CPLDs

[forte31]

I can't seem to find the Global Clk input risetime requirement in the datasheet. Is there another document I should use?

[gszakacs]

A quick site search for "rise time" got me to this answer record:

http://www.xilinx.com/support/answers/3226.htm

They're pretty loosey goosey about specifying rise and fall times.  Although they

say you can use a slower edge than recommended if it is "monotonic" you have to

watch out, because they mean "monotonic at the internal pad of the part" which may

differ from the pin if you have significant ground or power noise.

-- Gabor

[forte31]

Thanks gszakacs. I wish there was a more formal document i could use to show my management. I inherited a design using a 555 timer as a system clock. I have to assume this will cause problems as explained in the answer record.

[eteam00]

I inherited a design using a 555 timer as a system clock. I have to assume this will cause problems as explained in the answer record.

Fortunately, I had already set down my mug of coffee before I read this.  These are no doubt the funniest words I will read or hear all day long.  Is this a limited run of circuits, where you can jerry-rig a proper (schmitt trigger) buffer into the circuit?

Or perhaps you have an extra pin or two on the CPLD, to form your own R-C oscillator?

-- Bob Elkind

[gszakacs]

If you need more weight behind the answer, I would suggest talking to your Xilinx sales rep

or FAE.  As far as the 555 timer goes, I believe there are quite a few flavors of this part and

you might find one with better rise and fall times.  I must say though that it's been about

30 years since the last time I used one in a design.  I consider these to be more of a "hobbyist"

part.  I don't remember if its output is active drive or open collector.  If it's open collector, you

can always reduce the pullup resistance to get a faster rise time, but that would of course

eat more power.  If all you really need is a cheap and dirty RC oscillator, then you might look

at any number of CMOS inverting Schmitt triggers that have very good rise and fall time characteristics,

although the thresholds do not have the same precision as the 555 timer, so you'd need to

tweak the R and C values.  If you plan to build more than a handful of these designs, then

you should probably look into other clock sources.  There have been a lot of advances in

timing generation since the 555 timer.  You can buy a PIC micro in an SOT-23 package that

has an internal clock with 1% accuracy, and just program it to toggle an output.  It needs no

external timing components, and costs much less than $1 in small quantities.

-- Gabor

[eteam00 · Accepted solution]

As a followup to Gabor's comments...

The National (now TI) LMC555 has a datasheet rise/fall time of 15nS (typical).  There is no min/max spec.  This is a considerable improvement over the 100nS (typical) spec of the classic LM555.

The Philips (now NXP) folks have an improved ICM7555 part with a breathtaking 75nS (typical) rise/fall spec.

But seriously, it only takes one additional CPLD output pin to provide (more) hysteresis on the CPLD's clock input pin.  This should be sufficient, and a circuit simulation with the CPLD IBIS models should provide the solid data needed to convince a crusty old manager.

-- Bob Elkind

[forte31]

Thanks to all. I personally have been using the LTC6905 (0.5ns risetime). I will try to convince my management for a re-design. I know they will ask for the risetime requirement for the XC9500. I will use the answer record.

[forte31]

The hysterisis approach is a good suggestion too, thanks.

[eteam00]

If you can do a board turn, a proper clock source is much better than a two-pin hysteresis circuit.

Keep in mind that the LM555 runs up to about 1MHz before it super-novas into the circuit board.  Check the frequency range of the LTC6905 series.  You don't want to melt the poor CPLD with 100x clock frequency.

Maxim has an assortment of low-frequency (under 1MHz) oscillators ("silicon oscillators") which are less expensive than the LTC6905 parts.  The DS1099 is available in frequencies in the range that an LM555 might be comfortable.

Good luck to ya!

-- Bob Elkind

[forte31]

But the DS1099 has a 20ns risetime exceeding the PLD 10ns max.