[SOLVED] 31-Stage Ring Oscillator

Status
Not open for further replies.

deltaDG

Newbie level 2
Joined
Nov 17, 2014
Messages
2
Helped
0
Reputation
0
Reaction score
0
Trophy points
1
Activity points
26
Hi,

I'm learning IC design and I'm bit confused about ring oscillator. I look at the "Wafer Acceptance Tests" data found on MOSIS website and it seems to me that they always use 31 stage to characterize the ring oscillator parameters. For example, in a TSMC 0.18um process, the MOSIS file shows this

Ring Oscillator Freq.
D1024_THK (31-stg,3.3V) 302.91 MHz
DIV1024 (31-stg,1.8V) 377.13 MHz

My questions are

1. What is special about 31 stage. I think ring oscillator can be any odd number of stages, right? Why must it be 31? Why can't it be 3, 5, 7, or 29, 33? Is there something special about 31 stage from a parameter extraction / characterization point-of-view?

2. What are the transistor ratio of the inverter MOSIS use to make their ring oscillator? Are they minimum-sized or they are something else?

This is my first post to this forum. I'm sorry if I don't put things in the right format. Please advise me. Thank you very much!

Best regards,
DG
 


Nothing special, just a (sort of) standard to compare various processes. See this tutorial, 1st page, last but one paragraph.

2. What are the transistor ratio of the inverter MOSIS use to make their ring oscillator? Are they minimum-sized or they are something else?

For this case (frequency comparison) they are minimum-sized.
 
More stages can make it easier to measure, especially in
a high frequency technology it can be hard to get a good
enough quality power, ground, output connection-set to
get the real answer (bond wire inductance etc. can mess
with the on-chip rail span especially when the output is
heavily loaded, like even a 'scope probe). More stages
at lower frequency works better and only costs area.

I personally tend to make larger rings on the rare occasions
I need to to characterization vehicles. Most often I have
done 127-stage ROs.

You want to make it a prime number to prevent higher
order modes in the ring (3, 5, 7...) which can give you
kooky frequency results. You want only the fundamental
to be viable.
 
Status
Not open for further replies.
Cookies are required to use this site. You must accept them to continue using the site. Learn more…