EEVblog #512 - Rigol DP832 Bad Design Investigation

[Mikey]

This is awesome, even if it is a stupid mistake. That mistake just teached me how to read the datasheet on a heatsink!

[electrocat]

Hi

A better fix for the voltage regulator would be to change the transformer winding, from the hunt for the problem there looks to be +8v a drop across the regulator, no reason to have so much voltage drop, 4 volts would give lots of margin while dropping the power dissipation in half. With the larger heat sink things would be be just right.

 

[Zbig]

Hi

A better fix for the voltage regulator would be to change the transformer winding, from the hunt for the problem there looks to be +8v a drop across the regulator, no reason to have so much voltage drop, 4 volts would give lots of margin while dropping the power dissipation in half. With the larger heat sink things would be be just right.

If you mean lowering the currently used transformer tap output, that would mean re-winding the transformer (basically, making a new one). Hardly an optimal solution. If you mean using another tap which is already there used for the main output, well, the point is to have the logic supply (and the interface ports) separate and galvanically isolated from the main outputs. Anyway, most fans are rated 12V and I strongly assume this one is 12V as well.

[Zbig]

If I happened to have one of those earlier version Rigols, I think I'd switch the regulator for this fella: http://www.jm.pl/karty/AMSR178NZ.pdf

I'm already using this part in a one-off hobby project of mine and couldn't be happier with its performance. Well, it's 1A only so USB port loading with 500 mA would be a no-go but I'm pretty sure these Aimtec guys are making beefier ones as well (no longer 7805 pin-compatible but that's not that big of a deal given the available board space, I guess).

[salviador]

what a disappointment! rigol 

[H.O]

Hi,
I've currently got mine loaded down as follows:
Ch1: 23V, 3A  (E-Load)
Ch2: 23V, 2.95A (Resistive load)
Ch3: 5V, 1A (Resistive load)

It's been going for ~90 minutes so far, the air comming out the back is barely warm. I'll leave it untill either it shuts down, blows up, I get bored or need to leave the house.

Clearly the overheating voltage regulator is an issue and I'm sure glad it's been brought up. I've contacted my supplier regarding my particular unit and I think what happens next will say more about RIGOL than the actual issue itself. If they won't do anything about it now I'll be a bit dissapointed but I'll just use the thing and hope it breaks down or start acting up within the warranty period. If it doesn't I'll go in and replace the caps and either apply a bigger heatsink or a switcher when the warrant period is comming to an end (if I'll remember).

The power on glitch/spike MIGHT be an issue for SOME people. Personally I don't think it matters for me and now I AM aware of it thanks to Daves video.

The front panel layout is different, sure. But really, it's not that bad to actually USE.

[alm]

Loading the supply will likely increase the fan speed, making the 5 V regulator less likely to overheat than without load.

[ElectroIrradiator]

It's been going for ~90 minutes so far, the air comming out the back is barely warm. I'll leave it untill either it shuts down, blows up, I get bored or need to leave the house.

Any particular reason for not loading the 5V rail to 3A? May as well go for broke.

Also, according to Dave's video, the heat generated in the 30V regulators will be at a maximum at 22V output voltage, not 23. The first tap switched between 21 and 22V.

[lemon]

Just a clarification.
There was some mistaken communication that Rigol would be replacing units in Oz. That does not appear to be the case, or at least was an incorrect assumption at this stage. The local rep will simply come good with whatever Rigol come up with (board replacement or whatever). Nothing is confirmed yet, talks are still in progress with Rigol.

The messages at the thread drops like a rain!

I have order one of this from 10th August, via my local distributor in my country, but I haven't received yet.
Is there any information about from what serial number there is the correction?

[tom66]

Is Dave going to do a video on the start-up spike/voltage, or is that going to be missed in light of this serious problem?

[alm]

It would be good if Rigol put a sticker on every power supply and its box with the fix. Unmodified supplies may very well stay in the distribution channels for a while, and it would inspire confidence in the customers if they (or the seller) can easily verify that the power supply has been fixed.

[H.O]

Loading the supply will likely increase the fan speed, making the 5 V regulator less likely to overheat than without load.

Yes, but mine doesn't seem to reset without load and I beleive neither does others - as long as they aren't taken apart.

Any particular reason for not loading the 5V rail to 3A? May as well go for broke.

Ideally I'd like to put it at 0.5V and 3A or something but I didn't have any suitable load at hand.

Also, according to Dave's video, the heat generated in the 30V regulators will be at a maximum at 22V output voltage, not 23. The first tap switched between 21 and 22V.

Yeah, yeah but I wanted some margin to be sure I was on the correct winding.

It's still going (~2h), the ambient temp is ~21C and the air comming out the back is ~35C. I need to go out for a bit so I'll disconnect the loads but leave the outputs on, will see if it resets. Later tonight I'll see if I can find something to load Ch3 properly and lower the voltage of Ch1 and Ch2 closer to 22V, not that I think it'll make a huge difference.

[sync]

Also, according to Dave's video, the heat generated in the 30V regulators will be at a maximum at 22V output voltage

For a constant load. It's about 70W.

With a repetitive pulsed load it could be higher, e.g. 9ms shorting the output and 1ms no load. This will keep the big smoothing caps at high voltage, eg. 45V. Then the power dissipation will be about 120W.

[ElectroIrradiator]

Yeah, yeah but I wanted some margin to be sure I was on the correct winding.

Sorry.

It's still going (~2h), the ambient temp is ~21C and the air comming out the back is ~35C. I need to go out for a bit so I'll disconnect the loads but leave the outputs on, will see if it resets. Later tonight I'll see if I can find something to load Ch3 properly and lower the voltage of Ch1 and Ch2 closer to 22V, not that I think it'll make a huge difference.

The cooling air temperature rise is not a good indicator of what is happening inside, as most of the cooling air will not have been in contact with the hot heat sinks and the transformer. The only reliable method for testing whether there are other temperature related issues, is to measure the tab temperatures of the regulator pass devices. Your stress test would only reveal any catastrophic design issues, of which there are hopefully none.

There may also be components or heat sinks, which are not cooled appropriately due to the air flow being partially blocked by 'upstream' structures creating air 'shadows'.

Even if the capacitors or power semiconductors are run well above their recommended maximum operating temperatures, they may still last literally for months of continuous operation. High temperatures are usually a long term reliability problem, not something you'd expect would kill anything in a few hours or days (unless the problem is really, really bad).

[bktemp]

The worst situation for the mosfet used as regulator is, when the output is shortet at maximum voltage and the junction is already hot. If it does not fail then, it is unlikely that it will ever fail.
The mosfet is marked 80N15, I think it is a SUP80N15 which has a specified maximum power dissipation of 300W (@25°C) and is 175°C rated. This is why a single TO220 device is enough for such a power supply and should have no problems even at short circuit conditions at maximum input voltage. Unfortionatly the datasheet does not give any information for thermal derating.

[Rufus]

If I happened to have one of those earlier version Rigols, I think I'd switch the regulator for this fella: http://www.jm.pl/karty/AMSR178NZ.pdf

The 5v powers the RS232 interface, USB, and Ethernet. Stick cables into those ports and your power supply could double up as a broadband rf noise source.

[sync]

The worst situation for the mosfet used as regulator is, when the output is shortet at maximum voltage and the junction is already hot. If it does not fail then, it is unlikely that it will ever fail.
The mosfet is marked 80N15, I think it is a SUP80N15 which has a specified maximum power dissipation of 300W (@25°C) and is 175°C rated. This is why a single TO220 device is enough for such a power supply and should have no problems even at short circuit conditions at maximum input voltage. Unfortionatly the datasheet does not give any information for thermal derating.

The derating is 2W/K. Dave measured already ~70°C heatsink temperature at a medium load (3A@30V -> 48W dissipation). The transistor case will be considerably hotter.

The transistor is maybe enough. It's maybe barely enough. The poor LM317 works with a closed case too. I missed a temperature sensor near the transistor for thermal protection. It would not surprise me if you can kill it with the right load.

[Zbig]

If I happened to have one of those earlier version Rigols, I think I'd switch the regulator for this fella: http://www.jm.pl/karty/AMSR178NZ.pdf

The 5v powers the RS232 interface, USB, and Ethernet. Stick cables into those ports and your power supply could double up as a broadband rf noise source.

Not sure I got your point... are you suggesting that this DC-DC converter has so bad ripple, etc.? Well, I haven't measured those but then again I see no reason for it to be significantly worse than other DC-DC switching converters out there. And the +5V on the USB in your laptop/desktop is most probably provided by a switcher anyway, so... 

[SeanB]

That converter will work there, just need to do a little pin shuffling to get it to fit and give it a ground pin. You probably will need a 1.5A version though, as with 700mA for the onboard micro and 100mA for Ethernet along with 500mA for the USB this will be more than the 1A limit of it. There is enough capacitance there to keep the regulator happy though.

As to the pass devices, I would think they are running very close to the max allowable, any with poor mounting or a loose screw will fail rather spectacularly.

[Rufus]

And the +5V on the USB in your laptop/desktop is most probably provided by a switcher anyway, so... 

Yes and they paid a great deal of attention to filtering the switcher input and output and board layout and I/O filtering with lots of expensive EMC emission testing to see how they were doing and after all that they still had to mould a big ferrite bead on the PSU cable.

You can't just replace a linear regulator with a switcher especially when it powers I/O interfaces which are going to have antennas plugged into them.

[tom66]

You could replace it with a switcher, but it wouldn't be a drop in job. It would be a re-layout of that part of the board. I'd probably use a switcher which had a switch frequency above 500kHz, which simplifies some parts of the EMC design.  I've had some luck using LM2734's, but their output current is limited to around 1 amp.

[Zbig]

Yes and they paid a great deal of attention to filtering the switcher input and output and board layout and I/O filtering with lots of expensive EMC emission testing to see how they were doing and after all that they still had to mould a big ferrite bead on the PSU cable.

You can't just replace a linear regulator with a switcher especially when it powers I/O interfaces which are going to have antennas plugged into them.

Sure, I understand those are all valid considerations. But have you actually checked this regulator's datasheet? It's not your average SMPS driver chip requiring external inductors, etc. It's a hybrid, all-in-one solution with inductors and everything sealed in one package. It's not the cheapest thing around, either. I'd expect it to be an "end-to-end" solution, with proper thought given to the EMI and noise issues as well. But maybe that's just the wishful thinking on my side. I have neither the equipment nor the knowledge necessary to back my assumption. I only know I've designed it into my microcontroller based project and it's all dandy so far, some rudimentary scoping on the output haven't revealed anything worrisome, either.

[Rufus]

As to the pass devices, I would think they are running very close to the max allowable, any with poor mounting or a loose screw will fail rather spectacularly.

The SUP80N15 has pretty impressive SOA curves if that is what it is.

If you want to try to kill it set maximum output voltage and current and short the output. The pass device then has to dump most of the energy in the reservoir capacitors initially at about 50v and 3A (or worse if the current limit is slow).

[neslekkim]

I was in the mode of ordering the DP832 after Dave:s review. The price is right and the functionality is to my need.
Then I saw the follow up and decided to wait for answers.
I can´t buy a power supply who has a fault in its power supply.
I have now asked the dealer to verify that the units they have are of later edition.

I can´t wait for Dave to update with a video about the new boards performance.

Thanks!

Who is selling this in Sweden, and for what prices?

[H.O]

Wow, you guys are great, I'm sure that if we all give it our best we can come with a way to kill it. Seems like that's what we're trying to do anyway....

I'm not saying the thing doesn't have issues but I'll wait for RIGOLs response before making up my mind about what to really think about it. Apart from the LM317 issue I actually like it so far, especially considering the price - it's $400 for crying out loud... And, it's not like they (RIGOL) are the first ones to have issues with devices in the field, it's how they handle it that matters IMO.

Anyway, for those who might be interested. I came back roughly 5 hours after leaving the unit with its outputs ON but no load connected. The outputs where still on so no reset here. Again, I'm not denying the issue, just saying that unit works fine with the lid on - only time will tell for how long.

So I set Ch1 and Ch2 in tracking mode, parallelled the ouputs, set the voltage to 22.25V and the current limit to 3.2A. Then I set the electronic load to pull 6.3A. Ch3 is loaded with a couple of resistors and is sitting at 0.8V @3.2A. It's been going like that for over an hour now. Case is cold, air out the back is barely warm. The e-load is getting warmer though....