Re: No one has answered the original question
The last thing I heard was that production of their driver IC (of which the evaluation schematic is already on their site for some time now) would finally be starting this summer. That amounts to a delay of 12 months, something not unheard of in the development of class D ICs.Jocko Homo said:
Hi Jaka,
about B&O, maybe it's " there was mistake in the schematics" too? Who is haven't mistakes? (But if B&O designers will make appearance on diyaudio also, then...)
Hi Bruno,
that's really great event, that you are here!Incredibly!
(my congratulations to diiyaudio.com
I already wrote here about some my experiments with UcD like amp, i still impressed by very detailed sound of it. Do you have explanation for UcD's zero_feedback_sound?
about B&O, maybe it's " there was mistake in the schematics" too? Who is haven't mistakes? (But if B&O designers will make appearance on diyaudio also, then...)
Hi Bruno,
that's really great event, that you are here!Incredibly!
(my congratulations to diiyaudio.com I already wrote here about some my experiments with UcD like amp, i still impressed by very detailed sound of it. Do you have explanation for UcD's zero_feedback_sound?
IVX said:
Always the same. The loop gain is constant (around 24dB for the circuit known on this forum, around 32dB for the one in use in our products) up to well over 20kHz. The result is that the distortion is frequency independent.
When an amplifier "sounds like it has lots of feedback" it means that distortion is rising over most of the audio band, even though its maximum level may still be low. Mueta's THD characteristic is of that kind, but in their measurements THD at lower frequencies are mostly buried in noise (due to an odd preference of filtering post notch instead of pre-notch on the AP) so it isn't very obvious from the graphs. In B&O's case, the high-frequency distortion can actually become positively high (closer to 0.1%).
A rising THD response also shows up in the spectral makeup of the distortion. Since higher harmonics are attenuated less, their distribution will also tend to emphasise higher order harmonics as compared to low order ones.
So, even though a UcD with ridiculously low 1kHz THD figures is possible, I don't actively promote it because unless I can get a similar improvement at 20kHz, the sound quality will be worse.
Cheers,
Bruno
Hi Bruno
I am aware that its simplicity is one of the advantages of UCD. So maybe anything added to it will worsen its performance/cost ratio, which is definitely an issue for mass market applications.
But have you ever thought about using a double feedback loop in order to achieve higher NFB factors (and therefore even lower THD) without sacrificing the non-feedback sound ?
Regards
Charles
I am aware that its simplicity is one of the advantages of UCD. So maybe anything added to it will worsen its performance/cost ratio, which is definitely an issue for mass market applications.
But have you ever thought about using a double feedback loop in order to achieve higher NFB factors (and therefore even lower THD) without sacrificing the non-feedback sound ?
Regards
Charles
Hi Ivan,
everybody makes mistakes. But I did not apply for a patent, and because of mistake Bruno still has the chance to patent full bridge version .
Don't get me wrong, I think Karsten Nielsen has many very good ideas, but with this patent application they have gone a little to far (applying for amplifier directly connected to the voice coil and runing from nonisolated mains supply). Even in this forum people get banned for promoting such ideas.
Bruno,
I have noticed that allmost all digitally modulated class D designs use full bridge output stage, while most analogly modulated amplifiers use half bridge with split supplies. While it is clear to me why full bridge is preferable in digitally modulated designs, I don't know why it is not used more in analogly modulated ones. Maybe this has something to do with ease of implementation, but I woul like to hear your opinion.
Best regards,
Jaka Racman
everybody makes mistakes. But I did not apply for a patent, and because of mistake Bruno still has the chance to patent full bridge version
.Don't get me wrong, I think Karsten Nielsen has many very good ideas, but with this patent application they have gone a little to far (applying for amplifier directly connected to the voice coil and runing from nonisolated mains supply). Even in this forum people get banned for promoting such ideas.
Bruno,
I have noticed that allmost all digitally modulated class D designs use full bridge output stage, while most analogly modulated amplifiers use half bridge with split supplies. While it is clear to me why full bridge is preferable in digitally modulated designs, I don't know why it is not used more in analogly modulated ones. Maybe this has something to do with ease of implementation, but I woul like to hear your opinion.
Best regards,
Jaka Racman
Too lateJaka Racman said:and because of mistake Bruno still has the chance to patent full bridge version
... It should've been done before it got published here and before we actually made products using the circuit. According to our IP department, absolute secrecy (or consistent use of NDAs) is needed before a patent is filed. I'm not an expert in patent matters but it sounds reasonable when they tell me that.The problem with full bridge is that the (feedback input of the) control loop has to have very good CMRR and DC accuracy, requiring matched resistors and potentially benefiting from matched capacitors too. Half-bridge amps have no such complications.Jaka Racman said:
In addition, the reduced parts/pin count of half bridge amps is often a cost advantage.
Hi Bruno,
I was under an impression that producing a circuit actually gives you an advantage (priority) when applying for a patent. If I understand correctly, you can not patent full bridge, but it is not a problem for you since everyone producing it would infringe on your half bridge patent.
I was hoping that you could elaborate also on sonic differencies (if any) between half and full bridge amplifiers. ( I know about your multiphase DSD amplifier).
Best regards,
Jaka Racman
I was under an impression that producing a circuit actually gives you an advantage (priority) when applying for a patent. If I understand correctly, you can not patent full bridge, but it is not a problem for you since everyone producing it would infringe on your half bridge patent.
I was hoping that you could elaborate also on sonic differencies (if any) between half and full bridge amplifiers. ( I know about your multiphase DSD amplifier).
Best regards,
Jaka Racman
An additional advantage of half-bridge topologies is the possibility to be bridged by the user if needed.
It is not impossible to build bridgeable full-bridge amps but that wouldn't be very cost-effective (fully floating PSUs etc).
What I'd like to know is how Philips deals with the supply pumping effect. Are they simply using large PSU caps, is one channel inverted or is there any compensation circuitry in use ? Or do they simply not care at all ?
Regards
Charles
It is not impossible to build bridgeable full-bridge amps but that wouldn't be very cost-effective (fully floating PSUs etc).
What I'd like to know is how Philips deals with the supply pumping effect. Are they simply using large PSU caps, is one channel inverted or is there any compensation circuitry in use ? Or do they simply not care at all ?
Regards
Charles
Recently I've been toying with a full bridge UcD using only one comparator (ie. binary) and ran a comparison with a half bridge design with -as much as possible- the same specs. The filter coils and caps were of a different make, however. Well, the sonic differences were well within the range of what is to be expected with the component differences alone, so there doesn't seem to be any significant difference attributable to full or half bridge operation. One could theoreticise that the different regime on the power caps could make a difference, but the listening test didn't point in that direction.Jaka Racman said:
For the PPDSD, the choice was full bridge for the sake of having only one rail regulator. The regulator was capable of sourcing and sinking so pumping was not an issue there.
Depends. If the application can afford sufficient elcaps, we choose that. In cases where those few cents do matter (and sometimes they do), the channels will be wired in opposite phase.phase_accurate said:
The amount of buffer caps you need to keep tabs on pumping is not excessive. The caps you'd normally place after the rectifiers anyway are sufficient. A 2x100W/6ohm amp will have 2x4700uF on the rectifiers and maybe 2x1000uF on each channel locally. Even with 10Hz in phase on both channels, the pumping never adds more than 15% or so to the rails.
A cheapo multichannel receiver sometimes uses the same amount of storage for a "5x100W" amp (1-channel rating is used in this market) which is pretty much the limit of what you can do without taking some measures.
Philips Semicon's chips turn off when the rail voltage exceeds the +/-30V limit. For discrete amps we recommend the same, if only for in case a coupling cap starts leaking.
Bruno,
i infinitely respect your opinion, i just don't understand how it is possible so easy to hear such little changing of harmonic proportion if the class AB have THD@20kHZ <0.01%(THD@1kHZ 0.001% i.e. -100db!!).
For selftranquility i plan comparing of the FFT complex audio signal from UcD and from AB also.
ps: How much stable the frequency of the original UcD if load is OFF? (without zobel?)
i infinitely respect your opinion, i just don't understand how it is possible so easy to hear such little changing of harmonic proportion if the class AB have THD@20kHZ <0.01%(THD@1kHZ 0.001% i.e. -100db!!).
For selftranquility i plan comparing of the FFT complex audio signal from UcD and from AB also.ps: How much stable the frequency of the original UcD if load is OFF? (without zobel?)
Hi Ivan,
To answer the question about the stability with and without load:
The difference with and without a load is very little with the UcD!
Nice to mention, the UcD doesn't has a zobel
If you take a look to this link , http://www.hypex.nl/classd/classd.PDF on page 3 you can see the measurement with different loads. This is all original measurements and document written by Bruno to present the UcD IP model, (still many thanks Bruno).
We are now close to present the UcD in our Hypex 180W version as a test kit for everybody who likes to evaluate this marvelous Class-D design.
Regards,
Jan-Peter
www.hypex.nl
To answer the question about the stability with and without load:
The difference with and without a load is very little with the UcD!
Nice to mention, the UcD doesn't has a zobel
If you take a look to this link , http://www.hypex.nl/classd/classd.PDF on page 3 you can see the measurement with different loads. This is all original measurements and document written by Bruno to present the UcD IP model, (still many thanks Bruno).
We are now close to present the UcD in our Hypex 180W version as a test kit for everybody who likes to evaluate this marvelous Class-D design.
Regards,
Jan-Peter
www.hypex.nl
To be honest: I don't know. It keeps surprising me that such low THD still has an impact on sound, especially keeping in mind the distortion of the loudspeaker that's attached (>0.1% even at low listening levels). I've been noticing the same sonic trends on preamplifiers (opamp circuit) too where the distortion was well below -110dB, lower than either the following power amp and certainly lower than the speaker.IVX said:
phase_accurate said:
The particular format of the loop (nested loop, state-variable or what have you) has no implications on the shape of the loop gain vs frequency. Any implementation can be used to obtain any response. It keeps baffling me how people manage to write patents limited to specific hardware implementations of a generic loop function.
Using a high-order loop with a flat in-band response is certainly possible, and the maths involved in doing so for any order is well understood "here". I'm not in a position to dwell on any details for the time being, but I'll keep you posted when some hardware is ready that does precisely that.
For me this isn't understandable either, since most of it is well-known control theory.
I have once seen a Tripath patent describing a generic switching amp (i.e. from the description it could be PWM or sigma-delta or whatnots) using a double loop in a pretty standard fashion. I have not found it anymore recently. I guess they had to withdraw it because it was too much "prior art".
Ahhhh ! Somehow I was convinced that you will do so one day !
Regards
Charles
D11 is a schottky diode and is sufficient to keep the internal B-C junction from conducting. We have moved on to using two normal diodes instead of one schottky because at high temperatures the schottky leaks so much T11 starts conducting all the time.IVX said:
As for the switching frequency problem, analyse the loop phase with and without load and see if "anything" allows oscillation at Fr. Then you can see what to do in order to prevent this.
(In order to avoid disputes over me "giving away secrets" I have to hold back on the details. You'll find it's pretty trivial though). Typically the switching frequency changes no more than 10% from loaded to unloaded.
Class D has been revived in popularity
Hi Bruno,
I think it is good that you appeared here, because interest in class D was waning. Since the forum discussed the transistor differential input stage type of circuit, which I had called "N-champ" in the "My very first Class D pwm (switching) amplifier" thread last fall interest the topic concerning the elegant topology you made popular has not revived until now. Back then, at the beginning, I wasn't even aware of your circuit. Now, I feel that I can give the credit where it more belongs (patent wo03090343). The circuit diagram posted on that former thread simply did not have enough open loop gain and differential action to work fast enough. I had tried various schemes to form a comparator on the lower power supply rail out of transistors, but was unhappy with its performance as there was not enough balance yet snap in its response to eliminate dead spots in the oscillation process. I also have been unhappy with that old circuit's propagation delay.
Now I hope to maintain my interest in the UcD so that I can replace my class H circuit with the UcD as my favorite. A contender foremost requires simplicity to be friendly to diyaudio people, and for the sub woofer amp project, it still looks like a close race, which hopefully will change. I also require several hundred watts of peak power for driving a sealed but smooth responding speaker, so that is a mandate for bridging, in addition to my desire to use small filter caps so that during an output short, the stored energy will not blow the mosfets before the overload protection can kick in. The absence of bridging with such small caps would lead to extreme power supply pumping without some form of charge redistribution. A former experimental amp I made a couple of years ago which used 50v rails and driving an 8 ohm speaker blew some 63v 1000uF elcaps--yuk.
Hi Bruno,
I think it is good that you appeared here, because interest in class D was waning. Since the forum discussed the transistor differential input stage type of circuit, which I had called "N-champ" in the "My very first Class D pwm (switching) amplifier" thread last fall interest the topic concerning the elegant topology you made popular has not revived until now. Back then, at the beginning, I wasn't even aware of your circuit. Now, I feel that I can give the credit where it more belongs (patent wo03090343). The circuit diagram posted on that former thread simply did not have enough open loop gain and differential action to work fast enough. I had tried various schemes to form a comparator on the lower power supply rail out of transistors, but was unhappy with its performance as there was not enough balance yet snap in its response to eliminate dead spots in the oscillation process. I also have been unhappy with that old circuit's propagation delay.
Now I hope to maintain my interest in the UcD so that I can replace my class H circuit with the UcD as my favorite. A contender foremost requires simplicity to be friendly to diyaudio people, and for the sub woofer amp project, it still looks like a close race, which hopefully will change. I also require several hundred watts of peak power for driving a sealed but smooth responding speaker, so that is a mandate for bridging, in addition to my desire to use small filter caps so that during an output short, the stored energy will not blow the mosfets before the overload protection can kick in. The absence of bridging with such small caps would lead to extreme power supply pumping without some form of charge redistribution. A former experimental amp I made a couple of years ago which used 50v rails and driving an 8 ohm speaker blew some 63v 1000uF elcaps--yuk.