IOBDELAY's delay value

Hi,

In Virtex-II, the delay buffers are either ON or OFF. There is no other adjustment. In order to know the delay that has been introduced, you should run a timing report on your design. I say this because the magnitude of the delay buffers is not the same in each part -- e.g. 2v500 and 2v6000 will report different values.

The values are necessarily different because the purpose of these delay buffers is to add delay to eliminate positive hold times for input flip-flops when used with global clocks and no DCM. Larger parts have larger delay on the clock distribution, requiring larger delays for the input delay buffers.

Hope that helps, Eric

path?

value.

Virtex-4 (different from the earlier families) has a programmable IDELAY on every input, each programmable with 64 steps, each step about

Do I understand right that V4 contains an internal 200 MHz oscillator?

Can this ocsillator be used for other more general purposes?

Do I understand you right that V4 contains an internal 200 MHz oscillator?

Can this oscillator be used for other more general purposes?

Microchip/PIC says they get 1% at typical conditions and 2% over reasonable VCC and temp. (5% for wider VCC/Temp.)

That needs a calibration step specific to each chip. A magic number is stored in flash. You have to copy it over to a speed-tweak register as part of initialization. The factory initializes the flash location and/or you can rederive it with their tools.

That's from the data sheet for the PIC-12F675 which is several years old. I haven't checked their more recent chips. (Or newer versions of that data sheet.)

These things do keep getting better and better, but they are not simple, or trivial blocks.

Maxim and Linear tech have 'Silicon Oscillators', that claim these precisions, but in these devices all they have to do is oscillate, so they can target a process very specifically.

A very recent, and perhaps better example, is the Freescale MC9RS08KA2 tiny microcontroller. That is true mixed signal.

That claims 2% MAX, over voltage and temperature,(which I think means -40'C to 85'C and 1.8V-5.5V), and it has a 0.2% trim precision.

This costs ~140uA, for 32Khz, but they also have a 300nA/1KHz oscillator, with reduced precision ~30%.

They also have a 1% 1.2V bandgap reference in this device - AND it all costs under 50c !!.

All of this will be something of a culture shock to FPGA vendors :) [ uA ?! cough, what's a nA ? 50 _cents_ ?! , choke... ]

-jg