For those who want to try the UcD400 amps, please see my group buy:
http://www.audiocircle.com/circles/viewtopic.php?t=19614
Thanks
Ken
http://www.audiocircle.com/circles/viewtopic.php?t=19614
Thanks
Ken
This is really a question directed at Bruno, but I hope some others might offer an opinion as well.
Basically I'm really happy with my first UCD400 amp and have another pair of modules here to build a second. I want to try a more exotic powersupply this time, but it's a question of understanding which bits make a difference and where you are wasting your money. Some of this stuff has been covered in dribs and drabs before (I have been following this thread since day one), but I would like to get a concise big picture answer if possible.
So if we assume for a moment that a serious powersupply looks as follows:
http://www.tnt-audio.com/clinica/ssps2_e.html
So bearing in mind that there is a bunch of high freq filtering and decoupling in the amp modules already the question is which bits of this are actually required for a high end powersupply?
Lets break it down that you have:
1) Transformer
2) Decoupling into the diodes
3) Dual rectifier bridge (perhaps with snubbers on each diode)
4) Some big smoothing caps
5) RC filter network
6) Possibly some more decoupling caps on the output
So I'm looking to build a dual mono supply for the next amp. I already bought a pair of generic 500va transformers, enough schottky MBR10100 to make two dual bridges, and enough Aerovox 10,000 caps to get 20,000uF per rail per module.
This makes a pretty simple looking powersupply... I'm curious as to whether I will get any sonic benefits from adding any decoupling before and after, or using an RC filter network? Is this icing or should I expect a useful improvement? Anything else I missed?
These probably sound like simple questions, but to be honest I'm more interested really in building the speakers and working on the DSP code for the crossovers. I really want to get past the amp building stage and get on with the rest of the system which is still the weak link...
Thanks
Basically I'm really happy with my first UCD400 amp and have another pair of modules here to build a second. I want to try a more exotic powersupply this time, but it's a question of understanding which bits make a difference and where you are wasting your money. Some of this stuff has been covered in dribs and drabs before (I have been following this thread since day one), but I would like to get a concise big picture answer if possible.
So if we assume for a moment that a serious powersupply looks as follows:
http://www.tnt-audio.com/clinica/ssps2_e.html
So bearing in mind that there is a bunch of high freq filtering and decoupling in the amp modules already the question is which bits of this are actually required for a high end powersupply?
Lets break it down that you have:
1) Transformer
2) Decoupling into the diodes
3) Dual rectifier bridge (perhaps with snubbers on each diode)
4) Some big smoothing caps
5) RC filter network
6) Possibly some more decoupling caps on the output
So I'm looking to build a dual mono supply for the next amp. I already bought a pair of generic 500va transformers, enough schottky MBR10100 to make two dual bridges, and enough Aerovox 10,000 caps to get 20,000uF per rail per module.
This makes a pretty simple looking powersupply... I'm curious as to whether I will get any sonic benefits from adding any decoupling before and after, or using an RC filter network? Is this icing or should I expect a useful improvement? Anything else I missed?
These probably sound like simple questions, but to be honest I'm more interested really in building the speakers and working on the DSP code for the crossovers. I really want to get past the amp building stage and get on with the rest of the system which is still the weak link...
Thanks
ewildgoose said:
I tried to email you but it is blocked. I didn’t want to go so far off topic but am really curious about your DSP setup and what you are doing with it.
Your power supply plans sound fine and right in line with what has been recommended. One thing about using 2 bridges, this allows you to remove ripple current from the ground connection. From the bridge, connect directly to the first cap terminals. Connect this one on to the next cap then do the ground and power connection at the module. The module ground input becomes a star point of sorts. All this is good for noise. According to Bruno, additional decoupling would not be of much benefit, in fact could be harmful and cause ringing. For snubbing Schottky diodes a Zobel configuration was recommended. Unfortunately I don’t remember the values, they were something like .2 ohm and .o1 uf.
If you would like to discus your DSP project feel free to email me.
Roger
Pardon the probably dumb question, but isn't the Zobel network the same as my "5) RC Network", ie the bits on the very right of the diagram here:
http://www.tnt-audio.com/clinica/ssps2_e.html
I have a recommendation from another source for a 1 ohm cap in series with a 220 nF cap, and that lot in parallel with a 100uF cap. I assume this does as is said and just removes any high freq impedence from the powersupply circuit?
Assuming that above *IS* required (and I didn't misunderstand) then to recap: a high end PS for the UCD should simply be: rectifiers, big caps, and an output snubber? Seems too simple...
Is the quality of the components in the snubber essential? I'm just trying to figure out my parts list for another Farnell shopping trip...
P.S. Love to talk about DSP. I sent you a private message and you can read more about the kind of thing I am going here:
http://www.duffroomcorrection.com/wiki/User:Ed_Wildgoose
There are a whole bunch of things one can easily do using DSP, a spare PC and a decent soundcard. Personally on an A-B test I can't hear any difference between a fairly decent CD player (Meridian 508.24) and my RME 9632 PC sound card. So I now use a PC as my source and this gives me tons of flexibility to fiddle.
I find the digital switching amps essential for experimenting with DSP though. I found that some traditional amps seem to struggle to keep up with demands when you start sticking in 6dB+ of correction into the amp - it's obviously going to start seriously testing your amp and the Class D stuff seems to just shrug off that kind of abuse. With my older Meridian 557 power amp everything sounded nice but a bit lifeless with the correction switched in. No problem with the same experiments and a UCD400 in the loop!
http://www.tnt-audio.com/clinica/ssps2_e.html
I have a recommendation from another source for a 1 ohm cap in series with a 220 nF cap, and that lot in parallel with a 100uF cap. I assume this does as is said and just removes any high freq impedence from the powersupply circuit?
Assuming that above *IS* required (and I didn't misunderstand) then to recap: a high end PS for the UCD should simply be: rectifiers, big caps, and an output snubber? Seems too simple...
Is the quality of the components in the snubber essential? I'm just trying to figure out my parts list for another Farnell shopping trip...
P.S. Love to talk about DSP. I sent you a private message and you can read more about the kind of thing I am going here:
http://www.duffroomcorrection.com/wiki/User:Ed_Wildgoose
There are a whole bunch of things one can easily do using DSP, a spare PC and a decent soundcard. Personally on an A-B test I can't hear any difference between a fairly decent CD player (Meridian 508.24) and my RME 9632 PC sound card. So I now use a PC as my source and this gives me tons of flexibility to fiddle.
I find the digital switching amps essential for experimenting with DSP though. I found that some traditional amps seem to struggle to keep up with demands when you start sticking in 6dB+ of correction into the amp - it's obviously going to start seriously testing your amp and the Class D stuff seems to just shrug off that kind of abuse. With my older Meridian 557 power amp everything sounded nice but a bit lifeless with the correction switched in. No problem with the same experiments and a UCD400 in the loop!
snubbers or Zobel?
I was referring to using the Zobel as a snubber across the bridge diodes so there will be a bunch of them. The usual practice is to just put a cap across these diodes but it is recommended to use the "Zobel" for Schottkys.
Output Zobels can be of benefit and the values sound about right. Generally Zobel networks are to damp ringing at HF. Yes, this is just a simple series connected RC circuit but can do wonders. Probably should use 1-2 watt resistors and cheap polypropylene caps. These should be low inductance parts so the cap will be a stacked film MKP type and the resistor will probably be a metal oxide type. Not to critical as to value or other variables except the caps have to be high enough voltage. Yes, it is true sometimes simpler is better. Like check out how simple the UcD 400 is as to parts count, of course Bruno is a true genius (Lars won’t convince me otherwise!).
I think this is the wave of the future. Store your music from disk or download directly on your HD and create really nice play lists. This is what I do. Even the high bit rate MP3’s sound decent. The sound quality from these setups has surpassed the cheaper CD players and is becoming a threat to the high end ones. When a $700 computer set up can beat a $3500 player we will have arrived. Now about a dedicated Mac Mini……..
I don’t have a worthy sound card yet so won’t get too exited about the DSP aspects at this point. I will be emailing you later with some general questions on how to get started on this. At this point I have almost zero knowledge.
Appreciate the offer, thanks.
The lack of congestion and compression with the new class D amps are other reasons I have stopped designing A/B amps, no longer much point to it.
Roger
I was referring to using the Zobel as a snubber across the bridge diodes so there will be a bunch of them. The usual practice is to just put a cap across these diodes but it is recommended to use the "Zobel" for Schottkys.
Output Zobels can be of benefit and the values sound about right. Generally Zobel networks are to damp ringing at HF. Yes, this is just a simple series connected RC circuit but can do wonders. Probably should use 1-2 watt resistors and cheap polypropylene caps. These should be low inductance parts so the cap will be a stacked film MKP type and the resistor will probably be a metal oxide type. Not to critical as to value or other variables except the caps have to be high enough voltage. Yes, it is true sometimes simpler is better. Like check out how simple the UcD 400 is as to parts count, of course Bruno is a true genius (Lars won’t convince me otherwise!).
I think this is the wave of the future. Store your music from disk or download directly on your HD and create really nice play lists. This is what I do. Even the high bit rate MP3’s sound decent. The sound quality from these setups has surpassed the cheaper CD players and is becoming a threat to the high end ones. When a $700 computer set up can beat a $3500 player we will have arrived. Now about a dedicated Mac Mini……..
I don’t have a worthy sound card yet so won’t get too exited about the DSP aspects at this point. I will be emailing you later with some general questions on how to get started on this. At this point I have almost zero knowledge.
Appreciate the offer, thanks.
The lack of congestion and compression with the new class D amps are other reasons I have stopped designing A/B amps, no longer much point to it.
Roger
ewildgoose said:
1a) Rating of the transformer should be sufficient but not overdone, in order not to make the charging spikes any larger than they need be.
1b) Depending on application and ground structure, it could be nice to have the lowest possible capacitance from primary to secondary. In this respect, most toroids are created equal.
1c) What toroids are certainly not created equal in, is the "hole" in the windings where the terminations go. Toroid theory holds that the stray field from one winding is compensated by its neighbours, but this ideal breaks down when the windings are laid irregularly. The dominant problem source is where the lead wires come out, because clearly it's not possible to make a regular "lay".
2) Around 47nF straight across each diode will do nicely.
3) There is some improvement in the form of better balanced transformer loading, but it becomes mandatory if you want to get the most from T network caps (see a post by Pasi P somewhere). It also allows you to use schottky diodes, further allowing you to get rid of the 47n caps (see 6).
4) I'd say keep it reasonable, again not to make the charge spikes too narrow. 10000uF per side for a 4 ohm mono block and certainly not more than 20000uF.
5) If you're talking of damping the DC rail, a small electrolytic capacitor of 10uF or so can function as a zobel. The same trick is used on the module. Place this 10uF near to the UcD.
6) Adding a small cap across the power rails tends to make the sound quality worse, not improve it. The likely cause is that the power inputs of the UcD modules themselves already have 100nF caps on them as part of the EMI filter. Adding a high q cap to this produces a high Q resonant circuit between your caps and the 100nF on the modules.
In fact, the 47nF caps that one places across normal (=non-schottky) diodes are also parallel to the DC in some way, which is why schottkies can be an improvement as you can then get rid of the 47nF caps (instead one uses the aforementioned 1n+10ohms which clearly can't go into any resonance with anything).
Aha, this is interesting - thanks for that conclusive answer.
Can I just seen clarification on one little bit. By "5) RC Network" I was referring to the bits on the very right of the diagram here:
http://www.tnt-audio.com/clinica/ssps2_e.html
So this bit is a R + C in series with a parallel C. My electronics is very high school so I'm not clear how this filter network actually differs from a Zobel.
But reading back through your notes I conclude that this bit of circuit would be undesirable on the UCD due to "the extra small capacitance between the DC rails"? However, I just wanted to confirm we were talking about the same thing.
So in conclusion you appear to be confirming that a high quality UCD supply is just: dual Schottky bridge + big smoothing caps.... (and basically nothing else...)
Cool. Looks like I have enough bits after all...
Thanks for answering such basic questions (again.)
Can I just seen clarification on one little bit. By "5) RC Network" I was referring to the bits on the very right of the diagram here:
http://www.tnt-audio.com/clinica/ssps2_e.html
So this bit is a R + C in series with a parallel C. My electronics is very high school so I'm not clear how this filter network actually differs from a Zobel.
But reading back through your notes I conclude that this bit of circuit would be undesirable on the UCD due to "the extra small capacitance between the DC rails"? However, I just wanted to confirm we were talking about the same thing.
So in conclusion you appear to be confirming that a high quality UCD supply is just: dual Schottky bridge + big smoothing caps.... (and basically nothing else...)
Cool. Looks like I have enough bits after all...
Thanks for answering such basic questions (again.)
Bruno Putzeys said:
PSU built using these recommandations:
Double PSU with 20 000 UF per rail and MBR10100. Nothing more. Works very fine.
Bruno,
I was thinking about it in the past, and I start to conclude in a different direction.. Please find it explained in details here:
post33
Summing it up: the charge current pulses are defined by : both the transformer secondary leakage inductance & the reservoir capacitance.
Because of this, the bigger the transformer, the larger the leakage inductance is, the smoother the pulses; the same is valid for the bigger value filter capacitors. In the same time, the dynamic balance between discharge / recharge processes becomes more "diluted' in time, so this can have an effect, also.
The high value charge current situation is only valid at the beginning, before the system reaches equilibrium. [At the startup I suspect of other factors contributing as well, like core saturation]
Also, this - for me - explaines my experiences with EI transformers and their higher leakage inductance = softer sound.
Ciao, George
I was thinking about it in the past, and I start to conclude in a different direction.. Please find it explained in details here:
post33
Summing it up: the charge current pulses are defined by : both the transformer secondary leakage inductance & the reservoir capacitance.
Because of this, the bigger the transformer, the larger the leakage inductance is, the smoother the pulses; the same is valid for the bigger value filter capacitors. In the same time, the dynamic balance between discharge / recharge processes becomes more "diluted' in time, so this can have an effect, also.
The high value charge current situation is only valid at the beginning, before the system reaches equilibrium. [At the startup I suspect of other factors contributing as well, like core saturation]
Also, this - for me - explaines my experiences with EI transformers and their higher leakage inductance = softer sound.
Ciao, George
Joseph K said:
My discussion concerned toroid transformers that have very little leakage inductance. For EI transformers I agree with your analysis. Of course, adding a choke in series with a toroid has the same effect.
Cheers,
B.
Bruno,
I have made the tests exclusively with toroids [and even small, 160 V, 220 VA, and 300 VA] and inserted in my simulations the measured secondary leakage inductance values [100 - 200 uH /secondary] and arrived to the above conclusion in this way. My reference to the EI-s was just to point out that they would have this phenomenon even more emphasized.
And yes, chokes in the secondaries will have the same effect, though I don't know if they are an optimal solution or Pi filters [CLC] or something other. I just didn't try.
By the way, I've just read your lecture on the other thread. Thanks, you are better than a university class!
Ciao, George
I have made the tests exclusively with toroids [and even small, 160 V, 220 VA, and 300 VA] and inserted in my simulations the measured secondary leakage inductance values [100 - 200 uH /secondary] and arrived to the above conclusion in this way. My reference to the EI-s was just to point out that they would have this phenomenon even more emphasized.
And yes, chokes in the secondaries will have the same effect, though I don't know if they are an optimal solution or Pi filters [CLC] or something other. I just didn't try.
By the way, I've just read your lecture on the other thread. Thanks, you are better than a university class!
Ciao, George
Cool, I'm going to measure a few transformers here too. I hadn't realised the leakage inductances were that high. As you say: good news!Joseph K said:
Bruno Putzeys said:
Bruno,
Just a quick question, how are you going to make the measurements? I can see enough potential benefits that this whole inductor on the transformer output thing is worth looking into. This might be a nice way to clean up the sound without causing a bunch of other problems. I wonder how high the values have to go before we start seeing some swinging choke like effects on the voltage.
Roger
I'd simply short the primary and measure the impedance of the secondary over a sufficiently wide frequency range. This can then be modelled in spice with good confidence that there aren't any other surprises lurking out. If you then model the esr and self inductance of the tanks as well, you should be able to make a perfectly reliable simulation without having to try everything out.sx881663 said:
There is also a second, more funny method:
Looking for the ringing caused by rectifier reverse recovery.
It's frequency is determined by secondary leakage inductance [Lleak]; secondary stray capacitance[Cstray]; diode reverse capacitance[Cdiode]. So it's grossly undefined. but putting a higher value capacitor across the secondary, then measuring the ringing frequency, then putting an other value & measuring again will give well defined measurement points and leakage inductance can be precisely calculated.
This method is laughably more difficoult with respect to the previous [described by Bruno] but can serve as a cross check.
And yes, simulations can give a lot of insight. From playing with it, I got the following general idea:
Considering the transformer secondary / rectifier/ reservoir cap / load configuration. There are two main resonances noticable. In both of them the common element is the secondary leakage inductance.
The first is in the case of rectifiers non-conducting. The formed LC tank is as above Lleak // Cstray // Cdiode. It's resonant freq. is around a couple hundreds kHz; it is highly undamped [if no snubber is applied], because it is unloaded.
The second is in the case of diodes conducting. Here the tank is formed by Lleak // Creservoir. The resonance here is damped by the actual load [which is parallel across the LC tank], but also secondary DC R is helping in lowering the Q. [Capacitor ESR is usually much lower] A funny particular: the behaviour of this tank is "sampled" by 100 Hz! [or 120Hz]
Side note: in case of a light load [small output signal] also this tank will be less damped, ringing might even start to show up!
This resonance is the one which forms the actual shape of the current pulses. [In case of smaller / medium current load, I am dead sure. In case of higher loads the measured pulse shape starts to deviate from this - this is where also core effects are coming into the picture?]
When trying to find a resolution [through sims] for separating the load / recharge current paths, I found that a C-L-C configuration actually is able to do that. The question is that from which frequency upwards is it happening. [The recharge current pulses, being pulses, are having a relatively high frequency content - it is quite possible to "convince" them, by the help of some inductance, to "choose" the lower impedance local path across the first C.] The problem here is to "convince" [keep] the audio signal circulating exclusively in the "local" loop [second C] - even at lower frequencies. Playing with path impedances helps here.
And this all is not new at all. ALL tube amplifiers are working like this, and some better designed SS amplifiers are having such a power supply configuration, [Bartolomeo Aloia, in Italy, for example].
Ciao, George
Looking for the ringing caused by rectifier reverse recovery.
It's frequency is determined by secondary leakage inductance [Lleak]; secondary stray capacitance[Cstray]; diode reverse capacitance[Cdiode]. So it's grossly undefined. but putting a higher value capacitor across the secondary, then measuring the ringing frequency, then putting an other value & measuring again will give well defined measurement points and leakage inductance can be precisely calculated.
This method is laughably more difficoult with respect to the previous [described by Bruno] but can serve as a cross check.
And yes, simulations can give a lot of insight. From playing with it, I got the following general idea:
Considering the transformer secondary / rectifier/ reservoir cap / load configuration. There are two main resonances noticable. In both of them the common element is the secondary leakage inductance.
The first is in the case of rectifiers non-conducting. The formed LC tank is as above Lleak // Cstray // Cdiode. It's resonant freq. is around a couple hundreds kHz; it is highly undamped [if no snubber is applied], because it is unloaded.
The second is in the case of diodes conducting. Here the tank is formed by Lleak // Creservoir. The resonance here is damped by the actual load [which is parallel across the LC tank], but also secondary DC R is helping in lowering the Q. [Capacitor ESR is usually much lower] A funny particular: the behaviour of this tank is "sampled" by 100 Hz! [or 120Hz]
Side note: in case of a light load [small output signal] also this tank will be less damped, ringing might even start to show up!
This resonance is the one which forms the actual shape of the current pulses. [In case of smaller / medium current load, I am dead sure. In case of higher loads the measured pulse shape starts to deviate from this - this is where also core effects are coming into the picture?]
When trying to find a resolution [through sims] for separating the load / recharge current paths, I found that a C-L-C configuration actually is able to do that. The question is that from which frequency upwards is it happening. [The recharge current pulses, being pulses, are having a relatively high frequency content - it is quite possible to "convince" them, by the help of some inductance, to "choose" the lower impedance local path across the first C.] The problem here is to "convince" [keep] the audio signal circulating exclusively in the "local" loop [second C] - even at lower frequencies. Playing with path impedances helps here.
And this all is not new at all. ALL tube amplifiers are working like this, and some better designed SS amplifiers are having such a power supply configuration, [Bartolomeo Aloia, in Italy, for example].
Ciao, George
Do i must know anythink inportant bebore i buy ucd400.
Are there any special rules in operation like at zap pulse like:
Avoid DC on the input.
Do NOT remove the load at full Volume!
Startups with no output loads.
Monoblocks can be a problem.
or any other?
thanks.
Best regards,
a007dio
Are there any special rules in operation like at zap pulse like:
Avoid DC on the input.
Do NOT remove the load at full Volume!
Startups with no output loads.
Monoblocks can be a problem.
or any other?
thanks.
Best regards,
a007dio
The current ZAPpulse 2.3 fixes all or most of those problems. As for UcD I think that is OK as well.
a007udio said:
The UcD400 modules have coupling capacitors at the input to block DC. If you remove them (for better sound quality) and then apply DC to the input, the amplifier will go into protection (turn off), but will not get damaged.
You can run the amplifiers at full volume and remove the load. No problem. You can also start-up the amplifier with no load as well. Nothing will go wrong.
The UcD amplifiers can be used in any number of channels without reservation, be it on a shared supply or on separate supplies.