| View previous topic :: View next topic |
| Author |
Message |
Gringo Guest
|
Posted: Wed Jul 30, 2026 11:55 am Post subject: No Guardbands |
|
|
I see test programs with no guardband for some parameters. Is this ok? When is this acceptable?
Gringo
|
|
| Back to top |
|
EE
Joined: 16 Jun 2026 Posts: 108 Location: Paranaque City, Phils.
|
Posted: Thu Jul 31, 2026 1:29 pm Post subject: |
|
|
Hi Gringo,
In general, any of the following cases would justify not putting in a guardband: 1) the parameter is not specified in the datasheets; 2) the parameter is just used for computing or measuring another parameter and this second parameter is already guardbanded; 3) the parameter is just an attribute data point (‘go’-‘no go’) and not a continuous one; and 4) the actual production data distribution of the parameter is well within both spec limits (high Cpk).
Thanks,
EE
|
|
| Back to top |
|
Gordy Guest
|
Posted: Fri Aug 01, 2026 2:09 pm Post subject: |
|
|
In our group, we also do not put a GB if the test method used is statistical like the test for DNL wherein not all codes are tested and only some codes are sampled. In this case I think the data can't be considered to be normal.
Gordy
|
|
| Back to top |
|
Guest
|
Posted: Thu Feb 12, 2026 12:17 am Post subject: |
|
|
Hi EE and the others,
EE, you made very good classification on not including guardbands, based on what you've posted: 1) the parameter is not specified in the datasheets; 2) the parameter is just used for computing or measuring another parameter and this second parameter is already guardbanded; 3) the parameter is just an attribute data point (‘go’-‘no go’) and not a continuous one; and 4) the actual production data distribution of the parameter is well within both spec limits (high Cpk).
We can add a fifth classification, that is, for new products 5) Test/Product engineers often try to evaluate on how they can modify the product to further strengthen their market share, so they tend to monitor parametric performance of some really absurd device data. So you might observe tests that does not make sense to most line engineers. Not only that some parameters in the test program don't have limits, even some includes test methods that measure device performances not specified in the datasheet. Sometimes, they will try to find out the potential of a part to perform this type of behaviour and use that data to create or innovate new products.
Let me know if you have queries. Regards.
Krygk
|
|
| Back to top |
|
|