Questions and rumbles about solid state amplifiers

Started by tca, March 04, 2013, 10:46:48 AM

Previous topic - Next topic

tca

Quote from: KMG on March 14, 2013, 08:54:23 AM
I have different simulation results, solid lines - amplitude, dotted - phase.
Your resonance is about 70Hz, my - 120Hz (close to real guitar speakers).

ah, but I don't have a guitar speaker, just an oval bookshelf speaker!



P.S.
Ok, just corrected a 2*pi factor in the freq! Both calcs agree now.
"The future is here, it's just not evenly distributed yet." -- William Gibson

tca

Just found out that the oscillations that I mention in a previous post were simply bad connections to the breadboard, any bad contact makes the amp oscillate. Now it works marvelous with R6=10R although I still have some treble cut when connecting the guitar without a  buffer, it is not to drastic but sounds nicer with more highs.

Just forget to mention PRR's note:

> Your plan in #34 has a "horrible" flaw (which may be a happy discovery). The output gain and NFB is very different pull-up and pull-down. Wave is lopsided.

It is correct, the amps has an asymmetrical behavior for the positive and negative part of the signal. If overdriven it clips asymmetrical too :) .

P.S.

Some sims:



The frequency response was taken with the buffer.
Cheers.
"The future is here, it's just not evenly distributed yet." -- William Gibson

tca

So here goes some more notes. The schematic is frozen, with R6=10R, (see entry #34) that is the best sounding configuration.

Positive points:

- good compression
- sound best that any IC amp I've build
- lots of volume for clean and great when overdriven

Negative points:

- there is always a hiss at maximum gain (R6=10R) and it "disappears" with R6=1K
- there is also some fizz when playing a single note and it seems that the noise is modulated with the guitar signal, even at lower levels and because it also happens at lower volumes I think is not due to any clipping phenomenon
- when using a buffer, in this case a BOSS OD-3, the hiss/fizz reduces (but does not disappear), also if the emitter resistance (R8=4.7) is shunted  same happens hiss/fizz reduction

Taking in to account that I have the amp in the breadboard and the circuit is very sensitive to where the grounds, input and output speaker are connected and its relative position, what do you think about this? I'm going to make a metal boxed version to see if I can get any improvement on the noise. Any comments/suggestions?

Thanks.
"The future is here, it's just not evenly distributed yet." -- William Gibson

tca

What about a class a amp? And i DO know it is a waste of power!



Cheers.
"The future is here, it's just not evenly distributed yet." -- William Gibson

PRR

If the hiss vanishes *abruptly* at a certain gain, the amp is oscillating supersonics which alias-down into the audio range. ("Regenerative Receiver")

> What about a class a amp?

A is one thing. Resistor-coupled is a whole different thing.

A with a choke is up to 50% efficient. ("40%" is more likely in real-world.)

A with current-source is 25% efficient. (Again, really less.)

Resistor-coupled with no DC in the load is, at best, 5.7% efficient. (Yes, noted authorities give other numbers; AFAICT they have not thought it through because the case seems trivial.) 

Your plan is non-optimum efficiency. 27W in 0.9W (undistorted Sine) out is 3.3%. OK, not that bad.

Your output impedance is just 10 ohms. Compared to nominal 8, it has damping-factor which is suited to many guitar-amp speakers, cabinets, and styles.

I do not understand your biasing. If R3 is small and R1=R2, then T1 Collector is near 1.3V? Yet T4 Base has to be real close to 18V (say 16V). I don't see how this works (I must be misreading values) and doubt it is stable for supply and part variation.

Ya know, the old "FuzzFace" plan can stabilize the 2nd-stage bias current. The drop across the 2nd emitter resistor is equal to 1st stage Vbe plus drop in 1st stage base resistor(s). In any case you ought to have a buffer between gitar and a 2-tranny amp. So you can make that base resistor small, which makes hFE tolerance great. Q1 Vbe is 0.6V for any likely current, so 2nd emitter resistor is 0.6V/1.5A= 0.4 ohms. Yes, it drifts with temperature, but ~~10% for ANY temp you are likely to play in.

Here's my real objection (pure gut-work). Resistor does over half the work. What is the sound of the resistor? Bor-ing. Nothing like a Pentode, or a BJT/FET, or even choke. IMHO, this is good for a transparent headphone-amp, maybe not for a flavor-amp.

That concept alone is not as compelling as the GREAT improvement in efficiency possible with choke-load or push-pull. We are already fighting the 0.4%-2% efficiency of the "loud"speaker; multiply that by 3% amplifier efficiency and it's pretty pathetic (0.03% overall from DC juice to air motion).
  • SUPPORTER

tca

Hi PRR, thank you for your reply.

> If the hiss vanishes *abruptly* at a certain gain, the amp is oscillating supersonics which alias-down into the audio range.
> ("Regenerative Receiver")

It does... I suspected that would be the case. Back to the drawing board I guess, but that phenomenon does not appear on the sims... The amp is also noisy if nothing is connected to the input, the 100u cap on the collector of 1st transistor kills a lot of it.

> Here's my real objection (pure gut-work). Resistor does over half the work. What is the sound of the resistor? Bor-ing. Nothing like a
> Pentode, or a BJT/FET, or even choke. IMHO, this is good for a transparent headphone-amp, maybe not for a flavor-amp.

My feeling about class A amplifiers is that they are much more dull than a push-pull class ab (hard to do a straight comparison) although a high output impedance is easy to obtain in this case. Don't know why this happens? I've also tried a 18V version of a JLH type amplifier but the guitar sound was not fluid enough, kind of stuck sound. Probably lack of odd harmonics?!? Although my version of the JLH sounded great with regular music. I also tried this one



But the same happened, dull and stuck sound with a guitar but great with regular audio. I'm missing something? What?

Thanks again.

P.S.

Ref.: http://www.rason.org/Projects/transaud/transaud.htm
"The future is here, it's just not evenly distributed yet." -- William Gibson

tca

Quote from: tca on March 21, 2013, 10:17:16 AM
My feeling about class A amplifiers is that they are much more dull than a push-pull class ab ...

Just realize that this feeling I get from class a amps is just to much damping (I mean low), to high output impedance. What is the typical output impedance of a guitar amplifier 3R to 10R?

Cheers.
"The future is here, it's just not evenly distributed yet." -- William Gibson

PRR

Damping factor for guitar amps runs from >100 down to 0.1.

A basic chip-amp, or PA type amplifier, has damping in excess of 10 (may claim 1,000).

Naked no-NFB pentodes, from early Gibson one-6V6 to Ampeg VT-40, have damping near 0.1.

A broad range of classic Fenders, and some recent chip-amps adapted with series feedback, have damping factor in the general range of 1 (0.5-2).

A naked triode gives 3-5..... it is noteworthy that naked triodes have never been popular for guitar stage-amps, even though some audiophiles live them.

Damping factor is, roughly, the ratio of load Z to internal Z. Normalized to 8 ohms, the naked pentode is 80 ohms out, the solid chip-amp is <<1 ohm out.

Do NOT increase output impedance with dumb resistance, that's very wasteful. (Possible exception is bedroom amps where smaller is not cheaper, so waste power is not waste money.)
  • SUPPORTER

tca

So after a lot of sims I just realize that my attempts to build a typical class ab amp started to look much like a opamp! LTP on the input, current sources, etc... I guess that is what they were invented for! The small number parts amplifier that I've found was this Red Circuit (note: you can also find it on Teemu's book)



Ref.: http://www.redcircuits.com/Page164.htm

It is rather simple to understand: the R4 and R5 control the bias of the transistors and as the opamp requires more current to work so the transistors start to conduct more; R4 and R5 are determined such that the voltage drop, due to the  current consumption of the opamp, equals roughly .6-.7 V at idle. But my sims does not work... it seems that my simulator (TINA-TI) keeps the idle current of the opamp always constant! Still waiting for the transformer to arrive.


Meanwhile after that I've started to look for different amplifier topologies. There were two that catch my eye: the Taylor amp (check the  Scanned and cleaned up Wireless World Articles) and the Ciclotron amp (the solid state version, check http://www.tubecad.com/2013/01/blog0254.htm).

Concerning the Circlotron, this version is wonderful, but it eats a few amps in order to play some real music



Does not need an output transformer and with a 2:1 input transformer a simple single ended input version is an easy build.

Cheers.


"The future is here, it's just not evenly distributed yet." -- William Gibson

PRR

> my sims does not work... it seems that my simulator (TINA-TI) keeps the idle current of the opamp always constant!

Correct. Not the sim, but the model. Most chip opamp models do NOT simulate dynamic supply current; just a nominal idle current.

> a typical class ab amp started to look much like a opamp!

Well, you are well into a Cook's Tour (or maybe Magical Mystery Tour?) of amplifier possibilities.

The reference for "Good" power amps in a modern theme is Doug Self; though he doesn't teach the many-many alternate topologies.

However a Power Amp typically does NOT have to have all the features of an op-amp. It is invariably run at a medium and fixed gain. That means it does not need a wide input common-mode range; nor does it need "universal" NFB compensation. The DC gain can be zero (though in Hi-Fi there is a tendency to build for DC gain then chop subsonics).

In fact small opamps and large power-amps have converged to the same problem (with bigger parts). Sanitary NFB does generally imply simple-inverter gain structure. The LM101's topology does that, simple and neat.

> Circlotron, ...eats a few amps in order to play some real music ... Does not need an output transformer

Any Hi-Fi amplifier will need power (amps) to play amazing. Hi-Fi speakers are quite inefficient. Efficiency can be quite high in almost any sensical topology (though not if you have 6 ohm emitter resistors series with 8 ohm speaker).

The Circlotron is not an amazing invention. Look at a circuit as a string of pearls. You can move the pearls around the loop; the loop is unchanged. The Circlotron meshes the two loops of push-pull in a cat's-cradle-like way; it doesn't do anything grossly different.

Oh, there is a historic reason. McIntosh put half the load of the tube in the cathode, and patented it. This plus amazing build-quality got a good reputation. Electro-Voice rented the Circlotron patent to get about the same effect, with slightly lower build-quality, at lower price, and did well with the product. They just slid the pearls around the loop until it didn't conflict with McIntosh's "I Claim"s.
  • SUPPORTER

PRR

  • SUPPORTER

tca

Thanks PRR for your replies. I'll post some more ideas latter on.

Cheers.
"The future is here, it's just not evenly distributed yet." -- William Gibson

tca

More news! The prototype



and the schematic



It has current feedback taken from an 87's Elektor article.

Works great and loud as hell, at least for my standards, I'm just guessing my family comments... ;)

I guess is time to make a preamp...

Cheers.

P.S.

- the heaters don't need to be that big... but I'm glad I've used this ones (taken from a CRT). In case you are wondering, it DID make some smoke the first time!
- the power transformer is a 9-0-9 1A, so the amp uses +-14V
"The future is here, it's just not evenly distributed yet." -- William Gibson

tca

A few more notes about the previous design.

1. the input of the NE5534 should not exceed 0.6V, it breaks after that voltage (done that).
2. the BJT's do suffer from a thermal runaway/bad biasing, they start roughly at VBE=.6 and after some time, if you  push the amp a hard, VBE stays at .9V and stays there thankfully for the heat sinkers, but it will increase when heat sinks start to fail in dissipating all that heat. It will work with bigger ones. It starts as a good and low consumption class ab, and ends eating a lot of current (more than 1A at idle).
2.1 There is also the idle current of the NE5534, it drastically controls the bias of the BJT's
3. The current feedback introduces some fizz on the guitar signal.

Time to move on... not yet!
"The future is here, it's just not evenly distributed yet." -- William Gibson

PRR

> current feedback taken from

One thing of note: if the load falls off, the voltage gain goes to "infinity" (greater than the open-loop gain of the '5534). Generally, infinite gain finds a way to oscillate. Whether 14+V swings at possibly 10KHz-1MHz are a "problem" depends on the situation. (However the output devices will probably overheat.)

Most commercial amps with current-sensing also have a voltage feedback loop, which is non-dominant with nominal load but limits the maximum gain at light load.

> bad biasing, they start roughly at VBE=0.6

Since the idle current of the chip is in no-way similar to the bias needs, IMO this scheme needs to be biased deep-B, not even try for AB. Reduce the bias resistors. The large open-loop gain will pick-up most of the slack.

Overall this is a near-perfect amplifier with some major drawbacks. Not really what you were looking for (a music-specific amp).

Find Nelson Pass website. Look at his light-bulb amps. Nifty, single-ended, MOSFET, inefficient but inexpensive. You could get much of a Watt with 15V supply and a 12V car brake-light bulb idling around 10V. Lately he has been on a current-drive kick, so there's your output impedance.

> current feedback introduces some fizz on the guitar signal.

Not so much current feedback, I suspect. The speaker is a rising impedance. With infinite output impedance, stuff above 1KHz-4KHz can get brutal. In no-NFB pentode days it was universal to hang an R-C network across the speaker. 10uFd+10r is in the ballpark. (This was easier on the tube OT's primary side where 0.1uFd+1K would do.)
  • SUPPORTER

tca

> ...this scheme needs to be biased deep-B, not even try for AB

Lower the bias resistances to 15R, still works, and much cooler now!

> The speaker is a rising impedance. With infinite output impedance, stuff above 1KHz-4KHz can get brutal. In no-NFB pentode days it was universal to hang an R-C network across the speaker. 10uFd+10r is in the ballpark. (This was easier on the tube OT's primary side where 0.1uFd+1K would do.)

Very clever! Thanks. I'll try a cap in parallel with the speaker (220n ??).

> Nifty, single-ended, MOSFET, inefficient but inexpensive. You could get much of a Watt with 15V supply and a 12V car brake-light bulb idling around 10V. Lately he has been on a current-drive kick, so there's your output impedance.

I've done that also. ah, I'm sure you remember this: The Bulbamp: 1 Watt guitar class A amplifier with a light bulb (it does not have a 1W!).



It gets nicer and nicer as the voltage/amps increases, the biggest problem with these amps is the power source, big transformer, BIG caps. I'll try another run on this class a in the future. I even made a small version:



and a stereo version.
"The future is here, it's just not evenly distributed yet." -- William Gibson

tca

Just realize that the fizz is from bad biasing of the upper (PNP) transistor it simply does not conduct, I even tried to adjust the offset voltage (see datasheet) with no results. The circuit would work in class ab and not simply class b.

Just made an idiotic mistake... the output jack had a connection to ground (made because of the metal panel in which the input jacks were connected) and so the feedback did not work! It works great, no fizz, no high temperatures! Simply great!
"The future is here, it's just not evenly distributed yet." -- William Gibson

tca

So here is the final version of the amp:



The latter version still had some noticeable fizz... this version works better.

... time for the preamp. Because I have to limit the input of the amp to 0.6V I was thinking in doing something like the preamp project 27 from ESP (http://sound.westhost.com/project27.htm) changing the output diodes (to two and tweaking the resistor values). I'm still thinking about the tone control: a bass, middle and treble, or something like a one knob Princeton tone control...

There is in fact a very big difference in designing a preamp and a distortion box, one has to keep it simple in order to accept different stompboxes but it still needs, and it should, sound good on its own...  Any refs?


Cheers.
"The future is here, it's just not evenly distributed yet." -- William Gibson

tca

Here is the preamp, the tone control still needs some testing, but the clipping section works, although I find it a bit harsh when it is fully working, maybe the tone control section will tame it a little bit.



Cheers.
"The future is here, it's just not evenly distributed yet." -- William Gibson

tca

Some notes...

I've been playing guitar with this



One of the problems that it had was a noticeable fizz. This fizz varied in some degree and in quality as I vary the opamp, the transistors, the bias, the feedback loop (pure current, or mixed mode)... and also when changing speakers cabinets with different speakers! This fizz is very noticeable when playing between the lower B (123Hz) and lower F (175Hz) and the most accurate way of describing it is by saying that appears that the speaker has beach sand on the cone. I guess this is a pure speaker phenomenon due to the resonance frequency of the speaker or is the speaker badly tied do the cab (I think this is the correct answer, not sure yet) . Any thoughts? And BTW can this phenomenon be temperature dependent? It seems that the type of fizz changes during the day, not really sure about this either?

Cheers.
"The future is here, it's just not evenly distributed yet." -- William Gibson