Ampeg SVT 4 Pro 10V dc offset

[MarkusBass]

Very precise matching makes no sense - I told you that in the amps from the factory MOSFETs are matched with 30-40% accuracy. Higher accuracy is not needed in this amp. My suggestion is to match the MOSFETs "in-circuit". It's due to the fact that some parameters may be dependant on the supply voltage. So the transistors matched with the Nelson-Pass circuit may not seem to be matched when you use 70V power supply.

Mark

[J M Fahey]

I also connected my scope up, measuring differentially across the Source resistors (output buss vs each Source lead), the waveshape was psuedo-square-wave with variations.

Very good and important measurement, THANKS for sharing.

I expected that square wave behaviour and I think you somewhat confirmed it.

Many times I wondered about the behaviour of internal elements inside a complex gain stage, with NFB overall.

I always had the theory that the nice smooth final waveform we get at the end, thanks to NFB, actually comes by "brutal pushing and beating" inside, to keep the unruly element in line, literally.

I already checked it in the ADA microtube amps, which used a 12AX7 tube driving an SS amp, with gain (not simple voltage followers), and the whole inside an end to end feedback loop.

Which, by the way, completely negated the use of a tube as a "flavoring" element !!!!!

I scoped the tube plates, and when the amp approached clipping, I saw a weird waveform, looking like, for lack of a better example, the tip of a condom, or a sinewave with a smaller sinewave on top.
The main sinewave didn't change much, but the little one grew and grew when approaching clipping , and when flat topping was visible in the output, the small sinewave became a narrow rectanguklar pulse, reaching the tube +V supply.

So I noticed the NFB, when detecting the output stage could not keep perfectly tracking the input signal, pushed the tube as hard as it could to make it drive the transistors better ... until it could not physically do so any more and forced it to clip by itself.

I think a similar thing happens when using switchingb transistors such as these: even if "matched" (sort of), alwayds one will pstart pushing current before nthe others, and since load is not even (the othars heve not turned on yet) , the one tries to work for all and produces that local squarewave ... which we'll never see at the output.

Of course, somewhat later others also reach the threshold and do their own ... but you will only see it internally, and in a differential setup such as the one you did.

Thanks again.

MEF RULEZ !!!!!!!!!

[dwmorrin]

Great post, nevetslab. Thanks for posting that!

Mark, thanks for all your info as well. My questions weren't in pursuit of precise matching so much as to avoid very wrong matching. Many times on this forum I have read "did you match the transistors?" or "make sure you use matched transistors!", yet I found so little details as to how anyone is matching their transistors. So I just wanted to double check I wasn't fooling myself by using a hi-fi article to service a high powered bass amp.

[nevetslab]

I haven't had the pleasure to play with any of ADA's amps. So much gear out there that I've never had any hands on with, but those look intriguing. What you described in the pre-visual-clipping behavior often is seen in the distortion analyzer's monitor signal, after stripping out the fundamental, or in spectrum analyzer waveforms, as the non-linear behavior begins showing up. I've seen that 'tip of the condom' waveform before in gear, though at the moment, couldn't state on who's gear it was. Usually, when I come upon something unique or odd, it get's documented and added to my ever-growing database.

I always get a bit nervous on some of the amps having only one or two pair of output transistors, as though you're working with one pair or two pair of power tubes. Early SS combo amps, just using the aluminum chassis, or steel chassis with a small aluminum heat spreader. Eden WT-800's or earlier SWR SM900's being good examples of later low-pwr device count amps, though both are BiPolar, not MosFET based.. Yet, I rarely have to replace their power devices....at least in our rental inventory. I came from product development engineering for solid state power amps, and there was ALWAY's lots of silicon in the output stage, and substantial heat sink cooling. I've always liked the sound of a pair of channels in Bridge....lots of balls from the extra headroom, twice as much silicon, even though your min load impedance has to double....there's enough to share the load.

In general, the lower the power device 'count' in the output stage, the more you want device matching, so you can run closer to the limits. Thus far, I've only had to replace an output stage on an SVT3-PRO, and used a curve tracer to select the best Vgs matches from the bulk parts I had on hand. Got similar voltage/current distribution as i see on this SVT-4 PRO I just finished up with.

[nevetslab]

After running another couple SVT4-Pro amps thru the shop, each having 'spare parts' rattling around inside, it allowed me to tear each one down for inspection, and then check-out afterwards. One of them wasn't pulling much AC Mains current at idle (they typically run around 1.3 to 1.4Amps @ 120VAC AC Mains), so after checking the bias voltage across the source resistors, one Ch was nearly turned off, not measuring anything greater than 1mV, and the other, also turned down low, had bias readings similar to what dwmorrin was reading on his amp. After re-biasing (and just WHERE to you set Ch A's bias?), it gave me a chance to see what this yielded under load. I only recorded the N-Ch MosFET's, and found an average of 0.52 ohms as the source resistors:

Xstr..........Idle mV.............Idle mA............10W/8 Ohms mA..........100W/8 ohm mA
Q110........7.9mV...............15.2mA.............80.8m A......................277mA
Q112........4.3mV...............8.30mA.............61.2m A......................240mA
Q114........38.7mV.............74.0mA.............202mA. ......................452mA
Q116........9.5mV...............18.3mA.............94.2m A......................296mA
Q118........4.6mV...............9.23mA.............62.9m A......................240mA

Q210........25.0mV..............48.1mA.............123mA .......................329mA
Q212........24.0mV..............46.2mA.............112mA .......................308mA
Q214........16.0mV..............30.8mA.............87.7m A......................275mA
Q216........23.0mV..............44.2mA.............115mA .......................321mA
Q218........11.0mV..............21.2mA.............74.4m A......................256mA

I will end up replacing the set on Ch A, as there is a definite current hog in the set (Q114), but it was at least somewhat reassuring that the difference between devices dropped substantially under higher currents. Would Q114 fail prematurely? Well, prior to the amp being in the shop for preventative maintenance, the last time it was serviced was April 2010. I'd have to check the rental log to see how active it had been used, but, in spite the numbers looking a bit alarming, it hasn't failed. Good thing there are 5 pair of complementary MosFETE's per channel in the output stage.

[alansombr]

Of course MOSFETs installed in SVT4Pro are matched. Most probably not by Ampeg but by the manufacturer of the transistors. They are marked with a colour dot which most probably corresponds to Vgs(off) range.
The method they use is most probably similar to the one you used. This is the reason why in the amps that just left the factory there are discrepancies in currents of the MOSFETs (because the method is not that precise). But they are in a range of 30-40 per cent and not as high as you got. It seems that 30-40 % accuracy is sufficient for this amp.
If you want to know the current for the operation, you can calculate it. You have 0.47R resistor and 25mV - it looks like 100mA test is better than 200mA.
I was also using the method you mentioned. But there is much better method if you have a working amp. You just set average current, desolder one transistor, solder three long wires, attach tested transistor to a (isolated) heatsink and measure the transistor "in-circuit". You can even do it without soldering, with clips, assuming you are very careful when doing this. Of course, tested transistor has to be connected when the amp is turned off :-). With this method you can easily get 5 per cent accuracy - much better than at the factory.
Using transistors from one batch is not enough. I usually buy 10 transistors to get 4-5 of them matched.

Mark

Hello
Could you teach me how to do this separation?