One always discovers interesting "things" when perusing - further in that Engineering article, under Problem Solving -- Tacit knowledge - Wikipedia, the free encyclopedia ...
Yes, most people, most of the time, when they're trying to get something done,or to sort out an issue are doing this: Problem solving - Wikipedia, the free encyclopedia.
This is pretty universal, no certification or special training required ...
This is pretty universal, no certification or special training required ...
Or, you can do what some of us others do, which is to split the PSU lines and use higher voltage, fully regulated input stage and VAS supples.
How many more jumps through the hoop will you do until you come to the obvious, Nige?
It is a cheap and obvious upgrade . Even the mighty Naim Audio failed to see it . They gave a CD player 27 supplies if I am right ?
Some guy on a Naim clone kit thread had a slightly over voltage transformer . The outputs were fine at that voltage . I showed how to put in a simple amplified Zener regulator in to make it work ( VAS +/- 40 V ) . I was very surprised to get both reservations about the idea and advice on better regulators . So from no regulator to state of the art in a few sentences . In a nutshell how things always go . The zener had the filtering cap to remove hiss . An LM317/337 would have been more problematic . I used fast power transistors . I felt a bit sad about it . Also a guy who made a monster amp who was starting to tweak it . He had slipped a couple of diodes in from output to VAS . I suggested to remove them and try a nicer PSU . Nothing said by anyone .
If you want to try something as cheaply as possible try using a capacitor input voltage doubler ( Cockcroft Walton style ) . That allows the existing PSU to be used without conflict as the capacitor prevents short circuiting . The time constants should be so that the supply is just sufficient for needs as a starting point . A very fast transistor and zener good enough to remove harmonics . Use the spectrum analyzer to look at residuals and try increasing cap size . Even 220 uF starts to work if 15 mA used . After that you now can have whatever you want . Try 10 V below , equal . 10 V above . I suspect 10 V below will be favourite . Converting the zener to capacitance multiplier is worth a thought . I recommend not to use Darlingtons . If needed a FET . Thing to realize is a 220 uF 250 V is going to be an Audiophile device even if it isn't labeled that way . The high voltage dictates that . Most high voltage caps have better tan theta . It is a byproduct of resisting flash over . If using some soft recovery diodes and a few caps it is doubtful you will exceed $15 . My hunch is it will sound like $300 invested in more PSU caps . Usually that involves serious re engineering if it is not to have increased zero volt problems ( ill defined star point , and / or bad HF zero due to thickness of metal one is forced to use ) .
This is the problem with the engineers . Douglas Self team using 10 000 uF and a 40 year old Sinclair design ( Not RCA ) . Everything analyzed and no progress made . Then the brute force engineers . Like the Dodge Viper they can not be wrong . Where are the Lotus 7 's ? NAD 3020 is as near as we got I feel .
You feel wrong, Nige, but the REAL problem is you're too stubborn to take a nice, long look to see why.
How many times so far have I asked you to take a look at Harman/Kardon's work from say 1985 to this day?
How many times have I sent you examples of modern schematics, only for you to come up the next day with another 40+ year old variant of your own?
You don't have to agree, you don't have to like it, you don't have to use it, but you should at least see how far SOME have gone since the days of Sinsclair.
How many times so far have I asked you to take a look at Harman/Kardon's work from say 1985 to this day?
How many times have I sent you examples of modern schematics, only for you to come up the next day with another 40+ year old variant of your own?
You don't have to agree, you don't have to like it, you don't have to use it, but you should at least see how far SOME have gone since the days of Sinsclair.
My point is to breech the gap between the ones I admire most ( HK et al ) and cheap engineering .
The Sinclair brigade have been refining a reasonably good design for 43 years now . Having listened to all variants more or less the power supply stands out as the main difference . You might have guessed my day job is as a power supply designer . In my work I often have pennies to get a result . I get great job satisfaction from delivering the near impossible . When I see typical amplifiers I see the money very badly spent . I have absolutely nothing bad to say about brute force regulation and use it myself . I am convinced when it can not be used there are exciting possibility which if it entered my work I would be jumping for joy to solve . Audio designers seem to have read a religious text that says one way and one way only .
You know this makes me think . If I offered a service where any common place amp was brought to me I would do the upgrade I described . It would need a handful of components and no additional transformer . I would offer a small reduction in power with override. The formula is simple . Keep ripple bottom from signal top .
The Sinclair brigade have been refining a reasonably good design for 43 years now . Having listened to all variants more or less the power supply stands out as the main difference . You might have guessed my day job is as a power supply designer . In my work I often have pennies to get a result . I get great job satisfaction from delivering the near impossible . When I see typical amplifiers I see the money very badly spent . I have absolutely nothing bad to say about brute force regulation and use it myself . I am convinced when it can not be used there are exciting possibility which if it entered my work I would be jumping for joy to solve . Audio designers seem to have read a religious text that says one way and one way only .
You know this makes me think . If I offered a service where any common place amp was brought to me I would do the upgrade I described . It would need a handful of components and no additional transformer . I would offer a small reduction in power with override. The formula is simple . Keep ripple bottom from signal top .
Nigel knows this but for others who might read it:
You also have to leave room for the amplifier, above the signal and below the ripple, even if it's just Vceo plus the voltage across a 0.22 Ohm emitter resistor.
Some of the LMnnnn chip amps (e.g. LM3886 and LM3875) give the "Vclip" voltage as a function of supply voltage. It's typically 3 to 4 volts. It's the minimum voltage between each power pin and the output pin.
When the bottom of the ripple dips down into the amplifier's "vclip" voltage space, it tends to gouge ripple-shaped chunks out of the output signal. (Actually, it's whenever the bottom of the ripple and the top of the signal get closer together than the amplifier's vclip voltage.)
It's VERY similar to what happens to a linear regulator when the input voltage dips too low.
It's amazing how many people just subtract the estimated p-p ripple from the rail voltage and assume their signal can have that max amplitude.
Here's a recently-calculated one that looks correct:
http://www.diyaudio.com/forums/chip...gainclone-active-amplifier-4.html#post3554629
I usually look at it the other way, i.e. calculate the minimum C required for a given desired max output power rating, instead of the other way around. Just make sure that your caps' ESR is much less than your target PSU impedance!
---
If we take into account the ESR, using the approximation
ESR = 0.02 / ( C x Voltage_Rating_of_Caps)
then the overall (approximate) equation for the minimum value of the required capacitance is
Cmin (in uF) = 1000000(Vpk/(Rload(Vrail-Vclip-Vpk)))( (1/(2 fmains)) + (0.02/Voltage_Rating) )
---
I think that's the same as or similar to
C ≥ Δi / ( 2∙π∙fmains∙(Δv - (ESR∙Δi))), if Δv = Vrail - Vclip - Vpk.
We can see that if Δv - (ESR∙Δi) = 0, then C would be infinitely large.
To understand "why", we can write it slightly differently:
Δv = ESR∙Δi
or
Δv / Δi = ESR
But notice that Δv/Δi is the impedance we would need to stay below, in terms of the impedance that the active devices would "see", if the voltage should never change by more than Δv, for any Δi.
The ESR is the minimum impedance we can get, with a capacitor. So if the ESR is already as large as Δv/Δi, then we have already failed to meet a basic need of the power-distribution circuit, which would be required in order to meet our Δv/Δi specification.
That simply means that we need to try to make the ESR be much less than our desired Δv/Δi. Otherwise, the required capacitance value will be excessive.
Last edited:
In short use more than you think ....
Actually, the equation that I gave for Cmin is really the MOST you should need, because I structured it so that it will be enough capacitance to prevent clipping even if you pump out constant DC at the peak rated voltage and current levels, i.e. not the RMS levels of the current and voltage at the max rated power, the PEAK levels, instead.
Essentially, that means that if you use the calculated Cmin value, your amplifier won't clip with a sine wave until it reaches 1.414 times the rated max output power.
And it is guaranteed not to clip with ANY type of signal that doesn't exceed the max rated power, whereas with the usual method of using RMS sine values, you're only guaranteed to not clip with a single sine wave at the max rated power.
Edit: Of course, if your caps have really crappy ESR, and the approximation I used is better than the actual ESR you're using, then nothing will save you.
Edit2: Or, you could be saved, if you used the actual ESR with this equation:
C ≥ (imax∙Δt ) / (Δv - (ESR∙imax))
and set Δv (the max ripple) = Vrail - Vclip - Vpk, and set Δt = 1 / ( 2 fmains ), and set imax = Vpk/Rload.
The attached spreadsheet is very simple to use, and gives a range of capacitances and the max rated power ratings that result from them.
Last edited:
Note that the Vpk value is something that you choose. It's the desired peak output voltage for a sine at the max rated output power.
If you try to push it too high, before the equation explodes the required C value will become excessively large.
If you have in mind a desired max rated RMS output power, say Prms, you can find the corresponding Vpk value with
Vpk = √(2∙Rload∙Prms) , i.e. the square root of 2∙Rload∙Prms .
If you try to push it too high, before the equation explodes the required C value will become excessively large.
If you have in mind a desired max rated RMS output power, say Prms, you can find the corresponding Vpk value with
Vpk = √(2∙Rload∙Prms) , i.e. the square root of 2∙Rload∙Prms .
Last edited:
Last edited:
Any ESR that is enough lower than the desired PSU output impedance as seen at the point of load, could be called "low enough", since it will then not cause the need for an abnormally-large reservoir capacitance.
So you figure out the maximum ripple you can tolerate, based on your desired max rated output power and the Vpk output voltage peak that it implies, i.e. Vpk = √(2∙Rload∙Prms) and then Vripple = Vrail - Vclip - Vpk . I called Δv = Vripple, above.
So then, you can figure out the maximum current swing that the output device's power pin might require the PSU to provide at the point of load (the power output device). It would probably be when the output swings from zero to Vpk, for whichever rail we're talking about. So the current through the output device should go from 0 to Vpk/Rload, in the worst case. So the change in current would be Vpk/Rload. That's the Δi.
Using Ohm's Law, when a change in current causes a change in voltage, the impedance is their ratio, just like in a resistor. So in this case the impedance seen by the output device, looking back at the PSU power and ground rails, would need to be no higher than:
Z <= Δv / Δi
Z <= (Vrail - Vclip - Vpk) / (Vpk/Rload)
So your capacitor ESR should be "much less" than Z.
If the ESR is NOT "much less than" Z, then the term in the denominator of the equation that was given, Δv - ESR x Δi will be close to zero. And dividing by something close to zero gives a LARGE result, which in this case would be the required minimum capacitance.
We can also express Z in terms of the desired max rated RMS output power, Prms:
Since Vpk_max = √(2∙Rload∙Prms_max),
Z <= (Vrail_min - Vclip_max - √(2∙Rload∙Prms_max)) / (√(2∙Rload∙Prms_max)/Rload)
Then buy capacitors with
ESR << Z .
Or, if not, at least understand that the practical minimum ripple amplitude will be larger for caps with higher ESRs.
And thus the rated max output power will need to be lower, with higher ESR caps, if a practical (i.e. non-excessive) total reservoir capacitance is used.
If the required Z is too low, then instead of lowering the max rated output power, Prms_max, or raising the C value, you could simply raise the rail voltage, Vrail.
Cheers,
Tom
So you figure out the maximum ripple you can tolerate, based on your desired max rated output power and the Vpk output voltage peak that it implies, i.e. Vpk = √(2∙Rload∙Prms) and then Vripple = Vrail - Vclip - Vpk . I called Δv = Vripple, above.
So then, you can figure out the maximum current swing that the output device's power pin might require the PSU to provide at the point of load (the power output device). It would probably be when the output swings from zero to Vpk, for whichever rail we're talking about. So the current through the output device should go from 0 to Vpk/Rload, in the worst case. So the change in current would be Vpk/Rload. That's the Δi.
Using Ohm's Law, when a change in current causes a change in voltage, the impedance is their ratio, just like in a resistor. So in this case the impedance seen by the output device, looking back at the PSU power and ground rails, would need to be no higher than:
Z <= Δv / Δi
Z <= (Vrail - Vclip - Vpk) / (Vpk/Rload)
So your capacitor ESR should be "much less" than Z.
If the ESR is NOT "much less than" Z, then the term in the denominator of the equation that was given, Δv - ESR x Δi will be close to zero. And dividing by something close to zero gives a LARGE result, which in this case would be the required minimum capacitance.
We can also express Z in terms of the desired max rated RMS output power, Prms:
Since Vpk_max = √(2∙Rload∙Prms_max),
Z <= (Vrail_min - Vclip_max - √(2∙Rload∙Prms_max)) / (√(2∙Rload∙Prms_max)/Rload)
Then buy capacitors with
ESR << Z .
Or, if not, at least understand that the practical minimum ripple amplitude will be larger for caps with higher ESRs.
And thus the rated max output power will need to be lower, with higher ESR caps, if a practical (i.e. non-excessive) total reservoir capacitance is used.
If the required Z is too low, then instead of lowering the max rated output power, Prms_max, or raising the C value, you could simply raise the rail voltage, Vrail.
Cheers,
Tom
Last edited:
I got to think about this via the old Quad valve designs . There are many of the same problems . Quad had used the money in a very cunning way . The very high quality supply to the 2 x EF 86 in what valve people refuse to call long-tail pair ( the tail is so short I guess it isn't LTP ) . The 2 x KT 66 are not quite like the current amp output stage of transistor amps , not far off when output transformer is added . The system works well until about 10 watts . Then one hears dirty bass that is ripple modulated . At no point do measurement make it obvious . Up to 10 watts it is arguable the amplifier is like an all choke filtered supply and nasty capacitors avoided . A Siemens cinema amp also . 400 V to EF 41 x 2 ( ? ) 800 V 2 x EL 34 = 100 W pentode 13 dB feedback . The EL 34 on very raw DC ( 600 V stand by switched from projector ) .
With chip amps I wonder if some class C dumpers from a higher voltage supply could be used ? Simple feed-forward correction coming in high enough to do more good than harm ? In which case perhaps the same PSU for all with RC filtering to the chip . 5 watts should be a good switching level . Still simple enough for gain clones . I said once why gain clones can not quite equal home-brew amps as regards bias . I built what they show internally and think the bias is about 0.3 V below optimum . This prevents thermal problems . I was asked if Audiophile capacitors would change the sound after having identified why people find the sound mildly opaque .
With chip amps I wonder if some class C dumpers from a higher voltage supply could be used ? Simple feed-forward correction coming in high enough to do more good than harm ? In which case perhaps the same PSU for all with RC filtering to the chip . 5 watts should be a good switching level . Still simple enough for gain clones . I said once why gain clones can not quite equal home-brew amps as regards bias . I built what they show internally and think the bias is about 0.3 V below optimum . This prevents thermal problems . I was asked if Audiophile capacitors would change the sound after having identified why people find the sound mildly opaque .
- Status
- Not open for further replies.