Seems I cut through it with a dull hack saw blade, powdered iron, would that be at all useable?
If you can cut through it with a hack saw, dull or not, it's probably not ferrite...
Powdered iron cores are normally (exceptions are Toko inductors specially made for class D) slow saturating, which is good for DC but not for audio.
You can predict the difference from the circuit function. If the core is used in a choke (1 winding) it's bound to be p.i. If it's a CM choke it's ferrite. There must be a few of the latter sort as well in the PSU you're umm recycling.
Powdered iron cores are normally (exceptions are Toko inductors specially made for class D) slow saturating, which is good for DC but not for audio.
You can predict the difference from the circuit function. If the core is used in a choke (1 winding) it's bound to be p.i. If it's a CM choke it's ferrite. There must be a few of the latter sort as well in the PSU you're umm recycling.
I once used RM cores for output inductors. The airgap was made by the insertion of insualtion material into the outer "legs" of the core. The total gap length is then twice the insualtion thickness.
This is quite an easy method but I don't know how it behaves regarding distortion actually. The inner gap may possibly give some disadvantages.
For better understandability I added a simple cross-sectional drawing. The two core halves are in black, the winding in red and the insulation material in blue.
Regards
Charles
This is quite an easy method but I don't know how it behaves regarding distortion actually. The inner gap may possibly give some disadvantages.
For better understandability I added a simple cross-sectional drawing. The two core halves are in black, the winding in red and the insulation material in blue.
Regards
Charles
Attachments
Thanks for that tip. They're all P.I. ....one was plastic? I'll have to pick one up that's all.
phase_accurate said:
Spacing the cores creates a strong external field around the gaps (just like with a gapped toroid - the idea of which I'm not so fond, EMC-wise).
RM cores are very useful but you should grind the gap in the inner leg and keep the outside closed.
An RM10 will work quite perfectly up to about 400W (500 if you stretch it).
So my RM12 for 250 watts was a little exaggerated ?!
BTW: I got some EFD cores recently that are available from the manufacturer with an airgap.
Regards
Charles
Improved feedback loop
After the encouraging input from Bruno and Charles, I have made some updates to my prototype circuit to include an OP in the loop. It still oscillates nicely at a similar frequency (actually slightly higher; about 400kHz).
However, I have not yet achieved something like Charles' LAG filter, since there's an ideal amplifying element with A=100, and I have not tried to figure out how to translate that into a concrete circuit yet. Nevertheless, now I am more confident that it can be done; like you said Charles, it's a matter of doing the math to check the poles and zeroes...
The circuit which I now have on the breadboard is ucd_js3a in the attached archive. It already is an improvement, both subjectively and measurably. It can deliver a whole Watt into 8ohm with around, or less than, 1% THD over the audio range.
That said, I wonder if this is what Bruno had in mind with his mischieveous little comment that an integrator could be included (but not saying how).
An alternative is presented in the ucd_js3b circuit; there feedback is taking two paths; one is "normal" LF-feedback to the input OP/integrator, and one is oscillation-sustaining feedback to the positive input of the comparator (HF only).
Apparently, according to the simulation, it "works" (it oscillates, and appears to be doing something useful). I haven't tried it in real life yet though.
Question is: is this insane, or useful, or what?
Feedback please 🙂.
Regards / Johan
After the encouraging input from Bruno and Charles, I have made some updates to my prototype circuit to include an OP in the loop. It still oscillates nicely at a similar frequency (actually slightly higher; about 400kHz).
However, I have not yet achieved something like Charles' LAG filter, since there's an ideal amplifying element with A=100, and I have not tried to figure out how to translate that into a concrete circuit yet. Nevertheless, now I am more confident that it can be done; like you said Charles, it's a matter of doing the math to check the poles and zeroes...
The circuit which I now have on the breadboard is ucd_js3a in the attached archive. It already is an improvement, both subjectively and measurably. It can deliver a whole Watt into 8ohm with around, or less than, 1% THD over the audio range.
That said, I wonder if this is what Bruno had in mind with his mischieveous little comment that an integrator could be included (but not saying how).
An alternative is presented in the ucd_js3b circuit; there feedback is taking two paths; one is "normal" LF-feedback to the input OP/integrator, and one is oscillation-sustaining feedback to the positive input of the comparator (HF only).
Apparently, according to the simulation, it "works" (it oscillates, and appears to be doing something useful). I haven't tried it in real life yet though.
Question is: is this insane, or useful, or what?
Feedback please 🙂.
Regards / Johan
Bruno Putzeys said:
Thanks for the positive feedback!phase_accurate said:
However, "roughly" certainly fits well. I have modified the protoboard circuit to implement the 3b variant, but I'm finding that it doesn't work as "clean-cut" in real-life as it did in the simulated circuit. Problems with overall stability, or spurious "mini-cycles" caused by the comparator getting conflicting messages from two directions. I have got the circuit working, with slightly modifified feedback networks compared to the circuit I posted earlier. But then the nice load-independence turned out to be partly lost, so I have some more work to do before it's useful.
I will try to do the math, since there's many more variables involved now. (Two feedback filters, and the integrator constant). But hints & tips are gratefully received as well 🙂
The ones I posted are from PSpice.Bricolo said:
Shall I take the fun out of it ????? 😀 
Yes I will !!!
The UCD's response at the upper end is basically 1st order (at least in close proximity to the cutoff-frequency) due to the first order feedback function.
While we want a phase-margin of zero degrees at the switching frequency of the UCD itself, we don't want to introduce a 2nd outer loop with zero phase margin (in the worst case at a different frequency).
So you first have to build an ordinary UCD that has not only a cap for feedback but the usual cap/resistor combination.
If we add an outer loop with an integrator we have to watch out that the overall loop gain is a first order function.
We will therefore have to stop the integrator's gain from falling at the cutoff frequency of the original UCD. This is done by inserting a zero. Or in other words: The outer integrator's cap needs a series resistor !
This circuit is called a PI controller (Proportional-Integral). If you add a parallel resistor to said series RC you have the LAG filter again. And this is what I would use since I'd like to get constant loop gain over the audio range (or at least up to 5- 10 kHz).
Regards
Charles
Yes I will !!!
The UCD's response at the upper end is basically 1st order (at least in close proximity to the cutoff-frequency) due to the first order feedback function.
While we want a phase-margin of zero degrees at the switching frequency of the UCD itself, we don't want to introduce a 2nd outer loop with zero phase margin (in the worst case at a different frequency).
So you first have to build an ordinary UCD that has not only a cap for feedback but the usual cap/resistor combination.
If we add an outer loop with an integrator we have to watch out that the overall loop gain is a first order function.
We will therefore have to stop the integrator's gain from falling at the cutoff frequency of the original UCD. This is done by inserting a zero. Or in other words: The outer integrator's cap needs a series resistor !
This circuit is called a PI controller (Proportional-Integral). If you add a parallel resistor to said series RC you have the LAG filter again. And this is what I would use since I'd like to get constant loop gain over the audio range (or at least up to 5- 10 kHz).
Regards
Charles
Hello all.
I have retaken the forum after a few months out.
Please, could you tell me where to the the referred UCD pdf file in the Hypex website?
Thanks!
I have retaken the forum after a few months out.
Please, could you tell me where to the the referred UCD pdf file in the Hypex website?
Thanks!
Thanks.
And congratulations, Bruno.
BTW: I would like to hear you what do you think about "traditional" or oscillator based Class D amplifiers (as opposed to self-oscillating ones like UCD or Sigma-Delta). Do you think that its performance (in terms of THD, etc) can be better with the proper components).
Some say that self-switching is better for audio quality.
And congratulations, Bruno.
BTW: I would like to hear you what do you think about "traditional" or oscillator based Class D amplifiers (as opposed to self-oscillating ones like UCD or Sigma-Delta). Do you think that its performance (in terms of THD, etc) can be better with the proper components).
Some say that self-switching is better for audio quality.
There's no such thing as a simple truth I'm afraid.
Fixed frequency amplifiers are a bit easier in terms of stability at large signal swings, at least if you're making loops of order >2.
The drawbacks are that oscillator noise is a factor and that they don't lend themselves as easily to loops with such strong output filter control as UcD or Mueta. Self-oscillating amps have somewhat higher loop gain available compared to fixed frequency amps.
In both cases you need to know your control theory 🙂
When an amplifier is operated inside an audio system comprising an AM tuner, control over the switching frequency is also nice. We fix that by mixing a clock signal in the audio when the AM tuner is used. The amplifier locks to the carrier but I won't vouch for THD in that mode (loop gain is decreased).
Fixed frequency amplifiers are a bit easier in terms of stability at large signal swings, at least if you're making loops of order >2.
The drawbacks are that oscillator noise is a factor and that they don't lend themselves as easily to loops with such strong output filter control as UcD or Mueta. Self-oscillating amps have somewhat higher loop gain available compared to fixed frequency amps.
In both cases you need to know your control theory 🙂
When an amplifier is operated inside an audio system comprising an AM tuner, control over the switching frequency is also nice. We fix that by mixing a clock signal in the audio when the AM tuner is used. The amplifier locks to the carrier but I won't vouch for THD in that mode (loop gain is decreased).
Simulation results of differential ucd amplifier half bridge with audiophile quality 741 op amp inputs and 2n3904's/2n3906's+ irf511's.
At 100mV 1Khz
DC COMPONENT = 9.400898E-03
TOTAL HARMONIC DISTORTION = 1.907986E+00 PERCENT
TOTAL POWER DISSIPATION 4.94E-01 WATTS
HARMONIC FREQUENCY FOURIER NORMALIZED PHASE NORMALIZED
NO (HZ) COMPONENT COMPONENT (DEG) PHASE (DEG)
1 1.000E+03 1.107E+00 1.000E+00 1.785E+02 0.000E+00
2 2.000E+03 1.041E-02 9.402E-03 9.024E+01 -2.668E+02
3 3.000E+03 1.022E-02 9.230E-03 9.161E+01 -4.440E+02
4 4.000E+03 1.021E-02 9.225E-03 9.614E+01 -6.180E+02
5 5.000E+03 1.136E-02 1.026E-02 9.841E+01 -7.942E+02
1volt 1khz input
DC COMPONENT = 4.181206E-02
TOTAL HARMONIC DISTORTION = 1.933344E+00 PERCENT
HARMONIC FREQUENCY FOURIER NORMALIZED PHASE NORMALIZED
NO (HZ) COMPONENT COMPONENT (DEG) PHASE (DEG)
1 1.000E+03 1.222E+01 1.000E+00 1.792E+02 0.000E+00
2 2.000E+03 1.277E-02 1.045E-03 1.252E+02 -2.332E+02
3 3.000E+03 2.233E-01 1.826E-02 3.210E+00 -5.344E+02
4 4.000E+03 7.077E-02 5.790E-03 -1.330E+02 -8.498E+02
5 5.000E+03 2.897E-02 2.370E-03 -2.296E+01 -9.189E+02
Graphs attatched.
Not finished with it yet.
Chris
At 100mV 1Khz
DC COMPONENT = 9.400898E-03
TOTAL HARMONIC DISTORTION = 1.907986E+00 PERCENT
TOTAL POWER DISSIPATION 4.94E-01 WATTS
HARMONIC FREQUENCY FOURIER NORMALIZED PHASE NORMALIZED
NO (HZ) COMPONENT COMPONENT (DEG) PHASE (DEG)
1 1.000E+03 1.107E+00 1.000E+00 1.785E+02 0.000E+00
2 2.000E+03 1.041E-02 9.402E-03 9.024E+01 -2.668E+02
3 3.000E+03 1.022E-02 9.230E-03 9.161E+01 -4.440E+02
4 4.000E+03 1.021E-02 9.225E-03 9.614E+01 -6.180E+02
5 5.000E+03 1.136E-02 1.026E-02 9.841E+01 -7.942E+02
1volt 1khz input
DC COMPONENT = 4.181206E-02
TOTAL HARMONIC DISTORTION = 1.933344E+00 PERCENT
HARMONIC FREQUENCY FOURIER NORMALIZED PHASE NORMALIZED
NO (HZ) COMPONENT COMPONENT (DEG) PHASE (DEG)
1 1.000E+03 1.222E+01 1.000E+00 1.792E+02 0.000E+00
2 2.000E+03 1.277E-02 1.045E-03 1.252E+02 -2.332E+02
3 3.000E+03 2.233E-01 1.826E-02 3.210E+00 -5.344E+02
4 4.000E+03 7.077E-02 5.790E-03 -1.330E+02 -8.498E+02
5 5.000E+03 2.897E-02 2.370E-03 -2.296E+01 -9.189E+02
Graphs attatched.
Not finished with it yet.
Chris
phase_accurate said:
In my efforts to more thoroughly understand UcD behaviour, I took the formulas and all the component values from the patent text, and adapted them to
a) be plotted/examined using gnuplot, and b) augment with an extra loop filter.
Attached are two plots showing first a re-construction of one graph from the patent and more; closed-loop gain, open-loop gain, and phase-variation of the same. In ucd_loop.gif, it can be seen that the open-loop phase shift is 180 degrees at ~415kHz (you can't see it exactly, but I could by zooming in 🙂).
The blue line represents the amplification factor applied to noise generated from the switching stage. (So, the lower down that line is in the audio range, the better!)
Included are also the plot-script files (*.plt) that I used with gnuplot 3.73.
The other plot, ucd_js3a_plot.gif, shows the circuit augmented with Charles' LAG filter.
(An updated picture of the corresponding schematic is included.)
Curiously, this circuit also oscillates at ~415kHz, which is just a coincidence.
Here we can see some modest gains from inclusion of the extra loop filter; the open-loop gain is higher, resulting in more suppression of switching noise.
Charles: why do you feel that constant open-loop gain in the audio range is a benefit?
If we crank up the LF gain of the LAG filter, we get more open-loop gain, and better suppression of noise
at low frequencies. (The blue line can be suppressed to around 0.001 @ 1kHz). Isn't that a Good Thing?
This is as far as I've got today; I have not yet (seriously) addressed the 3b variant (the one with the more complex loop topology).
Bruno: to make that variant work well, is it not a pre-requisite that the outer loop's response has an absolute value well below the absolute value of the inner loop at the oscillation frequency?
In other words, should one be thinking in terms of LP-filtering in the feedback of the outer loop?
Because such an amp will be much closer to a non-feedback design soundwise.
Because the inner loop is already a part of the outer loop, its intrinsic lowpass function should suffice for that purpose. If you add another lowpass you would just have to compensate for it in another place, in order to keep the good phase-margin, which makes your additional LPF obsolete.
Regards
Charles
- Status
- Not open for further replies.