al draw conclusions such as "there are bothering oscillations in the square wave test, which can do &^@#$* to the signal..." But then, oh how does one read many things! Journalists [censored]!!
Regards
Thanks for reply (your #112). Yes, I had linear coils in mind (I use air-core, mounted such that I hope they act linear!). Also regarding testing with fast rise-time square waves. Mmmmm .... I suppose you are correct; even if no such signals are contained in the intended input, something can come along to upset the system. That, naturally, must not happen, stability-wise, slew-wise or whatever-wise. Only I sometimes read that reviewers et al draw conclusions such as "there are bothering oscillations in the square wave test, which can do &^@#$* to the signal..." But then, oh how does one read many things! Journalists [censored]!!
Regards
Bob, I looked at the phase correction network circuit, but you've obviously grossly underestimated my laziness. It's easier to fix the one or two parts that degrade the amp, than to use an input network. Good grief, if I had an input network, I'd want to put it in a box. Then, I'd want knobs on the pots. And connectors. Where would it all end? I suppose a more serious argument is that I'd also have to use capacitors that a tweako could claim negate the test. If I can show that a simple attenuated signal matches the amp output right down into the noise floor, with music as a signal, that's a pretty good argument that the amp can't possibly be affecting the signal.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.