Electronics > Microcontrollers

Is a revision A of Silicon reliable, if there is a B?

<< < (4/4)

AndreZheng:
Normally, the datasheet will lists all the differences applied on revision B. And another hidden fact is there might be some cost saving technicians also applied on revision B.

luiHS:

I have been reviewing the manufacturer's errata documentation, and I see that between version A and B, practically all the errata are kept in version B, except for one that is no longer in revision B.

In the document I see that almost all these errata are given solutions by software, that is, they are not corrected in the Silicon of the next revision of the chip, but it is the user who must solve it by software in his program according to the suggestions of this errata document.

So I do not see that it is so important to buy a revision A instead of a B, mainly because of the A there is stock in my country, and of the B I would have to access the Chinese market, with the risk that they will sell me product counterfeit or reused, which already happened to me with some MOSFETs and accelerometers.

Is it common that many errors are not corrected by hardware in the next revision of the chip, and the manufacturer simply suggests fixes by software in the errata document?

ataradov:

--- Quote from: luiHS on June 25, 2021, 12:19:14 am ---Is it common that many errors are not corrected by hardware in the next revision of the chip, and the manufacturer simply suggests fixes by software in the errata document?

--- End quote ---
Yes, very common. Often things are not fixed even if there is no workaround.

All erratas are weighted against the risk of introducing new errors. If the feature with an issue is not critical for majority of the customers, it may have erratas that are never fixed.

SiliconWizard:

--- Quote from: ataradov on June 25, 2021, 12:30:48 am ---
--- Quote from: luiHS on June 25, 2021, 12:19:14 am ---Is it common that many errors are not corrected by hardware in the next revision of the chip, and the manufacturer simply suggests fixes by software in the errata document?

--- End quote ---
Yes, very common. Often things are not fixed even if there is no workaround.

All erratas are weighted against the risk of introducing new errors. If the feature with an issue is not critical for majority of the customers, it may have erratas that are never fixed.

--- End quote ---

Yes. Also, what is considered is how effective the workarounds are. If they are indeed effective, there is little incentive to correct the hardware, as not only could it introduce new problems as you said, but it could require designers to handle the parts in a different way depending on silicon revision, which is a royal pain!

When designers have been relying on silicon bugs and implemented workarounds for them, they don't necessarily want these bugs to be fixed. Simple as that. Unless of course the workarounds are very annoying/not 100% effective/or add significant cost.

For issues that have no workaround, that could be the same: if designers have choosen a part despite of a bug with no fix, and have designed around it, then they won't care much if the problem is fixed in a later revision.

That's the general problem dealing with products that have been deployed, especially on a large scale. Once they have been, offering fixes that could actually make existing designs not work anymore is a tough decision to make. One approach that some vendors use is to finally release an "upgraded" part instead of just a silicon revision; for instance, TI sometimes releases parts with an "A" suffix; they are equivalent, but usually contain a number of fixes that make them not strictly equivalent. Since that's not strictly the same part number, and they often have a significant overlap in production with the two versions, existing customers can keep buying the old version, and take time to eventually upgrade.

Bassman59:

--- Quote from: ataradov on June 25, 2021, 12:30:48 am ---
--- Quote from: luiHS on June 25, 2021, 12:19:14 am ---Is it common that many errors are not corrected by hardware in the next revision of the chip, and the manufacturer simply suggests fixes by software in the errata document?

--- End quote ---
Yes, very common. Often things are not fixed even if there is no workaround.

All erratas are weighted against the risk of introducing new errors. If the feature with an issue is not critical for majority of the customers, it may have erratas that are never fixed.

--- End quote ---

Errata is plural.
Erratum is singular.

Just like data and datum.

So in the above, “erratas” is not correct!

Navigation

[0] Message Index

[*] Previous page

There was an error while thanking
Thanking...
Go to full version