Ampeg SVT4-Pro N-Ch MosFETs Is current leakage?

[nevetslab]

The Ampeg SVT4-Pro I bought from Pontiacpete a few months ago is finally on the bench, having acquired the replacement front panel controls, knobs and rear panel Cliff Jack mounting nuts. I pulled the front panel PCB assembly out, and replaced the Midrange Freq Select Switch & Line Out Pot, then one by one, removed all of the front panel controls to lubricate the resistance tracks with Caig Deoxit and exercised them, along with the front panel pushbutton switches. After replacing the handle screws with #10-24 x 7/8” Socket Cap screws, split lock and flat washers, I put that PCB assembly back in.

Finally pulled the main power amp assembly out for inspection. I found the Ch A P-Ch MosFET’s all measuring shorted, while all of the Source and Gate resistors of both channels measured ok. The MosFET’s looked to be original, all having the factory ‘color dot’ on them, indicating factory matching. After removal of the five P-Ch MosFET’s (IRFP9240), I found only two were hard short, while the other three I could turn on and off with my DMM. I still had matched/batched IRFP9240 P-Ch MosFET’s on hand, and installed those.

While I had the assembly on the bench, I checked the bipolar supplies, finding the Negative 15V IC Regulator not powering up. I swapped it out, still not getting -15VDC. This was the PCB I was raging about, caked full of RTV gooped everywhere. While ohmmeter check on the input and output caps didn’t show a short, it still wouldn’t power up. So, had to hack away at the RTV until I finally found replacing the output Electrolytic cap (47uF/35V) was the culprit, and installed 47uF/63V Nichicon PW series caps in their place. Now I had bipolar supplies again.

With the LV Bipolar supplies working, I checked the IC’s on the main power amp circuits, and didn’t find any surprises. While I had it out, I drilled a number of holes into the Output Busses and Circuit Ground, adding test lead ‘hoops’ to aid in routine maintenance. Finding a good test lead ground in this amp has always been something lacking.

I wasn’t ready to put the PCB back into the unit, as I didn’t have access to all circuits, so I set the chassis on it’s back, stood the power amp PCB up supported by the fan, with the board’s output jacks face up, so I had access to all of the circuits from the solder side of the PCB.



I had checked all of the semiconductors in the two channels, just to be sure I didn’t have failed Zener diode clamps on the gate busses, along with the diodes in the voltage gain stage, as well as those in the current limiting circuits. All ok via Diode Test Mode on the DMM.

Attached voltmeters to the two output busses, ahead of the relays, so I could monitor the DC output levels. And a third voltmeter to monitor the Source Voltages of the Ch A output stage, with the Bias control turned all the way down, so I could bring that up once I saw I had nominal DC output buss levels (typically less than +/-10mV). Both channels measured around 7mVDC on the outputs. I begin turning the bias pot up on Ch A, monitoring the P-Ch MosFET’s. Stopped with them at around 10mV Vs average, to see what I had on the N-Ch MosFET’s, as well as those for Ch B. Ch B’s Vs readings varied quite a bit, but nothing out of control. Ch A’s N-Ch MosFET’s had two reading around 47mV, with the others ranging between 12mV and 23mV. I watched the two higher-reading N-Ch MosFET’s increasing from 47mV, passing 60mV when I shut it down, and decided to replace the N-Ch MosFET’s all together, suspecting when the channel failed, the N-Ch devices may have been over-stressed. Though, usually when the channel fails, we loose Gate and Source resistors, while this was NOT the case on this amp.

With a fresh matched set of N-Ch MosFET’s (IRFP240) installed, powering back up with the bias pot reset to CCW, as I began increasing the bias, this time monitoring the N-Ch devices. Got them set at 25mV Vs, with a tight spread on all five, and saw around -22mV Vs on the P-Ch devices, I began seeing the N-Ch devices’ Vs increasing over time (within a minute or two). And, I was seeing the front panel Limit LED flickering. I kept turning the bias pot lower, and went as far as fully CCW, while the N-Ch devices’ Vs was still reading over 35mVDC. Shut it down. The P-Ch devices were turned off, having dropped the bias down. So, seems like there some device leakage in the circuit that’s turning the N-Ch devices on, taking control over the variable bias circuit.



I first replaced the Emitter resistors R119 & R126 (47R 1W MOS). Checked again, same problem. Replaced Q102, Q103, Q104 & Q105 in the voltage gain stage, and replaced the bias xstr Q107, along with it’s mate Q106, and C110. Tried again, but still have the same problem. Replaced the short circuit clamp xstrs Q108 & Q109. Still same problem. Replaced all the resistors in the Voltage Gain Stage R118, R120, R121, R122, R125, as well as in the bias circuit R127, R128, R129 & R130, and also removed the bias pot AP101 to seep in Caig Deoxit and exercised it, checking the DCR of it to make sure the pot doesn’t go open circuit. Tried again, same problem. Next, changed the diodes D106, D107, D108 & D109, along with current limiting diodes D110 and D113.

Tried again, and I’m still finding the N-Ch stage climbing up in Is current. Turning the bias pot fully CCW doesn’t turn that stage off, while it does turn the P-Ch stage off.

I don’t have any replacement caps for the Gate Buss to Drain Buss…C115 being the suspect part. 680pF/200V axial lead monolithic ceramic cap.

Earlier, when I had noticed the Limit LED flickering, I replaced IC1 (NE5532) and IC101 (TL074). The replacement IC1 resulted in increased DC offset, now around 12-13mVDC on both channels. I was still seeing the LIMIT Led Flickering, and reading the Vs voltages, they are also varying up and down slightly. No such behavior on Ch B.

I was able to reuse the Bergquist K6 insulator sheet, managing to get the MosFET’s off without tearing or leaving deposited material on the Drain slab on the back side of the used MosFET’s. I think if I had leakage issues with the insulators, I’d have DC on the heat sinks. They don’t get grounded until the Power Amp assembly is installed into the chassis, getting ground via Fan Mounting bracket.



I just powered it up again, and watched it for a few minutes. Initially, I was seeing both the N-Ch and P-Ch devices’ Vs reading around 20mV/-20mV. Slowly, the N-Ch readings increased, with the P-Ch readings decreasing. By the time I was seeing around 35mV Vs on the N-Ch devices, the P-Ch devices were approaching 0V Vs. I’m no longer seeing the Limit LED flickering.

So, I’m somewhat at an impasse as to what to try next. I’ll have to order replacement caps for changing C115 & C116. I haven’t yet replaced the four zener diodes. The Gate buss voltages are reading 3.65VDC (N-Ch) and -3.18VDC (3.65VDC). The Is current range on the two sets of devices are around 74mA @ 3.6VDC Vs & 31mA @ -3.5VDC Vs.

1600 Watt Power Amp PCB Schematics (428XXSCH_1).pdf
1600 Watt Power Amp PCB (428XXPWA_0).pdf

[glebert]

I am working on one of these right now. One of my all-time least favorite amps to work on. The wire harnesses are too short to put the incredibly huge PCB in a position that you can probe on things. On mine it was necessary to pull every output FET (even though half the amp was working) to get to the little bits hiding under the heatsinks.

[nevetslab]

I certainly understand and share in your aggravation on that. I've made a set of Xfmr primary wire extender cables to get around their limitations on where/how you can work on it. On this one I'm chasing, thus far, I haven't yet had to remove the two heat sinks. Of course, when you loose ANY of the 5W/0.47R Source resistors, no way around it. I've lucked out there so far. I did take a short cut on the Positive Current Limiting xstr, unsoldering the part under the N-Ch Heatsink 'shelf',and installed it's replacement on the foil side, bent over, and shook the spent part out. May have to do that with the N-Ch side back-to-back zeners, though thus far, I don't believe they're the cause of my leakage fault.

I've also worked on the amp assembly set on top of an insulated thin board, set on top of the chassis (I'll post photos of that tomorrow), where the harness wires do reach. I haven't yet made an extension for the 3-wire AC Secondary that feeds the power supply PCB.

A few years ago, I put together a test fixture to aid in screening MosFET's, to find the nominal Vgs voltage on N-Ch and P-Ch parts at 50mA Id current, with 85VDC Drain voltage applied. Then, use that Vgs for dialing in the setting on my Curve Tracer to record/mark the Bulk-purchased IRFP240 and IRFP9240 MosFET's so I have matched sets to work with. I was concerned about the differences in Id Current and Vgs when replacing the output devices, but, in operation, those being inside the feedback loop, they still balance out. I've never tried to tweak the voltage string in the Voltage Gain stage to try and compensate for that Vgs difference....doesn't seem to be necessary.

[nevetslab]

This morning, having found a pair of 430pF & a 220pF Mica, all 300V CDE parts, I removed the 680pF Axial Ceramic cap C115 and put the 220pF in parallel with one of the 430pF, measured 640pF on my bridge, and installed that for C115. Powered up to see the results. Nothing changed....after the amp woke up, with the N-Ch MosFET's starting out around 10mVDC Vgs, along with the P-Ch MosFET's starting around -10mVDC Vgs, the N-Ch devices' Vgs climbed back up passing 30mVDC Vgs while the P-Ch MosFET's dropped steadily. I shut it down, and again, seeing the DC output level on Ch A at 11VDC and falling, while that of Ch B was around 220mV and falling as the supplies decayed. So, it wasn't C115 leakage. What am I missing?

Where is the DC level at the junction of R111, R120 and R121 (output from IC1 feeding junction of the Voltage Gain stage emitter resistors)? That voltage from IC1A Pin 1 starts around -13mV, and continues to drop as the Vgs of the N-Ch MosFET's increase, while the Vgs of the P-Ch MosFET's fall. By the time Vgs on the N-Ch MosFET's reached 35mV, the voltage at R111, R120 & R121 had reached -220mVDC. Not yet sure what that means, but, it's tracking the problem. I had already replaced IC1.

[nevetslab]

Retracing my steps, the bases of Q103 & Q104 are fed from the +/-15V regulated supplies (coming off of the terminals of IC101). I had a failure on the -15V regulator circuit that I had to fix. I hadn't replaced R123 & R124, which feed Q103 & Q104 bases from the +/-15V supplies. I replaced those and checked again. For a while, after powering the amp back up, both N-Ch and P-Ch Vgs voltages were remaining stable at around +/- 9mVDC. I switched the amp; off to see what happened with the DC output level. It again went up to 11VDC and fell back down, while that didn't happen with Ch B.

Powered back up, and watched the Vgs levels remaining stable. I increased bias to set them around 25mV and watched. N-Ch Vgs again increased while P-Ch Vgs decreased, and it was back to the way it's been behaving, with Bias adjust not in control of the circuits. Sigh............

[nevetslab]

I tried swapping out zener diodes D116 & D117 (P-Ch Gate to Output Buss back-to-back zeners). They, as well as D114 & D115 on the N-Ch side had all measured ok, but, since it was the P-Ch side that failed, and I've swapped out all of the parts on the voltage gain stage following IC1, along with IC101, and C115 being changed hasn't stopped this loss of control of the MosFET's bias, figured it wouldn't hurt. Made no difference. I'm down to three 1N4740A's anyway, and don't think the N-Ch side of those zeners would be at fault here anyway.

All of the 5W 0.47R Source Resistors, and the 1/2W 47R gate resistors are all ok. And, the bank of replacement N-Ch and P-Ch MosFET's are all tracking each other tightly, while they all shift away from their pre-biased starting point when powered up. Something is causing the shift, as well as when the P-Ch parts are no longer conducting, I can't turn the N-Ch parts off. At that point, I just power down and wonder what is causing this?

Yesterday, I did check to see if both heat sinks were at any DCV level other than 0VDC. They remain at 0VDC, though while running in this open configuration, the heat sinks are NOT at ground. Would that make any difference? I just now grounded both heat sinks by clipping the two fan mounting frames together and grounding them to the chassis, so the system is fully grounded. But, this didn't make any difference either with regards to the N-Ch parts conducting and the P-Ch parts no longer conducting, once the N-Ch parts began increasing in conductance after being powered up for a short period of time..

One thing I found during all of the surgery on the main PCB.....as I'd always found on that board.....the solder pads are too small, holes too large, leaving very little surface left on the solder pads, and the PCB traces are WAY TOO THIN. I've had to patch some circuits back together, as seen with the 30AWG Teflon Wire jumpers on the bottom of the PCB. Guess I better spend more time there looking for faults in the circuit, not finding components responsible for this condition.

[nickb]

I suspect you are seeing normal leakage. At zero Vgs the drain current is not zero and rises with temperature. I think it's pretty safe bet that the P and N devices to not track exactly. The important thing is there remains zero volts on the output. It you are nervous about trying with the real load use a 100 ohm resistor and test again.

[nevetslab]

I don't think so, having serviced enough of these SVT4-Pro Amps and have never seen this particular behavior. It's not unusual to find 500-800mV difference in Vgs, with their nominal transconductance quite different, but not have one side slide upwards in conductance, while causing the opposite side to stop conducting. Thus far, I haven't tried running signal thru it. The DC output level DOES remain near 0V (13mV typical). The Ch B side has similar difference in Vgs.

I see I've been mis-stating the terms. What I'm monitoring is the voltage drop across the source resistor (0.47R), target being about 25mV which is around 53mA Id at idle.. Normally, that voltage drop will reman steady. Of course it will change with temperature, but, what I'm seeing that isn't typical is that Voltage, or Is thru the Source resistor dropping on the P-ch parts, and increasing on the N-Ch parts, and finally, not being able to turn off the N-Ch parts by turning the bias pot fully CCW. I CAN do that with the other channel, and both N-Ch and P-Ch devices turn off. at CCW on the bias pot.

When I get back to the shop in the morning, I'll swap out that bias xstr Q106. When I first changed that part out, I didn't see the I had trace failure on the solder pads, and for all I know, that xstr may have been damaged. Certainly can't hurt to swap it out, other than further damage to what's left of the solder pads and traces to them. As it is, I already have two 30AWG jumper wires restoring the circuit connections. It's the part being clamped against the heat sink, so I probably will end up with one more jumper wire to wire it in.

[nickb]

OK. I thought you were measuring a difference tn Vgs between the the N & P FETS and that was changing. That's what I get for skim-reading. . If it's the across the source resistor that is completely different. Do you have a load connected?

The limit led coming on is interesting as that circuit is AC coupled. Either the input to coupling cap (e.g C106) to that is leaky or the whole thing is oscillating.

[nevetslab]

At present, the Limit LED is no longer flickering, and I'm not finding the voltage readings across the Source resistors (to output buss) varying. I don't have a load on the output. Once the N-Ch stage had shifted upwards in Is while the P-Ch stage has turned off, output is still at 0VDC, but, as one would expect, turning power off, the DC output level goes high.....seeing about 11VDC then falling. The 'working' channel B only shifts upward at turnoff to a few hundred mV then falls towards 0V. I'll know more when I get into the shop in a couple hours.

[nevetslab]

Good call, nickb! It IS oscillating. Not right away....it takes letting the N-Ch and P-Ch stages to slowly drift apart, N-Ch stage conducting more, P-Ch stage turning off. With 100 ohms across the output, that sped that process up. With 8 ohm connected, it happens faster. I checked to see that the RF load network R168 (3.3R/1W) in series with C117 (0.1uF) is intact, though the path to ground on C117 is longer than that of the Ch B network.

IC1A is the Op Amp for the power amp circuit, it having local feedback C104 (100pF) between output Pin1 to Pin 2, while the system feedback is R112 (22k) & R110 (47.5k) with RC network C109 (56pf) in series with R109 (100R) across R110. I changed out C116 (680pF) between the Neg Gate buss to Drain supply buss, extracting that part from under the P-Ch heat sink with hemostats and placing the part removed from C115 on the bottom side of the PCB.

This is the first SVT4-Pro I've seen oscillating. Now, in the process of trying to cure this, I hope I don't loose MosFET's in the process. N-Ch parts want to turn on, though thus far, haven't seen any turn on hard. I just learned CenterStaging is shutting down for the holidays, so today will be the last time I really get a chance to solve this, as tomorrow noon, they go dark for the rest of the year.

I see there's no power supply bypassing around IC1, nor any of the quad BiFET's TL074's on this PCB. No ground plane, nor any local grounds near IC1. And, just to make life 'thrilling', Ampeg's brilliant PCB layout artist called out tiny pads, large ID's, so you're dealing with tiny annular rings, which is accompanied with poor quality PCB base material, so pad failure is a constant reality! Onward with the battle.

Any suggestions from those who've encountered this problem on the SVT4-Pro's would be welcome!

[nickb]

Let's put the oscillation to one side as I've a suspicion that is just another symptom of the actual issue that is also manifested by the current drift.

It's not spelt out, but it seems this happens with no load. Therefore the only current path is through the FETS. In that case the sum of currents through the N FETS must match the P FETS. Does it? If not there's another path. Your task is to find it.

[nevetslab]

This is happening both with a load (faster in time) as well as without a load. The sum of currents has NOT been equal thru both the N-Ch and P-Ch FET's. On Ch B, while the currents thru all 10 devices vary, the sum of N-Ch parts I had measured totaled 123mA, while that of the P-Ch parts measured 129mA. Pretty close to being equal there. I have NOT had anything approaching equal total currents on Ch A. At one point, when I first began, and had set the bias for seeing around 10mV avg across the source resistors, for a brief time, the N-Ch and P-ch parts were reading similar voltages (and hence, equal currents in N/P stages), but then the N-Ch parts began conducting more while the P-Ch parts less.

I had swapped out the two back-back zeners on the P-side yesterday, with no change. The N-side are still unchanged, though that's the side that's hogging the current. I don't think those are involved.

I added local 0.1uF mono-ceramic bypass caps between V+/V- on IC1, as well as across those on IC101, then, finding a ground path coming from underneath the phone jacks, I was able to add 0.1uF between ground and V+ & V- where the IC's are been fed. That appears to have helped, not seeing the oscillation I saw earlier, but, the current distribution problem hasn't changed. So, more digging to find the path.

Last move I made today was swapping out C114 10uF NP in the RC network feeding the P-Ch side's current limiter circuit. No difference. Spent more time inspecting foil side looking for suspect circuit traces, solder joints, errors, anything. I'll get one more shot at it before having to shut down before noon PST, then CenterStaging shuts down for the holidays.

Then, get back to my SVT-CL & SVT-VR Service Manual project and more reading, following the posts here. Merry Christmas to ye all!

Following nickb's suggestion of finding WHERE the increasing current is going when the N-Ch MosFET's are conducting, 35 to 45mV each across the 0.47R/5W Source resistors, and the P-Ch MosFET's have turned off.....only path I can think of is thru the RF network R168 (3.3R/1W), assuming C117 is leaky (0.1uF/100 film). I'll swap out that cap in the morning. Current has to be flowing someplace...everything else is DC blocked, and high Z. That N-Ch stage conducts high, with P-Ch stage off whether there's a load on the amp or not. It's NOT flowing thru the P-Ch stage, or I'd see it on any of the Source resistors there. Happens much faster when there's a load on the amp.

[nevetslab]

This morning, I replaced C117 0.1uF/100V mylar cap, part of the RF load on the output buss, parked next to the Relay for Ch A. It was locked in with RTV, so had to hack thru that. Lead spacing was 6.3mm, and what I had was an Evox/Rifa 0.1uF/250V MPP film, so had to get creative with the offset leads/body of the 3.3 ohm resistor, to tack-solder one folded lead to that resistor lead, then everything fit back together. But, it didn't change a thing.

Both the N-Ch MosFET's and P-Ch MosFET's started up with 12mVDC across the 0.47 ohm source resistors (25mA ea, across the five pairs). I was reading -3.1VDC on the P-Ch gates, and 3.79VDC on the N-Ch gates. No resistive load attached to the output, just my Amber 3501a Analyzer, with scope monitoring the output, wideband. Within a minute, the N-Ch Fets turned on harder, Vs jumping up to 35mV (74.5mA ea, total 372mA for the N-Ch stage). P-Ch stage Vs dropped down to 4.5mVDC. The Gate Voltage buss on the N-Ch side came up slightly, now reading 3.90VDC, while the P-Ch gate buss dropped to -3.09V. No oscillation on the amp, so the local bypassing I did around IC1 and IC101 appears to have helped.

But, I still am not finding WHERE that 372mA of current is flowing to. There's no DC load on the output that I've found. I looked at the clock, and saw it was now 11:15AM, so I had to shut down the bench, and pack up supplies to take home with me for the holidays. So, I guess I'll have to do some digging in existing service doc's on the SVT4-Pro or SVT5-Pro, that being nearly identical on it's two power amp channels, just lacking the fifth pair of MosFET's. Different preamp. I won't get back to the shop until Jan 2 to continue on the bench. As I had stated before, I've not seen this problem before on the SVT4-Pro or SVT5-Pro power amps.

[nevetslab]

This past few days, having been sheltered at home with the shop closed for the holidays, I loaded a simplified version of the SVT4-Pro Schematic into National Instruments Multisim 11 to see what I could learn, hoping I could find what is causing the shifting of current balance in the output stage at idle when you bring up the bias to turn on the output stage. I began with just one pair of IRFP240's and IRFP9240 MosFET's in the output stage. Having loaded four Simulated DMM's into the circuit to monitor current thru the N-Ch and P-Ch Source resistors, along with monitoring the potential of the Gate Busses, and the DC level following IC1 where it hits the center drive point of the Voltage Gain Stage, I was finding the current limiter circuits turning on, pulling the Gate Busses low. When I removed coupling diodes D110 and D111, I was then seeing high current flowing thru the Source Resistors R139 & R140. I never did achieve a stable DC state with the current limiter circuit disabled.

On the bench, I've never found the current limiter circuits turning on. When they do, they pull the two Gate Drive busses down, less than a volt across the pair. But, sure saw that on the simulation. So, I added the rest of the output stage, now being the same as the working circuit with all five pairs of MosFET's present. In the circuit I loaded into Multisim, I didn't include the front-end Optical Anti-Clipping circuit. I just made that stage another inverting stage with the same gain values seen around IC101B. I used the other half of the 5532 IC1B for that stage.

I've attached the Multisim 11 file, as well as the loaded circuit in pdf, in case those of you having NI's Multisim to see if you can find what I'm missing. I did at least find the current flowing thru Q102/Q105 being around 20mA on the simulator. The adjustable bias pot AP101 in the program can only be incremented in steps of 5% between 0 and 500 ohms. The program crashes often, so there's a lot I haven't learned in using it with this many parts of a circuit loaded into it.

In real world, I haven't yet had any failure of the output stage devices while attempting to get this circuit to work correctly. I was wondering if the half-wave rectifier circuit used to power the Voltage Gain Stage (using C2 & C3, fed from D3 & D4) isn't holding charge, such as the negative side at C3 is falling due to a bad 330uF/100V filter cap. Tomorrow, before I tear down the test setup left up since before Christmas, I'll have a look at those potentials. I did set up the simulated circuit in Multisim with separate +/- supply potentials so I could play with those voltages to see if that's involved.

I just tried to upload the Multisim 11 file, but the system won't allow loading that attachment for some reason. I'll try copying and pasting it here. That didn't work either. I suppose I could email the file for those who have Multisim 11 or later. File size is 477k.

Ampeg SVT4-Pro Simplified PA Sch_Multisim 11.pdf