Andrew
While the nominal phase margin will increase with gain, these amps are stable at any configuration over a gain of 10. The 3875 has 50dB of open loop gain at 20k (and that won't be every chip) so, with a gain of 33dB you are leaving it with only 17dB available to correct its inherently quite high distortion (there are huge thermal gradients across the chip for a start). By the time you reach 100k or so, you are asking the chip to do more than it can manage and it has completely run out of gain and therefore out of feedback. This is never a good idea. You usually want to demands on the amp to run out long before it reaches unity gain, this being a key element of avoiding instability.
Personally I doubt very much that higher gains lead to improved performance. Why would one want to put the gain in a chip that creates 0.1% distortion at 20x when you can have an op amp in front of it doing the extra gain for you at almost zero distortion cost.
Anyway, it's not a hard thing to replace a 100R with a 220R.
If the problem persists at x20 then there could be an argument for put a CR in parallel with the 4k7 (which I would anyway raise to 20k or so and increase the 220 to 1k or thereabouts) in order to reduce the closed loop gain. This requires care and it is probably worth just lifting the configuration from the data sheet.
Edit.
The graph for OLG on the 1875 doesn't go down this low, but I suspect it won't even be as good as the 3875 if they are not printing it. Use the 3886 data sheet for the parallel R and C values. IIRC it is about 50pF and 22k again.
Last edited:
Christian,
I don't understand your argument.
I refer to National's Open Loop Frequency Response on p12.
The graph shows:
~82degree phase margin @ G=+26dB.
~80degree phase margin @ G=+20dB.
~60degree phase margin @ C=+10dB.
and falling very quickly at lower gain.
80 degree phase margin should be in the right ballpark and allow gains of 10times or more to drive reactive loads without oscillating.
Reducing the gain significantly below 10times moves the amp into an area where it could easily become unstable if there is any reactance in the load or subjected to very fast signals.
Do National specify a limiting closed loop gain. I can't find it.
I had a look at the 1875 graph and had a wobble trying to figure out what National have shown. I can't make sense of it..
I don't understand your argument.
I refer to National's Open Loop Frequency Response on p12.
The graph shows:
~82degree phase margin @ G=+26dB.
~80degree phase margin @ G=+20dB.
~60degree phase margin @ C=+10dB.
and falling very quickly at lower gain.
80 degree phase margin should be in the right ballpark and allow gains of 10times or more to drive reactive loads without oscillating.
Reducing the gain significantly below 10times moves the amp into an area where it could easily become unstable if there is any reactance in the load or subjected to very fast signals.
Do National specify a limiting closed loop gain. I can't find it.
I had a look at the 1875 graph and had a wobble trying to figure out what National have shown. I can't make sense of it..
Last edited:
That graph is the Open Loop Gain and phase which I think you are misinterpreting. When you close the loop you will have just a horizontal line at say 26dB and what this graph tells you is how much available gain there is inside the amplifier above that. Ie the gap between the two. The phase response of your output does not follow this graph either, though it will be somewhat dependent on it. What one is trying to avoid is having the output phase having shifted through 180 degrees because then, in your feedback loop, what you think you are subtracting from the signal you are in fact adding. Hence you get oscillations. That phase shift can come from all sorts of places like stray capacitances and inductances or from things you have done deliberately between the input and output. (I realise that on a chip amp there isn't much between the input and output, since it is all inside, but on most amps there is.) The phase margin is how far away you end up being from that totally inverted signal after everything is done. The further you are away the more tolerant to load and so on the amp would be. The idea is generally to be on the high side of 45 degrees. But to do this the amp needs to have feedback available to it otherwise you will simply follow the OLG curve. So what must happen is that by the time you reach the 180 degree mark your demand on gain is actually negative, ie. that you have gone through unity gain much lower down.
National specify only a minimum stable gain, ie x10. This is one at which it should work but if you do the wrong thing, it may not. There is no access to the internal compensation so there is nothing you can do to move this, but if you stick within reason, their compensation (which I imagine is that 10pF capacitor plonked in the middle of the equivalent circuit) will do it for you. I would say that x48 is outside reasonable and that one can't expect the amp to work with at high frequencies so little loop gain available to it.
As for people saying it sounds better at higher gains, I suspect this partially comes from the idea that less feedback is a good thing (this is wrong in almost every instance,BTW) and from people thinking they are squeezing more out of the amp. The usual mistakes beginners make with amplifiers is to pick the highest voltages possible and the highest gains possible in the belief that they will then be getting the most out of the amp and, hopefully getting something for free that isn't in the spec sheet. Believe me, if they could get more out (without them blowing up) they would be shouting it from the spec sheet. Having said that, you can improve on their distortion levels by choosing a lower gain than they use and lower voltages, but then you don't get your 56W or 68W amp. See Bob Cordell's book or website for some ways this can be done.
Hope this helps.
Christian
It's a superb book. If it gets beyond you in places, go back to the Art of Electronics by Horowitz and Hill. Between them they should cover most things.
Just a quick question. Have you got an output capacitor (2200uF in the diag) on your amp? Without it you will just have DC coming out of the end of it, which would account for the wire getting hot, the resistor burning etc. This is a single supply application and those 22k resistors are meant to give half the voltage on the input.
Incidentally, because you are going directly into the + input - which has an impedance of several Mega Ohms, there will be no difference to talk of between a 1uF and a 100uF input capacitor.
I have a feeling this might be your problem.
Incidentally, because you are going directly into the + input - which has an impedance of several Mega Ohms, there will be no difference to talk of between a 1uF and a 100uF input capacitor.
I have a feeling this might be your problem.
Well, Christian was right, you skipped the output capacitor. That could also have killed your speakers, not only the resistor. Look at the Single Supply Application Circuit in the datasheet and add the 4700 µF cap that is drawn there.
It is hard to understand, how you can be building amps for quite some time now and still refuse to learn even the basics.
It is hard to understand, how you can be building amps for quite some time now and still refuse to learn even the basics.
No, a big cap is not needed with a split supply but you will need to have + and - 20V on each amp and op amp.
BTW, I'm not sure why you have such a complicated second section on the woofer. If you are doing a Butterworth and want to invert it (which may not be necessary and will be worth trying both ways round), then just have 10k going into the inverting input, 10k on the feedback loop with 10nf in parallel with it (ie in the feedback loop too). That gives you the real pole that you need.
Christian
BTW, I'm not sure why you have such a complicated second section on the woofer. If you are doing a Butterworth and want to invert it (which may not be necessary and will be worth trying both ways round), then just have 10k going into the inverting input, 10k on the feedback loop with 10nf in parallel with it (ie in the feedback loop too). That gives you the real pole that you need.
Christian
This is small size bi-amp active loudspeaker. For this I have used 2" Tweeter and 5" woofer driver. If you look at post # 39 photo, I have used cheap (low cost) components as possible.
First stage is input buffer with small gain. 2nd part is 12db/oct Linkwitz-Riley crossover. crossover point is 1591hz. At high-pass section RC shelving high-pass filter added. Whch equalise and remove peak for me. Last stage is power amp.
Woofer channel starts with 12db/oct LR crossover followed by biquad/6db diple equ. circuit. which boost low frequencies to closed box and remove peak for me. Shelving low pass also subside mid-frequency and drive woofer softly.
High-pass section is 4-6db more sensitive to woofer, so woofer gain is more than tweeter.
After your advice, I have decided to do some alterations to low-pass section. Power amp gain=19 (similar go tweeter amp) and at biquad 22k replace with 10k resistor.
Best Regards.
First stage is input buffer with small gain. 2nd part is 12db/oct Linkwitz-Riley crossover. crossover point is 1591hz. At high-pass section RC shelving high-pass filter added. Whch equalise and remove peak for me. Last stage is power amp.
Woofer channel starts with 12db/oct LR crossover followed by biquad/6db diple equ. circuit. which boost low frequencies to closed box and remove peak for me. Shelving low pass also subside mid-frequency and drive woofer softly.
High-pass section is 4-6db more sensitive to woofer, so woofer gain is more than tweeter.
After your advice, I have decided to do some alterations to low-pass section. Power amp gain=19 (similar go tweeter amp) and at biquad 22k replace with 10k resistor.
Best Regards.
HI, that's all fine - I hadn't realised you were doing EQ as well. You may find you don't need it. It's quite hard to do without measurements to back it up.
But the important question is whether you have resolved the issue of things burning up? I would look at the 3886 or 3875 application circuits and then adjust those to what you need. Just ignore the extra pins. Just make sure that you have fully understood the distinction between the single and dual supply circuits. It's a shame that the 1875 datasheet gives you so little. Or look at Bob Cordell's designs in the book. The inverted one should give some improvements over a standard design (and you have got an op amp to switch that back already if you need to.)
Best of luck
Christian
Also you may find that as bypass capacitors for the 1875 you might want to put 47u - 470u in parallel with those 100nF caps. That will be sure to get rid of any oscillation problem that is starting ii the big chip.
It can be 47k, but with the low values for your feedback circuit I would expect a much lower value to work better. Probably 4k7 for the tweeter and 6k8 for the woofer. Test several values and measure the output offset. Use the value that leads to the lowest offset voltage.
Without Rin the amp inputs have no reference to ground and the output will move to one rail.
Without Rin the amp inputs have no reference to ground and the output will move to one rail.
You can have it at 47k or 100k, or lower if you want. Put it in after the 1uF cap just like you have on the input to the 5532. The only thing this will affect is where the bass rolloff starts and you are in perfectly good territory with those values. If you go below about 22k you will want to increase the size of the cap, probably. But for the time being let's just let you get it working.
Oh, yes, PB is right on offset. A good rule of thumb is to have the values seen by each input be the same. That means the parallel of the two feedback resistors on the one side against what is seen on the other. But I wouldn't worry about this just yet. Check your offset and if it is below 30mV then march ahead and try it. It will probably be fine on higher values than that, (especially as you have already put 1/2DC into it, I guess <g>)
Oh, yes, PB is right on offset. A good rule of thumb is to have the values seen by each input be the same. That means the parallel of the two feedback resistors on the one side against what is seen on the other. But I wouldn't worry about this just yet. Check your offset and if it is below 30mV then march ahead and try it. It will probably be fine on higher values than that, (especially as you have already put 1/2DC into it, I guess <g>)
Last edited:
@ChristianThomas: Yesterday, I applied your advices GAIN=20x, now gain is 1.8K/100 = 19 at both channels. Sorry, Listening-wise I did not find any differences. But the satisfaction is I have build power amp to national datasheet specification.
@pacificblue: Thank you for detailed post for Rin. Pls. give some time to try. I did my recent alteration prior to read your post # 57.
@pacificblue: Thank you for detailed post for Rin. Pls. give some time to try. I did my recent alteration prior to read your post # 57.
I'm glad to hear it is working. Don't worry about not hearing any difference, there may not be very much, but you are at least in a more normal region than you were with gain x48 and the amplifier may thank you for it by living longer and staying cooler - though we never seemed to have gotten to the bottom of what your problem was.
CT
CT
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.