Gray code in asynchronous FIFO design

Why do we use Gray code in asynchronous FIFO? (I am aware that only
ONE bit changes in Gray code at a time, but what is the advantage of
that?)

On Apr 4, 2:57 pm, "anand" <writean...@gmail.com> wrote:
> Why do we use Gray code in asynchronous FIFO? (I am aware that only
> ONE bit changes in Gray code at a time, but what is the advantage of
> that?)


This sounds suspiciously like a homework problem, but I'll give you a
clue.  Suppose you wanted to sample the write address pointer using
the read clock?  What would happen if the address were changing just
as you sampled it?

On Apr 4, 12:19 pm, "Gabor" <g...@alacron.com> wrote:
> On Apr 4, 2:57 pm, "anand" <writean...@gmail.com> wrote:
>
> > Why do we use Gray code in asynchronous FIFO? (I am aware that only
> > ONE bit changes in Gray code at a time, but what is the advantage of
> > that?)
>
> This sounds suspiciously like a homework problem, but I'll give you a
> clue.  Suppose you wanted to sample the write address pointer using
> the read clock?  What would happen if the address were changing just
> as you sampled it?

Well the setup time would not be met, therefore, metastability will
occur?
No, this is not a HW problem, rather, I am trying to prepare for an
interview ;-)

"anand" <writeanand@gmail.com> wrote in message 
news:1175720120.262865.103380@q75g2000hsh.googlegroups.com...
> On Apr 4, 12:19 pm, "Gabor" <g...@alacron.com> wrote:
>> On Apr 4, 2:57 pm, "anand" <writean...@gmail.com> wrote:
>>
>> > Why do we use Gray code in asynchronous FIFO? (I am aware that only
>> > ONE bit changes in Gray code at a time, but what is the advantage of
>> > that?)
>>
>> This sounds suspiciously like a homework problem, but I'll give you a
>> clue.  Suppose you wanted to sample the write address pointer using
>> the read clock?  What would happen if the address were changing just
>> as you sampled it?
>
> Well the setup time would not be met, therefore, metastability will
> occur?
> No, this is not a HW problem, rather, I am trying to prepare for an
> interview ;-)

I almost think it's unfair to give you information if the new comany wants 
you for your knowledge and ability.  But perhaps the ability to ask the 
professionals in a newsgroup is underrated.

Keep in mind that propagation delays aren't the same for all sources and all 
destinations.  When you sample the write address from the read domain, there 
will be an area of uncertainty where the bits are transitioning between the 
previous and next values.  The metastability windows in modern devices are 
extremely slow; I think Peter Alfke's numbers from one of the most recent 
high-end families was sub-femtosecond.  Darn, that's small!  So 
Metastability isn't an issue, the uncertainty region between values is the 
issue.  If you're only uncertain about one bit, it doesn't matter if you 
have the previous or next value - both are valid for your "on the edge" 
calculation.  What isn't valid is getting a pointer that's off by 2, 4, or 8 
from either of the two valid values you're trying to sample.

Now, do you know how to properly pipeline?  How about state machines: can 
you design a soda dispenser state machine to count change?  And how many gas 
stations would you estimate are in the US (please explain your reasoning)? 
If you had a belt tight around the equator of the earth, how much length do 
you figure you'd need to add to raise the belt to an altitude of 1 foot 
across the entire distance and why?

Good luck with your interviewing. 

Quick typo clarification.  I meant small, not slow.

"The metastability windows in modern devices are extremely small; I think 
Peter Alfke's numbers from one of the most recent high-end families was 
sub-femtosecond."

I hate it when that happens.

"John_H" <newsgroup@johnhandwork.com> wrote in message 
news:13185gkp47lpq2c@corp.supernews.com...
> "anand" <writeanand@gmail.com> wrote in message 
> news:1175720120.262865.103380@q75g2000hsh.googlegroups.com...
>> On Apr 4, 12:19 pm, "Gabor" <g...@alacron.com> wrote:
>>> On Apr 4, 2:57 pm, "anand" <writean...@gmail.com> wrote:
>>>
>>> > Why do we use Gray code in asynchronous FIFO? (I am aware that only
>>> > ONE bit changes in Gray code at a time, but what is the advantage of
>>> > that?)
>>>
>>> This sounds suspiciously like a homework problem, but I'll give you a
>>> clue.  Suppose you wanted to sample the write address pointer using
>>> the read clock?  What would happen if the address were changing just
>>> as you sampled it?
>>
>> Well the setup time would not be met, therefore, metastability will
>> occur?
>> No, this is not a HW problem, rather, I am trying to prepare for an
>> interview ;-)
>
> I almost think it's unfair to give you information if the new comany wants 
> you for your knowledge and ability.  But perhaps the ability to ask the 
> professionals in a newsgroup is underrated.
>
> Keep in mind that propagation delays aren't the same for all sources and 
> all destinations.  When you sample the write address from the read domain, 
> there will be an area of uncertainty where the bits are transitioning 
> between the previous and next values.  The metastability windows in modern 
> devices are extremely slow; I think Peter Alfke's numbers from one of the 
> most recent high-end families was sub-femtosecond.  Darn, that's small! 
> So Metastability isn't an issue, the uncertainty region between values is 
> the issue.  If you're only uncertain about one bit, it doesn't matter if 
> you have the previous or next value - both are valid for your "on the 
> edge" calculation.  What isn't valid is getting a pointer that's off by 2, 
> 4, or 8 from either of the two valid values you're trying to sample.
>
> Now, do you know how to properly pipeline?  How about state machines: can 
> you design a soda dispenser state machine to count change?  And how many 
> gas stations would you estimate are in the US (please explain your 
> reasoning)? If you had a belt tight around the equator of the earth, how 
> much length do you figure you'd need to add to raise the belt to an 
> altitude of 1 foot across the entire distance and why?
>
> Good luck with your interviewing.
> 

hi thanks for your response(s).

"I almost think it's unfair to give you information if the new comany
wants
you for your knowledge and ability.  But perhaps the ability to ask
the
professionals in a newsgroup is underrated"

Well I would disagree with the above. Reason is, it is not that I
CANNOT design - I DO have a design background and have done a lot of
it in my last role (barring the async fifo that is!) . It is just that
the last few years I happen to be doing something else, and all I am
doing is brushing up. And for doing so, anyone will certainly use ANY
and ALL resources including groups at their disposal, isnt that
logical? Unless you use something on a daily basis, I dont think it is
realistic to show up at a tech company and be serious about landing an
offer.

Anyways, you have answered my question and that is what counts
Hopefully, my "read_ptr" is not even off by 1 bit :-)
Thanks!!!

>
> >And how many gas stations would you estimate are in the US (please explain your
> > reasoning)?
This is interesting. I dont know the answer, but the reasoning can be
tagged to the distribution of gas station /sq mile or per population
density.


> >If you had a belt tight around the equator of the earth, how
> > much length do you figure you'd need to add to raise the belt to an
> > altitude of 1 foot across the entire distance and why?
The earth's equatorial radius is 3,963.189mi, so elevate the radius by
1 ft, from the top view, it would be like drawing an outer circle with
a 1 ft border. From this we can calculate the new radius and hence the
addition in the radius gives us the new circumference and thus the
addition in length.....what do u think...am I way off???

> 
>>> If you had a belt tight around the equator of the earth, how
>>> much length do you figure you'd need to add to raise the belt to an
>>> altitude of 1 foot across the entire distance and why?
> The earth's equatorial radius is 3,963.189mi, so elevate the radius by
> 1 ft, from the top view, it would be like drawing an outer circle with
> a 1 ft border. From this we can calculate the new radius and hence the
> addition in the radius gives us the new circumference and thus the
> addition in length.....what do u think...am I way off???

You don't need the earth radius.

Let's says the original radius is r and expressed in feet, the
original belt would be 2*pi*r . By adding 1 feet, you now have
a belt of 2*pi*(r+1). So you added 2*pi feet to it no matter what.


	Sylvain

On Apr 5, 12:29 am, Sylvain Munaut <tnt-at-246tNt-
dot-...@youknowwhattodo.com> wrote:
> >>> If you had a belt tight around the equator of the earth, how
> >>> much length do you figure you'd need to add to raise the belt to an
> >>> altitude of 1 foot across the entire distance and why?
> > The earth's equatorial radius is 3,963.189mi, so elevate the radius by
> > 1 ft, from the top view, it would be like drawing an outer circle with
> > a 1 ft border. From this we can calculate the new radius and hence the
> > addition in the radius gives us the new circumference and thus the
> > addition in length.....what do u think...am I way off???
>
> You don't need the earth radius.
>
> Let's says the original radius is r and expressed in feet, the
> original belt would be 2*pi*r . By adding 1 feet, you now have
> a belt of 2*pi*(r+1). So you added 2*pi feet to it no matter what.
>
>         Sylvain

Yeah I knw....I have a habit of thinking out loud....atleast I wasnt
way off
Any idea on the estimation on the number of gas pumps in the US?

"John_H" <newsgroup@johnhandwork.com> wrote in message 
news:13185gkp47lpq2c@corp.supernews.com...
>
> Now, do you know how to properly pipeline?  How about state machines: can 
> you design a soda dispenser state machine to count change?  And how many 
> gas stations would you estimate are in the US (please explain your 
> reasoning)? If you had a belt tight around the equator of the earth, how 
> much length do you figure you'd need to add to raise the belt to an 
> altitude of 1 foot across the entire distance and why?
>
> Good luck with your interviewing.
Hi John,
I used to drive my buddy Kurt insane with these questions. The secret is to 
ask him about two pints into a Friday night Guinness session. After four 
pints he'll let you tell him the answer. The next few pints erase his memory 
so you can do it all over again a couple of weeks later. Eventually he 
bought himself a notebook!
For example, a brother and sister were born on the same day a few minutes 
apart. But they're not twins. How?
Cheers, Syms.