EEVblog #133 - Dodgy Digikey Components

[tyblu]

EEVblog #133 - Dodgy Digikey Components

This is a fantastic blog -- I didn't know you could do 4-term measurements with DMMs. Too bad you got burned! Must have been a 'Friday' batch from Burns!

Seems like D. Jones let out his frustration by following a crazy EE's train of thought:
1 - gah, 3.5% off!
2 - this resistor is 3.5% off!
3 - but wait, is it really? (measure 2X)
4 - ... is it really really? (measure 3X)
5 - ... you've got to be shitting me. (solder up 4term measurement setup and measure 4X)
6 - hey, i didn't have to solder that up at all (measure 5X)
7 - now that the camera is off ... (measure 6X, 7X)

Quite enjoyable!

[EEVblog]

 

I had actually already shot another blog about a dodgy Maxim component just the day before, so that's twice in almost as many days for me!
The Digikey one trumped it though as it's more pertinent.

Dave.

[mikeselectricstuff]

Dave - why are you ranting at Digikey when it's pretty obvious it's a manufacturer problem? Do you expect Digikey to test every reel that passes through their warehouse?
Have you contacted DK and/or Bourns yet?

[EEVblog]

I ranted at both of them, not that it was much of a rant anyway. I was mostly telling the story of what happened.
It is not entirely obvious who is at fault, it could very well be Digikey for all I know.
In any case Digikey are in the best position to investigate it, that's why I'm asking them what's going on.

No, of course I do not expect Digikey to test every reel, that is silly.
Yes, I have contacted Digikey about it.

Dave.

[Alex]

Oh well, it happened now. What about the units that were sent out? Any plans for those?

I remember in the past there were issues with fake components, especially power transistors (eg 2n3055). The die inside was half the size. The part would work fine when you test it but then blow up when installed. Are such parts still in circulation?

[Strube09]

Dave,

Just ignore the criticism of the video comments. Facts are you got bad components and it sucks. You never said it was anyone companies fault, but Digikey is where you bought them so they should fix it simple as that.

Thanks for telling the story and still keeping some education in there showing how the 4-wire works.

I would like to point out that I think this is the first time we have ever seen you use a bench meter... Maybe there are some pluses to a bench meter

Strube

[JohnS_AZ]

This video serves well as a 'real life' object lesson to all of the EE students who watch Dave's videos. I remember back in my Motorola days when we built about $125,000 worth of VME boards using diodes that had reversed markings from the manufacturer. It took 4 hours to build them, and a week and a half to rework them all.

Shit happens.

(BTW... VERY cool tip on the dmm/4-wire technique)

[TheWelly888]

I echo the others about the 4 wire resistance measurement - it made the concept so much clearer for me! Thanks, Dave!

It has just struck me, through hole resistors usually have a colour band denoting tolerance, SMD resistors certainly have no space for any kind of tolerance indication so no one can be sure exactly what has been despatched either from Bournes or Digikey!

[Fraser]

I found the blog very interesting and do feel sympathy for Dave on this one. I have come to respect the high quality that is now available in mass produced SMT and do not check the tolerance on every component I use but then mine are usually alpha prototypes and issues can be diagnosed during testing. What I find interesting though, are the comments on the YouTube Blog that say Dave should have tested every single uCurrent for performance before sale rather than batch testing. I find that view hilarious because Dave has produced these units at a wallet friendly price as a favour to the community... I really don't think he needs the money and feel sure he certainly doesn't need the grief that this matter has caused him. As I have stated in my Blog comment, time is money and if Dave had tested each uCurrent for performance I would expect the cost to rise in order to cover his time. Sadly that is why some relatively simple professional electronics cost significant sums of money, you are often paying for the testing that was done for specification compliance.

The uCurrent appears good value for money and I for one wish to thank Dave for making it available and also for raising this matter of component value deviation from specification. I found the 4 wire resistance testing very interesting as I own several bench meters that are equipped with that facility and I now fully appreciate it's usefulness. 

[EEVblog]

What I find interesting though, are the comments on the YouTube Blog that say Dave should have tested every single uCurrent for performance before sale rather than batch testing. I find that view hilarious because Dave has produced these units at a wallet friendly price as a favour to the community... I really don't think he needs the money and feel sure he certainly doesn't need the grief that this matter has caused him.

Correct on all counts. I never really wanted to do this extra batch, but a few people kept asking and I had some spare boards so I relented.
I originally only sold them to support the silicon chip article, and I originally did kits too which were a pain in the arse.
Not that any of that is an excuse for shipping some units out of spec.

If I ever got into a selling stuff like this full time I'd have some more time and enthusiasm for sustaining the products.
As it is, I generally come out with a new project and sell an associated kit because it is fun.
I actually think it's a bit unkind to produce a project and show it off and then when people ask "where can I buy it?" I have to say "sorry, I can't be bothered"
It was like that with the uCalc project.

I don't make much money out of this stuff. If I counted my time as money, I would make even less.

Dave.

[paulpthcom]

Dave,

Somewhat unrelated to the other comments, but how's that Keithley 225 you used?  Do you find that you use it often for things other then testing out the uCurrents?

[EEVblog]

Somewhat unrelated to the other comments, but how's that Keithley 225 you used?  Do you find that you use it often for things other then testing out the uCurrents?

Yes, it can be quite handy. LED testing is one example, so you know how bright a led will be for a given current, just dial it up.

I also have another Keithley current source that goes down to sub picoamps, not quite as useful, but nice!

Dave.

[Bored@Work]

Somewhat unrelated to the other comments, but how's that Keithley 225 you used?  Do you find that you use it often for things other then testing out the uCurrents?

If you want to build your own, simpler but useful version of a current source (partly doubles as load), see this article http://www.edn-europe.com/programmablecurrentsourcerequiresnopowersupply+article+2251+Europe.html
(the "no powersupply" in the article is wrong).

Just an LM317, a handful of resistors, a few BCD switches you can get off eBay http://shop.ebay.com/?_nkw=BCD+switch&_sacat=See-All-Categories plus enclosure, heatsink, terminals (and a power supply) gives you a nice retro-style current source.

[PeterG]

BoredAtWork, nice find, looks doable in an hour or so with parts in hand.

[Bored@Work]

BoredAtWork, nice find, looks doable in an hour or so with parts in hand.

It took me a little longer, including drilling of the enclosure and  front plate "art" work.

[paulpthcom]

I received my uCurrent today and the resistor is definitely off in the 10.2/10.3 range.  I didn't bother with the 4 wire test, since well doesn't seem like there's much of a point.

[mikeselectricstuff]

Of course being on the high side, you can fix it with a parallel trim resistor to bring it down to the correct value

[PeterG]

I have to say, its things like this that make me happy i have element 14 just 3km away.

Regards

[allanw]

Anyone else notice the high-pitched sound going on through the whole video?

[tyblu]

Not I, allan. Don't think the dog did either.

[ziq8tsi]

Of course being on the high side, you can fix it with a parallel trim resistor to bring it down to the correct value

You can not make a 0.1% tolerance resistor (with the required temperature coefficient and long term stability, etc) by paralleling resistors of higher or unknown tolerance.

Anyway, given that the dodgy batch of 10R0s appear to have values all over the place, measuring and calculating the required trim value for each unit would surely be more work than unsoldering the existing resistor and fitting one of the replacement 10R0s that digikey will doubtless provide.

[austyp]

I ordered the same part from Digikey to build my own uCurrent, P/N: CRT0805-BY-10R0ELFCD-ND, date on my packaging is 16-Nov-2010.  I measured a couple of resistors in my batch using my Agilent U1253A and they were 10.12 and 10.13, so I suspect that the parts I have are from the same reel.

I haven't contacted Digikey, but will be interested to know the outcome of Dave's investigation.

For the exercise I created my own schematic and pcb for the uCurrent.  If it's not some kind of infringement on copyright, I'm happy to give people the Eagle schematic and board files if they're interested in making their own.  I used BatchPCB, think each board cost about US$15 from them, parts quite a bit more once freight to Oz factored in.  This was my first board layout, and there's some things I'd do differently next time around.  I haven't even fully verified the operation of the assembled design but someone may find it useful.

Regards
Austin

[mikeselectricstuff]

Of course being on the high side, you can fix it with a parallel trim resistor to bring it down to the correct value

You can not make a 0.1% tolerance resistor (with the required temperature coefficient and long term stability, etc) by paralleling resistors of higher or unknown tolerance.

Yes you can, provided the trim is large relative to the nominal value. The trim required in this case is only a few percent, so any error and TC of, say a 1% trimming resistor will be less than 1% of  a few percent , i.e. insignificant.

Anyway, given that the dodgy batch of 10R0s appear to have values all over the place, measuring and calculating the required trim value for each unit would surely be more work than unsoldering the existing resistor and fitting one of the replacement 10R0s that digikey will doubtless provide.

You can calculate the require trim resistor from the current reading error. Or use a resistance box. The trim R will be several K so contact resistance won;t be a problem.
 

[alm]

You can not make a 0.1% tolerance resistor (with the required temperature coefficient and long term stability, etc) by paralleling resistors of higher or unknown tolerance.

Yes you can, provided the trim is large relative to the nominal value. The trim required in this case is only a few percent, so any error and TC of, say a 1% trimming resistor will be less than 1% of  a few percent , i.e. insignificant.

This is correct, but the issue is that the unknown tolerance/TC is the '10R0 0.1%' resistor, which might be a crap 5% thick film resistor for all we know.

[mikeselectricstuff]

Seems unlikely, given the identical markings and (ISTR) 3 digit mantissa.

[alm]

Some were off by more than 1% IIRC, so I wouldn't put much trust in the markings. They aren't 10ohm, 0.1%, that's fairly certain. No one (except possibly Bourns) knows what they really are, so I would assume the worst, and not simply 10.13ohm 0.1% (or whatever the measured value was). Tempco can be anything, they could even be high TC resistors designed for TC compensation (if they make those in these sizes and values).

[Icchan]

This just shows how important testing can be even in small batches of components/products. But what I really liked in this episode was that this one had a real problem and to solve that, Dave showed many basic techniques for finding, measuring and solving the problem. An inportant lesson for any  EE out there

[E.L.Brown]

This is exactly why i check every single part with a tolerance rating below 1% when it arrives at my place 

If you want 0.1% or better , you want it for a reason . . .

Trust me - the extra 10 minutes are worth it !

[Psi]

This is exactly why i check every single part with a tolerance rating below 1% when it arrives at my place 

So you're trying to avoid the following...

Unbelievable that this little piece of junk could be such a big problem.

[EEVblog]

I ordered the same part from Digikey to build my own uCurrent, P/N: CRT0805-BY-10R0ELFCD-ND, date on my packaging is 16-Nov-2010.  I measured a couple of resistors in my batch using my Agilent U1253A and they were 10.12 and 10.13, so I suspect that the parts I have are from the same reel.

I haven't contacted Digikey, but will be interested to know the outcome of Dave's investigation.

Thanks for the confirmation. Yes, likely the exact same reel.

I now have the replacements and will measure when I get home.
They are the exact same type and part number, but Digikey have assured me they have tested them and they are within spec.

Bourns have also requested one of my uCurrent's so they can analyze further in their Taiwan facility, and the left-over parts I have too.
How that is going to help them is beyond me, they have a whole bloody reel of parts, and my uCurrent is irrelevant!

Dave.

[Jimmy]

How that is going to help them is beyond me

Maybe they will copy it and put it on the market

[Fraser]

Bourns have also requested one of my uCurrent's so they can analyze further in their Taiwan facility, and the left-over parts I have too.
How that is going to help them is beyond me, they have a whole bloody reel of parts, and my uCurrent is irrelevant!

Dave.

May be they want to fully understand the implications of their 'out of tolerance' component on your design and the work required to correct it.... Financial Compensation assessment  ? Canon had a major reliability problem with some Sony CCD chips that they used. They repaired my Camcorder when it was over 5 years old....Sony had to pick up all associated costs for all such repairs including postage and labour.

[allanw]

How that is going to help them is beyond me

Maybe they will copy it and put it on the market

Haha, that's easy enough to do already considering Dave has published full schematics and layout and even described the entire design process in a video.

[mikeselectricstuff]

Bourns have also requested one of my uCurrent's so they can analyze further in their Taiwan facility, and the left-over parts I have too.
How that is going to help them is beyond me, they have a whole bloody reel of parts, and my uCurrent is irrelevant!

Dave.

My guess is that it's standard procedure to look at the customer's design to eliminate the possibility that the  application is not causing problems running parts out of spec. Current-sensing R's can be subject to all kinds of abuse, and users won't necessarily understand the limitations of the parts they are using.
 They probably haven't looked at the ucurrent design in any detail.

As regards testing, for a test instrument like this, I'd suggest that the minimum production test really should check zero and full-scale accuracy on all 3 ranges, as apart from the unlikely chance of bad parts, more common things like bad solder joints, PCB faults and wrong components can easily happen.
A full functional test pretty much eliminates the need for testing parts, as any part issues will surface.

[EEVblog]

My guess is that it's standard procedure to look at the customer's design to eliminate the possibility that the  application is not causing problems running parts out of spec. Current-sensing R's can be subject to all kinds of abuse, and users won't necessarily understand the limitations of the parts they are using.

No need for that when Digikey have confirmed they have found reels that are out of spec.

As regards testing, for a test instrument like this, I'd suggest that the minimum production test really should check zero and full-scale accuracy on all 3 ranges, as apart from the unlikely chance of bad parts, more common things like bad solder joints, PCB faults and wrong components can easily happen.
A full functional test pretty much eliminates the need for testing parts, as any part issues will surface.

That's exactly what I do.

Dave.

[Neilm]

My guess is that it's standard procedure to look at the customer's design to eliminate the possibility that the  application is not causing problems running parts out of spec. Current-sensing R's can be subject to all kinds of abuse, and users won't necessarily understand the limitations of the parts they are using.

They will also check that the customers process has not caused any of the problems - for instance the soldering might have been done at too high a temperature, or been held at that temperature too long. I once had a problem with an IR transever - they would work in the factory but fail prematurely. After sending off the boards from both failed units and boards just assembled, the supply came back with the conclusion that the unit was being damaged by the assembly process, but for some reason some were fine, other were not. It turned out that the failed boards had been reworked to fix minor faults. When that happened a soldering iron was held above the transever - it focused the energy onto the detector and damaged it. The fix - ensure they were covered up prior to any rework.

As regards testing, for a test instrument like this, I'd suggest that the minimum production test really should check zero and full-scale accuracy on all 3 ranges, as apart from the unlikely chance of bad parts, more common things like bad solder joints, PCB faults and wrong components can easily happen.
A full functional test pretty much eliminates the need for testing parts, as any part issues will surface.

Assuming that it doesn't have a wierd error curve that puts it out mid range.

Neil

[mikeselectricstuff]

Assuming that it doesn't have a wierd error curve that puts it out mid range.
Neil

True, but assuming the design itself works OK, a manufacturing fault causing mid-point errors is pretty unlikely. Except possibly  a stability issue, e.g. oscillating under certain conditions. 

[Bored@Work]

When you give them the left-overs they have the evidence and you have nothing any more. And then suddenly the evidence disappears.

[Neilm]

When you give them the left-overs they have the evidence and you have nothing any more. And then suddenly the evidence disappears.

When I sent off the IR transceivers I got back a whole series of X-ray pictures showing the receiving circuit of the various devices. It was that which allowed us to work out what had happened – instances of boards that had not been reworked were substantially less than non-reworked.

Neil

[Mechatrommer]

sorry for bumping this. but i got reminded when eetimes emailed me. here the link to bulk resistivity test (but i'm not sure what they mean by bulk), i thought how to measure massive amount of resistors, but turned out not be like that. but might useful to others (might not), check the linked pdf, adios:

http://www.eetimes.com/design/test-and-measurement/4212791/Measuring-the-resistivity-of-bulk-materials?cid=NL_TestMeasurement&Ecosystem=test-and-measurement

[MBY]

Sorry to bump this thread once more. But I tried to watch episode 133 again but something has happened.  The #132 and #134 works, but it seems that the file for episode #133 is missing?

[EEVblog]

Wow, so it is, no idea why.
Fixed.

Dave.

[MBY]

Great! Good episodes is always worth watching at least twice!

[pascal_sweden]

Maybe that company has to change their name into Dodgykey