Generating a 78MHz clock from a 100MHz base clock (VHDL)

[pierpaolo...@gmail.com]

I have to generate a 78MHz clock (duty cycle 0.5 or 0.7) from a 100MHz base clock (duty cycle 0.5) using VHDL language (so the ratio is 200/156). I know that I can use DCM, PLL or similar, but at this moment (unfortunately) I just can't.

Therefore I thought to use (excluding any DCM or PLL) a simple frequency divider, but in this case I also know that the frequency can be divided only by integer numbers (and minimum 2, because I would use counters to do that - and In my case I have to divide the base clock by 1,2820512820512820512820512820513...).

So I have no idea how to realize that without using any DCM or other stuff... I thought to divide the 100MHz clock in smaller frequencies (like 50MHz, 25MHz etc.) and adding them (50+25+3 for example), but is this the right way (logically I don't think so)?

So, have you some suggests?

[Gabor]

Whether you can do anything useful with counters depends on the clock
requirements. You can certainly use direct digital frequency synthesis
to generate the required 78 MHz - on average. Unless you use both
clock edges, the jitter will be 10 ns, which is very large for many
applications, but might be OK if you're just dividing this down to
clock a UART for example. Using both edges of 100 MHz you still end
up with 5 ns of jitter. If that isn't good enough you really need
to use a PLL or DCM to do the job.

-- Gabor

[Rob Gaddi]

You can't do what you're trying to do, or at least not reliably.

If you were SUPREMELY desperate you could build an NCO to drive a pin
or two as a poor-man's DAC (you're up above Nyquist, so you'd have to
be clever here), then use an LC filter to trap 78 MHz, a Schmitt
trigger to square it back up, then feed it back into the chip.

I feel like we should write this question over the door here: "What is
it you're actually trying to accomplish that makes you think you need
_____?"

--
Rob Gaddi, Highland Technology -- www.highlandtechnology.com
Email address domain is currently out of order. See above to fix.

[rickman]

On 11/12/2012 12:46 PM, Rob Gaddi wrote:
> On Sun, 11 Nov 2012 01:16:57 -0800 (PST)
> pierpaolo...@gmail.com wrote:
>
>> I have to generate a 78MHz clock (duty cycle 0.5 or 0.7) from a 100MHz base clock (duty cycle 0.5) using VHDL language (so the ratio is 200/156). I know that I can use DCM, PLL or similar, but at this moment (unfortunately) I just can't.
>>
>> Therefore I thought to use (excluding any DCM or PLL) a simple frequency divider, but in this case I also know that the frequency can be divided only by integer numbers (and minimum 2, because I would use counters to do that - and In my case I have to divide the base clock by 1,2820512820512820512820512820513...).
>>
>> So I have no idea how to realize that without using any DCM or other stuff... I thought to divide the 100MHz clock in smaller frequencies (like 50MHz, 25MHz etc.) and adding them (50+25+3 for example), but is this the right way (logically I don't think so)?
>>
>> So, have you some suggests?
>
> You can't do what you're trying to do, or at least not reliably.

I don't agree. You just need to think outside of the box.

> If you were SUPREMELY desperate you could build an NCO to drive a pin
> or two as a poor-man's DAC (you're up above Nyquist, so you'd have to
> be clever here), then use an LC filter to trap 78 MHz, a Schmitt
> trigger to square it back up, then feed it back into the chip.
>
> I feel like we should write this question over the door here: "What is
> it you're actually trying to accomplish that makes you think you need
> _____?"
>

How about of circuit that uses both edges of the 50% duty cycle 100 MHz
clock? That would give you a virtual 200 MHz clock with a resolution of
5 ns. This gives a base error of 22% worse case, not including
propagation deltas in the clock and I/O routing.

This is not an easy circuit to design, but it can be done I think. This
will require a double clocked counter with two registers and two adders.
Think of it as two phased...

The results of the two adders will be logically combined to produce two
phases of the output which would be combined in a DDR type output. If
the DDR output isn't supported, then the two phases would have to be
xored in a single LUT before being output. This should produce a clock
with 5 ns resolution and little edge jitter if manually placed
appropriately.

If this is not clear I can provide some HDL, but that might take a
little time and I don't have much today.

Rick

[BLue39]

Hi pierpaolo,

The method I'm suggesting is dependent on the FPGA you are using.

In the XILINX FPGA's there's a structure called "Carry Chain", which is a complete combinational circuit. You will have to generate a 100 MHz clock with 155 degree phase shift. (You will have to perform experimentations and find the number of the carry chain numbers you will require; this number actually is device dependent.)
Similarly you generate clocks with 78 degree and 235 degree phase shift.

Thus you have 3 clocks in the picture. All having frequencies of 100 MHz, but their phase 0, 78, 155, 235 degrees. Call them clk1, clk2, clk3 and clk4 respectively.

Suppose your final 78 MHz clock is named as clk_78.
Then you will have to toggle its value according to the below event order.

1. Toggle when rising_edge of clk1.
2. Toggle when rising_edge of clk2.
3. Toggle when rising_edge of clk3.
4. Toggle when rising_edge of clk4.
5. Again go to step1.

To follow this order, you'll have to do a clever clock muxing of clk1, clk2, clk3 and clk4.
Notice we are not changing clk_78's value at every rising edge of any of the 4 clocks.



I know this is a very complex way and you'll have to perform manual routing and take care of the routing delays.

But, I can think of this only way of multiplying/dividing your 100 MHz clock with such an odd figure.

--BLue39

[Andy]

Ummm... NO, this won't work.

Even if you solve the placement and routing, and overlook the temperature/voltage dependencies of the phase delays, the pattern has an additional phase delay getting back to step 1. Toggling the new clock signal four times every 100 MHz will still yield an average frequency of 50 MHz.

Just supposing it would work (which it won't), there are ways to toggle a signal on multiple clocks without muxing the clocks. Look up "Flancter circuit".

Andy