Alesis RA-100: strange power amp problem: sound quality, bias stability.

[Kaz Kylheku]

I've been driven crazy with bad done over the past week or so.  There has been a bout of wet weather so I blamed the humidity. I ran hot air against the speakers in the 4x12 cab, and put a bunch of silica gel dessicant inside the thing.  I cleaned jacks. I cleaned tube sockets in the ADA MP-1. I repositioned the cab in the room. But  I did the obvious last: try switching to the other channel on the power amp. Aha!

What is the perceived problem? Something harsh and objectionable in the tone in distorted programs, and a stiff lack of touch sensitivity on cleans, which are much more fluid and silky through the other channel.

This morning I cracked open the power amp, and checked the quiescent current in that channel based on the voltage on one of the 0R22 emitter resistors (with the amp completely unloaded: no speaker or resistor). It was around 27 mA. What? I had set it to 5 mA only weeks ago. How can it drift so far, and in such a short time? This is a sign of some trouble. I also noticed some quite marked instability. On power up, the voltage was 6.1 mV on the resistor, soon dropping to around 5.5 mV in less than a minute.  Change with temperature is expected, but I don't remember it jumping by such a percentage in a short time.

I haven't had time to do any more investigating than that, so far.

Maybe a driver transistor is going south?

There is no glaring problem with the amp, like pronounced distortion. It reproduces sound and plays loud.

I'm not going to have time in the next few days, but I will run some signal through it into a load resistor and scope it.

It's nice to have a second channel for backup and comparison.

[Roly]

Even otherwise quite well designed amps can have thermal stability problems; not go into runaway, just wander around, perhaps even do a very slow oscillation over minutes.  I had a Gibson repair that did something similar until I mounted the thermal sense diodes (as transistor E-B junctions) on the heatsink.

This amp appears to use a single transistor as its thermal compensation element, and it's often the case that this element is situated too far from the heatsink which it is supposed to be controlling.

Found the service manual.

[Kaz Kylheku]

Thanks for posting the schematic, Roly.

This was not obtainable on the net just half a year ago!

By the way, my earlier comment that I was surprised by the quiescent current was the result of a brain fart. About 5.5 mV across each 0.22 ohm resistor was in fact exactly what I had set it to in both channels. It's my preferred setting. So no mysterious drift had taken place.

Now I still have no explanation for the bad done which prompted all this. I believe it is a problem in my ears/brain. Sometimes I have episodes during which certain sounds bother me. Not only my guitar tone, but tones heard on records that are otherwise fine.  What was smooth and violin-like becomes irritating. It is a major frustration.

[Kaz Kylheku]

I have entered the RA-100 schematic into LTSpice and am getting meaningful, "life-like" simulations out of this thing.

I'm lacking models for the four Sanken transistors used in the Sziklai pair driver stages; for now I simply substituted 2N2904's and 2N2906's.  xP

I had to substitute a Zener diode for that thermal transistor whose base is left floating. That just does not simulate. Thermal shutdown can be simulated by varying the voltage on that manually. If you replace the Zener by either a short or open, the circuit will mute the input stage, so it needs an in-between voltage to work properly.

I also had to tweak a resistor value in the VBE multiplier to get a reasonable quiescent current, due to the substituted output devices.

Well, the thing is amplifying.  The JFET mute on power up reproduces in simulation (I shrunk a resistor for a shorter time constant than in real life)  The foldback current limiting behavior also reproduces, complete with clip LED being activated.

This is not just idle curiosity; armed with this, I'm going to design a better current feedback circuit that doesn't cause a DC offset like the one I'm using now.

[Kaz Kylheku]

This is not just idle curiosity; armed with this, I'm going to design a better current feedback circuit that doesn't cause a DC offset like the one I'm using now.

The above was no idle threat. I mulled over alternatives and then designed and implemented a solution (in one channel so far) several weeks ago.  It's been working beautifully.

The DC offset problem, previously 2.7V (!) (with a 4 ohm speaker load!) is completely cured.  I can dial in any mixture of voltage and current feedback without shifting the offset, and yet everything is DC coupled.  I even removed the CX13 capacitor: the one in the feedback loop that reduces gain to unity at DC. The amp now has gain all the way down to DC.

How did I do this?  I figured, since bias cancellation techniques are used in op-amps to null offsets, why not try it in a discrete power amp. So I designed a temperature-compensated input bias current cancellation circuit, consisting of a small (0.9"x1.1") circuit board with four transistors, seven resistors, trimmer pot and cap.  This circuit taps into the main board via four hookup wires: ground, -48V, and connections to the two bases of the input pair.

The rig is now producing some great tones! All I want to do is play guitar because the sound is just so good every time I pick up the guitar.  I've taken to sitting on the floor right in front of the 4x12, so I can just bask in the pure tone in my face.  And it's been consistently good. Turn it on in the morning or evening, hot day, cool day, doesn't matter: same great tone.

As far as the DC offset goes, when you now turn on the amp cold, it starts at around -30mV, and immediately starts ramping positive, until in a couple of minutes it reaches around +30 mV or so. It stabilizes there and then reverses, falling into the neighborhood of +12 mV and thereafter fluctuates as the amp is left on all day. I've seen it as high as around 20 mV, and as low as only 6 mV.  Given that I was prepared to be happy with anything below +/- 200 mV, I am very pleased.

I will post pics and schematics of the project later.

[Roly]

That is indeed very pleasing, and more so since it seems to have cured the harshness that was the original source of the problem.  I love it when the loop gets closed like this.

A few tens of mV's is okay, but 2.7 volts on 4 ohms sounds "busted" to me.  Given that the DC resistance of a nominal 4 ohm Z speaker will be about 2.6 ohms that's about an amp of unbalanced residual!    No wonder it sounds better.   :dbtu:

[Kaz Kylheku]

Pics!

Check out the way the resistors run under the trimmer pot, with a screw between them, and everything clears.
By placing the trimmer footprint over the resistors, I got it all to fit into a 1.1 x 0.9 inch outline.

[J M Fahey]

Very cool.
Congratulations.  :dbtu: