Don't miss the obvious source of pending disaster

[nevetslab]

Of the seven amps that came over after the Cochella Music Festival, the Ampeg SVT4-Pro I set aside for last, as it was stated to no longer have output. I got the other combo amps out of the way that didn't take much time. So, Wednesday afternoon, I removed the amp from it's rack-mount road case, moved it up to the bench, and removed the top cover. My initial scrutiny spotted a couple blackened resistors in the voltage gain stage ahead of the bias circuit and the output stage, then after checking the N-Ch MosFET's easily accessible, I found open gate resistors and shorted MosFET's. So, it was a good call, leaving it to last, as these always take a good deal of time to tear down.

Only the Ch A side died. Ch B's Output Stage measured ok, all of it's source & gate resistors intact, MosFET's could be turned on and off with the DMM's Diode Check circuit. So, unsoldered the MosFET's while the heat sinks were still mounted, standing the board up vertically. Completed the teardown, getting the fan removed, it's mounting brackets removed, after removing the mtg screw for the 25A bridge, removed the voltage regulator assy, then carefully broke the bond of the MosFET's to the thermal insulation pads. I was expecting to have to remove & replace them, as all too often, they self-destruct when breaking the thermoset bond after years of use. This time, that didn't occur.

So, got the heat sinks removed, removed the MosFET's, then removed all of the open source & gate resistors, parts in the current limiter circuit that failed, Voltage Gain stage parts, zener diode, etc. Installed the replacement parts, then broke out the two tubes of tested/batched MosFET's (IEFP240, IRFP9240), and selected one set of each. Re-mounted the heat sinks, fan, voltage regulator assy prior to installing the MosFET's. Cleaned up the solder flux, double checked my work, then decided to re-install the heat sink rather than dig out my bag of Xfmr Primary lead extenders in order to power the module up external from the chassis.

Regardless of how thorough I am, the initial power-up of these after the repair always makes me nervous. I clipped on DMM channels to monitor the Output Buss, and a source resistor of both N-Ch and P-Ch, to aid in setting the new bias. Walked up the variac, seeing all is coming up nominal, though once the output buss settled down to low mV of offset, I wasn't seeing any voltage on either source terminals. I nudged the bias pot and got no change. I moved to the opposite channel, saw that was reading fine. Went back to Ch A, and still not seeing any hint of change, and set the bias pot back to where it had been. What did I miss? I moved one of the voltmeter leads between output buss and source, to instead the gate on the front row. Powered back up, and heard a snap sound followed by a small plume of smoke as I turned off the power. Sigh.............

After a cup of coffee, I went to pull the amp assembly back out of the chassis, moved the chassis aside and turned it onto it's back to assess the damage. It appeared the entire P-Ch stage failed short, and while the N-Ch stage didn't measure shorted, none of the parts were behaving like those of Ch. B. So, stood the assembly back on it's end, unsoldered the MosFET's again, and then removed the clamps from the P-Ch side, beginning with Q111 & Q113. There was a physical crack in the case of Q111. I got the rest of the clamps off that side, and removed the five MosFET's. I double-checked Q113 thru Q119, and now, these are measuring ok, allowing me to turn them on and off with the DMM in Diode Test mode. I checked the N-Ch parts, and now they too are testing ok. All of the source & gate resistors are ok, as are the sense resistors, so I won't need to remove the heat sinks again. So, what caused Q111 to fail?

I checked to see if I had any similar-marked P-Ch parts that matched this set, but what remained was of a higher current reading on the curve tracer. So, I swapped the remainder of this set with a new set, and installed those. While installing the double clamp onto the first pair, I discovered the fault. The protruding 'finger' of the clamp that holds the TO-92 part up against the heat sink was in contact with the lead of Q111's gate resistor, which was standing proud of the surface of the PCB! Plain as day, yet somehow I didn't see that before!

I bent the resistor back to clear the finger, and proceeded to complete the installation of the clamps, then re-soldered the parts back into the board.



I also see that new gate resistor began overheating, seeing the black stripe has partially disappeared, which I've since changed. I haven't yet finished this relapse, needing to check the rest of the circuit. So, that gate resistor shorted to chassis ground must have been why the attempt to set the bias wasn't working. I had powered down prior to moving one of the voltmeter probes to a gate terminal of one of the N-Ch parts. When I had powered back up, THAT'S when the failure occurred. Not sure why that is, though that was the sequence of events.

Power Amp PCB Schematics (428xxh0_).pdf

[nevetslab]

I only lost one of the 10V zeners this time around from that mishap with the mounting clamp finger. So much for being 'thorough'. I let the amp idle for about 10 minutes before bumping up the bias for the nominal 25mV average across the source resistors. Initially, I had set that for 10mV, then watched to see where it went, then turned the amp on and off a number of times to watch the bias. As it was returning to the same value, I finally bumped it up to 25mV, and checked up and down the row of MosFET's to see what the spread was like. Very tight, having had good matching on the parts..about a 2mV spread. Then, after watching it come back up to 25mV from powering down, I applied signal to that channel, no load, ran it up to where the VacTEC limited its' output to about 40VAC, then moved signal to Ch B, and ran it up under load to heat up the heat sinks, while watching the bias on Ch A. Only changed by a couple mV while pulling over 7A under single channel load.

Finally moved signal to this repaired channel, ran it up to full power, pulling the signal plug back out several times to see where the bias went, and always returned back to the neighborhood of 25mV ea. Good enough. Shut it down, put the rest of the boards back in, top cover back on, spun it round to check all the functions, and, with everything working normal, back into the road case. My mind now at ease again.

[glebert]

Nice write up. I will probably use this when I go to fix the one that has been sitting in my closet untouched for the past year or so.

[nevetslab]

Nice write up. I will probably use this when I go to fix the one that has been sitting in my closet untouched for the past year or so.

When you DO get started with the repairs on it, go to the Advanced Search here, type in 'SVT4-Pro', and then Search. You'll find some 28 posts from all of us who service these amps, with great detail on the multitude of issues servicing the SVT4-Pro amp. I see them often, as we have many in our rental depot, along with steady clients doing the same. They, along with the SVT3-Pro, SVT5-Pro and SVT6-Pro all use the same IRFP240's & IRFP9240 MosFET's, which need to be matched. I buy bulk, usually 25 to 50 pairs at a time, and measure their Gate Voltages, mark them, then match on the curve tracer. Seems like I may have posted a thread on that here too. The N-Ch & P-Ch gate voltages are different in magnitude usually, and more often than not, their drain current curves, from which I'll select a 'reference curve as my bench mark in grading them, will be different between the N-Ch & P-Ch parts. Though, once in circuit and you're biasing them up, will still yield nearly common source voltages measured between output buss and source terminals. Only the polarity differs. Unmatched parts, you'll find the source voltages all over the map. That also starts trending that way over age.