I am in the process of building Aleph 5 and collected most of the "universal" parts, including 4x25.000/75V Sprague Powerlytic (type 36DX) and two Mallory caps (type CGS, 25.000uF/40VDC, max surge 50VDC) per channel. I am confident this will do but also am looking for information if it would be better to use CLC filtering. For this specific design, of course. If yes, does anyone know where to get the specific information. I don't want to hijack this thread, so, I give a thumb up for more capacity is better for the amp.
(As a hobbyist with very basic knowledge, but great love for music)
When I started building audio equipment, I always found electronics manual to be particularly silent on the issue of capacitance needed in a a power supply. In the best case, the authors refer to some "rule of thumb", but their respective thumbs vary in size too...
From the practice:
1) Have heard HIRAGA's "The Monster" 8watts class A on numerous occasions: uses 8 to 12 x 100,000uF caps (total 800,000uF to 1,2 F). Great schematic, very simple, well designed. Fantastic sound.
2) Some years ago, I built an opamp LM3886-based stereo amplifier, with a power supply built around 10x10,000 uF on each rail (200,000 uF in total), although it delivers 50W per channel only: stuning, and in the set up we blind-tested it (CEC top class CD player, Brinkmann converter, Klipshorns) better than very expensive PP 300B valve monoblocks connected to a MC Intosh pre. (no joke!).
3) I discovered a few years ago a great brand that produced in the 80s (real professional) class AB PA amplifiers ; the one I bought delivers 2x500W (4ohms) with a modest power supply built around one 10,000uF cap on each rail (20,000 uF in total). Works extremely well, dynamic, transparent etc. The circuit is based on a combination of bipolar and mosfet transistors.
Etc. etc.
So what? is the amount of power supply capacitance that crucial? maybe not. But it cannot harm to take great care of this.
Others have already pointed at the need for low ripple voltage.
In addition, I'd say that the power supply needs at least to provide the amperage needed for the amplifier, and these are a quite a few Amperes; if we take the 2x500W amp mentioned earlier, at 4 Ohms in stereo it requires a supply capable of producing about 16 amps (P=RI^2). The capacitors should be able to store this amount. Looking for example at this series of capacitors http://www.epcos.com/inf/20/30/db/aec_07/B41456__B41458.pdf, the 10,000uF/63 V version can deliver about 7 A nominal and 23A max. Seems a bit short to me to fit into the 2x500W amp in question... But another manufacturer might produce a 10,000 uF that can store more than 7/23A. That's when it becomes interesting to look at the capacitor datasheets...Including to find items with low ESR (Equivalent Serial Resistance) items, which, however can be decreased in proportion to the number of caps connected in parallel as everybody knows. Hence the superbanks of capacitors sometimes used. The HIRAGA 8W needs just 1A nominal per channel with 8 Ohm loads, 2A with 4Ohm loads. But ESR (and other factors like GRollins mentioned earlier) may need to be taken into account.
Now, I'm slowly building an Aleph something (duno yet which one : 0, 0,5, 1, 2, 3, 4, 5, 6, 7, 1.2, 1.3, 30, 60, 75, X, XA, XB, XC, J, JX, mini, XXL....), and therefore I ordered 4x150,000 uF (600,000uF) 63V caps to be on the safe side as regards both amperage (31A nominal each cap) and ESR, whatever the final option will be.
When I started building audio equipment, I always found electronics manual to be particularly silent on the issue of capacitance needed in a a power supply. In the best case, the authors refer to some "rule of thumb", but their respective thumbs vary in size too...
From the practice:
1) Have heard HIRAGA's "The Monster" 8watts class A on numerous occasions: uses 8 to 12 x 100,000uF caps (total 800,000uF to 1,2 F). Great schematic, very simple, well designed. Fantastic sound.
2) Some years ago, I built an opamp LM3886-based stereo amplifier, with a power supply built around 10x10,000 uF on each rail (200,000 uF in total), although it delivers 50W per channel only: stuning, and in the set up we blind-tested it (CEC top class CD player, Brinkmann converter, Klipshorns) better than very expensive PP 300B valve monoblocks connected to a MC Intosh pre. (no joke!).
3) I discovered a few years ago a great brand that produced in the 80s (real professional) class AB PA amplifiers ; the one I bought delivers 2x500W (4ohms) with a modest power supply built around one 10,000uF cap on each rail (20,000 uF in total). Works extremely well, dynamic, transparent etc. The circuit is based on a combination of bipolar and mosfet transistors.
Etc. etc.
So what? is the amount of power supply capacitance that crucial? maybe not. But it cannot harm to take great care of this.
Others have already pointed at the need for low ripple voltage.
In addition, I'd say that the power supply needs at least to provide the amperage needed for the amplifier, and these are a quite a few Amperes; if we take the 2x500W amp mentioned earlier, at 4 Ohms in stereo it requires a supply capable of producing about 16 amps (P=RI^2). The capacitors should be able to store this amount. Looking for example at this series of capacitors http://www.epcos.com/inf/20/30/db/aec_07/B41456__B41458.pdf, the 10,000uF/63 V version can deliver about 7 A nominal and 23A max. Seems a bit short to me to fit into the 2x500W amp in question... But another manufacturer might produce a 10,000 uF that can store more than 7/23A. That's when it becomes interesting to look at the capacitor datasheets...Including to find items with low ESR (Equivalent Serial Resistance) items, which, however can be decreased in proportion to the number of caps connected in parallel as everybody knows. Hence the superbanks of capacitors sometimes used. The HIRAGA 8W needs just 1A nominal per channel with 8 Ohm loads, 2A with 4Ohm loads. But ESR (and other factors like GRollins mentioned earlier) may need to be taken into account.
Now, I'm slowly building an Aleph something (duno yet which one : 0, 0,5, 1, 2, 3, 4, 5, 6, 7, 1.2, 1.3, 30, 60, 75, X, XA, XB, XC, J, JX, mini, XXL....), and therefore I ordered 4x150,000 uF (600,000uF) 63V caps to be on the safe side as regards both amperage (31A nominal each cap) and ESR, whatever the final option will be.
I remain modest compared to your amount of filtering, and for starters remain modest on the amp as well. Just doing the finishing touches on mounting up the copper rails for the filters on a Pass F5 clone, start with a 500VA toroid with 2x18 + 2x18VAC (four 18V secondaries at 7.1 Amps each) then, 35A/500V bridges (one per each channel) and 2x25.000uF, 0.22R/25W, followed with 4x33.000uF. In total 364.000uF for two channels, not counting the smaller 470uF on the PCB and the block ones over the big electrolytes. Should be fine I think...
For the image click on the link...
An externally hosted image should be here but it was not working when we last tested it.
For the image click on the link...
as elac310
tells us
Some great working/sounding amplifiers use a lot of CAPS
other great sounding amplifiers can use a fraction for psu caps filtering.
How come?
Well, amplifiers are different.
As if we did not know this, all of us
Hiraga fairly barbone and more simplistic amplifiers
can require more capactiance and very clean psu in order to perform as intended
Other amplifiers, with higher PSRR, power supply rejection ratio,
can get away with much less filtering,
and these can be more or less immune against psu low freq ripple
Besides PSRR, we also have higher/lower value of CMRR, common mode rejection ratio,
which is something similar
.. and tells how much amplifier can reject what happens at supply rails
Even with amplifiers having very high PSRR,
there is nothing bad or wrong with reducing psu ripple.
It can not ever be bad, even if some amplifiers will probably not perform much any better.
CLC filters have been used by Nelson Pass in some of his Class A designs.
They are very effective, in compare with only C, or CRC.
Nelson Pass has always cared for making very good, clean supply to his designs.
This is one of the 'secrets' to the good Pass Amps results.
In a number of cases.
tells us
Some great working/sounding amplifiers use a lot of CAPS
other great sounding amplifiers can use a fraction for psu caps filtering.
How come?
Well, amplifiers are different.
As if we did not know this, all of us
Hiraga fairly barbone and more simplistic amplifiers
can require more capactiance and very clean psu in order to perform as intended
Other amplifiers, with higher PSRR, power supply rejection ratio,
can get away with much less filtering,
and these can be more or less immune against psu low freq ripple
Besides PSRR, we also have higher/lower value of CMRR, common mode rejection ratio,
which is something similar
.. and tells how much amplifier can reject what happens at supply rails
Even with amplifiers having very high PSRR,
there is nothing bad or wrong with reducing psu ripple.
It can not ever be bad, even if some amplifiers will probably not perform much any better.
CLC filters have been used by Nelson Pass in some of his Class A designs.
They are very effective, in compare with only C, or CRC.
Nelson Pass has always cared for making very good, clean supply to his designs.
This is one of the 'secrets' to the good Pass Amps results.
In a number of cases.
lineup said:
The structure of supply filters is another topic, isn't it?
Nevertheless, from what I could take from the DIY forum, Nelson Pass himself seems to have evolved on the issue of filter structures. The topology of his early Aleph amps I saw on paper used a simple C structure, with generous capacitance, though: e.g. 4x22,000 on Aleph 2s, 8x22,000 on Aleph 3s; by the way, and interestingly enough, the A2 is capable of much more power and has a bit more quiescent current than the A3, but there were most probably capacitor size constraints, cost factors etc to take into account in the design process....
Then I saw forum discussions on the merits of simple C versus CLC/CRC filters, and CRC versus CLC which escalated but stoped (fortunately) shortly before reaching the point of discussing CRCLCRCLCRCCCCLLCC versus CLCLCRCLCRCRCLCCC filters etc.
It's true that NP (reportedly) does at present make use of CRC filters; I have not seen anything suggesting he is implementing CLC filters as well but there is so much information out there and I most probably missed something (I would welcome a hint to find the relevant info).
I'm just not entirely sure whether the rationale for NP using at present CRC/CLC structures is 100% the designer's conviction from the electronics standpoint or whether there might have been a need to also meet certain expectations of the audio community (Indeed, I found there was an overwhelm of people who opted in their DIY constructions for CRC/CLC configurations and a wide agreement that the lower ripple was beneficial to the sound quality)... Or both. It could also be that a C filter works better in an Aleph design but that X and other more recent designs are better suited with CRC/CLC (PSRR and CMRR considerations?). Finally, there's no shame in taking into account the positive experience of other "forumists" who report sound improvements deriving from options that are not necessarily justified from the point of view of electronics or cost/quality improvement factor. I presume that NP himself expects some benefits in return for spending a lot of time posting and answering questions, including taking stock of experiments from DIYers, many of which appear to be technically very knowledgeable in addition to being audio enthusiasts.
On a similar issue, I remember NP was also in favour of standard 25-35A rectifier bridges (in one of his papers available on passDIY.com I think he says he does not like too much fast rectifiers). But someone suggested that some of his post-Aleph designs do now make use of fast/ultrafast rectifiers.
Again, all this is probably another topic...
From my own (modest, hobbyist) empirical experience, I very seldom obtained a significant sound improvement by optimising only one factor/ element in an amplifier (or other) design that already made use of reasonably good options and components; instead, it was the accumulation of "small" improvements that made a difference. But I'm still in a learning process which makes me love this excellent DIYaudio website!
stein2 said:"I remain modest compared to your amount of filtering, and for starters remain modest on the amp as well. Just doing the finishing touches on mounting up the copper rails for the filters on a Pass F5 clone, start with a 500VA toroid with 2x18 + 2x18VAC (four 18V secondaries at 7.1 Amps each) then, 35A/500V bridges (one per each channel) and 2x25.000uF, 0.22R/25W, followed with 4x33.000uF. In total 364.000uF for two channels, not counting the smaller 470uF on the PCB and the block ones over the big electrolytes. Should be fine I think..."
Well, you say you end up with 364,000 uF for two channels, which is quite a lot. At these levels of capacitance, I don't think 360,000 or 600,000 uF make a difference. In fact, the last reason (I did not mention) why I chose the 4x150,000 caps is aesthetic. These 22cm (8 inches?) capacitors are the longest/highest I managed to find as they will fit well into, and fill up the future enclosure which will look rather empty otherwise
I very much like your idea of the 470uF decoupling caps mounted on the PCBs; this is also something I intend to experiment with, including with values available as film capacitors (MKP, MKT etc.).
elac310 said:Well, you say you end up with 364,000 uF for two channels, which is quite a lot. At these levels of capacitance, I don't think 360,000 or 600,000 uF make a difference. In fact, the last reason (I did not mention) why I chose the 4x150,000 caps is aesthetic. These 22cm (8 inches?) capacitors are the longest/highest I managed to find as they will fit well into, and fill up the future enclosure which will look rather empty otherwise
I very much like your idea of the 470uF decoupling caps mounted on the PCBs; this is also something I intend to experiment with, including with values available as film capacitors (MKP, MKT etc.). [/B]
I went for the maximal capacity I could fit into the future box because the F5 as mainly all the class A amplifiers are designed to work reliably and be as stable as possible with any load one attaches to the speaker out terminal...
My ESLs are nominally 4 Ohms, but like many of that sort, have a nasty tendency to drop below that nominal figure and the only precaution one can apply knowing the conditions during exploitation is to go a few Volts lower with PSU and increase the amount of current the amp can deliver when the impedance drops. That is of course, limited with many other factors, including the type and capability of the output device(s) to deliver the required "Umph, Kick and Bang" etc... I also opted for capacitors that go maybe halfway up the housing because once boxed and furthermore, when the components "settle in" after a few weeks I'll have to fine tune the amps again and filling the box all the way up would definitely not make my life easier when it comes to fiddling with the tiny screwdriver inside the 50C hot amp...
elac310 said:
This is correct.
Nelson Pass has never in the past to my knowledge advocated explicitly
- Fast Recovery rectifier diodes/bridges
- Special Black Gate caps anywhere for his Diy Audio Amplifiers
www.passdiy.com articles projects
On the contrary!!!!
In his info and articles you find references to ordinary
- Silicon diodes rectifiers of good current rating.
- Standard good quality Electrolytic capacitors for filtering
Now, even Nelson Pass has to follow the road his users takes.
At least take in to consideration.
If people are obscessed with using fast recovery diodes,
at AC Supply frequencies of 100, 120 Hertz
than he did listen a bit .. and probably decided to go with the flow.
If his faithful users likes Dark, Black or any other color of Caps,
why tell them not to
And same with those diodes.
Nelson has never been too dogmatic, about details of less importance.
Fast Recovery will work as well as Silicon ordinary low price diodes.
Why bother .. let them have it
And be happy
lineup said:Now, even Nelson Pass has to follow the road his users takes.
At least take in to consideration.
If people are obscessed with using fast recovery diodes,
at AC Supply frequencies of 100, 120 Hertz
than he did listen a bit .. and probably decided to go with the flow.
If his faithful users likes Dark, Black or any other color of Caps,
why tell them not to
Yes, a few years ago we switched our production over to fast rectifiers.
This was essential to meeting emi standards in some countries we export to,
aside from any sonic improvement.
As to capacitors, I recommend the Elna "silk" caps if you are using
electrolytics, otherwise we are using polyproplene where possible
and where caps are necessary.
But in general I don't like to take sides in the passive parts wars. There
are plenty of others with strong opinions.
The size of Electrolytic Caps? Well, I'm going off thread but I don't think anyone will actually notice...
The Diode thing is not anything to do with ripple @50-60Hz. Any reasonable component spec'ed for the I and V will work in that regard. The only somewhat less than sublte aspect beside the heat dissapation would be what happens when you turn on the amp from a cold state. The inrush current from charging 2.2 Gigafarad of caps will be enough to weld most of your PSU into one solid peice of ??? Those Diode bridges N.P. suggests are relatively inexpensive and generally rated at 35A+, and momentarily much higher, capable of the inrush current of something close to his suggested cap bank.
But, this is DIY Most of us are out to make all the improvements we can afford, figure out, get by the wife, etc. The subtle part... When the diode bridge crosses 0V, it causes a little problem somewhat like a class B amp's crossover distortion. It has to somehow jump over two Vbe's without anyone noticing and it has trouble doing that. For that short time, only during 0 crossing, the standard diode will make a bunch of high freq noise 
You can see sizeable spikes on your scope if you look. I beleive those spikes are going back through the transformer and out the AC line causing the EMI problem N.P. just mentioned. Well, what do you think happens in your amp? That noise burst goes flying right through your 2mH inductor due to parasitic/leakage L and C and goes right into your amp rails 
A CRC may be better at attenuating this noise than a CLC. What about your PSU filtering? Well, I'm sure it helps a little but you have ESR and ESL in your filter caps and possibly even more of a problem with 2.2 GigaFarads worth. Your power Supply works damn good at 50-60Hz but I havent herd anyone say anything about that high freq diode noise???
Basically all these fast, soft, Silicon Carbide, schotky etc. produce less 0 crossing noise Using TV Damper Diodes (tubes) is said to be the best but I shouldn't go there...
I suppose after all that, I could move on with MHO to the topic at hand
Maybe next post...
The Diode thing is not anything to do with ripple @50-60Hz. Any reasonable component spec'ed for the I and V will work in that regard. The only somewhat less than sublte aspect beside the heat dissapation would be what happens when you turn on the amp from a cold state. The inrush current from charging 2.2 Gigafarad of caps will be enough to weld most of your PSU into one solid peice of ??? Those Diode bridges N.P. suggests are relatively inexpensive and generally rated at 35A+, and momentarily much higher, capable of the inrush current of something close to his suggested cap bank.
But, this is DIY
Most of us are out to make all the improvements we can afford, figure out, get by the wife, etc. The subtle part... When the diode bridge crosses 0V, it causes a little problem somewhat like a class B amp's crossover distortion. It has to somehow jump over two Vbe's without anyone noticing and it has trouble doing that. For that short time, only during 0 crossing, the standard diode will make a bunch of high freq noise You can see sizeable spikes on your scope if you look. I beleive those spikes are going back through the transformer and out the AC line causing the EMI problem N.P. just mentioned. Well, what do you think happens in your amp? That noise burst goes flying right through your 2mH inductor due to parasitic/leakage L and C and goes right into your amp rails A CRC may be better at attenuating this noise than a CLC. What about your PSU filtering? Well, I'm sure it helps a little but you have ESR and ESL in your filter caps and possibly even more of a problem with 2.2 GigaFarads worth. Your power Supply works damn good at 50-60Hz but I havent herd anyone say anything about that high freq diode noise???Basically all these fast, soft, Silicon Carbide, schotky etc. produce less 0 crossing noise
Using TV Damper Diodes (tubes) is said to be the best but I shouldn't go there...I suppose after all that, I could move on with MHO to the topic at hand
Maybe next post...- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.