Similarly with the oversized xformer. The effect (and for me the only effect I can think of) is that the charging pulses into the caps get higher and narrower with a more powerful (lower secondary impedance) xformer. So the caps have to survive longer periods between charges. Apart from the increased spray of higher harmonics of the mains.
Whats the advantage??
Jan
Whats the advantage??
Jan
I to would like some clarification on this slowness and what is heard by that. Modern large caps have very low ESR well at least at 100k where I have tested lower than some film caps but not all
.
Nelson Pass has eluded to the short charging pules in low ohmage secondaries when he said he preferred slow diodes. For me the use of a choke to slow that current pulse and lower the spike is the diy way to go . Yes it expensive ,bulky and not high tech enough for some it is however effective.
There is also the thought it makes some amp sound more tube like in a positive way that i will leave to others.Sorry my friend it isn't true if a filter. In modern amps we don't need low ripple, it is totally irreverent. Deep down I agree with you and suggest we might be missing something. In theory we use big caps to avoid the limits of ripple current which might be considerable. 5 V ripple wouldn't matter if we can loose that voltage. If you say keeping it below 3 V gets me 2 V I will say OK. When I see amps full of caps with the lid off I think the worst. Happy to be wrong. Filtering "might" be true. Text books say not true. PSRR is the most often quoted spec. All this assumes class AB or D. What possibly is in question is how does a AB at high current differ form class A ? If A shows ripple hum at idle then that must be true of AB at that current ? If so low ripple might matter. It is not good to say hidden under music. What say you text books? I assume we are using an AB in A to test the theory for as long as the heat allows. As you run a bit of A it might cause you a bit of trouble,You and I battle this out . I suspect nobody reading this knows the definitive answer? The problem with my conjecture is it should show in measurements . Text books seem blind to it. Conversely if A at idle shows it AB at similar current must.
I am hoping someone sides with the books or Dvv. Myself I will slightly distrust the books. My cop out is I double the ripple current for 4 R full power.
Would a high loop NFB amp have better PSRR ? Instinctively it woulld. Does fighting ripple have consequences, inter-modulation of a subtle type?
I have built lots off PSU's where ripple has been a problem. It has been fun to eliminate it. It has always been at the lowest price. It focuses the mind to nail the problem and keep a tight budget. I suspect for that alone I earn my wage.
BTW. Your amps are hardly like the ones I criticize.
Same with slew rates. It is no good saying an amp that doesn't need > 6V/uS ( 100 W 8R ) is obviously better when 100 V/uS . It simply isn't true and no one on the planet can hear that difference ( 14 billion people who might have existed actually ). If it helps a different problem then it can be true. The popular one is RF is involved and JFET input works better if RF is present. . Ben Duncan thinks clipping is helped by high slew rates . Both PSU colouration and high slewing indicate amps being run into clipping routinely . Or at least where critical judgement says " that's much better" . Being honest I am guilty of that. The bass is more solid and the treble never gets harsh. The speakers need changing to have 10 dB headroom if so.
Last edited:
That's the best answer I could give . My brother thought it simple that the LC time constant of the PSU could be heard. He also said keep time constants in the amp as different as you can . He is gone now so can't ask him more. His big statement was the amp was in someways limited by the mains refresh rate. Thus you could argue that bass note was in the music. He felt amplifiers boom due to this.
My brother would happily use 1000 VA and 2 x 4700 uF if ripple current was OK. His motto was too many caps and not enough copper.
I do mean larger parasitic induction, Jan. Larger caps do filter better than same smaller ones, but also have a greater parasitic induction.
To be fair, this same parasitic induction is a problem one has anyway, with all caps. The current gain section of the power amp is only the most obvious and prominent place, since it requires greatest size caps.
The output stage current delivery problem is no different from the same problem for a high speed logic device. The loop inductance supplying current to the transistors, thence to the load is not limited just by the filter caps. The traces, wiring, layout all will affect the performance. If the output device layout is using TO3 transistors and wires all over the place even a big cap with an esr of 1 mOhm and an esl of 10 nH won't make a difference. Start with figuring what the peak current the load will require and what impedance is in series with that load (resistance, inductance etc.). Then work back to make sure you can provide that current. The high frequency current should not be coming from electrolytic caps 12" away with wires in a large loop, or going to speaker terminals with wires in a large loop if you need to drive 8 Ohms at 100 KHz with 10W. However, most likely, the output network will prevent the impedance from being so low at high frequencies. Still small loops, low ESR low ESL caps located close to the transistors with a good layout that doesn't compromise everything will make a big difference. This was the biggest enhancement Keith Johnson made to the Spectral amps.
Spreading out the conduction angle is always good. The rate of change of the charging current will define how much it penetrates into everything. My experiments with series inductors on the supply suggest that the usual concept won't work if you don't get to the critical inductance. The inductor needs to be conducting through the whole cycle to be meaningful. Otherwise its little more than another magnetic radiator causing problems.
Spreading out the conduction angle is always good. The rate of change of the charging current will define how much it penetrates into everything. My experiments with series inductors on the supply suggest that the usual concept won't work if you don't get to the critical inductance. The inductor needs to be conducting through the whole cycle to be meaningful. Otherwise its little more than another magnetic radiator causing problems.
Some say switch-mode helps. To my mind it adds complexity and the 100/120 Hz ripple is still there only worse when 470 uF 400V. I should do the maths as it probably is better than I think . The size alone says it is too small.
It seems we are getting endorsement for large local decoupling if I don't mistake what's being said. I said to add a very small inductor to make an RF Pi filter. It seems that will result anyway and I should have thought of it. Decouplers were always to offset the inductance leading up to the power dumpers, 220uF being typical. It seems it is a good thing. When I started I said 4700uF and 4700 uF decouplers . How do people see that? It is only to adapt a chassis I already have. Cheap PA amp PSU with 500VA and 2 x 4700uF. The weird bit is the amp only claimes 100 + 100 W 8R. Seems this Chinese outfit were an honest company. The toroid is larger than most 500 VA. Quad only fitted 300VA to the 405.
It seems we are getting endorsement for large local decoupling if I don't mistake what's being said. I said to add a very small inductor to make an RF Pi filter. It seems that will result anyway and I should have thought of it. Decouplers were always to offset the inductance leading up to the power dumpers, 220uF being typical. It seems it is a good thing. When I started I said 4700uF and 4700 uF decouplers . How do people see that? It is only to adapt a chassis I already have. Cheap PA amp PSU with 500VA and 2 x 4700uF. The weird bit is the amp only claimes 100 + 100 W 8R. Seems this Chinese outfit were an honest company. The toroid is larger than most 500 VA. Quad only fitted 300VA to the 405.
Last edited:
Sorry, Nigel, that's just not right - ripple in the voltage rails is bad stuff, unless one puts bulletproof rejection circuitry in place it will have major impact - many of the circuits posted on this forum by respected members can be easily shown to introduce distortion artifacts directly, via simulation, and distortion will also occur indirectly, from parasitic behaviours.
Frank I have been playing the other sides advocate. I don't agree with the " text books "and wanted to flush everyone out. I know for a fact when designing class A amps no amount of care substitutes for a good PSU. I have also learnt how cheap it is to power everything up to the drivers with ripple free power. I know class A to be different in that the PSU is more or less constantly doing the same thing, same ripple, day in day out. That means when current used is the same in AB it must be the same colouration. Turn the bias up for 10 seconds on your AB amp to know how good your class AB amp is. If hum is bad it must be doing something under the music at that curent. Move the ground and it might be better. You might only get 10 seconds. Should tell you plenty.
I did the switch-mode maths . Current will be very OK and ripple not. 1500 uF needed which I seldom see. Even then a puny PSU. They might just win more than they loose.
One the whole I think AB should sound better than class A. The PSU has lower ripple when it matters most. Also the transformer can swing a bit which is no bad thing, a storage choke assisted by the resistance and inductance off the incoming supply allowing it in bootstrap fashion rise up. Lights dimming in houses when amps go on must indicate something of that? Not so bad I think as long as not too high an impedance? The speakers and supply a sort of pendulum that class A can not replicate. I think John was hinting at this by his wording?
I would argue that the dumpers 10 V above the VAS will turn the dumpers into a series regulator as in the amplified zener of old. It also ensures softer clipping if driver clips way before the dumpers hit ripple . Simple RC filters will power the driver etc if dropping 10V. If output triples it still can be all stages up to the final one. If extra power needed the R of RC can be reduced in value to drop 2 V.
I worked out that Voyager must be about 24 light hours from Earth. That is mind boggling. Some are saying light speed might be different in space with reduced gravitational effects. So hard to know. If Voyager turned up at the next star 20 years from now we would be so shocked as it should be 40 000. Creationists hold that time has changed and will change outside the solar system. Please do not shoot this messenger.
I did the switch-mode maths . Current will be very OK and ripple not. 1500 uF needed which I seldom see. Even then a puny PSU. They might just win more than they loose.
One the whole I think AB should sound better than class A. The PSU has lower ripple when it matters most. Also the transformer can swing a bit which is no bad thing, a storage choke assisted by the resistance and inductance off the incoming supply allowing it in bootstrap fashion rise up. Lights dimming in houses when amps go on must indicate something of that? Not so bad I think as long as not too high an impedance? The speakers and supply a sort of pendulum that class A can not replicate. I think John was hinting at this by his wording?
I would argue that the dumpers 10 V above the VAS will turn the dumpers into a series regulator as in the amplified zener of old. It also ensures softer clipping if driver clips way before the dumpers hit ripple . Simple RC filters will power the driver etc if dropping 10V. If output triples it still can be all stages up to the final one. If extra power needed the R of RC can be reduced in value to drop 2 V.
I worked out that Voyager must be about 24 light hours from Earth. That is mind boggling. Some are saying light speed might be different in space with reduced gravitational effects. So hard to know. If Voyager turned up at the next star 20 years from now we would be so shocked as it should be 40 000. Creationists hold that time has changed and will change outside the solar system. Please do not shoot this messenger.
Last edited:
Nicely played, . The fact that class A and AB usually sounds 'different' is in large part due to design problems, as I tried, unsuccessfully , earlier to indicate - and here it has much to do with power supplies. Class A 'niceness' can be attained by giving a class AB circuit superb power supplies, but the latter is considered too expensive or overly complicated, not usually considered especially important - hence the typical class AB sound in real components.
. The fact that class A and AB usually sounds 'different' is in large part due to design problems, as I tried, unsuccessfully , earlier to indicate - and here it has much to do with power supplies. Class A 'niceness' can be attained by giving a class AB circuit superb power supplies, but the latter is considered too expensive or overly complicated, not usually considered especially important - hence the typical class AB sound in real components.
In my experience the kind of PSU that's needed with the lowest possible ESL and ESR cannot be built by traditional methods (some caps on the PCB, others on the chassis with huge bus bars). So yes, expensive because it's not amenable to mass production - it needs 3D hand crafting to optimize. I also don't believe its possible to get low enough ESL with multiple TO3 output transistors, the physical loop area is just too big. Chipamps are the way to go
That 2,000W into 1R would need that sort of 3D crafting, , to really get the job done - but with enough thought and perserverance it should be possible. Yes, a megapowerful unit just with chip amps is another approach, I considered it, but the expense of the multiple power supplies just didn't add up the last time I thought about it ...
, to really get the job done - but with enough thought and perserverance it should be possible. Yes, a megapowerful unit just with chip amps is another approach, I considered it, but the expense of the multiple power supplies just didn't add up the last time I thought about it ...
What were you proposing? A production design or a DIY special? Expensive multi-capacitor power supplies can be built with cheapo Chinese caps, the expense is the labour. I recently had one of my hexacaps costed up for manufacture by a local factory - the labour turned out to be 2 or 3 times the BOM cost. So alternatives are being sought....
The big bugbear is voltage swing - how do you get the rails to drive to the requisite peak voltage? Bridging gets you a bit further, then you enter the "interesting" world of floating power supplies - things become cumbersome.
Yes, this is only if aiming for high power into 8 ohm, the 2,000 into 1 should be possible conventionally - but the point of the design exercise was to obtain real world, measurable distortion figures that "looked impressive",. The normal chip amps won't do that using a straightforward topology, it would have to get pretty fancy to extract distortion figures one could boast about, .
Yes, this is only if aiming for high power into 8 ohm, the 2,000 into 1 should be possible conventionally - but the point of the design exercise was to obtain real world, measurable distortion figures that "looked impressive",
. The normal chip amps won't do that using a straightforward topology, it would have to get pretty fancy to extract distortion figures one could boast about, .
Ah, I'm not contemplating a chip amp design with anything like 2kW output, its beyond my wildest dreams (and requirements). A few 10's of W suffices at the moment which means the supply voltage in bridged is about 25V. Caps start getting quite a bit bulkier above 35V so 25-35V is the sweet spot for me at present. And I'm barely interested at all in impressive distortion figures, just engaging sound.
Yes, about 60W into 8 ohms is perfectly adequate, IME. My LM38xx gainclone pair was used to drive these speakers, Technics SB-5000 - Hi-Fi Database - Floorstanding Speakers - which we still have - in a pseudo active setup, ie. mounted vertically behind the speakers on a panel, with very short run to the tweaked crossover. These were capable of doing AC/DC to dance floor levels, and produced engaging sound - did the fully disappearing trick very nicely.
Last edited:
For my sins this is how I say thanks. Simple and I think a little better than usual. There are Sorbothane squash balls for Garrard 301 that might suit. The caps would be BBH or Rifa Evox if keeping the price sensible. The 4700 uF 105C Low Z. Screw terminal might be easiest. The star plate having many fixing points. You might try air-cored inductor to make a RF Pi filter.
The idea is to keep the magnetic field low whilst not lengthening the distance from toroid to caps. 2 x 500 VA as usually better and cheaper plus it can be mono-bloc in the brochure. The vibration reduction should reduce induced EMF due to magnetic flux lines cutting tracks etc. This is topological and not to any grand plan.Unlike most hype this might have legs.
- Status
- Not open for further replies.