Help with design for a simple low power audio amplifier

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
To factor out the protoboard influence, I'm going to build this circuit instead of doing more tests. I'll use local start ground scheme, and try to keep the layout clean.

john_ellis, that explains!
This confused me:
slew rate = I_input_transistor_max / C_miller

Is emitter degeneration improving I_max ??
I don't see why. The simulations I did yesterday showed no increase in colector current.

Also this relation for slew rate using emitter degenerated (by factor D) LPT.
slew rate = 4 * pi * fc * Acl * Vt * D
For more degeneration, more slew rate.
It seems that if D increases, slew rate must increase.

About the feedback,...you mean take the feedback between the the output capacitor and the speaker?
Right now I'm taking it from the 1/2 voltage point between the emitter's low-value resistors. This is how I thought it was supposed to be done.

Related, but a different topic...
I've just read this, "because this cap will charge through the speaker, the rate of change of voltage must be kept low enough to prevent speaker damage, so the amp has to settle to the ½ voltage rather slowly."

How can I control this??
 
Administrator
Joined 2007
Paid Member
The speaker coupling cap should not cause any damage to speaker as it charges. Amps were like this for decades and no one had problems apart from the annoying 'thump' generated.

If you look at the Europa circuit you will see those two transistors and an R/C delay network to reduce switch on thump. It allows the voltage supply to rise slowly to all parts of the amp except the output stage and drivers.

The feedback take off point should ideally be on a small spur taken off from the junction of the emitter resistors.

The ground points you return the feedback network to are crucially important.

See if you can follow this. Post #38 jumps is a good start.
3 stage LIN topology - NFB tappings?
 
Mooly, that about the capacitor confused me a little bit, as the author is Elliot from ESP, which I consider a person who knows what he is talking about...

I've read the thread about grounding. As I faced this with tube amplifiers, and they are all dead quiet, I think I'll manage to do it right. Just I don't understand why you say in post #44
"Also the speaker return was taken off the PCB and returned to the star."
Where is the main star ground? Isn't it the PCB ground bus connected to the main star ground via labeled point "power ground"?

john_ellis, I think I just answered myself about emitter degeneration and Miller capcitor: is it because degeneration lowers open loop gain? so, less compensation is needed, because 0 dB point is reached at a lower frequency?
So the relation
slew rate = 4 * pi * fc * Acl * Vt * D

If D is made higher, by increasing degeneration, fc and Acl is lowered, but by decreasing C_miller, they get back to the old value (prior degeneration) and you end up with the same values, but higher D.

Correct?
It was not immediately obvious for me.
 
Administrator
Joined 2007
Paid Member
I don't know what Rod actually says and the context of it, but I do know that we never used to worry over such issues in the past.

Your amp runs off a relatively low voltage which goes some way to mitigating the issue.

Ideally, any amplifier should be 100% silent at both power on and power off and that can really only be accomplished reliably with a speaker relay delay that is reliable and repeatable in operation.

With the grounding, the big issue was the problems that occured when connecting two identical boards (left and right) to a common power supply. Returning the speaker to an 'onboard' ground was fine for just one channel, but caused a serious 'loop' condition when the two channels were all connected up. The loop was completed by the input grounds having to connect together at some point, whether in the power amp or via the RCA connecting leads to a preamp.

The main star ground was a tapping or spur running off the main reservoir cap 'common' connection. This was a dual rail supply remember and so the high amplitude charging pulses for the caps flows in this series connection between the two parts. The spur is clean and becomes the main point of reference.
 
Here it is.
I soldered the amplifier on three boards, to get rid of the protoboard noises:

- Input stage + VAS + bias servo + drivers + feedback.
- Output stage + zobel + output capacitor.
- Power supply input + star ground main point.

I followed Mooly's recommendations on grounding, which is basically what I was doing in my tube amplifiers.

No hum, no oscillations, no noise whatsoever.
The 1uF capacitor connecting drivers' emitters is required though.

After a couple of hours of listening, I have to admit that I'm proud of it, specially it being my first solid state amplifier.
It sounds well, even at 5 watts output power.
The heatsink is enough to keep everything cool (below 40ºC).
Still using BD139/140, but I added some LTP emitter degeneration.
On the final board, I'll replace the OT with good beefer TIP31/32.

To be honest, I like both of my tube amplifiers more (one is class A single ended, the other is AB). I always thought that people who said "tubes sound warmer" were just technically ignorant (and in some way, I still think that, because there're so many factors), but this is exactly the conclusion of today's listening session.

Finally, I'd like to thanks once more everyone who has helped me make this real, sooner.
 

Attachments

  • proto_2_perfboard.jpg
    proto_2_perfboard.jpg
    516.9 KB · Views: 156
Last edited:
Administrator
Joined 2007
Paid Member
:up: You should be proud of it, and that looks quite a neat build. Its good that have got it noise and hum free as well.

One thing I would say though, and that is that not all solid state amps will sound like this one ;) This one will probably sound typical of the type of circuit configuration used, and the basic configuration used here is also used in many designs of varying complexity... but there are others to explore :)
 
Elerion
I agree with Mooly's comment that your perceived quality will be somewhat influenced by the Miller stabilisation capacitor.
You might like, when you have time, to modify your design to use phase lead compensation.
One possibility is shown in the attached diagram. The modifications are:
(1) Use 100 ohm degen resistors in the input
(2) use 1.3 or 1.5k base bias resistor for the VAS transistor
(3) Use a 4.7k tail resistor for the input stage
(4) Bootstrap this using the negative feedback to give a more constant current (cheap CCS)
(5) Change the feedback resistors for lower values (1k and 100 for a gain of 10)
(6) add a VAS degen resistor of 68 ohms
(7) also use 68 ohms for the CCS load of the VAS
(8) modify the bias stabiliser resistors to allow for the increased VAS current
(9) use a 220 pF phase lead compensation capacitor connected from the VAS collector to the feedback input
I have also added a 47pF input capacitor: this may protect against oscillation when the amplifier input is disconnected (as the two 100k resistors provide a high impedance) and used TIP41A/TIP42A output transistors. Simulations with the modifications using BD139/BD140 show too high odd harmonic (mostly 3rd) distortion (although only 0.12%) which reduces to 0.04% with the more powerful OP transistors.
It sounds a lot of mods, but I would be interested to see what you think of the sound if you manage to do this (and not have it oscillating!)
 

Attachments

  • 4W Mods.png
    4W Mods.png
    9.8 KB · Views: 173
John,
(1) I'm already using 100 ohm degen resistors in the input. This improves square wave response (eliminates a little bit of overshoot).
(3) Didn't you tell me not to use lower tail resistor than 10k?
(5) Change the feedback resistors for lower values? Why is this necessary?
(9) About the 220 pF phase lead compensation capacitor: could I just add all your improvements EXCEPT this one (in other words, still use my current Miller capacitor)?
And then, in a final step, replace the Miller cap with the phase lead?
 
What a lot of questions!
A lower tail resistor is possible because it is bootstrapped. Bootstrapping is not ideal, but because the feedback voltage is about the same as the input voltage, the feedback is transferred to the lower terminal of the resistor. So the voltages at each end of the resistor remain almost constant, relative to each other, so the current is probably more constant than even using a 10k resistor!
The phase lead capacitor performs two functions for the price of one. It provides phase lead compensation, but also loads the VAS stage which gives a phase delay. The capacitor needs to be about 220pF, so to keep the bandwidth, the feedback resistors are reduced. In this circuit which is AC coupled, the imbalance between the base resistors on the diff amp pair is not important.
You could in other words keep the feedback resistor high, but say that is 10k then the value would need to be 22 to 33 pF or thereabouts. That would require 220pF loading on the VAS separately, possibly with a phase correction resistor in series too (around 100 ohms).
This was a rather subtle use of phase lead, but I hope that is clearer now.
BTW the degen resistor in the VAS works with the 220pF capacitor to give a controlled roll-off too at around 10 MHz, so I don't think major component alterations are optional, but might be fine-tunable.
 
(4) Bootstrap this using the negative feedback to give a more constant current (cheap CCS)

This sort of design choice dates from the 1960s when transistors were expensive. It’s not the 1960s any more. I can buy a 2N3904 from Mouser or Digikey for under 10c. It takes one 10c transistor to turn that quasi-CCS with dreadful PSRR into a proper CCS which won’t result in the circuit humming. No bootstrapping needed.

A 10c transistor is orders of magnitude cheaper than the extra supply caps that you’d need otherwise.
 
Yes, what a lot of questions! And answers! So thanks.
It is a little bit more clear now, but I need to study this a little bit more too.

I understand about using transistors instead of bootstraping, as they are cheap, although it is nice to see how it can be done more simply. Helps to understand the whole picture also.

I don't really get the PSRR comment. I know what that is, and that in a common emitter stage, it is better as the emitter resistance lowers in relation to colector resistance (I look at this, as if it was a simple voltage divider).
 
I think Suzy's comment might have applied if the amplifier were running from dual supplies. But as it is running from one supply, I do not think that ripple from the supply rail could get into the amplifier, as the feedback side is grounded (as WG Ski noted). You can simulate for PSRR by adding a signal generator in series with the supply voltage and seeing what comes out (with normal input set to zero).

I would usually use a transistor as a current source - I have recommended this too, in these forums - but here I kept the resistor, bootstrapped, as it was simpler, and needed no extra components, which I thought was more in keeping with your "design of simple amp". Because, if you add a transistor, I also recommend adding its own bias to keep separate from other CCS transistors, which generally means 2 diodes + 2 resistors + transistor (or more if you add Zeners on a dual supply circuit to provide PSRR).
 
Oops, I didn’t notice the grounded tail. Too much time working with bipolar supplies.

Coming from IC design, where you design a really good reference and then use it everywhere, I don’t understand your rationale for replicating the reference. Surely if there’s something precluding the reference working properly (I recall you mentioned vas current robbing) then you’d fix that, not double up on references.
 
Suzy - in an IC design using MOSFETs the problem of current robbing is unlikely. In BJT amps I think that to use a common bias, then it needs to have a low impedance. But if a VAS stage clips/saturates the current it could take may be quite high. Some form of current limiting is beneficial. Some designers use a 1k resistor in series with the base, but that has a detrimental effect- it slows the VAS stage down through the Miller effect.
So I've used simple resistor + diode biasing so that the diodes provide a low impedance but the resistor provides current limiting. Separating the supplies for each stage means that if one were to hog current it would not affect the rest of the circuit (but of course the circuit would be clipped anyway).
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.