Re: hobby electronics

On 7/7/2024 12:55 PM, Martin Brown wrote:

On 07/07/2024 19:32, Don Y wrote:

On 7/7/2024 10:43 AM, Edward Rawde wrote:

When I worked as a designer I came across fake parts a few times. One reel
if devices was SO8 and the devices were marked with the correct number but
the manufacturer didn't make that part in SO8. So it wasn't hard to spot the
fake. Another reel of devices did have the correct part number on the
correct package but it was anybody's guess what chip was actually in the
package.

>
I had a client contact me about a problem he was having with some
other design (not one of mine).  Begrudgingly, I offered to help
(I don't like having to troubleshoot other designs as you never
know the quality of the design/designer).
>
It quickly became apparent that one device was bad in every sample.
On closer inspection, there was no die *in* the package!

 We once had a flashover event with a very expensive prototype 22bit ADC back in the late 80's where it vapourised the bond out leads. The chip was working really well right up to that point and we had a 6 sig fig linear ADC (intention was to replace bulk buy Solartron 7060s).

I did that while debugging a CPU ("chip") built from ECL SSI/MSI DIPs.
The system was "hot".  I slipped and let one of the DIP's leads short
the -5.2V power pin to the ground plane.  The 100A power supply didn't
even notice the change in load impedance -- when my eyes cleared from
the flash, the lead was gone.  Ooops.

We decided honesty was the best course of action and the chip makers liasson guy admitted that if *they* hadn't tested it themselves before shipping it to us they could have believed that they forgot that step!
 They gave us another chip and the things went into full production not long after they were able to produce them in bulk. Solartron were a bit surprised when one their best customers vanished completely.

I have always used integrating and self-calibrating approaches to get high
precision results.  Usually <something>-to-frequency where I can count the
per unit time to get high precision or measure time per count for lower
frequency results.  In some designs, I've had to address uncalibrated
tolerances of -50% to +300%.
[As long as you don't have to report dimensioned units, you can get away
with lots of "magic"]