Output capacitor: subjective and objective views?

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
The one and only
Joined 2001
Paid Member
I still believe that the signal does flow through the PSU caps and the PSRR is not relevant except for being a good indicator of feedback factor.

Of course signal has to flow through the capacitors, but it is
easy to build a no-feedback amplifier with a decent PSRR, and
if you have only the power supply caps to worry about, PSRR
is your best friend.

:cool:
 
retitle thread

I believe this thread should be retitled "are power supply capacitors really necessary?" Mooly's discussion of the output capacitor on single supply transistor amps at the beginning seems to me to be the only other one that discusses amps with speaker preserving capacitors in the output.
 
Disabled Account
Joined 2006
Mooly's discussion of the output capacitor on single supply transistor amps at the beginning seems to me to be the only other one that discusses amps with speaker preserving capacitors in the output.

indianajo, if you could add anything to one of my previous post (repeated below) I would be all ears!


Well guys, we certainly do have a good range of responses.

The opinion seems to be split bewteen:
*You can hear an output capacitor, but only if your system is resolving enough
*You can not hear it if it is big enough and of 'good quality'
*It must be placed in a global feedback loop in order to not hear it

Underlying my original question was the notion of designing a small and simple amp without a global feedback loop, and I think that using an output capacitor would help in these ways...
*Easy design of Single Ended stages
*Everything conveniently 'ground referenced'
*Relatively simple single polarity power supply
*No need for a DC servo
*No need for DC offset protection circuit

In post #44 tiefbassuebertr kindly mentioned some other points.

Do you think there are any other potential advantages?
 
I can think of a couple of snags:
A capacitor could fail short (unusual) and fry your speakers
The bigger the capacitor, the more the energy stored and likely the more spurious signal from vibration.
Most amplifiers do not take feedback from the output as there are big problems at switch on and low frequency stability
 
Disabled Account
Joined 2006
I can think of a couple of snags:
A capacitor could fail short (unusual) and fry your speakers
The bigger the capacitor, the more the energy stored and likely the more spurious signal from vibration.
Most amplifiers do not take feedback from the output as there are big problems at switch on and low frequency stability

Good points. I suppose the professionals would compare data to see which is most likely to fail short... an output cap or a push-pull transistor. I think most transistors are likely to fail open (...with a BANG!). I don't know about caps.

I didn't realise that susceptibility to vibration increased with energy storage. Perhaps mechanical damping might work.
 
how do capacitors fail

I just replaced about 60 electrolytic capacitors in a Hammond H182 organ manufactured 1968. All except one were low in value, that is the capacitance was not adequate to do the job. The one was the motor start capacitor, an AC electrolytic, that dried out, shorted, and fried the wiring harness before the house breaker tripped. (No fuse). I just bought another H100 for spare parts. The motor start capacitor, dated 1968, sits in a puddle of leaked fluid, but still started the motor one more time when I bought it. I have just left a job replacing three phase motor drives in a food plant. Many drives and power supplies are replaced because their electrolytic capacitors leak a puddle of slime, then fail to allow the unit to start on a cold morning. I take that as a symptom of low capacitance. I have seen one drive and one DC power supply with shorted blown up electrolytic capacitors.
My apologies to Gordy, his posts are on the topic as advertised.
I am interested in protecting speakers against amps, as I have retired 2 sets of speakers over the years due to open tweeters. Both were connected to the ST70, a low power tube amp not known for oscillating typically, and transformer coupled against DC. I view the output capacitor on transistor amps as one level of protection against DC faults, and possibly against the frequent bolts of lightning we get here in the midwest. One topic I am interested in, were the 3300 uf 80V capacitors I replaced the output capacitors of the ST120 with of adequate ESR? The vendor did not specify the ESR, nor did Dynaco. But the dynaco manufactured output capacitors were certainly MUCH bigger than the ones I replaced them with.
I looked over the schematic of the Peavey CS800S today. It has one input DC blocking capacitor on the input in the schematic, and no more that I can tell. The schematic dated 1995 is on eserviceinfo.com if anybody is interested in looking at it, I'm looking at the one with the pcb layouts also. I'm struggling to hook it up tonight to do an AB comparison vs the ST120 DJoffe modified in the same room. Bleah- I'm getting too stiff to hold wire between my toes to solder the connector on. If I can bend my knee enough (only one will do it) I need reading glasses to see the connector.
 
Last edited:
I believe this thread should be retitled "are power supply capacitors really necessary?" Mooly's discussion of the output capacitor on single supply transistor amps at the beginning seems to me to be the only other one that discusses amps with speaker preserving capacitors in the output.

Though, we know now that in any case the signal has to go through the caps.
So there is not to worry that much about output caps.
 
Administrator
Joined 2007
Paid Member
We are still unsure aren't we ?

Some real world pictures to illustrate.
Note the PSU and just 10uf caps to give masses of ripple. Note the 10 ohm in series with each (think of the e.s.r. with those) added to enable the current to be seen.

The "amplifier" is as bad as it gets... an unbiased buffer.
I used 100 hz as a test frequency to enable pictures to be photographed from the scope.

Image 1 shows the input to the buffer as the top trace at 10 volts pk/pk, the output across the load as the middle and the correspnding ripple component in the lower. The ripple is approx 1.5v pk/pk measured across the upper 10 ohm.
The output is perfect, even with only 10uf in the PSU and masses of ripple. In fact the ripple is larger than the signal... but the output is unaffected.
Due to the way the scope triggers the phase relationship between the ripple and the signal is not accurate... more on that later.

Image 2 shows the above but with a 10uf cap "in series" with the load... as we expect it's effect is dramatic as it forms a low pass filter.

The output current really does pass through this.

Image 3 shows a sinwave input... DC coupled, so no loss of LF.

Image 4 again a sinewave, now AC coupled via 10uf, loss of LF.

Image 5 is with the buffer fed from the 50 hz secondary voltage (via a divider of course) to show the true relationship between the ripple component and the signal. The waveform is distorted because that's exactly what it does look like on the secondary when the transformer is supplying the current to the circuit.

So I hope that explains what I am trying to put across.
 

Attachments

  • Ripple.jpg
    Ripple.jpg
    50.3 KB · Views: 302
  • IMAG0001.JPG
    IMAG0001.JPG
    75.4 KB · Views: 298
  • IMAG0002.JPG
    IMAG0002.JPG
    72.3 KB · Views: 297
  • IMAG0003.JPG
    IMAG0003.JPG
    77.7 KB · Views: 276
  • IMAG0004.JPG
    IMAG0004.JPG
    71.5 KB · Views: 275
  • IMAG0005.JPG
    IMAG0005.JPG
    74.9 KB · Views: 58
Last edited:
We are still unsure aren't we ?

Some real world pictures to illustrate.
Note the PSU and just 10uf caps to give masses of ripple. Note the 10 ohm in series with each (think of the e.s.r. with those) added to enable the current to be seen.

The "amplifier" is as bad as it gets... an unbiased buffer.
I used 100 hz as a test frequency to enable pictures to be photographed from the scope.

Image 1 shows the input to the buffer as the top trace at 10 volts pk/pk, the output across the load as the middle and the correspnding ripple component in the lower. The ripple is approx 1.5v pk/pk measured across the upper 10 ohm.
The output is perfect, even with only 10uf in the PSU and masses of ripple. In fact the ripple is larger than the signal... but the output is unaffected.
Due to the way the scope triggers the phase relationship between the ripple and the signal is not accurate... more on that later.

Image 2 shows the above but with a 10uf cap "in series" with the load... as we expect it's effect is dramatic as it forms a low pass filter.

The output current really does pass through this.

Image 3 shows a sinwave input... DC coupled, so no loss of LF.

Image 4 again a sinewave, now AC coupled via 10uf, loss of LF.

Image 5 is with the buffer fed from the 50 hz secondary voltage (via a divider of course) to show the true relationship between the ripple component and the signal. The waveform is distorted because that's exactly what it does look like on the secondary when the transformer is supplying the current to the circuit.

So I hope that explains what I am trying to put across.

Very interesting; now the next step must be the same checking by load between 2-4 ohms with the appropriate power buffer stage and supply; the
capacities now must be larger, 120ohms : 2ohms = 60 x 10uF ~0,5 til 1 mF and the external ESR? I would say 10 ohms : 60 ~ 100mR - 220mR.
By unsymmetric version ~250uF-500uF and for the couple cap? this is hard to say, because I have now clearly a typical high pass character. Check out 250 - 2200 uF.
 
Administrator
Joined 2007
Paid Member
Nice one Mooly!

If the output cap was suitably large so as to not produce a low frequency RC roll-off in the audio band, do you think it's distortion / effect when passing a few Watts of signal would be measurable?

The effect is easily measureable, but that's not quite the same as saying it's audible.

Distortion takes many forms, amplitude distortion, phase shift... all are forms of distortion... all are things the cap will add at LF... but I would go so far as to say it shouldn't be audible given suitably large values.
Whatever amp you have now, if it's DC coupled try adding a non polarised cap in the output and see if you feel it's audible.
 
The psu cap is fast charged through 10 ohms. it is discharged through 120 ohms + the output transistor "resistance", which takes a lot longer. It is maintained quite full. And, its is working one way. It is a reservoir.

In your example, the output cap is charged from the PSU through the load and the output transistor with a worse time constant than the psu cap is charged, at the signal level only, and this cap is emptied when the signal is crossing zero. It is not used as a reservoir, it's not kept full. it just passes a signal.

That's why i said previously that the psu caps are "in parallel" to the load, opposite to the output cap case which is in series.

Nevertheless, a big output cap solves the problem in the audio world.

NB: In a single supply amp, the output cap is polarised. Which is different from your example. It works differently.

Hope your cap is non polarised.
 
Last edited:
AB dual supply vs single supply amp with cap

Okay, I finally figured out how to get input into the Peavey CS800S split supply direct coupled fully complementary amp. The previous owner had cut the jumpers from the XLR connectors, which removed the input caps from the path, and the signal has to go in on the "thru" phone plug, which has no caps. This amp was purchased in 1998, and the schematic is on eserviceinfo.com. The Dynakit had the power supply and output (speaker) capacitors replaced in '88, and all the rest of the electrolytic caps replaced this spring. The Dyna has a 5 uf ceramic cap on the input, formerly tanalum. The Dyna has most of the "TIP" mod excepting the input transistors are still high gain BC108A for PAS2 compatibility. The outputs of the Dyna are NTA60's on one side and NTE181's on the other, much higher gain transistors than the 2N3772 as originally tested in 1966. The Dynakit had the Djoffe high bias current mod this spring, which improved low volume highs. With SP2 speakers, I really can't tell any difference. The first pass I had 40' of 16 ga on the Peavey, and that was clearly inferior to the 8' of 10 ga cord I usually use on the Dyna, so second pass, I made the fat cord fit the Peavey. The test record was Peter Nero "Warm and Wonderful" with piano, strings, and some percussion. The piano sounds were so good, both made me want to take the tuning key to my Steinway console, which has drifted off tune due to summer humidity. Both amps were superior to a Koss K/6x headphone run directly out of the Herald RA88A mixer driven by a 33078 op amp. The Herald has been modified so all op amps are 33078's, for less hiss.
If you try this at home, the Dyna capacitors are 3300 uf @ 100VDC. If you put some on a split supply amp, you will probably need 100 V rated caps (non-polar).
 
Mooly,
In a single supply amp, the output cap is charged at Vcc/2, discharged of the signal amount. It works one way. We are nearer from the psu cap case.
In your example it has no dc voltage across it, only the signal. And the voltage across it has to invert, crossing zero.
This zero crossing may add some distortion IMO.
We cannot really compare the two cases.
 
Last edited:
Administrator
Joined 2007
Paid Member
Bobo,
I added an output cap to demonstrate the effect of it's series reactance on the output current, and how that effect was not the same as DC coupling the output and transfering that "output cap" reactance to the PSU... the same old argument "that the speaker is in series with the PSU caps" which I thought the above 'scope shots showed what really happens without doubt.

Distortion in capacitors that pass a signal... what I think you are asking now is "does the DC bias across the cap" have an effect.

In a single rail amp there is always a constant DC voltage across the output cap... in my example above there is not. Any distortion or difference in distortion produced by not having a DC voltage would be miniscule, certainly totally invisible on a 'scope, you would need something like the Audio Precision to even begin to look at this.

You will find opinions divided as to whether electroylitic caps are audible or not, opinions are also divided as to whether it's better to have no DC bias across a the cap, or to have some bias.

As with anything, try it for yourself... I certainly haven't got all the answers, but I do question things and like to see for myself by experiment and test what really happens.

A good line to finish with is,
"So you have built the perfect amplifier, it has technical specs beyond reproach"

So why don't you like listening to it, and why do other amps "sound" better ;)
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.