EEVblog #861 - Rigol DP832 PSU FAIL & REPAIR

[EEVblog]

Dave repairs his Rigol DP832 Lab Power Supply.
Why did it need repairing? That's the FAIL part Rigol might want to look into...

[optoisolated]

I've had my 832 shorted many times with relays, and no issues to date. Definitely be keeping an eye out for that.

[DrGeoff]

Did you actually have 3A current limiting enabled?

Not saying that this is the cause, however interrupting full load DC can cause a large voltage spike (see DC circuit breakers and fuses). I've seen kV spikes when using interrupters on 48VDC/50A power supplies. If this was the case with the relay contacts then it may have been enough of a spike to damage the mosfet. A varistor on the output rail can clamp this kind of spike.

[Brumby]

A classic - and simple - example of troubleshooting.  I would classify this as recommended viewing for anyone starting down the track of equipment repair.

[EEVblog]

Did you actually have 3A current limiting enabled?

It's always enabled, you can't disable it.

[c4757p]

I'm dumbfounded at your annoyance by the hand-soldering around the pass transistor. How else are they to attach it to the heatsink?

[optoisolated]

The hand-soldering would be fine, if they cleaned up afterwards.

[BravoV]

Not an expert, but when everytimes I saw a typical switching power mosfet used in linear mode, it worries me on it's reliability, especially its not a special manufactured "linear" mosfet.

We had a discussion while back about this, and even the mosfet's rating is heavily derated, still its somehow a guessing work. 

Love to hear any experienced members here to comment on this topic.

[c4757p]

optoisolated, nobody ever does that for a small number of hand-soldered parts in a low impedance area. It's a waste of time. The flux doesn't do anything except annoy Dave.

BravoV, yeah, I worry about that too. I disagree with Dave's general assessment of "MOSFET = bad for linear", but if the SOA curves don't explicitly allow this operation, it should not be used that way. The datasheet for this one doesn't seem to have SOA curves, so... naughty Rigol. Edit: I stand corrected, it does. And they look fine. So sorry Rigol, not naughty for that, but something else isn't quite what it should be

I'd be interested in how this thing behaves as the contacts are closed and opened. I wonder how well they're doing to absorb the inductive kickback, and I also wonder how the current control loop responds - quickly shorting and opening is a good way to 'trick' the current loop in many PSUs into delivering more current than it's programmed to...

[Smokey]

Dave Says:
"...I would not expect a lab power supply of this price and grade to blow a pass transistor..."

Really?  It's a $450 power supply from Rigol.  That is pretty much the supply price and grade I would expect to randomly blow a pass transistor.  The Agilent/Keysight listed below is 80W and cost 3.2 time as much.  Now that is a supply of price and grade!

Some name brand 3 output supplies at tequipment now:

Rigol DP832 = $450 USD
http://www.tequipment.net/RigolDP832.html?b=y&v=7906

BK Precision 9129B = $695.00
http://www.tequipment.net/BK/9129B/DC-Power-Supplies-Lab-Power-Supplies/

Rohde & Schwarz HMC8043 = $1,300.00
http://www.tequipment.net/Rohde-&-Schwarz/HMC8043/DC-Power-Supplies/Lab-Power-Supplies/?v=7421

Keysight/Agilent E3631A 80W= $1,442.00
http://www.tequipment.net/agilente3631a/?v=7421

I'm not saying the DP832 isn't a good supply for the (small) money, but it's silly to pretend it's a high class instrument.

[c4757p]

Blowing a pass transistor because of an inductive/intermittent is faulty design, though. It's a general-purpose supply, it's meant for driving loads like that. I think it's perfectly reasonable to expect your device to be not faulty, whatever you happened to pay for it.

[amspire]

Not an expert, but when everytimes I saw a typical switching power mosfet used in linear mode, it worries me on it's reliability, especially its not a special manufactured "linear" mosfet.

The mosfet is rated for linear use - it has a DC Safe Operating Area curve and 38V/3A is within it. However, if you could get 100A peak for 1mS while waiting for the current limit circuit to start up, that could cause a cell in the mosfet to commence thermal runaway, and from that point, you can throw out the SOA curves. The MOSFET does have low channel resistance and low gate voltage, so it is certainly vulnerably to thermal runaway. Transistors tend to have self current limiting - they cannot put out any more current then the base current drive allows, and the higher the current, the lower the gain. This MOSFET is driven by voltage, and the device is probably capable of conducting 1000A under the wrong circumstance.

A more likely issue is that you have to keep the gate-source voltage under +/- 20V under all circumstances. Dave was ripping the source down from 30V to 0V in nanoseconds, possibly. If the gate voltage does not follow the source voltage down immediately, it can blow. The addition of the diodes looks like they were trying to add extra protection - probably to the gate. The trouble is that diodes still have a turn-on time, and if the gate is going over-voltage for a few nano-seconds, that may be enough to micro-puncture the gate, and then it is only a matter of time before it fails completely.

Bring back the 2N3055!

[dreaquil]

I'm thinking it was probably caused by a latent ESD fault.. maybe it had one foot in the grave already and all it needed was a good amount of current to push it over the edge. Either way, either bad design or bad quality control.

[BravoV]

BravoV, yeah, I worry about that too. I disagree with Dave's general assessment of "MOSFET = bad for linear", but if the SOA curves don't explicitly allow this operation, it should not be used that way. The datasheet for this one doesn't seem to have SOA curves, so... naughty Rigol. Edit: I stand corrected, it does. And they look fine. So sorry Rigol, not naughty for that, but something else isn't quite what it should be

I'd be interested in how this thing behaves as the contacts are closed and opened. I wonder how well they're doing to absorb the inductive kickback, and I also wonder how the current control loop responds - quickly shorting and opening is a good way to 'trick' the current loop in many PSUs into delivering more current than it's programmed to...

The mosfet is rated for linear use - it has a DC Safe Operating Area curve and 38V/3A is within it. However, if you could get 100A peak for 1mS while waiting for the current limit circuit to start up, that could cause a cell in the mosfet to commence thermal runaway, and from that point, you can throw out the SOA curves. The MOSFET does have low channel resistance and low gate voltage, so it is certainly vulnerably to thermal runaway. Transistors tend to have self current limiting - they cannot put out any more current then the base current drive allows, and the higher the current, the lower the gain. This MOSFET is driven by voltage, and the device is probably capable of conducting 1000A under the wrong circumstance.

A more likely issue is that you have to keep the gate-source voltage under +/- 20V under all circumstances. Dave was ripping the source down from 30V to 0V in nanoseconds, possibly. If the gate voltage does not follow the source voltage down immediately, it can blow. The addition of the diodes looks like they were trying to add extra protection - probably to the gate. The trouble is that diodes still have a turn-on time, and if the gate is going over-voltage for a few nano-seconds, that may be enough to micro-puncture the gate, and then it is only a matter of time before it fails completely.

Exactly like that, as always, when it comes to this particular topic, the discussion was and will likely to drift out to the driver and control mechanism, loop control & behaviour, etc .. etc that ended up way too complex, and from what I've observed, cmiiw, that it self made it less and less appealing than BJT.

Again, cmiiw.

Bring back the 2N3055!

Or even better, this reminds me back the General Purpose Power Supply Design of yours using "TONS"  of high hFE 2N2907As.  I have been patiently waiting for almost 3 years for you to finish it, at least the analog section. 

[EEVblog]

BravoV, yeah, I worry about that too. I disagree with Dave's general assessment of "MOSFET = bad for linear"

I neither said nor implied such a thing. I said they can have some potential drawbacks, but in fact they had a lot of positive aspects for such use.

[EEVblog]

Dave Says:
"...I would not expect a lab power supply of this price and grade to blow a pass transistor..."

Ok, I should have left out the word price.
The simple fact is Rigol pitch it as a fairly high end lab grade precision power supply.
http://beyondmeasure.rigoltech.com/acton/attachment/1579/f-01c1/1/-/-/-/-/DP800%20Datasheet.pdf

Typical Applications
? R&D lab general purpose testing
? Quality control and assessment
? Pure power for RF/MW circuits or components
? Power supply for automobile electronic circuit test
? Production automation testing
? Device or circuit characteristic verification and troubleshooting
? Educational experiment

If it's possible to blow the pass transistor by shorting the output then it's no such thing. Most ebay no-name cheapies won't fail by doing this.

[c4757p]

Most ebay no-name cheapies won't fail by doing this.

Yes they will

[Brumby]

There wasn't any apparent inductive load that I saw to cause back EMF on the channel that failed, so I'm wondering how that is being brought into the discussion?

The relay coil was energised from channel 2 - which didn't have a problem.

[amspire]

There wasn't any apparent inductive load that I saw to cause back EMF on the channel that failed, so I'm wondering how that is being brought into the discussion?

It could be something similar, but capacitive instead of inductive.

When the wires were being brushed together, if you could get the initial contact just long enough to drag the mosfet source down to zero, but not long enough to discharge the output capacitor via its ESR, then at the end of this initial contact, the source can jump back up to near 30V due to the output capacitor charge. If a protection circuit has just dragged the gate down to near zero, there may now be up to -30V applied to the gate after the source voltage rebound.

So it could all come down to getting exactly the right initial contact time when brushing the wires together. It might be that a short of just the right amount of time blows up the supply. Shorter or longer and it is fine. Sometimes, power supplies deliberately use high ESR capacitors on the output for stability reasons.

[bktemp]

Not an expert, but when everytimes I saw a typical switching power mosfet used in linear mode, it worries me on it's reliability, especially its not a special manufactured "linear" mosfet.

The mosfet is rated for linear use - it has a DC Safe Operating Area curve and 38V/3A is within it. However, if you could get 100A peak for 1mS while waiting for the current limit circuit to start up, that could cause a cell in the mosfet to commence thermal runaway, and from that point, you can throw out the SOA curves. The MOSFET does have low channel resistance and low gate voltage, so it is certainly vulnerably to thermal runaway. Transistors tend to have self current limiting - they cannot put out any more current then the base current drive allows, and the higher the current, the lower the gain. This MOSFET is driven by voltage, and the device is probably capable of conducting 1000A under the wrong circumstance.

The datasheet looks too good to be true:
My rule of thumb: Whenever you see a straight line in the DC SOA curve with no derating at higher voltages, you know it is wrong.
Every transistor is more susceptible to second breakdown (or whatever it is called for MOSFETs) at higher voltages, therefore the SOA curve always has a steeper slope towards higher voltages.
Here it shows a constant 300W limit right up to 150V.
And the SOA curve is only valid for Tc=25°C. In any real power supply, it will be much warmer because it is dissipating power.
And 300W in a TO220 package? You probably need liquid helium to make it work realiably.

The datasheet says:

Super high dense cell design for extremely low RDS(ON).

High density means it is less suited for linear operation, because the die is small.
It looks like the designer of the power supply simply trusted the specs in the datasheet, without thinking about if it is plausible.
There is nothing wrong with MOSFETs for linear operation, but you need to understand the limits, and it looks like the Rigol engineers did not.

Did Dave in one of his videos measure how fast this power supply goes into current limit? Then we could estimate the power dissipation in the pass transistor.
If Dave has not done it, it would be a nice follow up video to measure the power dissipation (measures voltage across pass transistor and current using the same scope and multiply + integrate the dissipated energy and compare it to the datasheet ratings).

[EEVblog]

Most ebay no-name cheapies won't fail by doing this.

Yes they will

Name one.
I've shorted countless cheap linear supplies over the years, never killed one that I can recall.
Only exception would be the Korad, but that wasn't a short with leads, it was oscillatory constant power mode on a load.
I wouldn't say you can't of course, but IME it's not common.

[Skimask]

Where have I seen that damn fuse before?
https://www.eevblog.com/forum/testgear/rigol-dp832-smoked-channel-1/msg580527/#msg580527
Good info all around though for future reference when I blow up something else on it...

[tggzzz]

However, if you could get 100A peak for 1mS while waiting for the current limit circuit to start up, that could cause a cell in the mosfet to commence thermal runaway,

100A through 1mS is 100kW, so I would have thought it would cause thermal suboptimalities!

Or were you thinking of milliseconds?

[hayatepilot]

Great video! 
Is this now the end of the EEVblog repair curse?
Let's hope for the best. 

[AndyC_772]

Maybe there's a follow-up video to be had here?

Hook up a scope to the MOSFET and measure Vgs and Vds.

Then, apply various conditions at the output terminals, and see what the MOSFET is subjected to in each case. Maybe stick a differential probe across that current sense resistor too, so we can see the drain current and calculate the instantaneous power dissipation in the die.

Then, see if there are, say, transient spikes on the gate which exceed its Vgs rating. See what the maximum Vds is, and what the power dissipation would be. Try and identify if there are load conditions which place undue stress on the MOSFET, and perhaps come up with a modification to protect it better.