EEVblog
HTML-код
- Опубликовано: 3 окт 2024
- Dave started out wanting to investigate the power-on spike on the Rigol DP832 Lab Power Supply, but ended up hunting down a reset bug that uncovered a bad thermal design mistake in the supply.
Watch Dave hunt it down step by step, and almost get duped by some marginal oscilloscope triggering.
How will Rigol respond to this?
UPDATE: Rigol have already fixed this issue: www.eevblog.com...
Followup video is here: • EEVblog #549 - Rigol D...
Forum: www.eevblog.com...
Initial user confirmations here: www.eevblog.com...
EEVblog Main Web Site:
www.eevblog.com
EEVblog Amazon Store:
astore.amazon.c...
Donations:
www.eevblog.com...
Projects:
www.eevblog.com...
Electronics Info Wiki:
www.eevblog.com...
I did a project with a Chinese/Hong Kong company a while back. During initial discussion I mentioned how important power supply design was to the project. Later on, I ended up in the factory in Hong Kong in the factory. The air conditioning was running full blast and everyone wore coats to keep warm. There was the product power supply under the blast from one of the air ducts. I did the usual check for hot stuff and found one of the switcher transistors was so hot, my finger slid right off. It vaporized the grease on my finger! When I questioned the designer, his comment was it had been working fine for two weeks, what's your problem? Further subtle hints were met with similar success. A month later, the sample they sent to England caught fire, along with the computer it was attached to. I so often see power supplies with the circuit board brown from overheating and it is so unnecessary.
Regards,
Jim
For absolut maximum performance, design with absolut maximum ratings !
LOL! (I hope that was a joke...)
@@power-max well not for rigol it seems
I always like watching Daves videos and listening to his rants. I don't see his long videos as a "waste of time", as you clearly do. I see this as entertainment. I don't watch TV anymore, I watch this kind of stuff.
Thanks for the video. I watched your video yesterday and I identified a potential overheat LDO issue in our design today. That's why I really like your channel. It solves real life problems.
Awesome, I love hearing that!
The minimum input voltage for a 5V lineal regulator is approx. 8 volts. A smart fix would be to put 1N4007 diodes in series to reduce from 12V to 8 volts at the input of the regulator. With a nominal current of 0.7 A this gives a total dissipation of approx 2 watts versus 5 watts on factory design. Another option is to use a ceramic resistor in series at the reg input. Regards.
They are seriously cheap, simple, and foolproof. Unless you don't calculate the dissipation properly.
:) nice video.
I had that heat problem with my first homemade circut. I solved the problem with a bit bigger heatsink and two diodes in front of the lm317 to spead the heat dissipation over 3 components.
Well I can assure you I have tried several times to watch Mike videos, but seriously, Dave work is so much clearer and pleasurable to watch. Also Shariar vblog is absolutely top notch. I will give Mike another try don't worry, once I get the courage to do so. I tried the Ipod nano video and just had to quit after 5 minutes as it all sounded as some background noise to me... As for the Disney Channel, since they purchased Ghibli and Ardman studios, they certainly have so interesting stuff too...
I watched you unbox this Rigol and from what I have seen with the problems you have had, this is a pretty good indicator to stay away from Rigol products.
Almost certain they didn't do it deliberately. If they did, they wouldn't already have the fixed board in the pipeline
Others have confirmed the temperature readings, so same current is implied.
Welcome back Dave! Its nice to see you getting back to your normal self with an outrageous rant at Rigol. Glad your feeling better. Great video, thumbs up!
Lesson learned: Don't start tracking down faults on stuff. You'll end up finding a huge heap of problems on the way.
If the Rigol comes with the new fixed board then it should be ok. Spend more for the Agilent? - your call. It's a good supply, but only single display.
It comes direct from the tap on the transformer, so it is what they designed it to be. 12V average with my mains input, and maybe 1V of ripple.
I've used LM317's very often over the years. It's my goto regulator for any project under 1.5 amps or so. Five watts is well inside it's design parameters however, I can think of no possible reason that a designer could possibly over look the thermal consideration of venting 5 watts. I find it hard to believe that this is a design mistake!!!! I smell money. I do wonder????
Dropout voltage is about 2V at that current , so it's not designed for 6V input and there's a more complicated design failure than simply doubling the input voltage.
Have you noticed if things aren't perfect Dave rants on and on. but that's because he takes a lot of pride in his work and wants to do things right.
Because is takes significant time, and I have to do it different ways for youtube and the blog site. And when I get virtually no positive feedback from people using it, why bother?
I generally don't bother with his videos unless they're 30 min+
Different strokes I suppose. Keep doing what makes you happy Dave, that's seems to have worked so far.
Did Rigol send you a new Keysight or Fluke anonymously hoping you'd use it instead? (smile) That takes balls and knowledge, finding the fault and blaming it on a poor design, far beyond the abilities of most technicians. Great video.
Doesn't matter. Even a genuine one would be operating outside it's max recommended die temperature range.
I got hardly any feedback at all on the bookmarks last time. Most people don't even read the description let alone expand the description to use the bookmarks.
Thank God for techs like you. I'm retired military and worked on the repair and caibration of test equipment.The manufactures have tried to send junk to the system. We wrote it up and recommended its return, I was thinking of purchasing this type product from Teleequipment.I'm glad I didn't (although I have purchased equipment there that was OK).
Not really. If they don't like it they generally just don't talk to me again!
Unfortunately, all the big name manufacturers are making those crappy thin tabbed TO220's now! ST and ON both are doing it. We use the tab for mechanical support and can't do it now. They call the new package "single gauge" and the old thicker tab "dual gauge" tabs in the datasheet, btw. I bet this saved the manufacturer a lot of money since most of the cost is the copper in that tab.
Why would I replace the heatsink? Why would I want to impromptu fix a brand new unit that has a long term design fault? The unit worked with the lid on, no need to fix anything right now. I waited to see what Rigol had to say about it and it turns out they are replacing all units. Replacing the heatsink would have been a stupid and pointless, if I even had a suitable size and shape heatsink in stock to begin with.
0x200th video! Great rant. Its bizarre how such precision complicated piece of equipment is let down by one of the simplest components in the entire box.
As an interim fix why not pre regulate the 12V down to say 9V with a 7809 or similar and spread the thermal load across two devices. The best solution would be to find a lower voltage secondary tap. Even the output supply taps seem rather generous 54V to get 30V out, A lot of wasted power right there, at least a few trips to the moon by Dave's standard.
This vid has restored my faith in Dave. I was thinking he was a bit of a Rigol fanboy. Guess that theory is blown.
Several remarks here :
-Rigol means "Laugh Out" in French
-Why not watercooling that thing with a bucket full of ice cubes and an aquarium pump ;-)
-Why do I give a F... designing my power supplies for proper thermal management if Rigol don't even care...
-I think the CEO will be glad about this "million dollar" video !! They now have a price on your head Dave !!
-I hope for the head designer that Rigol is not a Chinese company, otherwise he'd be recycled into spring rolls pretty soon !
Before that was the Korad blowing up!
The angry Dave is back. I was getting worried about you being so polite in your past videos. :)
Somehow it's getting common to use 7805 or just LM317 as a 5V rail regulators, but why on earth do you need to give it like ~12V, I saw even ~20V in some TV, the poor thing just unsolder itself from board. In few words they just look in data, ahh 1A current that's enough for us, but totally forget to make up simple math over it.
HA.... EEVBlog slaps the legs of badly behaved manufacturers..... AGAIN!
Just can't see videos like this from anywhere else. It's fantastic that years after starting this blog, you're still not afraid to point at the elephant in the room when necessary.
Nice one Dave....
Yeah, but doesn't explain why the reset fault goes away when cooling the reg with external fan.
Dave, tnx for all your great videos! I noticed the heat sink on the power supply was not even close to the el-cheapo stamped one that you are getting the specs off of. Back in my day if the part was too hot to touch than MORE heat sink !!!
I noticed engineers now days save a few pennies by not providing enough heat dissipation and basically just copy n paste from the data book.
Any ways, love your videos and keep em coming !
73 N8AUM Vidas
By the way dave described the architecture I believe the ADC is local to each PSU channel section so that would not have an impact. In general design terms, any sensitive noise devices such as ADC's would have guarding and filtering on the supplies (resistor can cap or indictor and cap). Cant be sure without looking in more detail but thats typically how its done - mind you if they can't even get an LM317 in spec who knows...terrible design flaw for a company wanting to compete with HP!. Gerry
It will not only be the DP832 that is affected, the 832A will have the same problem if they are sharing the same board.
The fix is easy. Use a switching 7805 replacement regulator like V7805-1500. The first time I used them they were a product of Power Trends, later bought out by TI. It was designed to be a drop in replacement for a linear 7805 (LM317) but with more efficiency and a output current of 1.5 Amps instead of the 1 Amp rating of the 7805. Simply pop out the 7805 with its heat sink solder the new part in and you're off and running.
I agree, putting a bigger heat sink, maybe, but why waste good electricity on heat, maybe just a super small transformer to just reduce a couple of volts off the regulator. I do not think any part should go over 140 degrees Fahrenheit.
Oh, yeah, didn't notice, thanks!
Yes there is a 90% switcher in parallel with the linear one, but you need to pay the fee to activate it!! Ahh the days when we will see designs like that ;)
Agilent supplies have always been the ducks guts if price is no object.
Dave, depending on how much the unregulated voltage drops under load, the full power output may not be the worst case for the pass transistor. The worst case should be at the lowest output voltage before the first tap switch happens, at maybe 80 to 90% full current, I'd guess.
The local rep has confirmed that Rigol will be exchanging all Australian units.
You know what? I think it was designed that way on purpose, so it will fail and most people, and specially big companies using this, will just buy another one. Designed obsolescence. You see it in the computer world all the time (that's my background), power supplies with obviously underspeced CAPS, super high speed HD's with little to no heat sinking (I'm looking at you HP and your 2.5" SAS drives), and the list goes on...
+Marvio Botticelli Yes. I found a 10V -rated cap on the 19V input of an Acer laptop. They do these things.
Marvio Botticelli planned obsolescence exactly.
The (incorrect) assumption is that the overheating LM317 is causing the 5V digital supply voltage to drop below some threshold and this caused the DP832 processor to reset. I have been able to replicate this exact scenario with my DP832A after upgrading to the latest 1.14 firmware version. Replacing the LM317 with a switching regulator does not fix the reset issue and neither does using an external 5V supply. Yes, I know the LM317 is getting quite hot, but this is not the cause of the resets - sorry Dave ;-)
Unplugging the flat ribbon cable from the top power board cause the software to report an error with the analog interface, but reboots also stops completely. The problem seems to be related to the firmware version coupled with an (as yet unknown) issue between the main control board and the top power board. The processor either detects an out of range condition while monitoring the analog board and resets (unlikely) or the comms between the two boards stalls and a watchdog kicks in, causing the main processor to restart. This is most likely a protection mechanism to prevent serious damage to the PSU or attached equipment.
I bought my DP832A some time ago from the South African Rigol agent (now bankrupt), so I don't think Rigol is going to replace it for me. Shipping the unit back from South Africa is also not practical, so my only option is to fix this bugger.
Has anyone managed to solve this issue on their own?
Also the 5V regulator input terminal can be cut to insert the resistor or diodes in series to avoid modification on the original PCB,
I was in Hong Kong working on a project. my part was not the ps but when I was shown the prototype I touched things. everything was OK until my finger slid off a to220. it was so hot it vaporized the grease on my finger letting my finger glide over the tab on the vapour cloud. The room temp was around 68f and air conditioning was blowing like a hurricane. tried to explain to designer this was not good with no luck. they put the case together and shipped it to England where it caught fire and burned up the computer it was attached to. some people don't get it.
It's about time you post this again. Seems like that happens about once a year.
@@xiretza3793 lolz
Great video...
Would love to see it running with a proper heatsink.
But then you are also cooling other stuff... To really get conclusive evidence, you would have to exclusively cool the regulator, either by changing the heatsink or perhaps by clenching the tab with a pair of beffy metal pliers with some thermal paste on the tips.
Of course you are still right that the design is totally borked either way.
Apparently Rigol couldn't be bothered to have a thermal design review. Yet they managed to have a meeting discussing just which chips needed their markings removed.
FAIL.
WOW! and I was just about to buy one of these PS. Good thing I waited. I'll wait and see how and when RIGOL responds to this particular issue.
I'd really like you to keep us updated on what happens :D
That's a real bad design flaw...
It was explained in video and not you can't rely on fan when you have 120 degrees on any component. Like Dave said it's just ridiculous bad design.
A quick and dirty fix, could be 3-4 1N4005 in series with the input on th lm317, the diode-drop could be enough to lower the power to be burned off in the lm317 :-) I think i would have prefered to fix this with a step-down swiching regulator, as i see it, that would be the cheapest way for Rigol of fixing the whole batch of DP832.
Huh. My guess is they tested it only using 120V mains, and that regulator is powered from the main transformer. So you're getting about twice the voltage they had it regulating from, and 5W rather than 1W. Seeing the thin tab, I also wonder if the regulator was replaced with one rated for lower current - say a 300mA one instead of 1A. That would cause it to heat up more easily. Still, with the oscilloscope measurements I expect something is really finicky down in the load also. Not encouraging.
Yes, I was testing best case. I should have tested worst case.
Thanks Dave! I'm looking for a PS and you just saved me a lot of money.
Agilent has some pretty crappy HW designs too.
A ~1000$ power supply had problems with rotary encoders, some of their stuff are made using cheap ass materials etc.
At the end of the day I'd say you can't be sure unless you have tried out the specific model you want to buy, regardless of the manufacturer.
Summary : Rigol tried to get away by running a STMicro LM317 Pass transistor at ridiculously high 5W the transistor junction getting hot much over the part maximum operating temperature. As per Steve's word, nobody in their right mind (with integrity) would design this. This is not a mistake it's a sweep it under the rug and sell it strategy. This regulator powers the digital section and is next to supply bypass cap that will dry up because of the excessive heat, all DP832 will fail in the field. How could this even get pass a design review meeting. This is a huge design oversight, Steve believes they had to know it and sweep it under the rug by running the fan loud all the time but the supply will still overheat and reset itself. Note to Dave : Even my 1985 Marshall tube guitar amplifier which has a digital DSP section is using a 5V tap on the power transformer to supply the digital section, none of this cheap rigol cutting corners with a bypass transistor dissipating 5W to scale the voltage down to 5V !
Rigol must do a recall and fix ALL the units they sold. As for me i will avoid this brand once you know they do these cheap tricks somewhere they will do other one across their products.
Hah, the airflow behind the PSU is running over some components which attracts lots of dust. In humid areas the dust acts like a sponge and helps the PCB to corrode relatively quick (it took 3 years here).
When fixing the primary issue by plugging the vias with solder the LM317T damaged the Fuse so the PSU won't start up anymore, I powered it via 5V USB charger and was able to recover the machine. I changed the linear regulator with switching regulator afterwards. The primary issue end up to add +5V to the actual output (measured with a fluke)...
Dave your videos are awesome... Thanks a ton.. Geeky stuff is awesome... Wish there were more channels like this
I watch every video all the way through, & love 'em! keep up the great work :)
Don't get why they wouldn't just use a 7805? The extra few cents they could have saved not needing the 2 voltage set resistors could have been put into a decent heatsink.
Should have split the voltage drop across two lm317's with seperate heatsinks. You don't drop all the way from 12 to 5 if you are going to be drawing any reasonable amount of current. Thankfully though, LM317's are self regulating and will turn themselves off before they burn themselves, but still.
Yeah, who would have thought. A lousy bridge rectified 5V regulator. Bread and butter stuff.
It does seem a bit ridiculous. Digital chips don't mind switch mode supplies at all and a nice small switch mode wouldn't produce much heat to supply 7-800mA.
Just earlier today I was looking on ebay for this supply, think I'll take my money elsewhere.
Thanks Dave!
Obviously it wasn't designed to have 12V at the input. Did you check to see if that was correct? You could get a cheap switching buck convertor to reduce the 12V to 7V and solve your immediate problem but I think the 12V from upstream is your real problem. As soon as you measured 12V at the input my alarm bells went off. It wouldn't take much current to overdrive the regulator. No engineer would allow it. Check the 12V and see if that is your upstream problem.
Let's get a bigger radiator and problem solved ? Yes but there is still more power dissipating on regulator than is needed for all front panel electronics 0,7 *5 = 3,5 W compared to 5 W they have to waste in that circuit. Very efficient solution Congratulations RIGOL...
I like Dave, he wants things done right. anyway I learned a lot here, not an engineer just a repair tech.
What do you mean exactly? It was shot over 2 days as mentioned in the video. No post editing voiceover or reshoots.
You're right I didn't notice it, but the transformer is set to right tap so that's not the point...
i'm not convinced the reset comes from the few-mV-high voltage spike (the one you had problems capturing). the stuff following that regulator also has to allow for some tolerances (like induced voltage or so), so a few mv shouldnt lead to a reset.
i'd say the voltage-spikes you've captured come from the changing current usage from the reseting electronics/mcus. then voltage will change with different currents. mcu is reseting from some other reason.
still, the super hot linear reg is unexcusable
I almost wanted to buy this unit ! The features looked great, specs are awesome, price is good for an expensive hobbyist kit. Meh, thanks for the investigation ! I'll guess I'll find something else or wait for Rigol to fix this.
Connect it to a pan and boil water!
A thought when it comes to possible hardware fixes for the the panel 5V rail:
Remove bridge, mount it on standing PCB mounted to the holes. Put a couple of series diodes (say, 2 or 3 of them) on the + side to take down the voltage to the input of the '317, so it doesn't have to work up such a bad sweat (9-ish V instead of 12-ish V). Kludge? Yes. Hack? Yes. But a kludge/hack that can work - I've had to use that trick when designing a simple +- 5 and 12V supply to keep the thermals in check.
Thanks for the interesting review. I liked how you called that bad design an "oversight" while appealing to Rigol for an explanation. Pretty diplomatic way of saying "I've spotted and exposed your designed-for-dumpster style surprise in a $400 worth equipment - please fix it". :)
I'm spotting such bad design issues too, only in different sort of equipment (DJ/stage/audio) and I find most of them intentional. They usualy occure where so called "scaling of the product" exists.
Please provide some examples
@@broklee If you'd know the term "product scaling" you wouldn't really need any specific examples. You could just find and see them for yourself. :)
This is so usual... bad design!!! That thing will be dead sooner than you think... 12V > 5V, I use this config myself for my logic and always use a DC>DC converter (TPS62143 5V), it's expensive but runs very very cool, it's just amazing stuff. Before I would use a linear regulator, now I don't look back for logic at least... At 5V@2A efficiency is around 90%, I can touch it with my bear hands, it's not even warm. I was thinking buying this unit, I won't now.. Keep it up mate!
I was thinking about for example read-back voltages will jump around etc. I dont know - It came to my mind becasue they used linear regulator for a reason....but it is possible that it was just cheaper.
I would check that as it makes the most sense.
No, you'd get the same 12V.
Absolutely fascination video. Your findings seemingly have some grave implications for Rigol design? There must be huge numbers of engineers and educational establishments that have thrown their lot into Rigol as an alternative to Agilent et al. Surely this must be a disconcerting issue for future investors in Rigol or do you think this is just a small hiccup which they will ride easily?
Yikes!! I was looking at this DP832A to maybe buy it. And another Rigol product. But these problems here (design overloaded regulator fry-pan, !red+ and black- wires!, etc) really do not give the the warm comfy feeling I was looking for from Rigol. What the heck is it there!?
I can totally image the "Dilbert's boss" scenario here with a "shut-it and go back to your bench" etc.... There always seems to be some genius knuckleheads "in-charge" who think they know better and with an overly sacred "deadline" to steamroll over what the real experts say, or to push push push the poor designers/testers faster than anyone could ever get it right. Nothing ever always goes perfectly right and on time - And when there is an issue; you can't just magically sweep it under the rug or simply say "get it fixed by morning or else" and pretend the issue will evaporate on it's own.
Hi Dave the "funny wave you see" at 37:53
They show the false peak in Vgs due to opposite side body diode reverse recovery (just before the Vgs plateau stage)
Fantastic rant Dave
"Who the hell designs a 5V regulator to run quiescent at 109Deg C? What a dickhead."
The best thing here is that Dave doesn't hold back when something should have been taken care of long before production runs.
Fantastic rant Dave
Thats amazing.. How did the board ever get off the engineers test bench. He never touched the heatsinks? Nobody ever tested the unit at its max rated temperature? No way that regulator runs with ambient 40C or more..
It must go through a rectifier circuit. You might have to add some series diodes to the full wave rectifier. What an ugly fix that would make! Adding an extra set would reduce the voltage across the LM317 of 1.4 Volts and reduce the power by 1W for every extra set you put in series.
My hope is that a software glitch has the current much higher than what its designed for. I don't believe that its designed that way and that something has was changed during production.
Does that mean that if it were plugged into 120VAC, the regulator would be getting around 6V input? (I don't know how different mains voltages are handled.) If so, the regulator wouldn't overheat here in Canada. But it also means they couldn't fix the problem by lowering the voltage of that transformer tap because then it would be too low to work on 120V mains.
Poor design indeed, no excuse for that !
My guess is that they were intending to use a much bigger (more suitable) h/s for that 317 but at some point in the manufacturing it got missed and the wrong one got used. It seems that there are a number of transistors using the same h/s so for whatever reason, they all got the same h/s. poor design & poor q/c leads to component failure giving the company a bad rep and costing them money !
Would you consider replacing that 317 and h/s yourself ?
Haha, thanks, yes I am cool calm and collected most of the time :) although if that snake would have had fangs i am sure it would have been a different story! I have had some recent experiences with comments relating to time domain stuff I have been doing recently that would make be rant on a bit - that might be fun :)
Oh, I wonder if one of the little DC/DC converters like Recom R-78C-1.0 or similar would help as a quick workaround. It will dissipate around 0.6 W I think, Looking at the datasheet it has an efficiency of up to 89%.
700mA is not all that unexpected. These fast ARM chips can easily use a watt or two also back lights for such large displays can eat a lot of power at full brightness and miscalculates high speed bits and bobs like USB and ethernet PHYs also take a bit of power..
Sure is nice to have such a hug community to perform design evaluation, isn't it?
also i like the length of your videos and i have no problem skipping ahead if i need to. but when i have time to kill and want to learn something i watch your videos :)
Good thing you've got your "warranty void if broken" sticker intact.
Oh, wait... :)
I wish Dave would speak up and say how he really feels