Generated Clock & master clock

LET’S MAKE IT SIMPLE

Kunal Ghosh
I get this one occasionally … not particularly about the concept, but about the ways we can create a generated clock
definition. Too many options make it too complex… But you know what, with a handful of images, this topic
becomes way too simple… Let me show how? Start with below simple divide-by-2 circuit and try to define it in all
ways possible

I believe, best way to understand any topic is the ‘graphical way’. So, let me show you how the input and output
wave-forms look like, in below image

If you start with master clock of say 1ns period, the output is expected of 2ns period (frequency divide-by-2). So at
the first rising edge of clock (assuming the initial state of circuit is ‘q’ = 0 and ‘d’ = 1), the data at ‘d’ pin will be
propagated at ‘q’ output (since it’s a rise-edge triggered flop), and ‘q’ becomes ‘1’, ‘q_bar’ becomes ‘0’, and ‘d’ is at ‘0’

VLSISYSTEMDESIGN.COM 1
(‘q_bar’ is connected to ‘d’). This is shown in above image. While, we keep doing this, the state of the circuit changes
at every rise edge of master_clock, and below is what you get

Now that was the digital hardware part of it. Let me tell you a secret of ‘static timing analysis’ of ASIC…We are
‘static’. We need definitions. We understand software language. And we will not change J . You are right in
thinking.. We are adamant on this J
Designer – Ok, so let me speak to you in your language. Let me write a define this new ‘gen_clock’ for you so you
understand it and use it
STA – you better do !!
Seriously, I have worked ~6 years in this area, and believe me, these analysis is very rude
Anyways, coming back, we need to fill up below table so that the analysis engine understands it and passes this
information to further flops clock pins, it might be driving

VLSISYSTEMDESIGN.COM 2
By the look of it, below seems to be most appropriate and valid values for ‘gen_clock’

Though this looks the easiest, there are other ways to define this (not recommended for this simple one). So if we
were to define the gen_clock based on the edges of master clock, below how it will like. We remove the ‘divide-
by’ option and use the edge values of 1,3,5 to define the new clock. This says, that at ‘1’ edge of master_clock, the
first rise edge of gen_clock arrives. At ‘2’ edge of master_clock, the first fall edge of gen_clock arrives, and at ‘3’
edge of ‘master_clock’, the second rise edge arrives.

VLSISYSTEMDESIGN.COM 3
Isn’t that simple…wait a sec, it was simple because I took a simple example. Let me come up with a more complex
circuit, break it down to little pieces and explain how to fill up this table.

Let’s take a divide-by-3 circuit and below is how its waveform at output will look like (assuming a non-50% duty
cycle). The output clock period is 3 times the input clock period and hence, frequency is divided-by-3. (It would be
a great idea to show the interiors of divide-by-3 circuit. Stay with me and I will do that in following posts)

So, now, if we need to write this gen_clock definition, its simple. Just fill out the table and below is what we get

VLSISYSTEMDESIGN.COM 4
This shows a perfect usage of ‘edge’ option. For circuits producing a clock waveform which has non-50% duty
cycle, the edge option can be very well used to define an output clock and can also be propagated throughout the
circuit. It just says, that 1st rise edge of gen_clock arrives at 1st edge of master_clock, 1st fall edge of get_clock
arrives at 5th edge of master_clock and 2nd rise edge (which completes 1 clock cycle) arrives at 7th edge of
master_clock
Pay attention to the words above. I am using ‘rise’ and ‘fall’ words for defining gen_clock (as these words are the
one’s that completes one clock cycle) and I am using just the word ‘edge’ for master_clock and ‘edge’ can be either
rise or fall, doesn’t matter from gen_clock definition perspective
Now, to explore other options of this table, let’s consider another circuit specification, where output is inverted and
divide-by-2 of input waveform.

Let’s assume that, and begin with below simple example of waveform table (by the way, this is a very classic
interview question in semiconductor industries. Also, this technique is used to debug complex clock constraints in
industries)

VLSISYSTEMDESIGN.COM 5
So, the above table, says, the first rise edge of the generated clock arrives at 2nd edge of master clock (remember, it
doesn’t matter whether 2nd edge of master clock is a rise or fall), the first fall edge of generated clock arrives at 4th
edge of master clock and second rise edge comes at 6th edge of master clock and below is how your generated clock
will look like

Let’s put this to test. How about everything else in above table remains the same, but the also the inverted is ‘ticked’,
then the same gen clock will look something like below. Also, there is a trick that can be done to define the below
gen clock. i.e…….(sentence continued after below image)

VLSISYSTEMDESIGN.COM 6
(…sentence continued from above image) the same gen_clock’ can be defined using just the edges, and removing
options of the table. And below is how the generated clock definition will look like for gen_clock’

And finally, let’s consider a more complex table like below, where we have the generated clock edges coming in
middle of clock edges at some ‘t’ time units

VLSISYSTEMDESIGN.COM 7
This is a classic example of how ‘edge’ and ‘shifted edge’ option of generated clock can be used in conjunction.
Let’s the first fall edge is in middle of 2ns and 4ns i.e. at 3ns
So this is how you will define the generated clocks

You will say, the first clock edge of generated clock arrives at 1st edge of master clock, and shifted by 0ns from 1st
edge (Hence you see the first element in ‘shifted edge’ at ‘0’). Next … (and quite important one) … the first fall
edge of generated clock arrives at 2nd edge of master clock, but shifted by 1ns from 2nd edge of master clock. So
2nd edge of master clock is at 2ns and generated clock is shifted by 1ns (i.e. arrives at 3ns) from 2 nd edge. Makes
sense…If not, I would suggest you read the above sentence again. And the second rise edge of generated clock
arrives at 5th edge of master clock. Hence you see the elements (0, 1, 0) in ns in the shifted edge row.

This marks the end of post related to ‘Generated clock and master clock – Lets make it simple’.

VLSISYSTEMDESIGN.COM 8