If you're so much into minimalising and decoupling, why not go for cap multipliers on the amp boards, next to the output stages.
Cheaper than electrolytic capacitors.
(e.g. Metaxas Solitaire 2002-2012 model, page 20 : http://www.metaxas.com/pages/___SOLITAIRE_MANUAL_FINAL.pdf )
Last edited:
And way more efficients...
At 10000uF/rail , ripple will be at least 1V at full power , commanding
60dB NFB for the amp to correctly reject it to 1mV at his output , still a mediocre value.
A basic cap multiplier will reduce such a ripple by 40dB at the expense
of about 2V rail voltage losses , quite a bargain....
I ask, again, a silly question, not think that I was wrong thread.
- What relation between the energy stored in the power supply and power output of the amplifier?
Not trying to be a good designer to "Undersizing" nor so ignorant to "Oversizing", the truth is I do not have plenty of money.
- What relation between the energy stored in the power supply and power output of the amplifier?
Not trying to be a good designer to "Undersizing" nor so ignorant to "Oversizing", the truth is I do not have plenty of money.
A "capacitance multiplier" is simple enough that it's faster to describe than to draw the circuit. Uses an NPN for the positive rail and a complementary circuit with PNP for negative. Collector goes to the + of the main capacitor reservoir, emitter to the + power rail of the amplifier, and the rail current flows through this transistor. Put a resistive divider between the main capacitor's + and ground, with the junction going to the base. Resistors should be scaled for (supposing a 50V rail) setting the base at 48V, and low enough resistance to supply enough base current to insure the transistor's current gain (or lack thereof) doesn't drop the voltage at full volume. The capacitance multiplication comes from connecting a capacitor between the base and ground. The capacitance will effectively be multiplied by the beta of the transistor.
The bottom resistor of the divider might go better to the emitter rather than ground, and of course be a proportionately lower value. (thus R1 would be between collector and base, and R2 between base and emitter rather than between base and ground).
Basically, in a few short words, it's an emitter follower with the input (base) AC-coupled to ground.
The post from Benb explain it well , i will just add that a darlington
is to be used.
One of the advantage is inherent soft start....
Here an exemple i simulated , with voltage unfiltered at the extremes
of the graph and the filtered voltages , of lower value of course , slightly
above/below .
The current is about 7 to 8A , we can clearly see the difference
in ripple level , as showed by a fourier analysis.
Attachments
Why not use the cap follower ONLY on the driver stages and let the output stages see the raw DC? It is what Matti Otala and I do. The cap multiplier is MUCH reduced in size, heatsink, and cost, AND the output stage can really drain the caps, with transients, giving MUCH MORE potential peak current (what you really need).
John, do I understand this properly that you run your predriver and driver transistors off a fully regulated power supply, and let only your output transistors have the juice from the capacitor filters only?
If so, good for me, because that's what I usually do; somehow, it seemed logical to me to do just so, and I can't complain about the results, it's a lot better than running only the voltage gain stages from regulated supplies. It makes the amp sound more sort of focused.
Please forgive my little drawings . These are my notes I keep . If wondering why 2 diodes ? The theory is the diode RF noise is minimized ( soft recovery diodes ) . This is for 4 op amps . It can be scaled up for higher current and / or voltage . I use a cascaded circuit as I feel is has slight stability advantages . This was not the final form of it , alas I forget what file name I gave it . Protection diodes should be fitted if not fitted inside the transistors ( also to transistor bases ) . Capacitors should be Rubycon ZL Panasonic FC or other high grades . Note the use of fast transistors . BD139/140 D44/45 and 2SA1943 2SC5200 suggested to cover all needs . MOSFET's are also suitable . There is a small amount of transistor hiss . If you look at graph No3 it fills in the bumps in the graph . If looking at the resistor divider only it is marginally lower . However it has bumps .
When we were discussing if very small hiss in power amps is desirable a few weeks ago I thought of this . It is just possible this sounds nicer when this low . A sort of dither .
The small caps were 10 nF . The 4R7's part of a protection system ( transistors added , BC550/560 , useless as 3000 uF caps charge too quickly ) . In the end I added constant current sources and removed the 4R7's ( 2 x BD135 2 x BD136 ) . This allows a very long cable to be fitted without RFI problems . The charging of 3000 uF was a problem ( hundreds of amps in theory ) . I gave up trying to suppress the didoes ( RC+C snubber ) as nothing was seen on the analyzer ( up to 16 MHz ) .
The subsonic spectrum was also very good . This is a pre-regulator so is fine at near 20 V . There was a 6 dB absolute improvement ( LM317/337 , not my choice ) . The other engineer was using 2 diodes also and 1000 uF + 47 R + 2000 uF + LM317/337 + 33uf + 2 uF ( 12 V ) . 1K load was not specific to the op amps , just a starting value .
Last edited:
I agree with Dvv . I had a Quad 303 go wrong where the regulator 2N3055 went short and survived . Suddenly 89 V was available ( usually 67 V ) . My friend disliked the repair . I should have fitted modern MJ transistors to replace 2N3055's for him and removed the regulator . The difference was interesting . Far more alive when broken . I preferred the purity of the standard 303 myself . If I had done a 75 V driver supply and 89 V dumper supply with MJ 15003 dumpers I suspect that would have been perfect ?
When class B amp it is perfect to use the current dumpers at a slightly higher voltage on the raw DC ( usually the other way around when chasing the last watt ) . The dumpers then become regulators in their own right if the driver voltage is held below the ripple levels ( as done in the Quad 11 tube amp and Siemens EL34 cinema amps 2 x EF41 (LTP ) 2 x EL 34 100 W pentode class AB2 820V fixed bias !!!! 1.3 % distortion 10 W , choke regulated driver , very raw DC to EL34's , 9dB feedback , 600 V standby on projector synchronized relay , only Siemens tube good enough I found ) . Then even 4700 uF is possible . As long as the ripple current rating is sufficient we should believe what Douglas Self says that no one can hear the difference . I suspect lots of caps in amps is about a feel good factor . Do I fit loads of caps ? Yes I do because with care of layout it does no harm and might just do some good . Do I hear better bass ? Not really when it gets to 15 000 uF . Class A is a totally different story .
I have to say crafty class B amps with distrotion cancelling ( like op amps ) probably sound better than many class A designs ? This is because the distortion is well under control and the power supply is miles away from saturation . Class AB better than any other ? Probably yes . Class A/C the best ( class G , AC hybrid ) ? I don't know . On paper that looks certain to be right ( put glitches at painful sound levels , 20 W will not be far off ) . Sliding class A ? hard to say , looks good on paper .
When class B amp it is perfect to use the current dumpers at a slightly higher voltage on the raw DC ( usually the other way around when chasing the last watt ) . The dumpers then become regulators in their own right if the driver voltage is held below the ripple levels ( as done in the Quad 11 tube amp and Siemens EL34 cinema amps 2 x EF41 (LTP ) 2 x EL 34 100 W pentode class AB2 820V fixed bias !!!! 1.3 % distortion 10 W , choke regulated driver , very raw DC to EL34's , 9dB feedback , 600 V standby on projector synchronized relay , only Siemens tube good enough I found ) . Then even 4700 uF is possible . As long as the ripple current rating is sufficient we should believe what Douglas Self says that no one can hear the difference . I suspect lots of caps in amps is about a feel good factor . Do I fit loads of caps ? Yes I do because with care of layout it does no harm and might just do some good . Do I hear better bass ? Not really when it gets to 15 000 uF . Class A is a totally different story .
I have to say crafty class B amps with distrotion cancelling ( like op amps ) probably sound better than many class A designs ? This is because the distortion is well under control and the power supply is miles away from saturation . Class AB better than any other ? Probably yes . Class A/C the best ( class G , AC hybrid ) ? I don't know . On paper that looks certain to be right ( put glitches at painful sound levels , 20 W will not be far off ) . Sliding class A ? hard to say , looks good on paper .
Last edited:
The problem with regulators in general is their speed, which sometimes leaves much to be desired. This is further compunded by the power factor of a power amp regulator. In essence, the regulator is really another power amplifier itself, and all problems related to speed apply to the regulator even more so than the power amp per se.
On the other hand, a regulator required to stabilize supply lines for EVERYTHING BUT the output stage has a much smaller power factor, and ultimately, you can always throw in fast output power transistors, such as 2SC5200/2SA1943 as the final stage and sleep soundly. At that level, even shunt is still in play, I doubt the driver will use more than 0.5 A of current, and that is still managable without breaking the bank.
I am fond of this system because it allows me to use the output transistors at a lower voltage, thus keeping it better inside its SOAR, without sacrificing the projected power. Also, as I stated previously, I feel a shunt regulator makes things better focused, so to speak, it simply makes the whole a bit more clear and precise.
Having made just 5 or 6 such power amps over the years, I cannot boast of a statistically meaningful sample figure, but it's clear enough to convince me.
On the other hand, a regulator required to stabilize supply lines for EVERYTHING BUT the output stage has a much smaller power factor, and ultimately, you can always throw in fast output power transistors, such as 2SC5200/2SA1943 as the final stage and sleep soundly. At that level, even shunt is still in play, I doubt the driver will use more than 0.5 A of current, and that is still managable without breaking the bank.
I am fond of this system because it allows me to use the output transistors at a lower voltage, thus keeping it better inside its SOAR, without sacrificing the projected power. Also, as I stated previously, I feel a shunt regulator makes things better focused, so to speak, it simply makes the whole a bit more clear and precise.
Having made just 5 or 6 such power amps over the years, I cannot boast of a statistically meaningful sample figure, but it's clear enough to convince me.
That s a good idea since the front end is in fact the part that will
benefit the more from a cap multiplier but then you wont elude
the necessary big cap at its output as the front end rails must
be very tightly coupled to ground in AC mode even at low frequencies.
I think for driver stage although it is slightly stupid to use slow regulators because it saves no money it probably does no great harm . This is not true of the output current dumpers and an extra reason to feed them raw DC .
Where a very fast transistor might work better is in keeping RFI out of a system as it's Ft is high . I suspect shunt is less good from this point of view ? I am happy to be told otherwise and do not know . People forget that driver stages are very nearly always class A and thus in themselves almost constant current devices . A small local capacitor should stop any TID or IM distortion of PSU origin . We overlook this at our peril . Solder an extra high grade ceramic across +/- of op amps ( 1 to 10 nF usually npo / cog , silver mica if wanting a feel good factor ) . Forget the 1 cm away advised in books as being OK .
Where a very fast transistor might work better is in keeping RFI out of a system as it's Ft is high . I suspect shunt is less good from this point of view ? I am happy to be told otherwise and do not know . People forget that driver stages are very nearly always class A and thus in themselves almost constant current devices . A small local capacitor should stop any TID or IM distortion of PSU origin . We overlook this at our peril . Solder an extra high grade ceramic across +/- of op amps ( 1 to 10 nF usually npo / cog , silver mica if wanting a feel good factor ) . Forget the 1 cm away advised in books as being OK .
Building on the above, I would really love an explanation why do the voltage gain stages, which almost always operate in true class A, still benefit from using a shunt in comparison with classic regulation schemes?
I take shunt regulation to be a sort of a pure class A regulator.
I posed a question on a thread about shunt regulators concerning how well they work at 100 MHz for something I was working on . It seemed that above 1 MHz they stop working and it is whatever decoupling we use that does the job . This makes me doubt that shunt regulators work in the ways people suppose . The series regulator seems to me to work in unison with the driver stage . The constant current source version even more so . The shunt regulator might work against it if not careful . Doubtless with practice these things are learnt and I respect that my position is one of prejudice .
People forget the very best shunt device is a zener . I seldom see them used . It's ills can be cured ( hiss ). Often the 5 watt ones are good enough for any sensible use . I notice tube designers are very happy with them . OK they are not LM317 accurate . Apart from a high power voltage switcher I built I have never needed that accuracy .
I can see many theoretical objections to using series regulators . However with the transistor I listed and the method I showed ( which can be converted to a classic series regulator ) I see no serious problems .
Funny thing is series regulators don't seem to oscillate unless tragically wrong . Tube amps seem to do it all the time . The latter seems to be forgiven and worked to a cure , then pronounced to be a pride and joy possession .
I think the reason people suppose shunt to be better is that it "seems " not to be in the serial signal path .
People also forget in a classic series regulator the zener is the shunt device . The transistor just a high speed buffer when ideal . as long as it doesn't oscillate where can it go wrong ?
When people do build these things they go complexly bonkers and make it as complicated as any amplifier . One zener , one transistor and a few resistors . Job done and perhaps 40 dB better noise ( or better ) . The amp probably has > 80 dB CMRR at 55Hz ( middle 50/60 ) to start with . 120 db seems a good standard . Doubt most people will get the earthing that good , especially if using many PSU capacitors ( star earthing problems ) .
Good points . If wanting to regulate the output stage then I totally accept that and about the inductor . Do we always want to clamp an inductor ? The impedance is debatable as both work well . In a driver amp I think ground current problems are easy to spot .
Would it be correct to say shunt for output stages is especially good ? Perhaps more so for class A .
Some will say class B is superior to class A as it can swing above it's rails ( not clamped ) . This was an argument put forward by the British press about 20 years ago . I notice no one says it now . The fact that the PSU is not saturated helps it do that ( and the series resistance/inductance of the mains supply , 21 A peaks were noted at clipping on a 100 W amp ) . When truly blameless class B arrived this became my point of view . I can honestly say amps like that when in class A or B sound mostly the same ( Yamaha made one years before the author of blameless amps ) .
I ask, again, a silly question, not think that I was wrong thread.
- What relation between the energy stored in the power supply and power output of the amplifier?
Not trying to be a good designer to "Undersizing" nor so ignorant to "Oversizing", the truth is I do not have plenty of money.
Crick, crick,....,crick
I've asked that before, too, and never gotten a real answer. Joules stored in the PSU vs output watts.
- Status
- Not open for further replies.