Author Topic: Warning- LM317K duds from Digikey  (Read 9479 times)

0 Members and 1 Guest are viewing this topic.

Offline floobydust

  • Super Contributor
  • ***
  • Posts: 7744
  • Country: ca
Re: Warning- LM317K duds from Digikey
« Reply #75 on: January 17, 2025, 08:08:18 pm »
So these are surplus IC dies? Scrounged and then amateur packaged for great profit I guess. But gotta save micropennies lol.
Many tech papers on fusing limits for IC bonding wires. Aluminum ain't like copper or gold.
 

Offline wraper

  • Supporter
  • ****
  • Posts: 18010
  • Country: lv
Re: Warning- LM317K duds from Digikey
« Reply #76 on: January 17, 2025, 08:17:25 pm »
So these are surplus IC dies? Scrounged and then amateur packaged for great profit I guess. But gotta save micropennies lol.
Many tech papers on fusing limits for IC bonding wires. Aluminum ain't like copper or gold.
All 3 EVVO ICs have the same die. Other pictures are from ICs from different manufacturers.
 

Offline floobydust

  • Super Contributor
  • ***
  • Posts: 7744
  • Country: ca
Re: Warning- LM317K duds from Digikey
« Reply #77 on: January 17, 2025, 08:29:08 pm »
That makes more sense. I was trying to compare bonding wire thickness between them, then saw the heat spreader missing and gave up.
I thought the wire fusing is the problem with the EVVO parts.

What pisses me off is search engines are returning results from Digi-Key to EVVO semiconductor parts.
Searched for "2SC1623" and it's a top hit. UGH. WOULD NOT BUY based on OP's experience.
 

Offline David AuroraTopic starter

  • Frequent Contributor
  • **
  • Posts: 489
  • Country: au
Re: Warning- LM317K duds from Digikey
« Reply #78 on: January 18, 2025, 11:17:30 am »
So these are surplus IC dies? Scrounged and then amateur packaged for great profit I guess. But gotta save micropennies lol.
Many tech papers on fusing limits for IC bonding wires. Aluminum ain't like copper or gold.

I don't know for sure, but I don't get the vibe they're surplus things like that. I've been chatting a bit with the manufacturer and they appear to be taking it seriously and wanting to gather info. I don't think a company just scavenging would bother with even replying to the first email I sent, but these guys are discussing it with me even though they're aware I only bought three of the things and therefore I'm clearly not an important customer

It's funny though, before this I've never given a shit about the guts of chips beyond the ideas of "Wow, it's cool that a chip can do this" or "The way they make dies is super interesting", but now I keep wanting a better microscope to look with more detail and I want to know everything about these bond wires  :-DD
 

Offline magic

  • Super Contributor
  • ***
  • Posts: 7517
  • Country: pl
Re: Warning- LM317K duds from Digikey
« Reply #79 on: January 18, 2025, 11:31:41 am »
It's from China, of course it has to be as shoddy as they can get away with, and sometimes a little shoddier to test the waters. Somebody makes the dice, somebody packages them, everybody cuts all imaginable corners. Then somebody else sells them as LM338 on eBay. Business as usual.

For each customer who blows them up at 1.2A there is probably a bunch of customers who happily use them at 0.6A. It's all about optimizing the overall profit, specs be damned. They usually don't even bother creating their own datasheets and copy paste plots from other vendors.

BTW, do you still have some remaining parts? Could you see what happens with 3.3V input, ADJ to ground and 1Ω load on the output, to absolutely eliminate any possibility of transient overcurrents due to overvoltage, maybe rails going negative because of opposite polarity regulators, stuff like that? Would they blow up anyway?
« Last Edit: January 18, 2025, 11:42:49 am by magic »
 

Offline rsjsouza

  • Super Contributor
  • ***
  • Posts: 6112
  • Country: us
  • Eternally curious
    • Vbe - vídeo blog eletrônico
Re: Warning- LM317K duds from Digikey
« Reply #80 on: January 20, 2025, 11:58:49 am »

I don't know for sure, but I don't get the vibe they're surplus things like that. I've been chatting a bit with the manufacturer and they appear to be taking it seriously and wanting to gather info. I don't think a company just scavenging would bother with even replying to the first email I sent, but these guys are discussing it with me even though they're aware I only bought three of the things and therefore I'm clearly not an important customer
I agree with you. Also, if they ask for the parts back to do a Failure Analysis, then you know they are even more serious than the average supplier. FAs are expensive and time consuming processes and done mostly for manufacturers that want to stand behind their products.

It's funny though, before this I've never given a shit about the guts of chips beyond the ideas of "Wow, it's cool that a chip can do this" or "The way they make dies is super interesting", but now I keep wanting a better microscope to look with more detail and I want to know everything about these bond wires  :-DD
Due to the high import taxes in Brasil, which makes sourcing parts from reputable distributors an expensive venture, my friends there have been looking at dies for years to try to identify and spread the word about fake parts from the typical sources (aliexpress, banggood, mercadolivre, etc.). We are very fortunate to have options...
Vbe - vídeo blog eletrônico http://videos.vbeletronico.com

Oh, the "whys" of the datasheets... The information is there not to be an axiomatic truth, but instead each speck of data must be slowly inhaled while carefully performing a deep search inside oneself to find the true metaphysical sense...
 
The following users thanked this post: David Aurora

Offline floobydust

  • Super Contributor
  • ***
  • Posts: 7744
  • Country: ca
Re: Warning- LM317K duds from Digikey
« Reply #81 on: January 20, 2025, 06:16:51 pm »
It's a blessing the 317's didn't fail shorted - going full output to damage the console. A lot of op-amps downstream...
I've killed a few 317's over the decades. They have a natural weakness to backfeed or reverse-polarity spikes but this is well known and easily solved by adding the datasheet two protection diodes. LT317 only needs one.

I'll bet you could cut open a new EVVO part and watch the bonding wire glow at load.
It would show the packaging fail with too skinny a wire, and also a testing fail that should have caught that. Although it is a PITA to load test when there's no heatsink you can add.
 
The following users thanked this post: David Aurora

Offline David AuroraTopic starter

  • Frequent Contributor
  • **
  • Posts: 489
  • Country: au
Re: Warning- LM317K duds from Digikey
« Reply #82 on: January 20, 2025, 11:32:29 pm »
It's a blessing the 317's didn't fail shorted - going full output to damage the console. A lot of op-amps downstream...
I've killed a few 317's over the decades. They have a natural weakness to backfeed or reverse-polarity spikes but this is well known and easily solved by adding the datasheet two protection diodes. LT317 only needs one.

I'll bet you could cut open a new EVVO part and watch the bonding wire glow at load.
It would show the packaging fail with too skinny a wire, and also a testing fail that should have caught that. Although it is a PITA to load test when there's no heatsink you can add.

Absolutely, I'd have been so pissed off if that had been the fault mode

If I had more spare time I'd have ordered a few more and played around with deeper testing/cutting open a working one and observing it/etc. I still might grab some next Digikey order and do it whenever I get to it
« Last Edit: January 20, 2025, 11:57:13 pm by David Aurora »
 

Offline Analog Kid

  • Super Contributor
  • ***
  • Posts: 1282
  • Country: us
Re: Warning- LM317K duds from Digikey
« Reply #83 on: January 20, 2025, 11:39:41 pm »
I'll bet you could cut open a new EVVO part and watch the bonding wire glow at load.

Hey, why doesn't someone here actually do that?
Not only would it confirm substandard construction, it would also be kewl to watch!
Noopy?
 
The following users thanked this post: David Aurora

Offline magic

  • Super Contributor
  • ***
  • Posts: 7517
  • Country: pl
Re: Warning- LM317K duds from Digikey
« Reply #84 on: January 21, 2025, 11:17:59 am »
I don't think there is any doubt that the wires fused, particularly on samples 2 and 3.

The only interesting questions are:
1. Are they simply not rated for the 1.5A current of LM317, or was it some transient due to shoddy external circuit.
2. What would happen to the chip if the wires didn't fuse.

Other than that, it's just a classic case of "buy Chinese, buy twice".
« Last Edit: January 21, 2025, 11:19:32 am by magic »
 

Offline David AuroraTopic starter

  • Frequent Contributor
  • **
  • Posts: 489
  • Country: au
Re: Warning- LM317K duds from Digikey
« Reply #85 on: January 21, 2025, 01:25:41 pm »
I don't think there is any doubt that the wires fused, particularly on samples 2 and 3.

The only interesting questions are:
1. Are they simply not rated for the 1.5A current of LM317, or was it some transient due to shoddy external circuit.
2. What would happen to the chip if the wires didn't fuse.

Other than that, it's just a classic case of "buy Chinese, buy twice".

1. I really think we've covered the circuit conditions about 10 billion times at this point. The parts were in no way abused, and the failures weren't even all under the same conditions (one failed in the unit the power supply was from, two failed on a dummy load). 2 other brands of the same part have both worked fine there. Also, none of the devices ever reached 1.5A of current anyway (again, something I've covered thoroughly here), they died around 1A. Like I've said over and over again here, NO ratings were even close to being pushed here.
2. That is definitely the interesting question.

The "buy Chinese, buy twice" thing is dumb (frankly I find the constant Sinophobic echolalia around here absolutely moronic). The LM317K is out of production. The options were to buy these from Digikey or buy from random sellers on eBay/etc. I assumed Digikey would be the more reliable source for genuine parts than eBay etc, this didn't turn out to be the case but I think it was quite reasonable to make that assumption.
 
The following users thanked this post: thm_w, Geoff-AU

Offline magic

  • Super Contributor
  • ***
  • Posts: 7517
  • Country: pl
Re: Warning- LM317K duds from Digikey
« Reply #86 on: January 21, 2025, 02:38:37 pm »
1. I really think we've covered the circuit conditions about 10 billion times at this point.
The question has been asked 10 trillion times actually, but never answered.

There is more than one LM317 datasheet and typically more than one application schematic in each of them. And in dual rail systems additional protection is required, which IIRC isn't mentioned in those datasheets.

The parts were in no way abused, and the failures weren't even all under the same conditions (one failed in the unit the power supply was from, two failed on a dummy load).
The latter is closest to a reasonable argument you have made so far, albeit still no details of the test circuit.

2 other brands of the same part have both worked fine there.
Not an argument. They may have worked out of spec. This may have contributed to the failure of the original part.

Also, none of the devices ever reached 1.5A of current anyway (again, something I've covered thoroughly here), they died around 1A.
I gather that you have measured inrush current with a scope and aren't just making things up?

Like I've said over and over again here, NO ratings were even close to being pushed here.
Because "trust me bro".

Life would be much simpler if only people who are right were confident that they are right.

The "buy Chinese, buy twice" thing is dumb (frankly I find the constant Sinophobic echolalia around here absolutely moronic). The LM317K is out of production. The options were to buy these from Digikey or buy from random sellers on eBay/etc. I assumed Digikey would be the more reliable source for genuine parts than eBay etc, this didn't turn out to be the case but I think it was quite reasonable to make that assumption.
Ain't libel if it's true.

If DigiKey starts carrying transistor brands like ICS or Savantic, they will still be the garbage they are. And it looks like power ICs are similar story. And then of course lots of ICs with completely nonsense/misleading names, such as CMOS LDOs sold as 1117. Which kinda works most of the time, until it doesn't because you run into some important difference between these technologies.
 
The following users thanked this post: Siwastaja, tooki

Offline wraper

  • Supporter
  • ****
  • Posts: 18010
  • Country: lv
Re: Warning- LM317K duds from Digikey
« Reply #87 on: January 21, 2025, 03:37:03 pm »
The "buy Chinese, buy twice" thing is dumb (frankly I find the constant Sinophobic echolalia around here absolutely moronic). The LM317K is out of production. The options were to buy these from Digikey or buy from random sellers on eBay/etc. I assumed Digikey would be the more reliable source for genuine parts than eBay etc, this didn't turn out to be the case but I think it was quite reasonable to make that assumption.
It's not out of production. The third option is to pay though the nose https://www.ti.com/product/LM317-N-MIL/part-details/LM317K%20STEEL/NOPB?HQS=ocb-tistore-invf-invftransact-invf-store-octopart-wwe
IMHO the most sensible option is just to use TO-220 part. You can make a terminal adaptor PCB if you want to make it look nice.

 

Offline David AuroraTopic starter

  • Frequent Contributor
  • **
  • Posts: 489
  • Country: au
Re: Warning- LM317K duds from Digikey
« Reply #88 on: January 22, 2025, 12:36:31 am »
1. I really think we've covered the circuit conditions about 10 billion times at this point.
The question has been asked 10 trillion times actually, but never answered.

There is more than one LM317 datasheet and typically more than one application schematic in each of them. And in dual rail systems additional protection is required, which IIRC isn't mentioned in those datasheets.

The parts were in no way abused, and the failures weren't even all under the same conditions (one failed in the unit the power supply was from, two failed on a dummy load).
The latter is closest to a reasonable argument you have made so far, albeit still no details of the test circuit.

2 other brands of the same part have both worked fine there.
Not an argument. They may have worked out of spec. This may have contributed to the failure of the original part.

Also, none of the devices ever reached 1.5A of current anyway (again, something I've covered thoroughly here), they died around 1A.
I gather that you have measured inrush current with a scope and aren't just making things up?

Like I've said over and over again here, NO ratings were even close to being pushed here.
Because "trust me bro".

Life would be much simpler if only people who are right were confident that they are right.

The "buy Chinese, buy twice" thing is dumb (frankly I find the constant Sinophobic echolalia around here absolutely moronic). The LM317K is out of production. The options were to buy these from Digikey or buy from random sellers on eBay/etc. I assumed Digikey would be the more reliable source for genuine parts than eBay etc, this didn't turn out to be the case but I think it was quite reasonable to make that assumption.
Ain't libel if it's true.

If DigiKey starts carrying transistor brands like ICS or Savantic, they will still be the garbage they are. And it looks like power ICs are similar story. And then of course lots of ICs with completely nonsense/misleading names, such as CMOS LDOs sold as 1117. Which kinda works most of the time, until it doesn't because you run into some important difference between these technologies.

Aaaand here we go again.

Get a life mate. You're the last man standing on this massive reach to look for anything but the obvious conclusion that the parts were duds. And you're still making the bold assumption that I am seeking or give the slightest shit about your opinion on it.
 
The following users thanked this post: 5U4GB, Analog Kid

Offline David AuroraTopic starter

  • Frequent Contributor
  • **
  • Posts: 489
  • Country: au
Re: Warning- LM317K duds from Digikey
« Reply #89 on: January 22, 2025, 12:47:32 am »
The "buy Chinese, buy twice" thing is dumb (frankly I find the constant Sinophobic echolalia around here absolutely moronic). The LM317K is out of production. The options were to buy these from Digikey or buy from random sellers on eBay/etc. I assumed Digikey would be the more reliable source for genuine parts than eBay etc, this didn't turn out to be the case but I think it was quite reasonable to make that assumption.
It's not out of production. The third option is to pay though the nose https://www.ti.com/product/LM317-N-MIL/part-details/LM317K%20STEEL/NOPB?HQS=ocb-tistore-invf-invftransact-invf-store-octopart-wwe
IMHO the most sensible option is just to use TO-220 part. You can make a terminal adaptor PCB if you want to make it look nice.



We're getting way into the weeds here if we're going to say it's not out of production because you can buy a version that costs more than 10 times the price of a standard one. Yes, the steel version exists, but over $120AUD for a single regulator IC is not the same as them being readily available. For all intents and purposes the LM317K is out of production.

It's easy to be an armchair expert with hindsight and someone else's experience, but if one of the most trusted major suppliers is selling new production parts for a reasonable price (i.e. these weren't $1 from eBay) why the fuck would my first instinct be to start hacking in a part with a totally different footprint? It's not at all unreasonable to have expected these parts to meet specs.
 

Offline thm_w

  • Super Contributor
  • ***
  • Posts: 7613
  • Country: ca
  • Non-expert
Re: Warning- LM317K duds from Digikey
« Reply #90 on: January 22, 2025, 02:16:26 am »
Dropout is about 1.1V at no load. Quiescent current ~2mA. With 650R R1, adjusted R2 so output ~5.1V.

Maybe I'll measure current limiting at low input voltage tomorrow. Bond wires don't yet show up on thermal cam.

Profile -> Modify profile -> Look and Layout ->  Don't show users' signatures
 
The following users thanked this post: David Aurora

Offline Analog Kid

  • Super Contributor
  • ***
  • Posts: 1282
  • Country: us
Re: Warning- LM317K duds from Digikey
« Reply #91 on: January 22, 2025, 02:18:14 am »
Jesus F. Christ, those wires look tiny.
 

Offline floobydust

  • Super Contributor
  • ***
  • Posts: 7744
  • Country: ca
Re: Warning- LM317K duds from Digikey
« Reply #92 on: January 22, 2025, 02:33:03 am »
Is that a visible kink in the middle of the output bonding wire? It's a long run and maybe they are damaged during assembly.

The LM317 has current-limiting and SOA protection for the pass transistor(s). It's not going to belt out a lot more current without its pass transistor failing. I won't fault the console's circuit here.
 
The following users thanked this post: David Aurora

Offline David AuroraTopic starter

  • Frequent Contributor
  • **
  • Posts: 489
  • Country: au
Re: Warning- LM317K duds from Digikey
« Reply #93 on: January 22, 2025, 02:34:43 am »
Dropout is about 1.1V at no load. Quiescent current ~2mA. With 650R R1, adjusted R2 so output ~5.1V.

Maybe I'll measure current limiting at low input voltage tomorrow. Bond wires don't yet show up on thermal cam.

(Attachment Link)

Certainly looks the same inside!

Same batch/date code (mine were all 2437)?
 

Online bdunham7

  • Super Contributor
  • ***
  • Posts: 8286
  • Country: us
Re: Warning- LM317K duds from Digikey
« Reply #94 on: January 22, 2025, 03:46:32 am »
Is that a visible kink in the middle of the output bonding wire? It's a long run and maybe they are damaged during assembly.

That brings up the question of why they placed the die there instead of closer to or between the pins.  They could use thicker or double wires and not spend any more because they'd be less than half as long.  Sometimes cheap products aren't so much overthrifted as they are poorly implemented. 
A 3.5 digit 4.5 digit 5 digit 5.5 digit 6.5 digit 7.5 digit DMM is good enough for most people.
 
The following users thanked this post: David Aurora, wraper, thm_w, Xena E

Offline electrodacus

  • Super Contributor
  • ***
  • Posts: 1867
  • Country: ca
    • electrodacus
Re: Warning- LM317K duds from Digikey
« Reply #95 on: January 22, 2025, 06:18:13 am »
Base on what I read and could see the fail was thermal fusing of the bond wire.
Voltage is fairly high 25V input and 17V output and there was a mention of DIY replacement of the capacitors.
If the new capacitors are higher capacity and or much lower internal impedance it cold be that fusing of the bond wire happens at power up due to transient  in charging the newly installed capacitor's on the output side.
If I understood correctly they failed few seconds after power up and in that case the bond wire is just incorrectly sized for the job.
The placement of the die seems fairly silly as having it closer to the input pin will have made that bond wire at least half the length reducing resistance and power drop on that to at least half.
So likely a product that does not meant spec and in that case Digikey should remove that component from sale.


Offline magic

  • Super Contributor
  • ***
  • Posts: 7517
  • Country: pl
Re: Warning- LM317K duds from Digikey
« Reply #96 on: January 22, 2025, 09:20:41 am »
It's easy to be an armchair expert with hindsight and someone else's experience, but if one of the most trusted major suppliers is selling new production parts for a reasonable price (i.e. these weren't $1 from eBay) why the fuck would my first instinct be to start hacking in a part with a totally different footprint? It's not at all unreasonable to have expected these parts to meet specs.
It's unreasonable to expect any specs from a manufacturer who can't even be assed to produce a datasheet other than copy-pasting chunks of various text form other vendors which don't make sense together.

Or maybe they just mean something else than you think they do. Because actually, you have presented no evidence that the part fails to meet its specification:


 :-DD

Of course it might still be that the wire will fuse at 0.5A and that otherwise the die would overheat and short out at 1.0A.
« Last Edit: January 22, 2025, 09:36:46 am by magic »
 

Offline David AuroraTopic starter

  • Frequent Contributor
  • **
  • Posts: 489
  • Country: au
Re: Warning- LM317K duds from Digikey
« Reply #97 on: January 22, 2025, 09:32:58 am »
Base on what I read and could see the fail was thermal fusing of the bond wire.
Voltage is fairly high 25V input and 17V output and there was a mention of DIY replacement of the capacitors.
If the new capacitors are higher capacity and or much lower internal impedance it cold be that fusing of the bond wire happens at power up due to transient  in charging the newly installed capacitor's on the output side.
If I understood correctly they failed few seconds after power up and in that case the bond wire is just incorrectly sized for the job.
The placement of the die seems fairly silly as having it closer to the input pin will have made that bond wire at least half the length reducing resistance and power drop on that to at least half.
So likely a product that does not meant spec and in that case Digikey should remove that component from sale.

As mentioned many times already, only the first failure happened in the unit that the owner had recapped, both other failures occurred on a DC load which rules out any capacitance issues downstream.

But yes, the units worked unloaded/lightly loaded but failed as soon as load was increased.
 

Offline David AuroraTopic starter

  • Frequent Contributor
  • **
  • Posts: 489
  • Country: au
Re: Warning- LM317K duds from Digikey
« Reply #98 on: January 22, 2025, 09:47:23 am »
It's easy to be an armchair expert with hindsight and someone else's experience, but if one of the most trusted major suppliers is selling new production parts for a reasonable price (i.e. these weren't $1 from eBay) why the fuck would my first instinct be to start hacking in a part with a totally different footprint? It's not at all unreasonable to have expected these parts to meet specs.
It's unreasonable to expect any specs from a manufacturer who can't even be assed to produce a datasheet other than copy-pasting chunks of various text form other vendors which don't make sense together.

Or maybe they just mean something else than you think they do. Because actually, you have presented no evidence that the part fails to meet its specification:
(Attachment Link)
 :-DD

The placement of the die seems fairly silly as having it closer to the input pin will have made that bond wire at least half the length reducing resistance and power drop on that to at least half.
It seems to be placed right on the OUT pin, perhaps for thermal relief.

Sometimes cheap products aren't so much overthrifted as they are poorly implemented. 
The opposite is often true - carefully engineered to be maximally cheap and dodgy.

Are you new to electronics? Do you have any idea how many data sheets are copy/pastes of old data sheets from other manufacturers? Jesus Christ. Either way, if we assume a manufacturer is providing inaccurate data sheets and Digikey is still selling the part then why are you obsessed with trying to pin the blame on the end user for *checks notes* trusting a manufacturer and a supplier to make/sell a jelly bean part that has been around for nearly 50 years and sold by damn near every IC manufacturer that has ever existed? Get a grip.

Evidence wise, why do you think I owe you evidence? Why do you think your opinion matters?

It's Shrödingers Regulator apparently- you claim the part is cheap Chinese shit, but simultaneously keep saying the part only failed because of some mystery misuse conspiracy. Pick a shit theory and stick with it, you absolute peanut.
 

Offline magic

  • Super Contributor
  • ***
  • Posts: 7517
  • Country: pl
Re: Warning- LM317K duds from Digikey
« Reply #99 on: January 22, 2025, 10:02:36 am »
Either way, if we assume a manufacturer is providing inaccurate data sheets and Digikey is still selling the part then why are you obsessed with trying to pin the blame on the end user for *checks notes* trusting a manufacturer and a supplier to make/sell a jelly bean part that has been around for nearly 50 years and sold by damn near every IC manufacturer that has ever existed? Get a grip.
I just suggest that you get a grip because this is the new normal now :D

It's Shrödingers Regulator apparently- you claim the part is cheap Chinese shit, but simultaneously keep saying the part only failed because of some mystery misuse conspiracy. Pick a shit theory and stick with it, you absolute peanut.
Shit circuit, shit part, smoke goes out. What's surprising about that?

Also, it's not me who blew this shit up so I have no vested interest in proving any theory about what went wrong. (Neither is it me who is on the hook for this thing continuing to work for another N years). I am equally prejudiced against audio gear and Chinese semiconductors, so whichever gets the blame I'm fine with it. But I'm curious what exactly went wrong and not accepting hand wavy non-explanations for this reason.
« Last Edit: January 22, 2025, 10:16:07 am by magic »
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf