DCM_STANDBY macro in Virtex-4

I was obliged to go back to Virtex-4 Stepping Level 1. So I'm now faced with the DCM problems. Should I use the DCM_STANDBY macro for each DCM on the FPGA or only for the instantiated ones?

Thanks in advance Mehdi

Reply to
GaLaKtIkUsâ„¢
Loading thread data ...

Mehdi,

The DCM Macro is to prevent the unlikely possibility of NBTI shift in the delay lines.

If the macro is used, or is the DCM is used with a clock that is present, then the DCM will never experience NBTI shift.

NBTI shift requires: circuit is static (does not toggle), and power is applied, and a long time. NBTI can be accelerated by very high temperature and high voltage. Since the delay line supplies are regulated, there is no way for a customer to apply a higher voltage to the delay lines.

It was thought that in some cases, a pcb is inserted, and not provisioned (such as a card in a large network). The card might remain, powered up, at a fairly high ambient temperature for a few weeks before it is needed.

The NBTI shift will cause the delay line to have duty cycle distortion, which means that the maximum frequency will drop from ~700 MHz (where it normally breaks) to ~500 MHz.

In order to stay inside the +/-5% duty cycle specification for distortion, we placed the "autocal" macro in the software to automatically get placed.

If you use, or don't use, the DCM's, and never plan to change, or use them all below 250 MHz, then it is pretty much a don't care.

If you don't use a DCM, and need to use that DCM later, and it needs to work >250 MHz, then the autocal block is a good idea (guarantees operation to specification).

If you use the DCM for awhile, then the shifts are equalized, and subsequent loss of clock doesn't do anything at all (eg a 168 hour active burn in would prevent any further issues).

NBTI can be baked out by powering the device off, and annealing it at

125 C for ~72 hours. Or, you can just exercise the DCM, and the NBTI shifts will equalize, and it will (eventually) work.

As annoying as the NBTI was, it was a terribly obscure and minor issue, one that has never occurred in any report from the field, and one that we only saw in our qualification tests where we intentionally run things at temperatures and voltages outside of the absolute maximum specifications.

For V5, NBTI was accounted for, so you do not need to worry.

Austin

Reply to
Austin

Another question on the subject. Is it interesting to try changing the macro to control all the unused DCMs? Another question: I extensively use the DRP port to change the frequency during the development phase of my projects. As I see it's impossible on Stepping Level 1?

Mehdi

Reply to
GaLaKtIkUsâ„¢

Mehdi,

The autocal block should be automatically inserted for all DCM's, used or unused by the software...

At least, that is what I had heard.

As for the DRP port, if it is listed in the errata for that stepping, then that is what that device (does not) delivers.

Which device? What steppings?

Austin

Reply to
Austin Lesea

  1. I think that autocal is not inserted automatically for STEPPING1.

Since DRP is used by autocal than it's not accessible by the user!

Reply to
GaLaKtIkUsâ„¢

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.