Out of spec resistors

[robrenz]

I am working on a decade resistor box project and  I bought a bunch of 1206 0.1% tolerance resistors.  I had to go to four of the main distributors and use 4 different well known top manufacturers to get things that were in stock.  These range from 0.1 ohm to 1M and I bought 500 of each size(.01 & 1.0 are 1%).  So I just got my new fluke bench meter that is freshly calibrated and I thought just for the fun of it I will test some to see what they look like.  I picked the 10K ones and started measuring them by using the kelvin probes poked thru the back side of the tape strip.  I got 17 out of 100 were out of spec on the high side.  I called the distributor  and in about an hour two techs from the manufacturer call me about the problem.  The ask me about my instrument and measuring technique and find nothing I am doing wrong.  Since I don't know how good the Fluke kelvin probes were I volunteered to get regular kelvin clips and measure them again.  Same out of spec readings.  By now all the other values have showed up and I measure a bunch of each value, all are in spec.   And two of these values were the same brand as the out of spec 10K resistors.  Per the manufacturers request I sent the 17 with my measurements for them to measure but it sure seems like these will be out.  I am somewhat new to electronics so Is this common? Do you have to inspect all your product when it comes from top name suppliers and manufacturers?

[qno]

this is very interesting.

I remember Dave having problems with his microcurrent probe and the 0.1% resistors.
Is something going on at the resistor manufacturer.
Or are they all sourced from the same factory?

[saturation]

It does happen.  Dave talked about his bout with it almost a year ago:

https://www.eevblog.com/forum/index.php?topic=1970.0

[robrenz]

I watched eevblog #133.  I was not going to name names, but since my supplier was Digikey and the brand was Bourns I think the names are now relevant. The two other Bourns values I got were from other suppliers and they checked out OK.  I made up a set of Kelvin clip leads using the Meuller gold kelvin clips. I also made a switch unit to reverse the polarity of the drive current.  This allows you to get rid of thermal EMF errors by taking a reading with one polarity and then reversing the current direction and taking a second reading then average the readings (at 10K this is overkill). This drops out the voltages generated by the Seebeck effect of all the connection points. I also set my Fluke 8846A up in math mode to average 40 readings. for each measurement direction.

[Conrad Hoffman]

With the fresh meter cal you should be in good shape. It's always nice to have a standard to use, just to be sure everything is still OK. I always keep a calibrated resistor around so I can sleep soundly at night. You don't need Kelvin probes to measure a 10K. They come into play at lower values, depending on your meter resolution and accuracy. I'll usually go for the Kelvin probes below 1 kohm for high precision stuff, and 100 ohms for 4 digit level stuff. Out of spec parts are rare today, but it does happen. It can be quite interesting to record all the measurements and make a bell curve. I've done that for thru hole parts and what's interesting is the curve is sometimes not bell shaped at all. It makes me think there was a selection process going on and part of the population was abducted!

[jackbob]

With 1% you pay the extra money for a reason, I would be a quite angry if I bought them and that many were off spec.  Resistors are weird I have had 5% carbon resistors that are right on the mark not even 1% off while the true 1% are even further off. If one or two are a little off then ok but 17 wow. If it was the low values that were off such as 1 ohm resistors or less I find them to be the most far off. I do mostly large power electronics where 5% is good but I am doing a project and using 1% resistors I will give it a look. And one more thing to keep in mind is the resistance can vary with temperature but under normal conditions shouldn't vary too much.

Jacob

[robrenz]

I used the Kelvin clips and current reversal because I was also measuring 1K, 100, 10, 1, and 0.1 Ohm resistors.  These are 25ppm TC and I did compensate for measuring at 25 deg.C instead of 23 deg.C.  For the 10K that were bad that is only 0.5 Ohm correction.
Because of availability I ended up with Bourns, Vishay-Dale, RCD, and Panasonic resistors.
The RCD were crazy accurate with most reading less than .01% off nominal.
Vishay-Dale and Panasonic were next with most using about half of the available tolerance.
Bourns was the worst, even the two values that were in spec. were all over the place and constantly skirting the high low limit.
One thing I learned is that having a good bench meter has its benefits.  I didn't think I needed one, I was just lusting after one for for a long time and finally gave in.

If you are using pick and place how would you measure all the resistors on a tape and still be able to use them?

[saturation]

Yes, noticed that too.  Kind of like this:

http://dangerousprototypes.com/2010/07/01/actual-values-of-10-tolerance-resistors/

With the fresh meter cal you should be in good shape. It's always nice to have a standard to use, just to be sure everything is still OK. I always keep a calibrated resistor around so I can sleep soundly at night. You don't need Kelvin probes to measure a 10K. They come into play at lower values, depending on your meter resolution and accuracy. I'll usually go for the Kelvin probes below 1 kohm for high precision stuff, and 100 ohms for 4 digit level stuff. Out of spec parts are rare today, but it does happen. It can be quite interesting to record all the measurements and make a bell curve. I've done that for thru hole parts and what's interesting is the curve is sometimes not bell shaped at all. It makes me think there was a selection process going on and part of the population was abducted!

[robrenz]

Since these metal film resistors are laser trimmed to spec (according to the two Bourns techs I talked to) I am not sure the same thing applies. from looking at all the manufacturers I have parts from, it looks like Bourns laser trim process is crap compared to the other manufacturers.  These were produced in Taiwan.

[saturation]

When I was reading about counterfeit items in the supply chain, I was dismayed to read that even component manufacturers can outsource manufacturing to subcontractors, or that authorized distributors may get name brand components from other distributors or agents, all vying for lowest price.  This has led surprisingly to counterfeit items penetrating the authorized distributors parts chain, albeit far less than outside it. 

http://distributioninsider.electronicproducts.com/articles/counterfeits-continue-flooding.php

"If the supplier knows in advance that the customer has a formal inspection process in place and will not only refuse to pay for, but basically destroy, counterfeit materials, it really removes the motivation that the supplier may have to try to pass off counterfeit goods. If the supplier cannot make money, why do it?" says Hamiter, who also chairs a "Counterfeit Electronic Components Avoidance" workshop series in conjunction with Custom Analytical Services Inc., a failure-analysis lab based in Salem, N.H.

Thus, my speculation, its very possible to have quality control problems with say, resistor tolerances, from some makes even if those parts aren't  considered counterfeit, if said company oursources their manufacturing.  Tolerance is one thing, but what if you need them also to conform to proper wattage, or parasitic inductance or capacitance?

I don't build manufacture in quantity, so I can afford to test every part I use, but if one were to do this for a high volume consumer item, it would lead to higher production cost unless some robot can do the testing.  Right now a common tactic is limited to testing batch samples.

Sadly, six sigma is supposed to avoid that.  With faults at 3ppm, one can project failure rates and its over all costs more consistently without testing every component used.

I used the Kelvin clips and current reversal because I was also measuring 1K, 100, 10, 1, and 0.1 Ohm resistors.  These are 25ppm TC and I did compensate for measuring at 25 deg.C instead of 23 deg.C.  For the 10K that were bad that is only 0.5 Ohm correction.
Because of availability I ended up with Bourns, Vishay-Dale, RCD, and Panasonic resistors.
The RCD were crazy accurate with most reading less than .01% off nominal.
Vishay-Dale and Panasonic were next with most using about half of the available tolerance.
Bourns was the worst, even the two values that were in spec. were all over the place and constantly skirting the high low limit.
One thing I learned is that having a good bench meter has its benefits.  I didn't think I needed one, I was just lusting after one for for a long time and finally gave in.

If you are using pick and place how would you measure all the resistors on a tape and still be able to use them?

[robrenz]

I sent 17 of the bad resistors with my measurements back to Digikey for Bourns to evaluate. When pressed for when I would get good replacement product, Digikey suggested I return them all and pick another manufacturer.  The unspoken but implied message was, if you need real accurate smd resistors pick a company that actually produces them.

[saturation]

Yes, such tactics to save money is self defeating.  Components are all about quality and accuracy to its spec sheet.  I would return the whole order, avoid Bourns in the near future, unless the 2 other better makers are out of stock. 

I sent 17 of the bad resistors with my measurements back to Digikey for Bourns to evaluate. When pressed for when I would get good replacement product, Digikey suggested I return them all and pick another manufacturer.  The unspoken but implied message was, if you need real accurate smd resistors pick a company that actually produces them.

[robrenz]

I did return them all, and Digikey was very courteous and professional.  From my measurements RCD used the least amount of the tolerance band and were very competitive on price so I used them to replace the Bourns product.

[westfw]

The unspoken but implied message was, if you need real accurate smd resistors pick a company that actually produces them.

Or "if our stock of these is compromised (ANYWHERE in our supply chain), you'd be getting replacements that might be compromised as well."