High voltage capacitors in audio

[ectoplasm]

Actually, what is the performance of the TIP29, 30 in terms of noise;
I mean, is it a good one for audio purposes?

How about taking the same transistors as in the headphone amp I
mentioned in previous message? Like the 2SD1763, 2SB1186 pair?

 

[John Popelish]

I see. So use a film capacitor there instead of electrolytic.
Polyester would be fine then I guess, as Wimpie suggested (this one
might be too small for that perhaps).

Polyester would probably be good enough that its flaws are
unhearable. My point is, that if the capacitor in question
is not in the signal path (e.g. power supply storage or
decoupling) or in the signal path, but having essentially no
voltage change during signal swings (e.g. coupling
capacitors) or are involved in the signal path but only for
high frequency, ultrasonic functions (like opamp stability)
then their imperfections cannot be heard in the output.

For the few, rare capacitor applications that are in the
signal path and also have significant part of the signal
voltage appearing across them, then the imperfections in
their operation will impress alterations on the signal.
Tone control and other filter functions fall in this
category. Here is a good capacitor tutorial site that
compares and contrasts the various kinds of imperfection
various dielectrics suffer.

http://my.execpc.com/~endlr/index.html

On the other hand, MooseFET says the 10uF at the input is the only one
worth changing. But as others mentioned, electrolytic does fine for DC
coupling (zero bias coupling) & no filter purposes, which is what this
10uF cap does... MooseFET?

I think this cap falls in the category of, "in the signal
path, but seeing insignificant voltage changes with signal
swing" so its dielectric flaws are not important. You just
have to make sure it has enough capacitance to remain in
this category. So the only reason to consider another
dielectric would be if the environment (high temperature or
low pressure) or long lifetime requirement might cause this
capacitor to dry out and loose capacitance.

What value would you suggest then?

I withdraw that comment. I was assuming the output stage
was operating, class AB, where the base capacitance would
have to be charged and dumped each signal cycle. But now
that I have read the description and see that the amplifier
is biased for class A operation, the base capacitance
remains essentially charged and those caps are big enough.
It was also a comment made more with hunch than calculation.
Sorry.

 

[John Popelish]

There is a 10uF after the pot in the path to the following op-amp. No
DC signal can get from the pot through the amplifiers.

I stand corrected. I saw the bottom of the pot connected to
the 10k resistor to the inverting input of the last opamp,
and thought this would be a way for offset to be injected
into the final stage. I missed the fact that the bottom of
the opt also connects directly to the output, which totally
determines the voltage into that resistor.

I appreciate the help.

 

[John Popelish]

Actually, what is the performance of the TIP29, 30 in terms of noise;
I mean, is it a good one for audio purposes?

How about taking the same transistors as in the headphone amp I
mentioned in previous message? Like the 2SD1763, 2SB1186 pair?

I doubt the noise contribution for any functional
alternatives is hearable, let alone, measurable.

I just chose the TIP29, 30 pair because they have fairly
flat gain out to more then the expected peak current, where
the MJE340, MJE350 pair does not. This would lower the
output current requirements of the opamp driver and improve
the linearity during the peaks. But those high gain Zetex
transistors would really lower the opamp load, but could not
handle the heat of full class A operation, without being
embedded in a heat sink (perhaps being inside a heat pipe).
Their die area is about the same as the TIP29-30, so their
safe area is fine, if you can keep their temperature down.

As to the suitability of the 2SD1763, 2SB1186 pair, here are
the data sheets:
http://rocky.digikey.com/WebLib/Rohm/Web Data/2SB1zzz, A and 2SDzzzz, A.pdf
With a 1.5 amp rating and designed as complementary pairs,
they would work just fine. They each have a minimum current
gain of 60, maximum gain 200 at 1 amp, and I suspect might
have a bit more at the half amp expected peak current. But
from this data sheet, I would prefer the narrower gain range
and higher minimum gain pair 2SD2400A, 2SB1569A (minimum
gain 100, maximum gain 200 at 1 amp). I would like to see
the actual gain versus collector current and safe operating
area curves, though. All of these are much better choices
than the MJE340, MJE350 pair in the original design, in my
opinion.

 

[Rich Grise]

You all like kicking class A, eh!

Take a look at this headphone amp:
http://www.headwize.com/projects/showfile.php?file=jorgen2_prj.htm
(Kumisa III, by Benny Jørgensen)

The proof is all there...

(not that I am naive that this is only due to class A... it must be an
excellent design & build)

Man, that's a lot of parts! ;-)

With the right headphones, couldn't you just plug them into line out?

Thanks,
Rich

 

[Eeyore]

You all like kicking class A, eh!

Who said that ?

Take a look at this headphone amp:
http://www.headwize.com/projects/showfile.php?file=jorgen2_prj.htm
(Kumisa III, by Benny Jørgensen)

The proof is all there...

It's absurdly over-complicated for a headphone amp !

And why +- 24V ?

Graham

 

[Eeyore]

So the 470pF / 15pF together

The 15pF isn't part of any filter. It's playing the role of an internal
compensation cap for the 5534. See the 5534 data sheet.

would provide a wrong filter when the AOT
resistor is not zero? Like when it would be 3.3k (as MooseFET
mentioned the cross over point being well above 1 MHz in that case.

So that means if the AOT is set to another value, these two capacitors

One capacitor, the 470pF.

will require different values?

That's why I suggested simulation. I suspect it'll just affect the damping
factor of the filter. Then again, you don't actually *NEED* a filter ! It's
simply over-design.

Graham

 

[Eeyore]

Actually, what is the performance of the TIP29, 30 in terms of noise;

Especially low noise performance is totally irelevant at this point in the
signal chain.

I mean, is it a good one for audio purposes?

What's a bad one ? Do you know the difference ?

How about taking the same transistors as in the headphone amp I
mentioned in previous message? Like the 2SD1763, 2SB1186 pair?

I've used those. I like the fact they have more current gain than TIPs. Also
they have fully isolated packages which makes heatsinking nice and easy (no need
to use thermal pads etc).

Graham

 

[Eeyore]

I doubt the noise contribution for any functional
alternatives is hearable, let alone, measurable.

Surely measurable and hearable should be reversed in that sentence ?

Graham

 

[Eeyore]

Man, that's a lot of parts! ;-)

With the right headphones, couldn't you just plug them into line out?

Here's a simple headphone amp !

Take say, 10 high-performance audio op-amps per channel and 'combine' their
outputs via 82R each.

That'll give you an 8 ohm output with ~ +- 300mA drive capability.

Graham

 

[GregS]

Here's a simple headphone amp !

Take say, 10 high-performance audio op-amps per channel and 'combine' their
outputs via 82R each.

That'll give you an 8 ohm output with ~ +- 300mA drive capability.

But its not class A !

Otherwise its sounds like a good amp system, but a bit much. I guess you need about 10
to do it right, but would work OK at low volumes with a lot less.

I like the simple designs. Hey heres one..............

greg

 

[Eeyore]

But its not class A !

You can easily make it class A by adding a 300mA current sink ! It'll run hotter of course.
Interestingly, with suitable DC feedback, you could use this to switch between class A and AB
operation instantaneously.

Graham

 

[Tim Williams]

Brain damage appears to be synonomous with audiophoolery.

Tim

 

[Tim Williams]

Sure, if it was there to begin with.

Well, presumably, LPs are rich in 0.55 Hz noise.

If you never play records, it's probably not an issue, no.

Tim

 

[Eeyore]

John Popelish wrote

Sure, if it was there to begin with.

There is often plenty of LF energy that the average loudspeaker can't reproduce.

Graham

 

[Jamie]

In a pre-amplifier, I am replacing all relevant capacitors in signal
lines by 100V or 250V polypropylene film capacitors (i.e. big yellow
ones). These are much better for audio than standard alu elcos.

My question is:

1) the signal is only a couple of volts in amplitude. Is it bad that
the capacitors are way beyond this value? I mean, could it be that
these big caps handle low voltages not as well?

2) is it actually a good idea to replace *all* of these capacitors by
these big ones? (if you'd like, look at the circuit at <http://
sound.westhost.com/project24.htm>) Or would some good, small tantalum
capacitors do better in some locations?

If you had to ask that kind of question, how did you come to the
conclusion that polypropylene was a better choice?

 

[Jamie]

Subsonic energy being produced by your headphones or speakers is a waste
of amplifier power that can cause them to distort the stuff you can
hear. It is usually a good thing to get rid of.

Yeah, but what if you were super man? Would you want your
ears to be denied the range of full quality noise ?

 

[me]

Well, presumably, LPs are rich in 0.55 Hz noise.

If you never play records, it's probably not an issue, no.

Tim

I played few records with it, however my other amp (without the maketing
gimmic non hearable filter) played lots of records while sounding as good
or better...

I can't think there could be that much input below 20Hz from a 33.33... rpm
record. Clicks, pops, and rumles certainly but I can hear those...

 

[MooseFET]

I see. So use a film capacitor there instead of electrolytic.
Polyester would be fine then I guess, as Wimpie suggested (this one
might be too small for that perhaps).

On the other hand, MooseFET says the 10uF at the input is the only one
worth changing. But as others mentioned, electrolytic does fine for DC
coupling (zero bias coupling) & no filter purposes, which is what this
10uF cap does... MooseFET?

The 10uF at the input and the about 22K it is loaded with makes for a
cut of frequency of about 0.7Hz. At reasonable volume settings, this
defines the low frequency cut off of the whole circuit. If any
coupling capacitor has an AC voltage on it, this is the one that does.

The 22uF, which I said served no purpose, ends up setting the low
frequency limit when the gain is all the way up.

 

[MooseFET]

You all like kicking class A, eh!

Take a look at this headphone amp:http://www.headwize.com/projects/showfile.php?file=jorgen2_prj.htm
(Kumisa III, by Benny Jørgensen)

The proof is all there...

Check out what they call "well damped" gag.