The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design >> Analog Design >> Current mirrors (Number of fingers, current through each finger does it matter?)
https://designers-guide.org/forum/YaBB.pl?num=1438889350

Message started by curious on Aug 6^th, 2015, 12:29pm

Title: Current mirrors (Number of fingers, current through each finger does it matter?)
Post by curious on Aug 6^th, 2015, 12:29pm

Hi all,

If I have designed current mirrors with same ratio but the current that I am mirroring is small that it pushes the transistor close to subthreshold region.

Question is I draw the two mirroring transistors with multiple fingers, each finger will have a very less current, how will it affect?. For example let's say mirror current is 1000nA and there are 10 fingers, each finger will have 100nA. Does this affect the mirroring action?

I know that they are biased closed to the threshold voltage of transistor, the Vth-mismatch will affect it alot, apart from this is there any concern having a transistor operating at 100nA.

Regards
ch

Title: Re: Current mirrors (Number of fingers, current through each finger does it matter?)
Post by loose-electron on Aug 9^th, 2015, 11:20am

Check to see if the semiconductor simulation model was done using currents down to that level. Many models do not test and model at low currents.

Title: Re: Current mirrors (Number of fingers, current through each finger does it matter?)
Post by rfmagic on Aug 9^th, 2015, 12:41pm

It really depends on the process you are working on. In many CMOS process nodes the leakage current is of the same order as you mention in your question when the channel length is approaching the minimum. In that case you have to make sure that the device size is chosen very carefully.

Title: Re: Current mirrors (Number of fingers, current through each finger does it matter?)
Post by Tako on Aug 13^th, 2015, 3:27am

loose-electron wrote on Aug 9^th, 2015, 11:20am:

Check to see if the semiconductor simulation model was done using currents down to that level. Many models do not test and model at low currents.

What are low currents values exactly? Is it <0.5 uA or < 1 uA or <10 uA ?

If I remember correctly, I met with a lot of IEEE articles where a current through single transistor is ~2.5 uA or 5 uA.

@curious
Wouldn't it be better to use just one finger with 1 uA current flowing through it?

Title: Re: Current mirrors (Number of fingers, current through each finger does it matter?)
Post by RobG on Aug 16^th, 2015, 7:35am

curious wrote on Aug 6^th, 2015, 12:29pm:

Not really at those current levels - you have half the current but also half the width so it is basically a wash. At higher current levels the gate and source/drain resistances can be an issue.

Quote:

I know that they are biased closed to the threshold voltage of transistor, the Vth-mismatch will affect it alot, apart from this is there any concern having a transistor operating at 100nA.

As Tako said, you have to consider leakage. Make the widths as small as possible.

If it is a static mirror then you should be able to take it out of subthreshold by making the device longer. 100nA isn't that low of a current for a properly sized device.

Title: Re: Current mirrors (Number of fingers, current through each finger does it matter?)
Post by loose-electron on Aug 18^th, 2015, 12:56am

Tako wrote on Aug 13^th, 2015, 3:27am:

loose-electron wrote on Aug 9^th, 2015, 11:20am:

Check to see if the semiconductor simulation model was done using currents down to that level. Many models do not test and model at low currents.

If you want to run at a particular current range, make sure that the model was developed with data in that current range.

Title: Re: Current mirrors (Number of fingers, current through each finger does it matter?)
Post by Tako on Aug 21^st, 2015, 8:00am

loose-electron wrote on Aug 18^th, 2015, 12:56am:

Tako wrote on Aug 13^th, 2015, 3:27am:

loose-electron wrote on Aug 9^th, 2015, 11:20am:

Check to see if the semiconductor simulation model was done using currents down to that level. Many models do not test and model at low currents.

If you want to run at a particular current range, make sure that the model was developed with data in that current range.

I see. I was curious, because I have never come across a minimum current remark in foundry documents. So, we should always ask foundry what current range the model support?

Title: Re: Current mirrors (Number of fingers, current through each finger does it matter?)
Post by loose-electron on Aug 21^st, 2015, 11:02pm

Tako wrote on Aug 21^st, 2015, 8:00am:

loose-electron wrote on Aug 18^th, 2015, 12:56am:

Tako wrote on Aug 13^th, 2015, 3:27am:

loose-electron wrote on Aug 9^th, 2015, 11:20am:

Check to see if the semiconductor simulation model was done using currents down to that level. Many models do not test and model at low currents.

If you want to run at a particular current range, make sure that the model was developed with data in that current range.

I see. I was curious, because I have never come across a minimum current remark in foundry documents. So, we should always ask foundry what current range the model support?

Yes! I have seen people run models where the transistor empirical data was done with much larger currents.

Here read this:

http://chipdesignmag.com/display.php?articleId=438

Wrote that back in 2006 for Chip Design

Title: Re: Current mirrors (Number of fingers, current through each finger does it matter?)
Post by Tako on Aug 22^nd, 2015, 11:20am

I remember that a foundry shares its empirical data in one of its document for a chosen technology. I'll check currents there on Monday.

Very interesting article! Thank You Jerry aka loose-electron! :)

Title: Re: Current mirrors (Number of fingers, current through each finger does it matter?)
Post by Tako on Aug 24^th, 2015, 1:26am

I checked a foundry document. There are verification plots where model is compared to a measured data. I think this is what I was looking for, because a foundry rather does not give current range directly in for example a table, writing:
minimum supported current by a model: xxx nA
maximum supported current by a model: yyy mA
?

Title: Re: Current mirrors (Number of fingers, current through each finger does it matter?)
Post by loose-electron on Aug 27^th, 2015, 11:55pm

Tako wrote on Aug 24^th, 2015, 1:26am:

That sounds correct. If the information is uncear contact the foundry and ask them to answer your questions.