Lars Clausen said:
And that is the, whatever.
If you have the equipment to do it then it might be nice to see what happens at various levels for say a 1KHz plus 5KHz input for an air cored inductor and a poorly designed inductor.
I've just made those frequencies up but I'd expect that, as the level increases, you'll see more intermodulation products at the output.
The plot you give 'later' is for a single tone?
Mind you, things might get mixed up elsewhere because there might be something else going on with the available slew rate from the amplifier..... but I would 'hope' there would be a measurable difference.
I mean the 'good' inductor would show something but the 'bad' inductor would show something worse.
You are right.
If I did not think there was something in what I am trying to say I would ignore myself as well.
DNA
I thought that it was absolutely obvious that the phenomena shown in these plots was not in any way saturation. You don't seem to be familiar with inductor measurement and characterization.
Saturation manifests itself as a progressive reduction in inductance with time, and thus, an increase in the current versus time slope, while these plots show exactly the OPPOSITE effect.
That's hysteresis, which manifets itself as an unexpectedly reduced inductance after each polarity inversion of the voltage applied to the coil. Then, inductance increases progressively until the steady state value is reached, which is shown as a reduction of the current versus time slope.
By the way, ripple current was barely +/-1A in these measurementes (0.5A per division), there was no DC bias at all, and the measured saturation current of these inductors is around 10A, so saturation was just not possible anyway.
I will try to introduce a strong DC bias to cause saturation and I will post the waveforms, so that you can learn how to identify both phenomena with the oscilloscope.
Saturation manifests itself as a progressive reduction in inductance with time, and thus, an increase in the current versus time slope, while these plots show exactly the OPPOSITE effect.
That's hysteresis, which manifets itself as an unexpectedly reduced inductance after each polarity inversion of the voltage applied to the coil. Then, inductance increases progressively until the steady state value is reached, which is shown as a reduction of the current versus time slope.
By the way, ripple current was barely +/-1A in these measurementes (0.5A per division), there was no DC bias at all, and the measured saturation current of these inductors is around 10A, so saturation was just not possible anyway.
I will try to introduce a strong DC bias to cause saturation and I will post the waveforms, so that you can learn how to identify both phenomena with the oscilloscope.
Eva: If you measured with a load resistor, and a square wave as input, it could also be the normal output of an LR filter we are looking at. It's hard to say, since you don't really tell us exactly what your measurements show.
quote:
Me too i didn't really look at the plots before i burt out in a comment 😀
quote:
Me too i didn't really look at the plots before i burt out in a comment 😀
So, are you saying I'm not stupid.....
I wish you would sort youselves out.
I'll settle for confused.
DNA
I wish you would sort youselves out.
I'll settle for confused.
DNA
All my measurement setups are more or less compensated to cancel the inductance of the measurement resistor and other errors, so the current waveforms that I show can be considered quite accurate. Also, yhe class D circuits from the previous pirctures was idle (no signal) and with the output unloaded).
I can't cause the inductors to saturate with my class D prototype because the voltage amplifier is currently designed to clip with a +/-10A command, so that it can't ask the current amplifier to produce more than 10A output, and the inductors saturate above 10A.
However, I have prepared these two figures with my "poor-woman inductor curve tracer", I hope them to help making things clear.
And:
This may also help to understand why THD rises for higher audio frequencies with most gapped ferrites (not to talk about iron powder). It's just hysteresis kicking in 🙂
I can't cause the inductors to saturate with my class D prototype because the voltage amplifier is currently designed to clip with a +/-10A command, so that it can't ask the current amplifier to produce more than 10A output, and the inductors saturate above 10A.
However, I have prepared these two figures with my "poor-woman inductor curve tracer", I hope them to help making things clear.
An externally hosted image should be here but it was not working when we last tested it.
And:
An externally hosted image should be here but it was not working when we last tested it.
This may also help to understand why THD rises for higher audio frequencies with most gapped ferrites (not to talk about iron powder). It's just hysteresis kicking in 🙂
OK now i understand your curves 🙂 Thanks!
But if hysterisis is worse at low current crossings, shouldn't the resulting THD rise (according to your theory) at higher freq's be more noticable at low powers, and less at high powers?
All the best from
Lars
PS Do you make a ClassD amplifier as a product?
But if hysterisis is worse at low current crossings, shouldn't the resulting THD rise (according to your theory) at higher freq's be more noticable at low powers, and less at high powers?
All the best from
Lars
PS Do you make a ClassD amplifier as a product?
According to the waveforms that I have seen, hysteresis does not seem to be particularly dependent on DC bias (if we can consider the LF audio current component as DC while analysing things in a switching cycle per cycle basis). As I mentioned, hysteresis effects appear just every time the polarity of the voltage applied to the coil is reversed, and that causes the inductance to be unexpectedly low for a few microseconds. Then inductance reverts progressively to its normal value. These few microseconds of non-linearity become negligible at low frequencies, but important at 10Khz.
I'm currently working on my first class D commercial project, but it's not audio related. Further projects may include audio amplifiers, though.
I'm currently working on my first class D commercial project, but it's not audio related. Further projects may include audio amplifiers, though.
No, in fact not, well maybe or I don't know.
I can see that a low level signal riding on a high level one might have something done to it at the top end of the plot.
In the same way I can see how a similar thing might happen at low levels. From your plots there is a non-linearity at low levels.
The problem for me is that you are plotting a 'full' excursion for the material.
If you operate at 'low' levels then you are using a minor loop of the material and the effect you are seeing from your plots should be reduced.
OK, it's still there... but....
Things are more complicated than I might be able to understand.
Anyway
I'm off to bed.
Have a nice one yourself.
DNA
I can see that a low level signal riding on a high level one might have something done to it at the top end of the plot.
In the same way I can see how a similar thing might happen at low levels. From your plots there is a non-linearity at low levels.
The problem for me is that you are plotting a 'full' excursion for the material.
If you operate at 'low' levels then you are using a minor loop of the material and the effect you are seeing from your plots should be reduced.
OK, it's still there... but....
Things are more complicated than I might be able to understand.
Anyway
I'm off to bed.
Have a nice one yourself.
DNA
Eva: Ok makes sense. I'm not so sure how it would affect an audio signal, but i will make some more compares of ferrite vs air core in the next couple of days, and post here.
Just so i understand correctly, you are saying that the contributing THD should be the same at high and low powers alike? But would only be depeding on the LF signal frequency? (Say 10 kHz vs 1 kHz).
Best regards
Lars
Just so i understand correctly, you are saying that the contributing THD should be the same at high and low powers alike? But would only be depeding on the LF signal frequency? (Say 10 kHz vs 1 kHz).
Best regards
Lars
I would worry about the stray field accompanied with air core....
Back to AD1994, I don't see the datasheet is cheating by measuring THD in 1W. To me the THD+Noise vs Power plot suggests very low noise floor. In fact other chips like TA2022 and TDA8420 measure worst in low power.
However the AD1994 was measured with internal PGA set to 0dB gain, that means the overall gain is only 14dB. This might be the reason for such a good noise performance 🙄
Back to AD1994, I don't see the datasheet is cheating by measuring THD in 1W. To me the THD+Noise vs Power plot suggests very low noise floor. In fact other chips like TA2022 and TDA8420 measure worst in low power.
However the AD1994 was measured with internal PGA set to 0dB gain, that means the overall gain is only 14dB. This might be the reason for such a good noise performance 🙄
Agreed. 800 pixels wide would be fine. 1300 means not being able to see the whole thing easily.
Thanks
Thanks
I haven't done any analysis of the subject "class-d and core hysteresis". But my gut feeling tells me that there might indeed be a difference in susceptibility to hysteresis effects between carrier-based and delta-sigma class-d amps.
The carrier-based ones might be less susceptible to it since you have an intrinsic constant (well - more or less) RF bias signal a bit like the one used in tape recorders.
Delta-sigma OTOH has a very irregular switching pattern which might be less useful in this respect.
But one advantage of delta-sigma is that it is less susceptible to timing-errors.
For those interested in the working-principles of delta-sigma modulators (and not willing to read Norsworthy, Schreier & Temes) there was once a very good article in the JAES: "One Bit Audio: An overview" p 166, vol 52, March 2004
Regards
Charles
The carrier-based ones might be less susceptible to it since you have an intrinsic constant (well - more or less) RF bias signal a bit like the one used in tape recorders.
Delta-sigma OTOH has a very irregular switching pattern which might be less useful in this respect.
But one advantage of delta-sigma is that it is less susceptible to timing-errors.
For those interested in the working-principles of delta-sigma modulators (and not willing to read Norsworthy, Schreier & Temes) there was once a very good article in the JAES: "One Bit Audio: An overview" p 166, vol 52, March 2004
Regards
Charles
I can't resize that kind of pictures because some lines will disappear and the plots will become unreadable. I don't have trouble watching them at 1024x768 as I use the cursors to scroll. Note that the part of the picture that reflects the oscilloscope screen fits well in a 1024x768 display.
Eva said:
I don't seem to be able to post pictures but here are your originals resized to 600X483 using Xnview and saved with 64 colours. Don't know if you can update your posts but here they are....
t_sine21.gif
t_sine22.gif
t_sine23.gif
t_sine24.gif
DNA
Thanks DNA. I may have to get that image software as it seems to include some special algotithm to resize pictures with lines and fonts.
By the way, does anybody know an easy way to machine ferrites? I have just found out the hard way (measuring!) that a big 6mm air gap in the center leg of a pair of E42/20 cores produces about 10 times less stray magnetic fields than the smaller 3.8mm gap in all the legs required to keep the same permeability (and similar saturation current). Furthermore, the usual 2:1 criteria for center leg gap versus global gap is only met for energy storage figures (3mm in all legs produces similar I^2*L product than 6mm in center leg), but not for permeabilty.
I have also found out that ground planes seem to provide only a low-pass-like magnetic shielding (by means of acting as a shorted turn coupled to the leakage inductance of all the PCB tracks on the other side), and flux bands provide a similar effect but less noticeable (quite marginal actually).
Damn, I have to find a way to create custom center-leg air gaps...
By the way, does anybody know an easy way to machine ferrites? I have just found out the hard way (measuring!) that a big 6mm air gap in the center leg of a pair of E42/20 cores produces about 10 times less stray magnetic fields than the smaller 3.8mm gap in all the legs required to keep the same permeability (and similar saturation current). Furthermore, the usual 2:1 criteria for center leg gap versus global gap is only met for energy storage figures (3mm in all legs produces similar I^2*L product than 6mm in center leg), but not for permeabilty.
I have also found out that ground planes seem to provide only a low-pass-like magnetic shielding (by means of acting as a shorted turn coupled to the leakage inductance of all the PCB tracks on the other side), and flux bands provide a similar effect but less noticeable (quite marginal actually).
Damn, I have to find a way to create custom center-leg air gaps...
Hi Eva,
Bruno was saying before he ground his own 1mm air gaps with the use of a dremel mounted in a drill stand, and taking his time with it. Maybe two cuts to his one will get you your 6mm. Careful with those cutting wheels though, they blow right apart easily enough.
Regards,
Chris
PS: you can use MSPaint to resize, just click on "image" in the menu bar, stretch/skew, type in a percentage for vertical and horizontal, save as gif.
Bruno was saying before he ground his own 1mm air gaps with the use of a dremel mounted in a drill stand, and taking his time with it. Maybe two cuts to his one will get you your 6mm. Careful with those cutting wheels though, they blow right apart easily enough.
Regards,
Chris
PS: you can use MSPaint to resize, just click on "image" in the menu bar, stretch/skew, type in a percentage for vertical and horizontal, save as gif.
- Status
- Not open for further replies.