I will, thank you - pretty interesting.
Maybe it's just a matter of personal preference. I like DC servos. Though capacitor is a much simpler solution and in many cases does the job.
The easiest solution is to do nothing to the amplifier, and delay the turn-on of the output relay until after it 'thumps'.
The problem is the opamp is inside the feedback loop of the amplifier. You have no control over the turn-on behavior of the opamp, and the data sheet tells you absolutely nothing about what happens when you turn it on. It will be nearly impossible to fix because there are no bias currents or voltages shared between the internal opamp circuits and the output stage like there is in a fully discrete amplifier.
You might get it lower by fiddling with the relative turn-on times of the output stage vs the opamp, but I doubt it will ever be low enough for headphones, and why waste the time when a relay delay will eliminate it 100%?
The problem is the opamp is inside the feedback loop of the amplifier. You have no control over the turn-on behavior of the opamp, and the data sheet tells you absolutely nothing about what happens when you turn it on. It will be nearly impossible to fix because there are no bias currents or voltages shared between the internal opamp circuits and the output stage like there is in a fully discrete amplifier.
You might get it lower by fiddling with the relative turn-on times of the output stage vs the opamp, but I doubt it will ever be low enough for headphones, and why waste the time when a relay delay will eliminate it 100%?
Nothing wrong with servos.
Properly implemented capacitors offer a lot of advantages outside of lower cost and complexity. Understanding how to implement them is how you avoid their pitfalls.
And don't think that commercial designs are always optimized either. I've seen terrible implementation of electrolytics in consumer audio circuits.
If I remember correctly, Bateman's research indicated that an electrolytic introduces negligible distortion when biased with a DC voltage and is configured so its impedance is negligible in a circuit. In a split supply circuit (where they are typically employed without DC bias) their performance is different; but they can be reverse biased to a certain voltage before introducing distortion (I don't remember how the threshold is determined). But even with this constraint, we see that we can still pick an electrolytic that will provide good performance; just choose one large enough so the AC voltage across it is virtually zero at all frequencies.
I still use some kind of film capacitor (I bought a bunch of them with a lot of capacitors and they're physically large but work great) in a lot of coupling applications. But electrolytics are fine in single supply applications.
I still use some kind of film capacitor (I bought a bunch of them with a lot of capacitors and they're physically large but work great) in a lot of coupling applications. But electrolytics are fine in single supply applications.
What sort of sound ?
Oh Mooly tell vzaichenko. please!
Watch this space
We 'aint done yet.I seem to remember that Bateman in his latest report said that adding a DC bias made distortion worse.
Single Electrolytics were poor performers compared to the plastic film types.
In order of bad to very good.
polarised low voltage electrolytic
polarised high voltage electrolytic
non polar electrolytic
back to back polarised electrolytic
back to back non polar electrolytic.
This last was almost as good as plastic film caps.
This is from memory. Go read his latest very detailed report.
Back to back non polar may well be bigger and may be more expensive that a good MKT and similar to a good MKP.
Single Electrolytics were poor performers compared to the plastic film types.
In order of bad to very good.
polarised low voltage electrolytic
polarised high voltage electrolytic
non polar electrolytic
back to back polarised electrolytic
back to back non polar electrolytic.
This last was almost as good as plastic film caps.
This is from memory. Go read his latest very detailed report.
Back to back non polar may well be bigger and may be more expensive that a good MKT and similar to a good MKP.
Last edited:
But this is for distortion testing when a significant AC voltage exists across the capacitor.
When a correctly sized coupling cap is used, the AC voltage across it becomes insignificant, probably around -60dB to -80dB ref the signal passing to the receiver circuit. Now determine the distortion on the AC Vdrop across the correctly sized coupling cap. If it is -40dB ref the Vdrop then that distortion becomes -100dB to -120dB below the passing signal.
If a very good back to back non polar electros were to be used and the distortion is -80dB ref the Vdrop, then the distortion added to the pass through signal becomes -140dB to -160dB.
The distortion created by a correctly sized coupling cap, borders on the unmeasurable (even by Bateman's newly developed test apparatus) and many consider completely inaudible.
When a correctly sized coupling cap is used, the AC voltage across it becomes insignificant, probably around -60dB to -80dB ref the signal passing to the receiver circuit. Now determine the distortion on the AC Vdrop across the correctly sized coupling cap. If it is -40dB ref the Vdrop then that distortion becomes -100dB to -120dB below the passing signal.
If a very good back to back non polar electros were to be used and the distortion is -80dB ref the Vdrop, then the distortion added to the pass through signal becomes -140dB to -160dB.
The distortion created by a correctly sized coupling cap, borders on the unmeasurable (even by Bateman's newly developed test apparatus) and many consider completely inaudible.
Last edited:
Which won't happen if you select your values correctly.
Correct.
Bateman provides actual measurements, which are surprisingly low.
That's how we hope it works out.
You post some good stuff sometimes, Andrew.
I keep posting references to Bateman's work because it separates fact from fiction in an objective way. A lot of people seem to design fully DC coupled stuff, thinking that they are getting some benefit. While this is OK for the DIYer that understands what's going on, it is not OK for consumer or pro grade equipment. If you put just one coupling capacitor in the audio path (at the input of the power amplifier), you might save yourself an expensive and unfortunate incident.
Alright, I want to end the debate about the offset voltage being the real issue here. I connected a cap and a diode in series to the output of the amplifier to see how high this spike actually is. I measured 7V on one channel and about 5V on the other channel, adding about 0.7V diode drop, there's no wonder I find this thump annoying (good thing I haven't broken my headphones). Unless there's something I don't understand about offset voltages, a 5-10mV offset does not cause a 7.7V spike, does it?
I must admit that a delay followed by a reed relays 'click' sounds wonderful, feels like the amplifier is 'warming up', getting ready before it says 'click' to show that everything is settled I must say that it's a tedious solution though, I was hoping for a different fix Then again, I keep hearing delayed relays clicking in all kinds of hifi-gear so I assume that this is a pretty common thing regardless of amplifier design?
I must admit that a delay followed by a reed relays 'click' sounds wonderful, feels like the amplifier is 'warming up', getting ready before it says 'click' to show that everything is settled
I must say that it's a tedious solution though, I was hoping for a different fix Then again, I keep hearing delayed relays clicking in all kinds of hifi-gear so I assume that this is a pretty common thing regardless of amplifier design?
I agree 100% the problem is not dc offset.
The problem is opamps have no turn-on pin or soft turn-on feature. You could pull a hundred 5532's out of a tube and every one could turn on different. That is why Pro and Home audio signal processing equipment all have relays or mute circuits in the output - because if they use opamps, they all have turn-on/turn-off pop in the signal path. The mute circuit is always designed 'slow-on, fast off' so when the equipment rack is turned on, it has a delay to wait for all the dc offset to settle, then it closes the relay/unmutes. When the power is turned off, it mutes fast so no turn-off pop will get through to the speakers.
The problem is opamps have no turn-on pin or soft turn-on feature. You could pull a hundred 5532's out of a tube and every one could turn on different. That is why Pro and Home audio signal processing equipment all have relays or mute circuits in the output - because if they use opamps, they all have turn-on/turn-off pop in the signal path. The mute circuit is always designed 'slow-on, fast off' so when the equipment rack is turned on, it has a delay to wait for all the dc offset to settle, then it closes the relay/unmutes. When the power is turned off, it mutes fast so no turn-off pop will get through to the speakers.
You can build a "soft start" circuit into your design. Of course this is much more effective with single ended supply circuits.
Relays are cool too; but of course increase complexity and cost. And it's one more thing to fail; I have just has several receivers in a row in my shop that had burned up relay contacts. That's much less of a problem with low level circuits, of course.
I always try to avoid relays in my designs, but I still use them sometimes.
That's what I was thinking of, but I have never implemented it.
Has anybody ever devised an electronic shunt to take care of thumps? With a series resistor, it ought to work for a headphone amp. I have heard of "crowbar" circuits employed as speaker protection. What do you think?
Test results
Here we go. Ran some tests - results are here: Cap vs DC servo.
The cap is good enough, but not as good as DC servo. Depends on particular requirements and... attitude, if you want.
Here we go. Ran some tests - results are here: Cap vs DC servo.
The cap is good enough, but not as good as DC servo. Depends on particular requirements and... attitude, if you want.
Hmm, removing the input capacitors seemed to completely remove the pop sound. I'm not sure of what to make of this. I understand that these input capacitors are charged at start up, but I don't understand how they can cause the output to give a 7V spike. With a gain of 13 it means that something like 500mV is present on the non-inverting input at start up
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.