Odd Oversampling

Why do you need any tricks? You have at least 7 cycles of your fast clock in which to sample each data point. The samples will be either 7 or 8 clocks apart. So, you look at the 16MHz clock and use it to create a clock-enable in the 125MHz clock domain.

If you don't have access to the original 16MHz clock, you can either try to recover it by locking a digital PLL on to the data transitions (7x oversampling is *just* about enough to be able to do this easily) or you can use dead-reckoning by using a numerically-controlled oscillator (NCO) to generate

16MHz sampling pulses with a phase jitter. And yes, an NCO is effectively an implementation of Bresenham.

--
Jonathan Bromley, Consultant

DOULOS - Developing Design Know-how

There's a simple one if you have both clocks. Write a synchronous process using 125MHz as the clock and 16MHz as an input named "rate" Synchronize "rate" and use it to generate a rate_rising clock enable pulse.

-- Mike Treseler

It's easy enough with just 4 times oversampling. XAPP224 shows how to do it at 400Mb+ data rates. Cheers, Syms.

[...]

OK. I was recalling bad memories of an attempt to do DPLL with only

4x oversampling in a system where the cheap-and-nasty fibre optic transceivers introduced rather a lot of pulse width distortion, which made it a whole lot more difficult.

Thanks for the reference.

--
Jonathan Bromley, Consultant

DOULOS - Developing Design Know-how

Please correct me if I got the problem wrong.

Is it right that I could define a counter (clocked with 125MHz) and define one counter position as the sample point which is located in the middle position neighborhood of the bit to sample ? The counter would be quasi started if the first transition is recognized. This transition could be recognized with a two stage flip flop chain in the 125Mhz clock domain. The problem would be that the following sample points would walk because

125/16 is a fraction number. Is that right ?

Could that sample point walk even then if the oversample factor 125/x would be an integer number for example 120MHz/12MHz when 120MHz having 0.05% tolerance and 20MHz having 0.02% tolerance ?

If I have 7 times oversampling would I need then an input stage with seven stages (XAPP224)? What additional clocks would I need then ? Or are four clocks sufficient?

Thank you in advance.

Rgds André

One more question:

The quasi "oversample" solution in XAPP224 does imply that the sample clock is the same as the clock from the incoming data stream.

But in my situation I have a clock that is 7-8 times faster.

Please clarify ...

Thank you.

Rgds André

Hi André, I'd probably do it like this. Have a 7 bit counter that counts modulo 125 on your 125MHz clock. Call it 'counter'. If you get a transition on the data (which you've already sampled into the 125MHz clock domain, right?), reset it to 0. When the counter says

4 | 12 | 19 | 27 | 35 | 43 | 51 | 58 | 66 | 74| 82 | 90 | 97 | 105 | 113 | 121 sample the data. (I'd check those numbers for yourself, my arithmetic isn't what it used to be..) Of course with half the bits there will be a transition so you'll only be sampling at big counts if you get a lot of consecutive equal bits. Remember if the count gets to 124 the next count should be 0, i.e. modulo 125. Have fun, Syms.

Here is a circuit that generates 16 MHz from a 125 MHz clock: Use a Xilinx DCM with simultaneous multiply by 16 and division by 25. That gives you 80 MHz, which might be convenient for 5x oversampling. Peter Alfke, Xilinx Applications

Please reread my posting. The idea is as clear as I have time to make it.

-- Mike Treseler

The 16MHz stands for an external data stream. I have no direct clock recovery of that data stream. The 125MHz comes from a PLL. The input clock of the PLL is not the source clock of the data stream.

I will think about your suggestions.

Thank you to all.

Rgds André