Well, good luck Charles. If you are successful, you MIGHT be able to bypass ADI, which they kind of deserve, not giving you 'right of return' privileges anymore.
I had a similar problem with the Toshiba 2SK147, in 1990. They were drop shipping me 100 units/mo when this happened. They ALSO changed the recipe, and made a bad batch that was very noisy. The usual performance was shown by graphs on the data sheet, and they maintained it for more than a decade. In fact, it became almost a 'waste of time' to measure them on the QuanTech, before assembly, but then the 'process' changed and every device was different at 10 Hz. I still have a whole set of graded devices, picking the best ones for the input, next best for the second stage, and third best for an upgrade, still further down the line. Still, the worst case spec. was almost 10 times worse than was typical for the previous decade and HUNDREDS of devices didn't make even that spec, so I was allowed to return them for exchange stock. I am pretty sure that Toshiba was embarrassed about this, as later devices got better again. And so it goes! That is what we have to do, to make 'world class' audio products, right Charles?
I had a similar problem with the Toshiba 2SK147, in 1990. They were drop shipping me 100 units/mo when this happened. They ALSO changed the recipe, and made a bad batch that was very noisy. The usual performance was shown by graphs on the data sheet, and they maintained it for more than a decade. In fact, it became almost a 'waste of time' to measure them on the QuanTech, before assembly, but then the 'process' changed and every device was different at 10 Hz. I still have a whole set of graded devices, picking the best ones for the input, next best for the second stage, and third best for an upgrade, still further down the line. Still, the worst case spec. was almost 10 times worse than was typical for the previous decade and HUNDREDS of devices didn't make even that spec, so I was allowed to return them for exchange stock. I am pretty sure that Toshiba was embarrassed about this, as later devices got better again. And so it goes! That is what we have to do, to make 'world class' audio products, right Charles?
Assembly usually handled by an independent sub con. That's totally different to a tech with a second hand machine.
"I believe failures after burn in are no longer called "infant" mortality by definition. At the single digit production scale, burn in is pretty easy."
Perhaps in a strict definition, but the nightmare is failures on the customers production line and early failures in operational service life in the first few months.
Perhaps in a strict definition, but the nightmare is failures on the customers production line and early failures in operational service life in the first few months.
We do chip on board, here in the lab about as DIY as you could ever want. Epoxy die attach and hand bonding that anyone could get the hang of. Gooped over with UV set goop, Masterbond has a new hermetic one that is quite good. I have no reason to believe there would be any reliability issues in a commercial grade (0-70C) environment.
We have also prototyped bare-die optical fiber connections with laser drilled alignment holes, at $35 a hole (2 required) it was not economical, but reliability was not bothered by this. The demo left a couple of customers with dropped jaws.
Last edited:
Come on this is the "highend" marginal reliability comes with the territory, it proves how close to the edge we are operating.
There's your noise problem, it's UV rays knocking electrons out of bonds and the free radicals from the photoinitiator causing erratic charge carrier flow. You need virgin Teflon!
Keep at it, Scott, one day you'll get the hang of this IC stuff.
"Come on this is the "highend" marginal reliability comes with the territory, it proves how close to the edge we are operating."
After paying 10 or 20 grand for high end gear I would want to know it would last.
Doing this stuff to demo is ok "engineering samples".
I have a few stories about bare die as well . . . But I'll not pursue that here.
After paying 10 or 20 grand for high end gear I would want to know it would last.
Doing this stuff to demo is ok "engineering samples".
I have a few stories about bare die as well . . . But I'll not pursue that here.
I guess I forgot the smilie, then again there are more than one review that starts with the first unit dramatically going up in smoke.
You might be surprised how many people are
more adventurous (willing to support research)
than you .
The thing that's funny about it is that the part is pretty old. Barrie Gilbert designed it in the early '90s. They have a partial schematic in the data sheet. I was wondering about a resistor value in one of the places and it may have even been Scott Wurcer who answered my question. He said that the part was so old that they didn't have the documentation anymore. It was done before PDF's were invented and nobody ever scanned it. So he couldn't look it up anywhere. The tech support guy (Scott?) said, "Let me call you back."
So he found one in a ceramic package and popped the lid off of it. Then he looked at it under a microscope with a calibrated graticule and by measuring the size of the resistor he could tell me that it was a 10kohm part. I thought that was pretty good service!
Anyway one day I'm getting worried that they are going to discontinue the part. I got a hold of some guy who said it would never be discontinued. He wouldn't say exactly why, but I got the impression that it had been spec'ed in to some sort of weapons system -- you know, "A Matter of National Security".
But I guess that the smart bomb can hit its target even with a noisy chip. They just got noisy overnight. The admitted right up front is was because they switched from a 4" wafer to a 6" wafer or something like that. I figured that they would just refine the parameters until they got the process dialed in.
That was about 12 years ago....
For the first couple of years they were all concerned and acted like they would fix the problem. Then they must have got a new facility manager. He was like, "It means the spec. Go screw yourself."
I got all excited when National did their big audio push six or eight years ago with all of the super low distortion audio products. They cloned so many parts.
BUF634? No problem.
World's lowest distortion op-amp? No problem.
But when they cloned the AD844 current feedback op-amp, they forgot to connect pin 5. It would have cost them $0.001 to add one single bond wire to an 8-pin SOIC. Absolutely no reason in the world not to. But they didn't.
So we just have to keep screening the ADI parts, throw 20% of them away and hope that they keep making them. I e-mailed Mark Brasfield who was leading the group at National. He said, "It measures better and sounds better than all the AD current feedback parts." But it doesn't have pin 5 connected... can't use it.
So he found one in a ceramic package and popped the lid off of it. Then he looked at it under a microscope with a calibrated graticule and by measuring the size of the resistor he could tell me that it was a 10kohm part. I thought that was pretty good service!
Anyway one day I'm getting worried that they are going to discontinue the part. I got a hold of some guy who said it would never be discontinued. He wouldn't say exactly why, but I got the impression that it had been spec'ed in to some sort of weapons system -- you know, "A Matter of National Security".
But I guess that the smart bomb can hit its target even with a noisy chip. They just got noisy overnight. The admitted right up front is was because they switched from a 4" wafer to a 6" wafer or something like that. I figured that they would just refine the parameters until they got the process dialed in.
That was about 12 years ago....
For the first couple of years they were all concerned and acted like they would fix the problem. Then they must have got a new facility manager. He was like, "It means the spec. Go screw yourself."
I got all excited when National did their big audio push six or eight years ago with all of the super low distortion audio products. They cloned so many parts.
BUF634? No problem.
World's lowest distortion op-amp? No problem.
But when they cloned the AD844 current feedback op-amp, they forgot to connect pin 5. It would have cost them $0.001 to add one single bond wire to an 8-pin SOIC. Absolutely no reason in the world not to. But they didn't.
So we just have to keep screening the ADI parts, throw 20% of them away and hope that they keep making them. I e-mailed Mark Brasfield who was leading the group at National. He said, "It measures better and sounds better than all the AD current feedback parts." But it doesn't have pin 5 connected... can't use it.
If a part meets spec and its running in volume production, there is just no incentive to go back and change it. If the particular parameter that you are interested in is not on the data sheet, the vendor is not obligated to maintan that parameter from device to device. That unfortunately is the economics of volume production, and if you are a small user (anyone in any branch of audio proper except for maybe 3 companies would be a small volume semiconductor user) you especially are unlikey to get things changed.
Some vendors do special selections for customers, but there are minimum volumes and the parts cost extra. You might want to explore that route with your supplier.
Some vendors do special selections for customers, but there are minimum volumes and the parts cost extra. You might want to explore that route with your supplier.
Last edited:
That's the unclear part- do the parts meet the published spec or not? The implication is that they don't, but the wording here is vague. If the parts meet spec, I don't see what the problem is- if the specified noise is too high, the engineer chose the wrong part for the application. If the parts don't meet spec, the manufacturer is obligated to take them back.
Well, this is what separates the 'men' from the 'critics'. '-) Isn't this wonderful when manufacturers tell you to: "Go screw yourself" rather than addressing the problem?
Toshiba FIXED their problem, that I gave them feedback on.
Personally, National was pretty slouchy in the 1980's, essentially citing 'worst case specs' rather than the typical noise, on the spec sheet.
That is when I met the fresh and eager Scott Wurcer who had made an LF411,412 'clone', and the noise specs were more consistent, so I paid a little more to use them for servos in the Vendetta Research. Now we have a different Scott Wurcer, I presume.
Toshiba FIXED their problem, that I gave them feedback on.
Personally, National was pretty slouchy in the 1980's, essentially citing 'worst case specs' rather than the typical noise, on the spec sheet.
That is when I met the fresh and eager Scott Wurcer who had made an LF411,412 'clone', and the noise specs were more consistent, so I paid a little more to use them for servos in the Vendetta Research. Now we have a different Scott Wurcer, I presume.
- Status
- Not open for further replies.