[rant]Questionable Quality of eBay Electronic Parts

Go To Last Post
23 posts / 0 new
Author
Message
#1
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I have eBought a bunch of THT resistors recently.

eBay seller wrote:
Tolerance: 1%, power: 1/4W, metal film, we ship from HK, welcome you.

Ok, so I am picking two random ones of same nominal value, connect in series, apply a 5xAA battery pack (~7.7V) and DCV measure both voltages of the divider with my dmm on a 4V range (40Mohm input impedance).

I inspected several batches from 1k-100k range, no touching etc.
For some pairs the ratio of above voltages is ~1.06

I did not measure the nominal values of those resistors yet, I just wanted to check tolerances (all have "Chinese Brown" tolerance stripe).

Concluding: onepercentmyass. The values are all over the shop, not even 2%.

No RSTDISBL, no fun!

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

A bit like the infineon igbts i bought. They did 'look' like igbts but they didn't measure the same as the originals.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

If batches from a European factory fail the quality control, they get destroyed. I presume that Chinese factories will have the same policy.

All the same, it is quite possible for products to leave by the side door. A 5% resistor will work like a 1% resistor. Ebay is designed just for this purpose. The punter is very unlikely to measure the value. Even if she does notice, she was happy with the attractive price.

I bought some 20ppm 32.768kHz crystals from Ebay. They are actually 40ppm when you measure them.
I bought some 10ppm 32.768kHz crystals from Farnell.
Strangely enough, they measure 10ppm. But they cost £0.13 inc VAT compared to the Ebay ones that cost £0.10

The practical difference between Ebay and Farnell is that Ebay often comes with "Free Shipping @ 10-40 days".
Farnell needs a minimum order for "Free Shipping @ 1 day". And costs a little more.

As a general rule, you get what you pay for.
If a MAX7219CNG+ costs £9.02 from Farnell, how come you get the same chip for £0.77 from China?

David.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Your calculations are flawed...taking two resistors in series you could have, worst case, +1% on one and -1% on the other....Do the maths and calculate the error. Add to that the error of your measuring instrument... Probably another 1% at best. Then add that error on.Your resistors probably are within the stated 1%...

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I think that Brutte has used a reliable method.

Whereas the absolute accuracy of hobbyist instruments is probably not very good, the short term stability will be fine.

Even a £5 DMM will be able to compare resistors. Brutte found that his samples did not even lie within a 2% window. e.g. +1% -1%

Likewise, I can measure the frequency of a batch of crystals.

Electronic components will work perfectly. If the accuracy is out of spec, they are not the genuine article. They are still resistors or still tuning-fork crystals. However a 5% resistor or a 40ppm crystal is not what is being sold.

David.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

chartman wrote:
taking two resistors in series you could have, worst case, +1% on one and -1% on the other....

Ok, so in the case these two were equally distant from nominal value, the ratio could have been 1.01/0.99~1.02, worst case.
chartman wrote:
Add to that the error of your measuring instrument... Probably another 1% at best.

I didn't measure the nominal value yet, the accuracy of the meter is irrelevant. The quantization error of a 4000 count meter at full range is half of around 1/4000 which is about 125ppm. Also the nonlinearity and offset does not matter as I measure two similar values each time. Also the dvm impedance does not matter as it is 40Mohm and sampling current is ~same at both measurements and the resistor is of kohm range (4 orders of magnitude smaller).
I can safely assume a 250ppm worst case voltage error (relative value), but that is unlikely.

So, one cannot get the error of that ratio greater than (1+0.000250)/(1-0.000250) with such dmm (~1.0005) measuring at full range.

You are suggesting that the fraction is indeed 1.02 but the dvm injects error which results in 1.06 fraction (error of around ~1.039)? As you can see your suggestion is of two orders of magnitude (about 78 times) beyond the reality.

No, I didn't lick resistors and I didn't spark and it is not the Earth's magnetic field.
It is just eBay.

No RSTDISBL, no fun!

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

There are thousands of excellent honest Ebay vendors.

Most commercial companies will sell genuine full-spec parts.

However, the price of some goods appears too good to be true. Of course, surplus inventories and bankrupt stock can be both genuine and cheap.

How much did you pay for your "Tolerance: 1%, power: 1/4W, metal film" resistors?

It is perfectly reasonable to just use the Resistance scale on your DMM. You need to check the absolute accuracy with known precision resistors. But for comparison purposes, you just need a short term high and low value.

After all, you are only measuring 1%. 10000ppm is hardly laboratory precision.

David.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Quote:
How much did you pay for your "Tolerance: 1%, power: 1/4W, metal film" resistors?

$12 for (50*50 values). Including shipping. Arrived within 5 business days!

Quote:
There are thousands of excellent honest Ebay vendors.

Feedback 99.0%, ~30k feedbacks, on eBay from 1998 etc. No rants about the quality of passives.

No RSTDISBL, no fun!

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Brutte,

I agree that your method is 100% valid. The absolute accuracy of your DMM isn't a factor, its short term stability is. Short term stability in this case would be revealed by a drifting reading, which I assume you don't see.

To further prove your method is correct, repeat the experiment with known quality 1% resistors of approximately the same value.

david.prentice stated:

"If batches from a European factory fail the quality control, they get destroyed. I presume that Chinese factories will have the same policy."

Laugh it up, David ! Nothing gets thrown out in Chinese factories, it all gets shipped.

Besides, resistors are not manufactured to a specific 1%, or even 5%, value. They are manufactured in batches from which a range of values is expected. The resistors are then individually automatically tested and binned into the standard 1% and 5% values. The only ones that get rejected are those which are so far outside the expected range, they are suspected of a fatal manufacturing flaw.

At least tht's the way it's been done in civilized industrial nations for decades. In China all bets are off ( IMHO ) where anything goes (literally: goes out the door).

What is amazing with all of these CCC items sold from China on Ebay is that they arrive in 5 days with free shipping.

I had a similar experience with a bogus computer product I ordered from a major on-line distributor last week. I got the "Shipment Tracking" information link in an email from the US distributor about a day after I ordered the item, showing that it left some unpronouncable city in China a few hours prior. Expected delivery was stated as "5 to 15 business days." I expected it to arrive sometime during the Spring Thaw. Nope, less than 5 calendar days later it appeared in my mailbox bearing a return address in California. All for free!!! How do they do that?

BTW, the item was mis-described on the US distributor's web site. A key feature was not as stated. The processing of the returned item will probably cost the distributor 10x of the selling price.

Ahhhh, the enduring illusion of low cost quality, how we cling to its allure!

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Chuck-Rowst wrote:
its short term stability is.

How a short term stability can influence two similar voltage measurements that are taken 1s one after another??
I get 3.632V with one measurement and 3.849V with the other. I reverse the leads and get -3.632V and -3.848V
No matter how I connect those the fraction is around ~1.06
Are you suggesting that in between those 1s measurements the gain of the dmm changed by 4% to get 1.06 out of 1.02 ratio??

Quote:
Short term stability in this case would be revealed by a drifting reading, which I assume you don't see.

I can see the drifting reading because my 5xAA pack gets discharged during measurements. The readout drifts perhaps 1mV per hour, nothing even close to the error I get.

And, I do not have any reputable other 1% resistors.
If you do then take a ~kohm range batch, pick bounding pair with ohm meter and then apply the voltage of below 2x the meter full scale at some range (<8V with 3999 count, <4V with 1999 count etc) to the divider. Neither current nor voltage nor even the nominal value influences the fraction when you measure tolerances of resistors.

I do not even have the ohm-meter that could measure the nominal values but nominal value is a different subject that I didn't investigate (and I do not have plans for that). Even if I did use the 0.0001% error ohm meter then the only additional information I could get would be that the resistors are not even within +-3% from the nominal value. But there is no way these can be tighter than the 3% even when you use nomatterhowprecise ohm-meter.

I suspect these are "restriped" 5% resistors.

No RSTDISBL, no fun!

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Brutte,

You confirm what I am saying. Your measurements are NOT drifting, at least not fast enough to affect the 1% test you are performing. I reiterate: I believe your test method is 100% valid.

So bottomline, what-if question:

If this order had gone well and all of the delivered resistors were perfect 1% units, how much money would you have saved over buying them from a main-line distributor like DigiKey, Mouser, Arrow etc?

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Quote:
Your measurements are NOT drifting, at least not fast enough to affect the 1% test you are performing.

OMG, I do not know what kind of dvm you suspected me to use but even the el-cheapo +-1999 count $3 ones do not have the absolute error anywhere near the missing 4%. These usually have 0.5% + 3 counts total error on VDC range and are more than adequate for investigating 0.1% resistors tolerance tests. For 1% resistors tolerance tests I did such el-cheapo is an overkill.

Quote:
how much money would you have saved

That is an interesting question.
But how to compare that HK stuff with Digikey??

I bought 50x50 values of "metal film" 1/4W THT resistors.
I assume I could buy 25*100 or 100*25 sets at ~similar price.

Anyway, lets pick Digikey.
E96 range, one decade, 480 total, $29

You cannot buy 1% THT resistors at Digikey at quantities < 5k (that is 100 times more than I require). At 5k the price is ~$0.01 which is twice as high as the "restriped" ones from eBay.

The problem with those resistors is that it is hard to buy a metal film 1/4W 5% resistors so I think I am going to check that as well. You can buy carbon film ones that are 5% but these have high temperature coefficients (metal film are below 100ppms/K, carbon film are around 400ppms/K).

Carbon film set of 72*5 = 360 for $17

The Digikey shipping cost is prohibitive (at least for me).
y1jwv

No RSTDISBL, no fun!

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

[snarky comment alert]

You get what you pay for!

[/snarky comment alert]

Jim

 

Until Black Lives Matter, we do not have "All Lives Matter"!

 

 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Farnell have a 3100 resistor kit http://uk.farnell.com/multicomp/mf0w4ffe006kil/resistor-kit-0-25w-1-e6/dp/9342362 for £80.30

This is about $0.03 per resistor compared to your $0.0048 per resistor.

Somehow, I reckon you should pay somewhere in between these two prices.

Mind you, why should you worry about 1% in the E6 or E12 values?

Surely you would be ok with 5% or 10% tolerance.
Then buy a handful of precision resistors.

David.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Quote:
Farnell have a 3100 resistor kit

Thanks, this is a nice kit, (TCR<50ppm/K) but I am afraid the price of the kit + shipping would have ruined my resistor budget.

I am going to check some other alternatives (CCCP resources).

Now it is time to estimate temperature coefficient (TCR) of those "Chinese metal film" resistors (0-70)degC.
If these are beyond +-100ppm/K temp coefficient then they are going back home.

Quote:
Mind you, why should you worry about 1% in the E6 or E12 values?

It is a selection of E24.
Indeed, more important is the TCR. I need those for strain gauges conditioning circuit and I have to set same (~800) gain for active gauge and for compensation gauge (so that the influence of temperature of both gauges nulled out).

No RSTDISBL, no fun!

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

It will be interesting to hear exactly how 'accurate' your cheapo resistors are.

99% of electronic components require little precision. So I would happily use them if within say 5%. If they are worse than 5% I would bin them. i.e. they are duff manufacture.

Then buy the very expensive precision resistors for the strain gauge.

Looking at other Ebay vendors. e.g.

http://stores.ebay.com/CAPPRO/Assortment-Kit-/_i.html?LH_BIN=1&_fsub=1706584010&_sid=1012861370&_trksid=p4634.c0.m322

Some kits are advertised as 1% and others are conveniently opaque in their description.

Chuck has probably hit the nail on the head. Chinese factories may not destroy their out-of-spec products. They either go to Ebay or to assembly houses.

David.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Quote:
It will be interesting to hear exactly how 'accurate' your cheapo resistors are.

I do not own an ohm-meter that would allow measuring absolute accuracy within 1% of the tolerance value. And, even if I had such stuff, that would require at least the testing rig as in EEVBlog #215/216 to plot whole 2k5 batch.

But, I have found a "batch" of 5 "decent" resistors in my inventory. I know, 5 is not enough to infer any statistical properties, here they are:

1/4W, tinned copper leads ~0.6mm, the body itself is magnetic??

The stripes are:
violet/brown/green/gold/brown
Which means these are 71.5 ohms, 1% resistors.

I have measured all pairs, calculated 20 fractions (I know, it would have been enough to measure 10).

The biggest fraction I got was 1.0059 the maximal measurement error that could get involved was same as before, that is +-0.0005
As you can see that is an order of magnitude tighter tolerance than of those from the "eBay score".
I understand that this is only a set of 5 resistors so in theory it could happen a volume has wider tolerances...
OTOH these are picked random and something tells me that the problem lies not in the measurement method and quantity but in eBay (and me).

I have also measured/estimated the temperature coefficient of "eBay score" (0-50degC) and it is bigger than 280ppm/K somewhere in this temperature range!

The "decent" ones have the TCR higher than 27ppm/K but I am not able to measure that exactly because TCR is so small that the quantization error of dvm results in a bigger error than the TCR induced fraction change itself.

That eBay score was a total failure.

Resistors, pack yourself, you are returning home.
jwrga

No RSTDISBL, no fun!

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Brutte,

Be careful when measuring low value resistors like this with a steady current as you are doing. The heat gain in the resistor itself ( you know, I-squared R stuff )will heat the resistor above the ambient temperature and make your tempco look worse than it is.

Precision resistance measurements use a pulsed current or voltage and on-the-fly measurement scheme to avoid this self-heating error.

This would be quite an undetaking in your situation. So, I would recommend you stick with higher resistor values and use a lower voltage. Calculate the nominal wattage created in the resistor-under-test and use a test voltage which keeps this nominal wattage at say 1/20 th or 1/50 th of the resistor's wattage rating. And apply the voltage and make the reading as quickly as practical.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Quote:
Be careful when measuring low value resistors like this with a steady current as you are doing.

I did try to investigate a "decent" batch with 0.22W dissipated power at first but then I switched to a 400mV range (dissipating 2.2mW) - no measurable fraction change as it was still beyond the capabilities of the method I used so I didn't bother playing with these five fellows any longer.
The "eBay score" were dissipating only <1.3mW (12k resistors) under test and after heating/cooling a pair no more than 0:50degC/50:0degC the fraction changed by over 2.8%!

No RSTDISBL, no fun!

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Was there a permanent change in the room temperature measurement caused by the tempeature cycling?

Taking your numbers at face value:
2.8% = .028000 = 28,000 ppm
.028000/50 = .000560 = 560 ppm/degC

Agree?

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Quote:
Was there a permanent change in the room temperature measurement caused by the tempeature cycling?

I do not really understand what you meant by that.
Quote:
Taking your numbers at face value:

No, of course not.
Step by step, bit by bit, here is what I was doing:
    1. Room temperature ~25degC (irrelevant) 2. Pick two random resistors Ra and Rb of same nominal value
    3. Connect in series
    4. Apply almost 8V (irrelevant)
    5. Ra in >0degc (I used a plastic PE bag with water and ice)
    6. Rb in <50degc (I used a plastic PE bag with water and K thermocouple)
    7. Measure the fraction to get F1.
    8. Ra in <50degC ...
    9. Rb in >0degC
    10. Measure the fraction to get F2.
    11. Calculate dR_during_le50K_changed_twice=1-F1/F2
    12. Divide that by dT (2*50K = 100K)
    13. Post on avrFreaks that : "(..)temperature coefficient of "eBay score" (0-50degC) and it is bigger than 280ppm/K somewhere in this temperature range! "

Quote:
Agree?

No.

No RSTDISBL, no fun!

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

What I mean is do these marginally performing resistors change permanently if they are heated, cooled and returned to room temperature?

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

No, these work just like PTCs.
You heat them and the resistance goes up, when cooled resistance goes down (I am not sure how linear that is but the relation is about proportional, no surprises within 0-50degC).
When these get back to room temperature, give about the same fraction as they had before the experiment.

No RSTDISBL, no fun!