EEVblog
Vložit
- čas přidán 18. 03. 2013
- Inside the Atten PPS3205T-3S 3 output precision laboratory bench power supply.
EEVblog Main Web Site:
www.eevblog.com
EEVblog Amazon Store:
astore.amazon.com/eevblogstore-20
Donations:
www.eevblog.com/donations/
Projects:
www.eevblog.com/projects/
Electronics Info Wiki:
www.eevblog.com/wiki/ - Věda a technologie
Its good to know I am not the only person who has to do complex operations twice. Thanks for the vent...makes me feel better.
Double thumbs up for shooting that teardown twice. Thanks for the effort.
Excellent video Dave, I learned a lot. Thanks. You recovered from the camera disaster pretty quickly... sign of a true pro.
Nice review, just as a tip: you could try to short thermostat contacts to see what happens once heatsink(s) reach 80C point.
Thanks for your efforts Dave ! Another tops video.
Thanks for taking the time to re-shoot the videos Dave. They were very informative.
Haha, thumbs up ! I'm laughing at your reaction on video clip failure :) Love EEVblog, realy nice job, i'm learning tone of stuff, thanks for your uploads..
Love the teardowns!
I'm pissed for ya too, Dave. I can only imagine how much damn time it takes to shoot these. Still, vastly helpful videos so we appreciate all your hard work. Thanks, mate!
Great job on the reshoot - thanks for all the extra effort! Must of been very frustrating.
I liked this before watching it, been looking forward to seeing inside!
Enjoyed the video, thanks!
I used to be a lead technician on a electronic manufacturing line for over 10 years and I've seen soldering messes like what you show in your videos. One thing that made this seem to me to be more of an isolated case is that you had other boards in the PS which showed decent soldering. I've seen every cause of soldering problem issues from dunking boards in the WAVE solder to just a little chatter on the guides shaking components out of the pre-paste. All those defective cards get put aside and eventually wind up in a re-work and then it can be" solder sucker to the rescue" scenarios. I've seen the "tale-tale" signs of these defective cards in your videos, they had whisps and slivers of hair fine solder on the boards and also solder balls. So I think these cards were re-worked and were just not cleaned properly after the re-work. That left-over solder on the surface is just an accident waiting to happen. Over time, heating/cooling/vibration can dislodge those hair-fine solder strings and they can short out between component leads, etc.
I agree with you, it was a pretty crummy job of quality control on that soldering, I can understand why it is that way, however it is unacceptable. My point is that I feel the soldering is more of an isolated case and not respective of the continued over-all process. But I've seen stuff like this get thru and with all the manufacturing quality control safeguards, it only takes one person out of hundreds not doing their job for something like this to slip thru.
Remember, copper traces (or wire even) has resistance and inductance, so if you chain the grounding, then all the components connected to it will get different impedance's and inductance's to ground.
The difference between one component's ground and another might be several hundred mV, and the inductance causing ground oscillations (ringing).
If you connect them all at a common point (aka a star, with leads radiating outward) then it is almost guaranteed they are all seeing the same ground!
Dave, thanks for the effort, even if it took you ages to film everything again, thanks
Looks like a decent power supply for the money! Solder is a little home-made but it's cheaper than the others, survived a shorts, no overshots and is quite precise what is most important! I did not expect this much!
10:10 "I'm mad as hell and I'm not gonna take it anymore!" Sounds like a lot of us at work doing our day jobs; you felt our pain.
Dave, the relay on the serial board switches the RS232 RX/TX lines between the FT232 and the MAX202. The FT232 has an output which in this case is used to drive the relay when a USB cable has been connected to a host
I had problems with several SMPSs when the load is pulsed/oscillating (e. g. PW modulating LEDs/motors).
The PSUs weren't even close to their theoretical limits (e. g. drawing 100W from a 500W PSU), but using PWM on 100Watts worth of LEDs totally freaked them out. And that example was with a dedicated 24V/20A SMPS and not an "abused" PC-PSU.
Same setup with a "simple" linear 200W PSU: no problems at all.
Good enough depends on the application. SMPS are good enough for basically everything even digital oscilloscopes use SMPS and onboard DC-DC regulators without issue. But if you want to test something without needing to worry about your PSU being good enough for something then a precision linear supply is pretty much as good as you can get normally. You want to test things one at a time not all at once in a lab setting.
Wall adaptors are almost all SMPS due to modern energy efficiency standards.
I think the linear regulator will help with noise to a degree but linear regulator's ripple rejection rating is frequency dependent. So while you will reject the LF ripple, higher frequency noise (the switching noise) will still get through the regulator and depending on your application this can be undesirable.
I think you can make a ultra-low noise SMPS supply it will just require a lot of filtering, post-regulators, better components, cost ... (And then making it adjustable and stable)
All type of interference, noise and such from all the different parts are all gathered at one point instead of several since resistance here to ground is the lowest in the apparatus. Since that point will have the lowest resistance to ground so will that force any noise (electricity only goes in the line which have the lowest resistance) to got there instead of anywhere else if there were different points to ground.
This grounding technique is the main ground standard in audio equipment.
The reason for the 2 paralleled outputs. It's not just the current, it's the watts. When you multiply the rated voltage with the rated current often exceeds the power dissipation in watts. By paralleling 2 transistors you double the watts of dissipation.
Linear supplies have a much better noise characteristic than switch-mode stuff.
It might actually be to make it have a slow response. The part is rated to a maximum junction operating temperature of 150C so there is likely still some margin of safety. If they put it above the transistor it would have to use a higher temperature cut off. Ideally you would have those directly shutting down the supply relays and a temperature sensor monitoring it actively to do fan regulation and have a soft-cutoff for redundancy.
You can also just get power transistors with thermal limiting
For extra backup you could try using one of those HDMI streaming things so you have a live backup to one of those live stream sites.
Big cap - not black version but "Negative black" - a black mark on the -ve tag
>where good coders can squeeze out enough
>performance gains to make it worthwhile
I doubt the many people bother optimising standard code with inline assembly these days.. anything that could be optimised should already be optimised in your toolchain. You might see it in places where special processor features like SIMD are used but even GCC can do automatic vectorisation. With your "byte swap" example I would actually check the code GCC generates. You would be surprised at how much GCC does.
It helps to stop ground loops.
If the power supplies are designed to be floating then you need isolation from the building ground referenced (earth) circuits or you'll tie the floating circuit's reference ground to building ground which isn't desirable in many applications.
Another big reason to have opto-isolation is for protection in case something horrible happens in the isolated section, the isolator has high withstand ratings and will protect you/other circuits/parts from damage.
25:55 is it possible the doubled pass transistors are using time division? That is, they alternate passing to reduce duty cycle? Or do you think they both run at the same duty cycle all the time?
The Channel 3 transistor at 14:35 - so that pink pad is a good thermal conductor and also a good electrical insulator? If so, I had no idea that such a material existed. You learn something every day on this channel!
About the two over temperature sensors on the top of the heat sinks. I would have thought that they would be normally closed when the temperature was low and then open when the temperature was too high, causing the units power to be interrupted. But I see that the two sensors are in parallel. So that wouldn't work as both would have to open at the same time. Thus they must be normally open when temp was low and closed when high causing the unit to trigger a failure message but not power off.
What about radiant heat between the 2 heatsink? If all channel are fully on, the temperature should be higher a bit right!? And as it turns out the transformer is cooled by the same airflow before the heatsinks, so the air can't be the coldest possible as it arrives to the heatsinks when the case is on, and what about a higher ambiant temperature? Does it hanle that well?
Say dave.. using that rs232 interface round the back, could one rig a better user interface for controlling the BPSU, perhaps using an arduino?
What is the current accuracy with high currents for a long time? These current shunts didn't seem like they are very temperature stable.
By using that transistor, They are actually saving a few cents per board because of bulk buy discount.
At 28.49 I see dodgy solder connections on the TLC568/ U7 I would swear they are broken.
At least the boards are glass epoxy. Frame, cover, front panel and other mechanics are just fair.
Design is heavily dependent on fan. A mixed blessing.
Did you see the soldering?
Easy, easy... :D... just take it apart too. I know that feelings and its mind blowing. Hats off to the not-destruction of the camera in such cases.
What about the spikes in the output that did not originate with the power supply?
I think the vital parts this power supply is missing is it hasn't got a current shunting valve entering a resistive thermionic flange :)
So you can control the voltage/current remotely.
It would be cool to see what happens if you purposefully trip the thermal cutoffs.
Dave, whilst apart, did you remove the 'eyelash'? If I saw that, its like seeing the line on a trinitron tube... you can't take you eye off it after...
Is that TLO74C at 28:05 toasted?
I did a build-up of my new computer and shot it with my Canon HF-S100. Only half of one video file worked and the rest were corrupt! I formatted the card and it recorded perfectly from then on. It has only done it once again, so I make a habit of formatting the card on the camera after every singe video import session. Never had a problem since! *touch wood*
Interesting they would use Nippon caps for the primary filters and cheap chinese for the rest which are more likely to fail. what brand is the fan?
I don't think there is an issue with the cap that is almost 2mm away from one of the big heat sinks. As you mentioned, the air flow is going by the cap, then through the heat sink, towards the back of the unit. This means that the air flowing by the cap is way cooler than if it were flowing the other way. Does that make sense?
You are so right there
You might misunderstand a part thermodynamics. The convection process does cool everything, but thermal radiation will still heat that cap up lots. Think of a big fire in a fireplace. Air is being sucked into the fire and the up the chimney (exactly like a heatsink), but you're still nice and warm through radiation, and exponentially so the closer you are. :)
Duan Coetsee
i don't think the radiation will be so high but yes it will warm up a little because of it, but capacitors can handle some heat.
smbrob Not only is the cap being heated, but that cap will warm up by itself through being the main filtering cap for that channel! Heated unevenly on one side! So no matter how you slice it, it's not good design. Just swap the relay with the cap. ;)
on my unit, one of the D1047 power transistor failed because of poor placement on the heatsink. i notice that the main power board was designed to hold four D1047 power transistor. so I added the twoo missing D1054 and the one that failed.
my question is, is it right to add the additional 2 Power transistor? i cannot see a power supply on the internet that has 4 D1047. what do you think? im talking on a power supply only one channel
I know it is 4 month but by any chance you tried 4 transistor?
yup.. i tried 4 transistors and it worked... and i analyzed the circuit and its way better to add those transistors to double up the power capacity. if you want i can send you some pictures of what i did.
Why do all the through hole components look like they haven't been soldered at all? Example, the opto isolators at 22:31.
ps: no software review for remote programming?
Does the camera have built in flash storage, or does it record to SD cards, if it does, then make sure you are using cards with decent performance, and write cycles and a life time warranty. (I personally use the sandisk extreme). Most cameras with built in flash will eventually start corrupting the files when the flash memory is nearing it's end of life (for SD cards you just RMA them, but for the built in stuff, you have to replace the camera)
i had a d1047 in an aplifier wich only had npn transistors,the ballast resistance make sense since they were no exteral resistors in that amp,i would have thought it because of sheer cheapness
Is that cap going to get that hot? That part of the heat sink should be pretty cool because it's getting fresh air, and the cap itself should also have some airflow over it.
Hey Dave can you do a video on converting a Desktop Computer ATX Power Supply into a Bench Power Supply please. I've seen a few on here but non are PRO and you're the BOSS!
i thought the same, maybe putting them in series, normally closed would have been better but there must be a reasson, because usually cheap chinese stuff dont do stuff that's unnecessary, like having an interrupt in the micro to handle that and then cut the power via code (bugs!!) when you can have them in series and just detect when the transformer gets disconnected, shut them down and wait. From a security point of view, the less code you need to protect something, the better and safer it is.
Is the yellow wire visible around 06:36 in as bad a condition as it looks ?
I wanted to ask at #439 why you turn it on before you take it apart. Tadaaa... there is the teardown. Thumbs up.
You Should Check the Temp by the capacitor by the heat sink :D
I think Intel's IPP library is some very high performance hand tuned libraries that are both compiler and hand optimized to take advantage of specific aspects of their processor lineup.
Although C is usually close enough to assembly unless your doing something special. Bare metal is obviously the route for maximum performance if you know what your doing otherwise we would all be using very high level languages and relying on their compilers to figure everything out.
hmm that cap right next to the heat sink. i suppose the fan blows outwards actually cooling the cap???
Isn't Atten just a lower grade Siglent? I've seen so many products that look identical.
but what about a SMPS followed with a linear regulator? that way you wouldnt need the relays, just drop the voltage a few volts above the regulator and that's it. not much of a difference but you get a lot better efficiency and a less noisy fan.
Not sure about the transient response thought.
IIRC (some of) the Aim-TTI supplies work like that and have a quite nice reputation.
Lack of sufficient solder for through hole and excess on the SMD parts. (The bottom side likely has a blob of solder)
You should put some hot glue on those big filter caps.
Very good video btw =)
That whole heatsink area, with fan on, was almost 70C so it's radiating quite a bit of heat on to the capacitor, which will heat up increasing ESR causing more internal heating. The whole situation is very bad for the lifetime of the capacitor because it's at an elevated temperature 90% of the time, which could mean it's going to drop under spec in 3000-5000 hours of use.
Not sure but usually a micro interface to an SD card is going to be using SPI mode while the cameras need the higher bandwidth of the full interface. The controller on the SD card might misbehave if you remove it at the wrong time with a higher data rate while the slower SPI mode may be less prone to glitching on removal.
Camera's need maximum performance and likely do not do much error handling as your PC does.
@32:21 - Is R5 supposed to be missing?
I think the next tear down Tuesday will be of a Canon HFG10 or you could start a new segment called smashup Wednesday ;0 Great video as always Dave.
Maybe the use the little relay on the serial board to switch between usb an rs232.
They just power the relay strait from the usb to switch the data lines.
incredible you can have a decent implementation of FAT for playing with a sdcard on an atmega but canon (and many others) cannot get it right on their processors....
I thought he was going to throw that power supply through my monitor at me for a second.
what does the ru on components mean
I think some Canon cameras has twin SD card slots or SD/CF combo so this kind of shit does not ruin your day.
I shoot photos as a hobby and always use the camera's second SD slot as backup - sloppy card writing ruined so many of my photos before it wasn't funny.
Short of that I think Dave could do with a dual camera solution.
radiant heat is not dependent on advection. sure, the cap will be cooled by the advection provided by the fan, but it would be a whole lot cooler if it weren't snuggled up against that 60°C heatsink. this is just one more indication of planned obsolescence
Great video, pity you lost the footage...
SMPS won't have the low noise a linear supply has. If you have something critical that is sensitive to noise you don't want a SMPS. Plus normally SMPS are used to power digital circuits which don't care as much and you can just use a PC power supply and get upto 1500W of "clean" power.
agreed, Lab power supply's are linear because they have cleaner power output.
Is it the camera at fault or just the memory card?
Why is opto-isolation needed?
If the fan blows air OUT of the case then the filtering capacitor for the channel 3. is not in such danger, but this is not good construction practice anyway.
unsafe disconnect of the cam + ms fat = crosslinked files - happens quite often.... Hint: check filesystem once unsafe removal from pc or power-outage happened
Plastic front is a bad thing? Isn't it safer than a conductive front panel?
How is a plastic front panel safer than a metal panel?
Dave, thought you were going to get yourself a nice thermal camera? ;)
Thumbs up!
I think that from a user's point of view having the micro provide an "over temperature - please wait" error message is better than simply shutting off. This would happen if the temp sensors were used to abruptly cut off the mains and the thing just goes dark. That's OK for a space heater but you don't want the user thinking that the power supply had died completely! My $0.02.
ROFLMAO. Nice rant about the failed clips, Dave. I HATE it when that happens. Nice review, too. But frankly, I'd put the money towards something that wouldn't frustrate me all day long. Thumbs up on the video.
@EEVblog Dave, everyone has lost files they've put lots of work into. Go play with Sagan, kiss the wife, walk around the block, and when you get back, blow the living crap out of a cheap DMM. It's the little pleasures in life that really matter. We can always watch it twice whenever you get it done so you get twice the Ad revenue. We all appreciate the pains you go though for us. Thank you Dave.
I've always see these Chinese manufacturers would use both Silpads and thermal grease together on the TO case, Wouldn't that be a bad idea since they'd increase the thermal resistance?
bacause noise. If all ground was in same net (series) noise in each point there be added.
Saweet!
wow. D5 at 28:30 wouldnt pass qc at my facility... thats some pretty bad work
Time to change your slogan - you (almost) always turn the stuff on before taking it apart :D
Nice video. 17:45 I think it means: Negative = Black!
Regards...
David, you know what. As a Chinese a do gradually love what have china dedicated to the electronic worlds through your Blog!
Chinese products = cheapest price + (possible) very unique design + (70%) not very good quality.
thumps up!!!
I thought the file was damaged on YT. It aborted 17 minutes in. Now I see the whole thing. I'll watch it twice anyway. Doing my small part for EEVblog.
Regarding the comment that they probably have programmed it in assembly: Why would you think that? I guess it's because of your "only 128 bytes of RAM" remark. However, modern C compilers (read as: Any C compiler which is newer than, let's say, 5 years or so) optimize better than we humans could write it in assembly, even with program space optimization only (-Os on GCC).