speaker dc offset question

[AndrewT]

But at low frequencies the capacitors impedance rises and so does the a.c. voltage drop across it. Using a large valued capacitor will keep it's impedance low at low frequencies. Although I haven't seen people do this by design, it might be worth limiting the low freq. bandwidth of the amp earlier, say at the input where a non-electrolytic cap is used, so the output cap never experiences frequencies which might produce distortion.

This is precisely the lesson I learned from the experts on this Forum. I have been promoting that same message ever since, possibly for >5years now.

Set the passband with the two input filters.
Ensure the NFB RC time constant exceeds the input DC blocking RC time constant. The only discussion left is by how big a factor the NFB RC should exceed the Input RC.

Similarly the smoothing capacitance RC must exceed the NFB RC.
In a dual polarity supplied amplifier the caps are in series. In a single polarity supplied amplifier, one of those caps is relocated to be in series with the load.
It is those same two capacitors whose RC must exceed the NFB RC.

 

[DF96]

Although I haven't seen people do this by design, it might be worth limiting the low freq. bandwidth of the amp earlier, say at the input where a non-electrolytic cap is used, so the output cap never experiences frequencies which might produce distortion.

I assume all competent designers do this, although they might not draw attention to it.

 

[system7]

This particular amp is a rather different ZCA MOSFET zero feedback design, AndrewT.
DIY Class-A 2SK1058 MOSFET Amplifier

Any non-linearity in the bass is as likely to be down to the single MOSFET itself, or power supply, as the capacitor. I'd be careful what I conclude here.

I remember the Celestion Ditton 15 with ABR (that passive driver at the bottom, which acts as a reflex) which IIRC had a different tweeter from your 15XR (pictured here):

[URL="

An externally hosted image should be here but it was not working when we last tested it.

What I was getting at with remarks about output impedance, is that a 15 ohm output amp will be tonally different from a pure voltage amplifier with speakers. The bass will be less tightly damped, and the treble will rise at high frequency as discussed in that Arpeggio article. For all I know, those vintage speakers are designed for valve amps, so you may have got lucky here. But interestingly, you can work round the impedance issue anyway with crossover design.

 

[AndrewT]

Any non-linearity in the bass is as likely to be down to the single MOSFET itself, or power supply, as the capacitor. I'd be careful what I conclude here.

Is this message implying or accusing me of supplying the wrong information to Bigun and Hameay?

 

[system7]

Is this message implying or accusing me of supplying the wrong information to Bigun and Hameay?

No criticism implied at all. Just don't want hameay fretting uneccessarily about an output electrolytic capacitor that may well be working just fine!

The devil is often in the detail with these sort of things.

 

[AndrewT]

Are you saying that I am misinforming Hameay?

 

[system7]

Are you saying that I am misinforming Hameay?

Andrew, if I understand what you are saying here, you are talking about a NFB amp. This isn't a NFB amp. It has an inherent distortion as output rises.

But it certainly makes sense to limit the amplifier input low frequencies too. Everybody is agreed on that.

Let's not go off-topic into a squabble.

 

[AndrewT]

Now that you are being specific, I can help you understand what my post was about and why I am giving Hameay the correct information.

I gave three separate conditions for good amplifier passband practice.
1.) set the pass bands limits with input filters.
2.) set the NFB RC > Input RC
3.) set the PSU/Output RC > NFB RC.

Where there is an open loop amplifier, then the NFB RC does not exist.
A combination of conditions 2 & 3 gives
2a.) set the PSU/Output >>Input RC.

What is wrong with that advice? as implied with

I'd be careful what I conclude here.

meaning you conclude to something different to what I was recommending.

 

[system7]

OK, I think I follow. You want to make the input filter RC of the amp smaller than the output RC on the speaker/output capacitor...I'd hope the designer has done that and haven't worked it out, but here's the circuit:

An externally hosted image should be here but it was not working when we last tested it.

DIY Class-A 2SK1058 MOSFET Amplifier Project

I wasn't planning to redesign the amp, as it goes...

 

[hameay]

Thanks for all the information guys. It's really appreciated and definitely broadens my knowledge of electrolytics and their characteristics. Given my current level of electronics expertise it may well take me a few months to digest it And all from a question about DC offset at the speaker outputs I'm looking forward to exploring the issue more with my signal generator and oscilloscope when I get time, and will report back any interesting findings.

system7: yes - those are the speakers. My tweeters "look" the same as those in your picture. They seem to work well with the amp, which the designer recommends sensitive speakers for. I've been looking into redoing the crossovers (still the originals) and will be interested to see how they change the quality of the sound when I finally get round to it.

 

[AndrewT]

The input RC is 1M * 0.47uF = 0.47seconds.
The output RC is 15 * 4710uF = 0.07seconds.

I recommend that you change both filters to achieve the conditions for near zero AC voltage across the output capacitor.
You may also have to consider changing the PSU smoothing capacitor to a similar value to the new output capacitor, unless you have modified the PSU to show a different output impedance, i.e. a regulator, or cap multiplier, etc.

 

[system7]

An externally hosted image should be here but it was not working when we last tested it.

Yes, AndrewT, I got much the same results. That input capacitor could be 1/10 the size and still work at 20 Hz. And I noticed the power supply capacitor was smaller than the speaker capacitor too.

There is a thread about this amp:
http://www.diyaudio.com/forums/solid-state/111775-class-mosfet-amplifier-ps.html

There is a thing in engineering we used to call "The Law of Unexpected Consequences" though.

Small changes can have unexpected effects. You might, for instance, change hum rejection, overload power, effectiveness of fuses and input linearity too.

hameay, this amp of yours is very nice indeed. I want one!

 

[AndrewT]

Thank goodness we are agreed.

 

[system7]

Thank goodness we are agreed.

Quite. Harmony restored. Shouldn't you have allowed for the input impedance of the MOSFET though...just sayin'...

 

[Bigun]

for those of you who like this simple amplifier design you should look on this forum for the threads discussing the SEWA and OTA amplifiers. The SEWA was highly regarded.

 

[AndrewT]

.........you have allowed for the input impedance of the MOSFET ...........

Is it effectively 1M or 1G or 1T.
I guess I must have erred towards 10M

 

[system7]

For sure there's a small input capacitance on a MOSFET gate. Can't remember how much it is, but it particularly matters if you have no feedback...

@AndrewT, I like your idea of considering the RC constants at various stages of the amp. For maximum power transfer IIRC, source and load should have the same impedance, so I'm thinking, particularly with power supply RC and speaker/output capacitor RC, they should be the same. It's nicely economical too, and all the electrolytic capacitors get equally non-linear at low frequencies that way...

BTW, just fell in that 470 uF on the power supply is really much smaller than you'd use on most amps. My little Rotel has (poly bypassed) 2X 6,800 uF on the supply. An odd design decision.

 

[hameay]

I wasn't very happy with reports on the performance for the power supply so I used this cap multiplier instead (courtesy of Kinnja) given in this thread:

http://www.diyaudio.com/forums/solid-state/111775-class-mosfet-amplifier-ps.html#post1403783

 

[hameay]

And actually, I was wondering if the lack of the 470uF and 100nF caps before the load resistor in the cap multiplier compared to the original amps schematic would make any difference?

 

[DF96]

In an audio amp we rarely want maximum power transfer, but efficient power transfer. As Andrew and I have said, you don't want any signal voltage across electrolytics - no advantage in ensuring that they all have the same small signal voltage.