Magnatec class d amp sulution

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
The most common misconception is that people are believing that the response to a pulse can't immediately be seen at the output of a filter. And this is WRONG.

If it really were like that, mankind wouldn't have flown to the moon, there were no modern aircraft with autopilot and much less of electronic assistance in cars etc etc

Regards

Charles
 
mikeks said:
...how do you class-d folks measure loop gain in your systems?
It's quite pointless to measure loop gain. Due to the sampled nature of a class D amp, you design for an exactly known and desired loop gain.

First you calculate what loop gain characteristic you want. Then you design the loop filter to do precisely that. Then you build your amp. If the closed-loop gain, THD and/or output impedance aren't what you expected, you've done something wrong. Conversely, if the measurements do match the calculations, loop gain is certain to match the design within a fraction of a dB.
 
phase_accurate said:
Maybe it is the gain of the modulator/switching stage combination that Mike is curious about.
For the classic PWM amp it is simply the ratio of output_supply_voltage/peak_voltage_of_triangular_reference for instance.
If the rule is generalised to use dV/dt (at the comparator inputs) and switching frequency instead of pp amplitude of a triangle wave, it goes for non-classic-PWM amps as well. Again it's not something you measure. Mathematical truths don't need physical verification.
 
Bruno Putzeys said:

It's quite pointless to measure loop gain. Due to the sampled nature of a class D amp, you design for an exactly known and desired loop gain.


Assuming your closed loop class d system takes an analog input, and outputs a nominal analog signal, then surely the applied feedback is necesarily voltage (shunt) derived, voltage (series) applied (AKA 'voltage' feedback).

Why would you consider it pointless to measure loop gain in such a system...?

One would have thought that in view of the significant phase shifts introduced by the output filter, (for feedback sampled after the filter), judicious measurement, or at least SPICE simulation of the loop transmission path would allow one to compensate for the filter's poles with precisely placed zero's in the loop..

...This should allow feedback to be maximized without compromising stability....
 
mikeks said:

One would have thought that in view of the significant phase shifts introduced by the output filter, (for feedback sampled after the filter), judicious measurement, or at least SPICE simulation of the loop transmission path would allow one to compensate for the filter's poles with precisely placed zero's in the loop..
The output filter is an integral part of this mathematical study, as it is an integral part of loop gain. It is surprising that apparently you were assuming I forgot about this.
Poles in the filter become zeros in the error transfer function as a matter of fact!

Ergo:no such thing as explicitly compensating poles by zeros need be done. If it were, it would be impossible to do, as merely changing the load would obliterate the compensation scheme. So, we must thank mathematics for the fact that it happens automatically.
Class D amplifiers with correctly designed loops are extremely insensitive to changes in load or filter values.

I would like to invite you to perform simulations to see this for yourself. You will also appreciate the fact that to try and find the correct transfer function, by measuring actual hardware or even by simulating it, is a hopeless endeavour. The reason why so relatively few amplifiers with good post-filter feedback are made, is precisely because designers are afraid of doing the analysis first.

You will understand therefore that the phrase "to compensate for the filter's poles with precisely placed zero's in the loop" suggests you yourself are rather new to the subject of class D loop control. This is not a problem - I welcome anyone willing to become acquainted with the subject - but to try and teach me (or charles) how to roll a control loop while obviously being unexperienced at it oneself is not expedient to a fruitful exchange.

I hope you will understand my mild irritation and rest assured that no offense is meant.
 
Bruno Putzeys said:


You will understand therefore that the phrase "to compensate for the filter's poles with precisely placed zero's in the loop" suggests you yourself are rather new to the subject of class D loop control. This is not a problem - I welcome anyone willing to become acquainted with the subject - but to try and teach me (or charles) how to roll a control loop while obviously being unexperienced at it oneself is not expedient to a fruitful exchange.

I hope you will understand my mild irritation and rest assured that no offense is meant.

Actually, a working analog/digital class-d design was the subject of my dissertation many years ago....(scored 80% overall by the way...which was the highest mark attained on that course for such since it's inception)....

This included an indepth quantitative analysis of control loop (DSP and analog) schemes......including practical verification.....

Indeed, placing singularities in the feedback path to ameliorate those generated by the filter in the foward path has been quantitatively and practically demonstrated in JAES....I just cannot remember the author of the article at the moment.... :scratch:
 
mikeks said:
Indeed, placing singularities in the feedback path to ameliorate those generated by the filter in the foward path has been quantitatively and practically demonstrated in JAES....I just cannot remember the author of the article at the moment.... :scratch:
Indeed, the degree of complication people are willing to go through while missing the obvious is a constant feature of class D design. Especially published papers tend to bear this out.
 
Hi Bruno,

A simple low THD design would of course be interesting....

Complexity versus relatively poor linearity is the most discouraging facet of full-range class-d arrangements...

Simpler concoctions may be implemented for sub-woofer applications of course....

In respect of audio power amplifiers, i am a linear class-B/AB person really, but would be interested in your class-d design philosophy....and the sort of linearity obtained by its implementation...
 
mikeks said:
A simple low THD design would of course be interesting....
In respect of audio power amplifiers, i am a linear class-B/AB person really, but would be interested in your class-d design philosophy....and the sort of linearity obtained by its implementation...

Have a look around the threads that bear the word "UcD" in their title and check out www.hypex.nl who sell modules based on the concept. A spec sheet with plots is available there and will show the smoothest and most load-invariant frequency response you'll have ever seen since long. Effective output impedance is 20mohm+500nH, significantly lower than that of the boucherot networks typically found on linear amps.

Lower THD figures (-110dB) are readily realised, but so far the improvement on sound quality produced by this has proved zip (not a surprise considering loudspeaker distortion), so for the time being there are no plans offer these "improved" modules commercially.

Cheers,

Bruno
 
Bruno Putzeys said:
....Lower THD figures (-110dB) are readily realised......


You'll forgive me for taking this with a vast pinch of salt...

....at what frequency is this figure quoted?

...and of even greater importance, what was the measurement bandwidth?

.....If the measurement bandwidth was the typical 80KHz used for linear amps., and we assume you mean -110db at 1KHz, then this class-d amp. outperforms the majority of linear class-ab designs...

Even assuming that residual carrier signal at the class-d amp's output did not cause the input stages of of the distortion analyzer to slew limit, this figure is unlikely in the extreme....
 
Mikeks,

Hereby a measurement of a UcD Class-D amplifier with an extra integrator loop. Measurement bandwith was 22kHz.

This is not implemented in a product, but a test to take a look how low we can get the THD......

Regards,

Jan-Peter

www.hypex.nl
 

Attachments

  • graphic1.gif
    graphic1.gif
    24.2 KB · Views: 333
mikeks said:
You'll forgive me for taking this with a vast pinch of salt...

....at what frequency is this figure quoted?

...and of even greater importance, what was the measurement bandwidth?

.....If the measurement bandwidth was the typical 80KHz used for linear amps., and we assume you mean -110db at 1KHz, then this class-d amp. outperforms the majority of linear class-ab designs...

Even assuming that residual carrier signal at the class-d amp's output did not cause the input stages of of the distortion analyzer to slew limit, this figure is unlikely in the extreme....

This figure was quoted at 1kHz. Measurement BW is 22kHz, but you will recognise that going to 80kHz would make noise dominate the measurement, but not add any further harmonics to the measurement. In fact, for this reason, the 110dB pertained to THD, not THD+N.

Concerning slew-rate limiting on distortion analyser inputs, you will understand that this would require a very large residual.

The AP analyser will analyse at least up to 200kHz full-scale audio signals, The carrier is at 400kHz. It should not elicit any slew rate problems as long as it is better than 6dB suppressed. It is suppressed by better than 40dB.

AP's do potentially have a problem with the presence of a carrier, in that the autoranging circuits base their decisions on the full-bandwidth signal. At low amplitudes the carrier will force an unnecessarily high gain setting, adversely affecting the device's resolution at low signal levels. The AES17 pre-filter used to measure D/A converters (most of which have large outband noise levels causing similar problems) remedies this.

On the AP2 this problem is not overly dramatic. JP's plot was made without the use of an AES17 filter.
 
mikeks,

You seem to doubt that Class D amps can perform technically if not sonically. I’ve posted FFT’s on the UCD180 thread and others, that show the performance of both my Digital and Analogue Class D designs that are limited by the performance of the AP2 and R&S UPD05 test systems.

I’m in the process of designing a new measurement system that should allow me to measure the full performance of my designs.

John
 
Jan-Peter,

You say: "At the moment we have no plans for a SMPS. SMPS are still less reliable as with a traditional toroidel transformer and not to forget the price is much higher."

If this is the case, why do all the 300-500W Class D DVD receivers coming out of China have SMPS? Or are you only considering the case of low-volume home build?

Regards,

Skippy.
 
skippy said:
Jan-Peter,

You say: "At the moment we have no plans for a SMPS. SMPS are still less reliable as with a traditional toroidel transformer and not to forget the price is much higher."

If this is the case, why do all the 300-500W Class D DVD receivers coming out of China have SMPS? Or are you only considering the case of low-volume home build?

Regards,

Skippy.
We've had tons of these over here for analysis. If you run all channels at full power, they shut down within seconds. Some of them will deliver 2 channels of full power for prolonged periods of time, but not all.
Therefore, a 6x50W receiver is not a 300W device, far from it.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.