The Pioneer M22 as a head amp? (and recapping guide)

Moving this to the right thread.

I too am befuddled. Resistors were 1% to start with but I matched resistors to ~0.05% or better on the two halves of the push-pull architecture. One thing is my new resistors are rated at 0.6W whereas the data sheet spec is only 1/4W, ...but the new ones are nonetheless half the size compared to the old ones. I wondered if thermal effects might play a role. They may be able to handle 0.6W dissipation but the dT/dP may still be higher. It’s also possible that there is just one nasty resistor in there somewhere giving me problems. I did not test after each replacement.

 

Definitely not a matching issue then. @purr1n is probably on to something....

 

I’m fairly sure they were carbon film (not carbon comp) as they had hourglass style cases. This possible aging effect (which I think is really a composite issue) was nonetheless another reason I decided to replace them, but I measured randomly about 20% of those I pulled, and they were all in spec.

I’m not really giving you the whole story. There are a few wirewound resistors in output stage. These were replaced and may also be suspect. The new wirewounds measured better on my LRC-meter at several audio frequencies, but that doesn’t prove they’re innocent. A few really critical resistors directly in the audio path were replaced with z-foils but to no audible effect.

 

I doubt it's going to be the wirewounds. There's very little you can do wrong with them.

Do you have an oscilloscope? It'd be interesting to feed both channels the same signal and compare the feedback node of the two on the screen....

 

Just looked up a schematic. Didn't realise the M22 was the Hitachi topology. (I've circled the feedback node in red.)



Here's a version of it that I've been working on:

 

Well, thank you, @JeffYoung!
I do have a 4-channel TeX analog. But I’m not sure what to look for on the Oscope? Distortion signals are pretty small according to analyzer. What do you suggest I might find (Especially when operating in closed-loop).

 

I just had lightbulb go off. I also replaced the voltage drop diodes on driver pair (Q6 and Q7 in the diagram you posted)... maybe they are now biased badly? I did go for same forward-voltage *spec* but maybe the transistors are quite sensitive to actual voltage?

 

Interesting findings.

Seeing as you have 2 sides of the circuit to compare, are your voltages comparable left to right?

 

I recently acquired another M22. It's for speakers, but I'm going to explore it's ability as a headamp upon it's return.

It's currently with a tech in NorCal. This one is a Japanese model in excellent cosmetic condition, and comes with a step-down transformer. So far he's replaced some intermittent relays, rebiased from 750mV up to the default 850mV, and replaced the AC line fuse to one of correct spec.

I wasn't going to do further upgrades, but asked him to give me suggestions on what he thought would make the biggest sonic difference. He said that replacing the original electrolytic caps on both PC boards with ELNA Silmic II types would be good, and install Mundorf M-Cap Supreme bypass caps after the rail fuses on the PC boards (rather than across the big filter caps, which is what I did on my first M22). Lastly, he recommended replacing the input coupling cap with a modern substitute.

What do you guys think?

It already had some modern output transistors installed. He mentioned that it's a shame that the original 10Mhz Sankens are rare, because modern types (usually Motorola) are inferior.

 

PTS,
I would say replacing the coupling cap will give the biggest impact on sound. That made a huge difference for me. You won’t be able to fit any of the good ones on the boards, you’ll need to have them wired up with leads and squeezed in somewhere or used outboard. At 3.3uF, I found I had to skip Miflex and many other good recommendations that were just too big.

I don’t think Mundorf caps makes sense as power supply filters, but that’s me... I just go for lowest ESR caps I can find. Likewise, I also am not sure I’d use ELNA silmics. The electrolytics are largely noncritical to sound, but certainly you should replace many or all of them because I found many had high leakage currents from age. If memory serves, there is maybe one electrolytic near the the audio path that matters IMHO, and it’s bypassed by a PP film cap anyway.

When you said you bypassed the power supply after the rail fuses—I don’t quite know what fuses you speak of. The only fuses I’m aware of are on the A/C side of the rectifier bridge. Caps there are surely a bad idea. There are three power supplies in each channel, two regulated and one unregulated. My sense is the unregulated supply needs to have low output impedance, which is why it’s fed from the 33,000uF caps directly. Bypassing those big electrolytics with film caps, however, made no measurable or audible difference for me. (note: I had already replaced the big electrolytics with new ones.)

I found the most effective fix for A/C line noise in the output was to get the switching energy of the rectifier diodes down. I found some soft recovery Schottky diodes that helped tremendously. This is a debated topic but I have experimental results showing 15dB reduction on the 60Hz fundamental. Just understand that you’re getting a lower 60Hz peak in exchange for more HF (Fourier transform of a steeper shutoff function), but done well and the overall noise floor can still be dropped.

 

My working theory is that the only thing that could change the distortion specs that much is a change in feedback amount. While resistor values elsewhere could alter the open-loop gain, the elephant(s) in the room are the two resistors that set the closed-loop gain. Everything between the two is feedback.

So you'd just be looking for an increased signal on the feedback node. (It occurs to me that you could also simply measure the closed-loop gain at the outputs. If the higher-distortion channel is slightly hotter, then the other channel is lower distortion simply because it's getting more feedback.)

 

The diode drops are a pretty fundamental characteristic of the way they're made, so I doubt they're different. However, resistor values elsewhere could affect the bias spread. (My version has a lateral MOSFET output stage, so I'm not sure how sensitive the BJT output stage is to bias. But I've certainly seen designs that are hyper-sensitive.) In any case, I'd definitely re-set the bias (the trimmer under the voltage drop diodes). The service manual should spec a voltage that you measure over the emitter resistors of the output transistors, but if it just specs the bias current then you can use Ohm's law to calculate what the voltage drop over those resistors should be.

Keep us posted....

 

@JeffYoung how do you feel about the hitachi architecture with single IXYS hockey puck output stage?

 

I have the original Sanken. But wondered if you might do alright with some non-TO3 package options.

I replaced my input diff pair and I think the new arrangement gives an improvement in speed. (Installed on both channels now after the resistor fiasco).

 

based on the plot I posted, looks like the closed-loop gains are darn close between the two channels: 0.01dB difference.

 

Be forewarned that the topology lends itself to marginal stability. But I recommend it if either (1) you're good at compensation methods, or (2) you want to learn a lot about compensation methods. (I'll confess to being only in the second camp.)

One of my issues with it was trying to keep it to 3 stages. The second differential with the current mirror can't drive 6 pairs of lateral (or vertical) MOSFETs. But it might be able to drive a pair of pucks. (If not you'd need to add a driver stage.)

I also found it easier to stabilise with a JFET input differential. (They'll probably sound better too.)

I've attached my latest SPICE simulation if you want to play with it.

Cheers,
Jeff.

PS: mods, can we add .asc to the list of allowed file-types for uploads?

 

Ahh, yes, I didn't see that at the top of the charts. So it's not the feedback resistors. Your left channel could still have more feedback, but it would have to be because of higher open-loop gain rather than lower closed-loop gain. Sadly that's a lot harder to measure.

There are still some (smaller) elephants in the room. The three resistors circled in blue will be the dominant factors in determining open-loop gain.

 

Thanks Jeff. I’ll measure and find out. I know very little about design, so I find this most helpful.

 

@JeffYoung, an update:

1. The feedback resistor (your red circle) is a factory special thick film and one of the few that I opted not to change on either channel.

2. The resistors on the long-tailed pair are in spec on both channels. The channel with new resistors was matched using my 6.5 digit 4-wire HP ohmmeter. Matching is excellent.

3. The resistor on the current mirror (your blue circle) is a 2W wirewound. Both channels measure in spec. The new and old are 0.009% different in value.

4. Diodes biasing the output transistors have slightly different forward voltages: new ones are 0.523V & 0.511V, versus 0.538V & 0.567V for the old set.