ClassD4Sure,
Do we not all see the bigger picture with the whole synchronised UCD's here. We all know that Bruno is a very competent engineer, and had he been able to do UCD as a full bridge at the time, I'm sure he would have done. (No PSU pumping, low PSU rails, lower distortion etc). However, he didn't. He did a half bridge and the simple reason why, is that you are stuffed on the UCD approach.
If you use differential feedback on one modulator you need an input amplifier with a CMRR of 140dB at Fs to get the required performance; not really an option! Or you can use two half bridge modulators and risk the IMD problem that they are not synchronised and you have two Fs components that are constantly changing. I've got a pair of UCD400's linked together on my bench. The problem here, is that when you use the capacitor to sync. method, the modulators do stay sync'd up to about 60-70% modulation, then it all falls apart and with a music signal (or a multi-tone test signal) you can see the effect of the IMD effect appearing on the noise floor, and I'm sure that this will be audible.
There isn't a way around this with UCD and this is pretty obvious, otherwise there would be a patent in the name of Bruno Putzeys telling exactly how to do it.
F.
Do we not all see the bigger picture with the whole synchronised UCD's here. We all know that Bruno is a very competent engineer, and had he been able to do UCD as a full bridge at the time, I'm sure he would have done. (No PSU pumping, low PSU rails, lower distortion etc). However, he didn't. He did a half bridge and the simple reason why, is that you are stuffed on the UCD approach.
If you use differential feedback on one modulator you need an input amplifier with a CMRR of 140dB at Fs to get the required performance; not really an option! Or you can use two half bridge modulators and risk the IMD problem that they are not synchronised and you have two Fs components that are constantly changing. I've got a pair of UCD400's linked together on my bench. The problem here, is that when you use the capacitor to sync. method, the modulators do stay sync'd up to about 60-70% modulation, then it all falls apart and with a music signal (or a multi-tone test signal) you can see the effect of the IMD effect appearing on the noise floor, and I'm sure that this will be audible.
There isn't a way around this with UCD and this is pretty obvious, otherwise there would be a patent in the name of Bruno Putzeys telling exactly how to do it.
F.
Hello, guys.
I have been following this thread with great interest. IMHO, all asynchronous designs like UCD, etc, have this problem potentially when several modules are put together. This can happen with other modules like LCAudio, etc.
This is the reason why I like much more synchronous designs. When properly designed and good quality components are used, they are much more controllable and whistling problems can be easily avoided by synchronising several modules with a single master clock.
Just an opinion.
I have been following this thread with great interest. IMHO, all asynchronous designs like UCD, etc, have this problem potentially when several modules are put together. This can happen with other modules like LCAudio, etc.
This is the reason why I like much more synchronous designs. When properly designed and good quality components are used, they are much more controllable and whistling problems can be easily avoided by synchronising several modules with a single master clock.
Just an opinion.
Hi lads,
Due to my being rather busy I don't follow the forums much, but JP drew my attention to recent posts on this thread.
I have UcD's both in full bridge (two-state=using one comparator) and in half bridge and both work equally fine - similar performance, sonically indistinguishable. The step from a half bridge to a full bridge is so trivial it isn't worth patenting. The feedback network in practical UcD's is differential anyway (it senses both the signal and the ground at the speaker terminal) so you can use these 2 lines to sense differentially across the output of a full bridge amp.
Most UcD implementations are half-bridge for the simple reason that it is cheaper. Pumping is found to be a fringe effect in practice. In many cases it is only used as an argument by people who urgently have full-bridge amplifiers to sell.
But here you are - if for whatever reason you need a full bridge UcD, no problem.
Synchronising UcD's is unnecessary if you know how to make a board layout that doesn't produce whistles. I have stereo versions of the UcD400 and any whistles are well below the noise floor (the worst I've seen was still <10uV).
If the "competing" modules need syncing to prevent whistling, a likely cause is the output coil which has an odd external air gap arrangement (two un-gapped RM core halves spaced 2mm apart), which produces a comparably large stray magnetic field, and which could interfere with nearby modules. I have an inkling this was done because ferrite suppliers are not quick to deliver small quantities of cores with nonstandard air-gap lengths ground in.
You would want to synchronise a UcD to an external clock (achieved simply by mixing a small clock with the audio) if you are using the amp in a "receiver" type box and listening to AM. We typically have two frequencies to choose from, so you can always stay clear of the reception frequency (and the mirror). When not receiving AM, the set turns off the oscillator and the amps run free again (better THD and noise)
Sometimes there is a practical need for using two mono half-bridge UcDs in a full-bridge configuration (say you just want to make a bridgable stereo amp). When the load is tied across the two amps, synchronicity is automatically achieved but is lost quite rapidly. The 47nF cap across the load in bridge mode insures that synchronicity is maintained up to near clipping level.
Otherwise, use the amps without synchronisation, you will get the best performance.
Due to my being rather busy I don't follow the forums much, but JP drew my attention to recent posts on this thread.
I have UcD's both in full bridge (two-state=using one comparator) and in half bridge and both work equally fine - similar performance, sonically indistinguishable. The step from a half bridge to a full bridge is so trivial it isn't worth patenting. The feedback network in practical UcD's is differential anyway (it senses both the signal and the ground at the speaker terminal) so you can use these 2 lines to sense differentially across the output of a full bridge amp.
Most UcD implementations are half-bridge for the simple reason that it is cheaper. Pumping is found to be a fringe effect in practice. In many cases it is only used as an argument by people who urgently have full-bridge amplifiers to sell.
But here you are - if for whatever reason you need a full bridge UcD, no problem.
Synchronising UcD's is unnecessary if you know how to make a board layout that doesn't produce whistles. I have stereo versions of the UcD400 and any whistles are well below the noise floor (the worst I've seen was still <10uV).
If the "competing" modules need syncing to prevent whistling, a likely cause is the output coil which has an odd external air gap arrangement (two un-gapped RM core halves spaced 2mm apart), which produces a comparably large stray magnetic field, and which could interfere with nearby modules. I have an inkling this was done because ferrite suppliers are not quick to deliver small quantities of cores with nonstandard air-gap lengths ground in.
You would want to synchronise a UcD to an external clock (achieved simply by mixing a small clock with the audio) if you are using the amp in a "receiver" type box and listening to AM. We typically have two frequencies to choose from, so you can always stay clear of the reception frequency (and the mirror). When not receiving AM, the set turns off the oscillator and the amps run free again (better THD and noise)
Sometimes there is a practical need for using two mono half-bridge UcDs in a full-bridge configuration (say you just want to make a bridgable stereo amp). When the load is tied across the two amps, synchronicity is automatically achieved but is lost quite rapidly. The 47nF cap across the load in bridge mode insures that synchronicity is maintained up to near clipping level.
Otherwise, use the amps without synchronisation, you will get the best performance.
Hi,
JP or Bruno, am I correct when I think that your implementation of full bridge UcD still uses split power supply instead of a single one? Otherwise FuriousD would be correct about very high CMMR requirement for error amplifier. IMHO AD8130 seems a good candidate for this job. Second, what is the problem of using larger than 47nF cap for synchronising? In my simulation I used 330nF with good results.
Best regards,
Jaka Racman
JP or Bruno, am I correct when I think that your implementation of full bridge UcD still uses split power supply instead of a single one? Otherwise FuriousD would be correct about very high CMMR requirement for error amplifier. IMHO AD8130 seems a good candidate for this job. Second, what is the problem of using larger than 47nF cap for synchronising? In my simulation I used 330nF with good results.
Best regards,
Jaka Racman
Hi Bruno,
Thanks for coming out of hiding to respond, A little more information on the full bridge UCD though would be good. I'm not even going to entertain the 47nF capacitor approach as my test gear is good enough to see what is really going on on the UCD400's when this happens. I agree that their bloody good modules for what they've been designed for, and if you put a discrete input stage onto them that doesn't limit the performance, they really kick some ****, but I'm after perfection!
Full bridge UCD using a single comparator with differential feedback leads to a problem in that you get a common mode error signal generated due to the lack of CMRR at Fs. It becomes the limiting factor on the design. You need a comparator with "DC Typical" CMRR performance at 400kHz. Thats just not possible. How have you got around that? If you have an answer to that, then I can see why you speak with such an assured nature on the subject, and i'd love to know what it is.
F.
Thanks for coming out of hiding to respond, A little more information on the full bridge UCD though would be good. I'm not even going to entertain the 47nF capacitor approach as my test gear is good enough to see what is really going on on the UCD400's when this happens. I agree that their bloody good modules for what they've been designed for, and if you put a discrete input stage onto them that doesn't limit the performance, they really kick some ****, but I'm after perfection!
Full bridge UCD using a single comparator with differential feedback leads to a problem in that you get a common mode error signal generated due to the lack of CMRR at Fs. It becomes the limiting factor on the design. You need a comparator with "DC Typical" CMRR performance at 400kHz. Thats just not possible. How have you got around that? If you have an answer to that, then I can see why you speak with such an assured nature on the subject, and i'd love to know what it is.
F.
Jaka Racman said:Hi,
JP or Bruno, am I correct when I think that your implementation of full bridge UcD still uses split power supply instead of a single one? Otherwise FuriousD would be correct about very high CMMR requirement for error amplifier. IMHO AD8130 seems a good candidate for this job. Second, what is the problem of using larger than 47nF cap for synchronising? In my simulation I used 330nF with good results.
The comparator has a differential input (and in the case of discrete UcD's the construction is 100% symmetrical too). You really need good component matching on the feedback network. After all, CMRR becomes PSRR. That's a compounding factor on the cost of full bridge. Once you've got that bit right, DC offset will again be dominated by the input pair it doesn't matter whether the supply is split or single.
There's no law stating you can't use bigger "bridge" caps. 47nF is sufficient in most applications and is cheaper too (yeah sorry I work for a consumer electronics firm).
The power stage is balanced (two state control, not 3 state) so there is no 400kHz common mode component to speak of. Actually the FB version has less 400kHz HF common mode on the comparator inputs than does the HB version.FuriousD said:Full bridge UCD using a single comparator with differential feedback leads to a problem in that you get a common mode error signal generated due to the lack of CMRR at Fs. It becomes the limiting factor on the design. You need a comparator with "DC Typical" CMRR performance at 400kHz. Thats just not possible. How have you got around that? If you have an answer to that, then I can see why you speak with such an assured nature on the subject, and i'd love to know what it is.
Note: If you are using a true differential output filter, you should also have some common mode filtering to prevent really high-frequency hash getting out. This hash, apart from being an EMI hazard would indeed upset most comparators.
Using two separate LC filters works well and removes these worries completely, but admittedly takes up a bit more space.
The CMRR to worry about is LF, requiring good component matching in the feedback network. This can get pricey (in a relative sense). Precise component matching on the output filter is not required.
Bruno Putzeys said:
The power stage is balanced (two state control, not 3 state) so there is no 400kHz common mode component to speak of. Actually the FB version has less 400kHz HF common mode on the comparator inputs than does the HB version.
Note: If you are using a true differential output filter, you should also have some common mode filtering to prevent really high-frequency hash getting out. This hash, apart from being an EMI hazard would indeed upset most comparators.
Using two separate LC filters works well and removes these worries completely, but admittedly takes up a bit more space.
The CMRR to worry about is LF, requiring good component matching in the feedback network. This can get pricey (in a relative sense). Precise component matching on the output filter is not required.
Think that we're missing each others points here. Using two LC filters and Diff. feedback from either half means that you are summing the diff feedback into an SE error signal after comparing with the incoming audio signal. This requires a huge CMRR to do.
I am not making an SE error signal. What should I be needing that for? The whole circuit, feedback network and all is differential. Turning it into a ground-referred signal somewhere is an unnecessary complication which, as you can imagine, I cheerfully forfeit.FuriousD said:
Think that we're missing each others points here. Using two LC filters and Diff. feedback from either half means that you are summing the diff feedback into an SE error signal after comparing with the incoming audio signal. This requires a huge CMRR to do.
Bruno Putzeys said:
I am not making an SE error signal. What should I be needing that for? The whole circuit, feedback network and all is differential. Turning it into a ground-referred signal somewhere is an unnecessary complication which, as you can imagine, I cheerfully forfeit.
OK, I know the configuration of the UCD system well enough.
But in a differential implementation, lets say you have a switching error on the bridge, you take the differential feedback to the diff input of the comparator and regardless of whether the comparator output is differential in nature, you are still measuring the difference between it's inputs generating a SE error signal. The fact that the thing has a differential output just means that the error signal becomes common mode, which is of no real benefit.
Any common mode crap (i.e. switching noise etc) on the feedback lines will need to be rejected or will be included in the error signal generated, so the CMRR of the comparator becomes the limiting factor of the system.
If the output filter is worth its salt no common mode of significance will reach the comparator. Again, if you make a differential-only filter without further common mode filtering that's a different thing. And then it's not 400kHz common mode we're talking about, but 10MHz and higher.
Believe me, either you are indeed experiencing "a" problem in a circuit you're building and you're attributing it to the wrong phenomenon, or you don't have the problem and you're merely speculating on a phenomenon that may just as well not exist when left to actual electrons to try out.
Either way it's working fine here but I cannot vouch for other's implementations.
Back to hiding mode.
Believe me, either you are indeed experiencing "a" problem in a circuit you're building and you're attributing it to the wrong phenomenon, or you don't have the problem and you're merely speculating on a phenomenon that may just as well not exist when left to actual electrons to try out.
Either way it's working fine here but I cannot vouch for other's implementations.
Back to hiding mode.
Bruno Putzeys said:If the output filter is worth its salt no common mode of significance will reach the comparator. Again, if you make a differential-only filter without further common mode filtering that's a different thing. And then it's not 400kHz common mode we're talking about, but 10MHz and higher.
Believe me, either you are indeed experiencing "a" problem in a circuit you're building and you're attributing it to the wrong phenomenon, or you don't have the problem and you're merely speculating on a phenomenon that may just as well not exist when left to actual electrons to try out.
Either way it's working fine here but I cannot vouch for other's implementations.
Back to hiding mode.
Thanks for your help Bruno. Enjoy the silence.
Hi,
Will these improve the sound of the UcD? Will it help with heterodyning issues?
http://www.amusicdirect.com/products/detail.asp?cat=380&sku=AAQEV8
Has anyone tried them?? lol.... 7.6K PER FOOT!!!
Hey and if you think that was good... Don't forget to get half a dozen of these to compliment them:
http://www.amusicdirect.com/products/detail.asp?cat=280&sku=AELEV
I thought that last link was funny, even more so when I realized it's their first item under RFI/EMI Noise control, then I saw this:
http://www.amusicdirect.com/products/detail.asp?cat=280&sku=ASHAKSTN
.... it's a ROCK LOL! But not just ANY rock, a fully "passive" rock, "requiring no cables or electrical hookup. "The result is clearer and sweeter instruments and vocals, with more liquidity, detail and focus"
So, I'm thinking... group buy?
Anyone?
Will these improve the sound of the UcD? Will it help with heterodyning issues?

http://www.amusicdirect.com/products/detail.asp?cat=380&sku=AAQEV8
Has anyone tried them?? lol.... 7.6K PER FOOT!!!
Hey and if you think that was good... Don't forget to get half a dozen of these to compliment them:
http://www.amusicdirect.com/products/detail.asp?cat=280&sku=AELEV
I thought that last link was funny, even more so when I realized it's their first item under RFI/EMI Noise control, then I saw this:
http://www.amusicdirect.com/products/detail.asp?cat=280&sku=ASHAKSTN
.... it's a ROCK LOL! But not just ANY rock, a fully "passive" rock, "requiring no cables or electrical hookup. "The result is clearer and sweeter instruments and vocals, with more liquidity, detail and focus"
So, I'm thinking... group buy?
Anyone?
http://www.amusicdirect.com/products/detail.asp?cat=280&sku=ASHAKSTN
.... it's a ROCK LOL! But not just ANY rock, a fully "passive" rock, "requiring no cables or electrical hookup. "The result is clearer and sweeter instruments and vocals, with more liquidity, detail and focus"
lol...Can I carry this rock around with me? Maybe into a crowded/noisy venue. As I walk through the crowd, people will be amazed with the clarity that suddenly appears.
I'll got some bricks in the backyard, I'll give those ago.
I like my rocks all active. In fact, back in '96, I was the first guy in New York with an all-active rock system. Now everybody's doing it and I gotta find something new to keep my edge.
Does the modulator or comparator circuitry in a class D amp suffer from jitter-related timing errors?
Hi,
Yes and no. If they're clocked they're susceptible to it, if they're self oscillating, there's no clock. Neat huh?
Yes and no. If they're clocked they're susceptible to it, if they're self oscillating, there's no clock. Neat huh?
Does the oscillation circuit suffer power supply disturbance problems? I'm a little uncertain how these things work, but am wanting to know the point of greatest susceptibility in the UCD circuit.
serengetiplains said:Does the oscillation circuit suffer power supply disturbance problems? I'm a little uncertain how these things work, but am wanting to know the point of greatest susceptibility in the UCD circuit.
It doesn't mention PSRR in the data sheet and so I can't quote you a number, but It isn't so bad that anyone here has ever built a highly regulated supply for it, properly sizing the PSU caps should be enough.
Point of greatest susceptibility? Well that's easy, same with most things, it's the user!
Seriously though I would think your two biggest concerns should be a proper layout (both wiring and placement, nothing new or very different when it comes to audio, typical amp stuff), and making sure it has some cooling. They're highly efficient but of course not a 100%, and do require some airflow (no fans, just don't pack a bunch of them into a sealed box) and a nice chunk of aluminum to bolt them to.
You can find enough info in these threads to answer most any question you could possibly have.
Read this entire thread, read the entire UcD 180 thread (take you a week or two but it's seriously worth it, that thread should almost be printed and sold as a book) and check out the thread on optimal power supplies for UcD and ZapPulse. I don't think you'll be left wanting after that.
Regards
Hi cd4s
I spent the last few days reading the 400 + 180 + PSU posts through, in their entirety. Not having a schematic at any point to which to refer, I found it difficult to picture the overall circuit and of the possible power supply related vulnerabilities. If anyone has provided an additional regulated circuit for the oscillator circuit, I'd appreciate hearing their experience. Seems to me that PSU capacitor quality would affect the overall sound of this amp, as it clearly does, not solely due to the direct effect those capacitors have on output current and such, but perhaps even more so (?) due to the effect they have on vulnerable digital and oscillator circuits.
I have a Panasonic SA-XR50 I've been fiddling with. I've found the amplifier to respond quite well to, among other things, cleaning the AC supply feeding the amp with an Elgar line conditioner.
I spent the last few days reading the 400 + 180 + PSU posts through, in their entirety. Not having a schematic at any point to which to refer, I found it difficult to picture the overall circuit and of the possible power supply related vulnerabilities. If anyone has provided an additional regulated circuit for the oscillator circuit, I'd appreciate hearing their experience. Seems to me that PSU capacitor quality would affect the overall sound of this amp, as it clearly does, not solely due to the direct effect those capacitors have on output current and such, but perhaps even more so (?) due to the effect they have on vulnerable digital and oscillator circuits.
I have a Panasonic SA-XR50 I've been fiddling with. I've found the amplifier to respond quite well to, among other things, cleaning the AC supply feeding the amp with an Elgar line conditioner.
- Home
- Amplifiers
- Class D
- UcD400 Q & A