Groundplane in poweramplifier PCB design

What about a reactive load? Isn't any speaker a reactive load?

As you can read in my reply to Ban's message, the absence of a current limiter is intentional. I also don't plan to include one. In all the time I have used the amp, and all the others that where built, it has not been necessary.

D1 will be 2V +/-10%, because the LED is operated at 1/10 its rated current and the current through Q3 will be between 2 and 2.6mA. You are right it is stable. But I was talking about Q4, which has a beta of 100. Without adjustment and temperature compensation it will take from 0.9 to 1.7mA when idle. I call that at least 1% THD. Don't forget, the input pair is the reference and when the current ratio is 1:2, distortion will blossom, no matter how high the gain will be.

Well normally the bias current would be 1.3mA/Beta = 5uA and the voltage on the cap -120mV, but when the O/P goes to +40Vp 20Hz, the current of 1.3mArms

*25ms / 100uF will let it have +200mV. When the input overvoltage scenario happens, it might be a multiple. Also note the author talks about 1V offset without the load, realistic indeed.

Don't be too sure about it. just plug a preamp into it and then switch it on or off with the amp still connected. That preamp might sit on its rail for a while, the O/P cap is for sure a few hundred uF or it is DC-coupled, half a second overload with at least 5V happen almost everywhere. Or you have a mixer and some idiot switches on the Phantom-power, the output will go high or low and hang a bit on the rails. Nothing special, the mixer survives it, but this amp? And the expensive speakers?

You are right. A nice 20mH coil + subwoofer, a negative overload. Q4 switches off, Q5 and Q7 too, and then the other side will be damaged from voltage reversal.

It won't help at all, 5A is the max current for 100W load. 40V steady are

200W. I can not understand why people make such an effort just to blow their speakers. :-(

--
ciao Ban
Apricale, Italy

I still don't understand one thing. If you add the wiring resistance to the ground resistors of those two devider networks, then you wil have something like 22.000,001R for the input, and 1000,001R for the feedback. Because the input devider has a very small devision, and the feedback a very large one, won't the ratios be influenced in a completely different matter by the wiring resistance? (I must be driving you mad here...)

BTW, a question. What exactly is it that causes an amp to oscillate if you connect the input ground on the PSU's 0V, instead of on the PCB. And, in case of the P3a schematic, I would conclude that it's not about the signal-R1-R2-gnd path being as short as possible, but the signal-LTP-q3-gnd path. Is that correct?

you could trim the offset with the SIM connections or by adjusting the base resistor of Q4

No, it is bipolar. A normal electrolytic cap can be destroyed when its voltage is reversed and more than 1.6V.

See my other post. I answered everything. Maybe for your home use it is OK, but certainly *not* for stage or studio monitoring as claimed. I would recommend an IC called TDA7293 which will have a guaranteed sound quality and protection incorporated and is in the same ballpark of power. In fact to drive lower impedance speakers, you can even parallel several. They are around 5 Euro.

It's outdated technology. If you like to understand more about power amplifier design, read some scientific books about it. Actually IMHO all competent amps sound the same if operated in their intended range, better invest the money into speakers.

--
ciao Ban
Apricale, Italy

I don't understand the reasoning behind this, but the measured THD of the amp is 0.04% typical at 1W to 80W.

Can one consider the instantanious AC voltage as DC? I mean, by that token, there is also DC on my speaker, but we can all agree that there's not.

No, he says that with 22 safety resistors on the rails, you shouldn't read more than 1V DC on the output. At full voltage, it's less then 100 mV. In my particular case, it's about 50 mV.

I have included a speaker DC protection in the amp btw, so any DC on the output, and the speakers will be shorted to ground (note: speakers, not speaker_out of the amp), within about 3-4 ms if I'm not mistaken.

That is what says the website. Also note the "typical". Did you measure it on your amp? You can do that with a good soundcard and some program like audiotester.de free downloadable. If the ratio of the two collector currents is Ic1/Ic2, the difference in Vbe will be: e^((Vbe1-Vbe2)/Vt) = Ic1/Ic2 with Vt= 25.5mV and Ic1 = (Ic1+Ic2)/2 (1+tanh(Vbe1-Vbe2)) now we develop the tanh in a series up to the third power with Vo=Voffset: tanh= 1 + Vo/2Vt - Vo^3/24Vt^3 and substitute Vo with Vo_peak*sin(wt) w=omega and get finally the ratio of harmonics to fundamental THD approximately 1/48 * (Vo_peak/Vt)^2 with THD=1% Vo_peak would be 0.7Vt around 18mV. Since the amp has a DC-gain of 1, but an AC-gain of 23, you would have 1%THD with 400mV offset. 0.1% with 123mV offset. This is only the first stage, the others will add more. You asked for it. Hope you don't mind. Post it to the group.

The speakers are meant to work with AC, but when you put a polarized cap even momentary on it, it might explode, if the normal voltage get reversed. It is more like your walkman, when you put an AC wallwart instead of DC. The maximum reverse voltage *is* mentioned in every datasheet of these kind of caps. Look what capacitors are used in loudspeaker crossovers, those would be better.

As you could see from before that 50mV should be nulled out. Actually it is better to measure it across the two SIM resistors. now you can put another resistor to positive rail, where the value was lower. with 50mV 18Meg are good. That is quite a high value, but you can divide it down.

That is very smart indeed.

--
ciao Ban
Apricale, Italy

Look at it like this:

Speaker ------- R(wire) 8 Ohms +in----! Magic !---/\\/\\/-------O---/\\/\\---O -in----! box ! ! ! !---/\\/\\/------------------ -------

Lets say, for round numbers, you want the "magic box" to make 10 times Vin appear on the speaker or at least as close to that as you can given the practical wiring.

Since the "magic box" does not have ESP, you need to bring the speaker voltage back to be measured.

------- ! !--------------- Speaker ! ! R(wire) ! 8 Ohms +in----! Magic !---/\\/\\/-------O---/\\/\\---O- -in----! box ! ! ! ! !---/\\/\\/------------------ ! ! !---------------------------- -------

Now consider how to apply these feedback points inside the amplifier. Both lines should go to somewhere near the input section to be added and subtracted from the input signal correctly.

Think about the differential op-amp circuit I gave you and I think it will start to become clear.

The wire from the PSU 0V to the PCB has a voltage drop in it. If this is large enough, feeded one times that voltage into the input causes enough output to sustain oscillation.

--
--
kensmith@rahul.net   forging knowledge

A speaker has a significant fraction of its impedance as resistive, within the audio band. If you hook a large valued good capacitor across the output of the amplifier, you can cause a large AC current to flow. The capacitor, being a good one, doesn't burn up any power, so obviously, all the power has to go up in heat in the output stage of the amplifier.

Many an amplifier has been smoked this way.

[...]

I've never needed my seat belts but I still wear them.

--
--
kensmith@rahul.net   forging knowledge

In article , Ban wrote: [...]

Its more because it is a green LED than the low current. I got the wrong number when I said 3V.

I disagree with this. The input stage current and offset voltage will be what moves. If Q4's collector current changes, the circuit will servo it back.

I don't see how you got the 1%. Could you expand on it?

The supply is only 35V. I don't see the output swinging more than 33V or so.

Huh????

--
--
kensmith@rahul.net   forging knowledge

In article , Ban wrote: [....]

I think you grabbed the wrong transistor(s) to look closely at for distortion. It looks to me like Q4 is more important than the input pair. I haven't worked the numbers to see.

--
--
kensmith@rahul.net   forging knowledge

In article , Ban wrote: [....]

OR Better yet make Q4 into a current mirror by adding a resistor in its emitter leg and a diode or transistor in series with its base resistor.

You could make Q4's collector load a current source too.

If you do that, chances are no trim pot is needed. You just have to make it scale so that Q4 has the right current when Q1 has 1/2 the input section current.

--
--
kensmith@rahul.net   forging knowledge

YES, you are right. I simplified the whole thing by assuming the load of Q4 a pure current source with infinite resistance, the reality looks much worse. Now we can use the Early-voltage Vy of Q4 to determine the gain. Note, the collector of Q4 has to supply the full voltage swing of the output (even more with the

0.33R and the loss in the transistors). I want to consider a Vy of 150V, which is a high value for a PNP. With Vy/Vt we get the max. gain 5770. For 30W into 8R we need 15.5Vrms or 22Vp. this will be 3.8mVp at the base of Q4. Now our differential amp at the input has only a miserly gain of 14, whereas th feedback attenuates 23 times. So the input voltage variation needs to be 6.3mV. This means a K2 of roughly 6%. WOW And in reality it will be higher. Even with 1W output we are already above 1%. That is worse than *anything* I have seen so far.

--
ciao Ban
Apricale, Italy

It is very difficult to analyse the circuit, since the collector load of Q4 is not linear. First I tried it with a constant current load, but that was leading to a very high distortion (4% at 1kHz). This bias capacitor is really a smart thing, because with the 3.3k resistors it somehow compensates for the exponential transistor characteristic. So I plugged the whole thing into Spice and got with some tinkering a minimum of 0.51% with 12V out *without any feedback*, exept the cap connection. The distortion stayed also almost identical with different supply voltages. The voltage gain of this stage is quite high 1600. But the output impedance is very high 15k, so this node is should be layed out with care, to avoid high freqency loss and oscillation problems. The given values are actually well chosen. With the input pair gain of 14 and the output stage of 0.94 we have a feedback factor of around 900, which helps to reduce distortion. we still need to linearize the output stage, which will contribute most to the final distortion value.

--
ciao Ban
Apricale, Italy

Now that I have looked a little longer at this amp. I like *very* much the way this bootstrap capacitor is incorporated. Not only allows it strong current drive near the negative rail, but it also linearizes the exponential transistor behaviour of Q4. So the power stage is OK.

But whoever frickeled that differential pair on the input, should be shot, it cannot be the same who made the power stage. It sacrifices the common mode range and rejection completely and consequently even worse mistakes are added. The cap on the inverting input to gnd will cause an awful overload behaviour and induce phase shifts at low frequencies. Which are aggravated because the input low frequency limit has to be much higher.

The main problem is that the resistor R6 that drives Q4 has also to carry the bias current of the inverting input. When Q4 gets hot and the bias needs to be less, the pair gets more and more unbalanced, thus changing the gain of the transistors, increasing distortion and causing huge output offsets. The designer tries to reduce DC gain with the blocking cap C3. Without the cap there would be more than 500mV of offset drift on the O/P when the temperature of Q4 rises only from 25 to 60 deg. There is no way to adjust a stable offset. And an overcurrent protection is not possible because the amp recovers with a time constant of 0.1s.

To get rid of *all* the problems the input bias current has to be separated from the signal current. Just an old diagram of the 741 shows you how to do it. A current mirror to the pos. rail and an emitter follower to drive that infamous Q4. This will allow a nice gain on the input pair (not only 14), DC-coupling of all the stages, a separate stable offset adjustment, and even a current limit that doesn't sacrifice sound. DC noise gain can be adjusted with the bias resistor of Q4.

And the clueless OP Wiebe who put my first answer to this forum without asking my permission, can finally connect his input gnd to the rest. Instead of the gnd-plane he should have better made an output-plane though, at least under the power stage. Well time for another layout. :-(

THX to Ken, who was also on the right track.

--
ciao Ban
Apricale, Italy

I found where he got it from! AoE Chapter 4.27. Unfortunately it is not listed under "Bad Circuits", but as an example with a whole page of analysis. :-(

--
ciao Ban
Apricale, Italy

Why exactly is that unfortunate?

And about the page of analysis, does it say anything about it being good or bad?

Read it yourself. Maybe Win replies, he has been a bit rough with me, so I do the same. For your knowledge this book will be helpful, I recommend it.

--
ciao Ban
Apricale, Italy

I'll see if I can add it to my list of electronic books I plan to read. Does is begin with the very basis of electronics?

BTW, if I have given you the impression of disrespect or anyting in my posts, I apologise. I didn't mean to.

In article , Ban wrote: [....]

Yes, this sounds right. It will only be true at low frequencies. I'm much to lazy to do the actual calculations but you've got an assumed impedance over 20K and power transistors in the same circuit.

This bit sounds wrong to me. The total open loop gain of the amplifier is about 80,000, at low frequencies. I would not expect 6.3mV on the input.

I would expect Q4 to be adding somewhat more than 3% of distortion in the open loop. This will get reduced by the usual 1/(1+GH) in the closed loop case.

--
--
kensmith@rahul.net   forging knowledge