The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Modeling >> Semiconductor Devices >> device local mismatch
https://designers-guide.org/forum/YaBB.pl?num=1517852930

Message started by dog1 on Feb 5^th, 2018, 9:48am

Title: device local mismatch
Post by dog1 on Feb 5^th, 2018, 9:48am

Hello, I'd like to ask if device local mismatch is on the order of +-10% or +-0.1%.

I saw many PDK provide description of mismatch of analog devices (poly cap, resistor and transisor current) on the order of +-10%. For example, one document I have now states around +-10% of transistor current mismatch for 3 sigma, when Vg=Vdd.

However, when I search for PRNU (Photo Response Non Uniformity) in google, I noticed that the PRNU is normally on the order of 0.1%. I attached one table I found with google as an example. As I understand, PRNU is the mismatch of the response from different pixels in a pixel arrary in image sensor. This include many mismatches, such as mismatchs between pixels, mismatch between floating diffusion capacitance, mismatch between readouts chains... How come that the mismatch here is so small compared with the device mismatchs?

Can anybody give me some hint? Thanks.

Chen

Title: Re: device local mismatch
Post by sheldon on Feb 5^th, 2018, 1:11pm

Chen,

It is possible that they are talking about pixel as a circuit not as a
device. At the circuit level, the circuit may compensate for the device
variation.

Sheldon

Title: Re: device local mismatch
Post by Tako on Feb 5^th, 2018, 2:44pm

You should stay to what is claimed by the foundry. In case of doubts ask them directly. As a customer you should have a foundry support.

In foundry's documentations there are many reports for mismatch. You have 3 sigma vs 1/sqrt(A) plots, where A is arena of a given device (transitor, resistor etc.). Hence you may see what is Id or Vth mismatch for given transistor's size.

Title: Re: device local mismatch
Post by dog1 on Feb 6^th, 2018, 3:43am

Hello Sheldon and Tako,

Thanks for your answers.

Sheldon, I think they are indeed talking about pixel as circuit. But I've never seen a image sensor able to compensate for the gain mismatch from the pixel, which is directly related to the mismatch between floating diffusion capacitance. And the PRNU should be even larger than this as it contains also gain mismatch from other devices in the readout chain. That's why I am confused how it can be so low.

Tako, the document I have at hand doesn't say how the 3 sigma is measured, if it is die to die, wafer to wafer, batch to batch, nor if it is local or global on the wafer. I don't have the previous PDKs now. But Seeing from the small PRNU, it seems that the local mismatch should be much smaller than 10% order. Thus I was confused. I will ask the fab as you suggusted.

Thanks very much again for both of your comment.

BR
CHEN

Title: Re: device local mismatch
Post by Tako on Feb 6^th, 2018, 4:05am

I opened the document for the technology I have currently and for Vth mismatch (the way of measuring is described in the document, so you will know whether it is local or global, but this one is local) for a transistor it start form ~30 mV for one sigma and goes as low as single mV.

The same for Ids mismatch which is in %. Starts from ~6% for 1 sigma and goes to 0.X %.

Title: Re: device local mismatch
Post by Geoffrey_Coram on Feb 6^th, 2018, 8:55am

And typically the mismatch is worse for small (eg, minimum-geometry) devices, but probably they use larger devices for the pixel array.

Title: Re: device local mismatch
Post by dog1 on Feb 7^th, 2018, 2:41am

Thanks very much for all your answers.

Geoffrey_Coram, I think pixel is indeed large, often on the order of serveral um. The floating diffusion, however, which is related to the gain of a pixel, is normally smaller. It is not too much largerer from a well capacitor. That's why I was guessing that it should not be orders of magnitude smaller than normal device mismatch. I could be wrong. though. I also noticed that in Tako's reply, 1 sigma can indeed goes to 0.X% for Ids. Maybe this indicates that for um order size device 0.X% order mismatch is normal.

Tako, this is very helpful. Although I still don't understand how PRNU is normally so small, I have confidence that I am comparing apple to apple (local mismatch to local mismatch). I will, as you suggusted, try to find answers from some fab staffs. Maybe they have relavent image sensor knowledge to clear my confusion.

Thanks again.

Title: Re: device local mismatch
Post by Tako on Feb 7^th, 2018, 6:21am

Happy to help you in some way. :)

When you find out something, please share.

Good luck!