Is there anybody who had ever taped out chip which size was larger than 3cm*1cm?

[nksunmoon]

Is there anybody who had ever taped out chip which size was larger than 3cm*1cm? I don't know is there any restriction to tape out so large chip.Is there anyone who has the experience?

 

[keith1200rs]

I haven't done one that large but have spoken to a fab about it for one project I was doing. There is some limit to do with the mask sizes I think, but they can go above that limit with some technique ("stitching"?). The result is a more expensive mask set (I am not sure if it needed two full mask sets) and more expensive processing. It is probably best to speak to the fab.

Keith.

 

[erikl]

Is there anybody who had ever taped out chip which size was larger than 3cm*1cm? I don't know is there any restriction to tape out so large chip.Is there anyone who has the experience?

Camera producers have: 50MP (Mega Pixel) CMOS sensor chips are 49mm*37mm, see here. Hubble space camera chips are said to be even larger.

 

[nksunmoon]

I also have heard that there is some restrictions when prducing the mask.Thank you for your advice

 

[erikl]

I also have heard that there is some restrictions when producing the mask.

Right. The max. chip size for a single mask set depends on the process lithography (I remember 15mm*15mm for a 180nm process). For larger chips "mask stitching" is needed as Keith mentioned above. This of course is a special and rather expensive service, as 2 or more mask sets and additionally very special lithography steps are needed (the overlapping area of 2 adjacent masks must receive an adapted light blocking pattern, see e.g. this patent).

 

[keith1200rs]

Erikl,

Thanks for the clarification. The chip I was looking at a while ago was 0.8um and the limit was around 25mm I think and we needed 30mm. The cost jumped and pretty much killed the project.

Keith

 

[erikl]

Re: The cost jumped and pretty much killed the project.

... we needed 30mm. The cost jumped and pretty much killed the project.

I guess! :-( Couldn't you try and use a tighter process?

 

[keith1200rs]

I should have mentioned it had a photodiode array on chip and we had to produce a certain optical size - hence the length.

Keith

 

[eld03]

I have taped out sensor 6cm by 6cm Silicon strip detector , my foundary supports contact mask . In your case find the reticule size the fab is using , it is usually 30mm by 30 mm, but this doesnot mean that you cannot beyound it, A mask stiching process is done for designs higher than reticule size. The yeild would not cross 70%

 

[RCinFLA]

You must have a big pocket book. Look up Poisson distribution for yield on such a large chip.

You should average about 5 to 7 good parts out of a 300 mm wafer, in range of about $400 per I.C. (not countng mask charges)

 

[nksunmoon]

You must have a big pocket book. Look up Poisson distribution for yield on such a large chip.

You should average about 5 to 7 good parts out of a 300 mm wafer, in range of about $400 per I.C. (not countng mask charges)

yeah, Maybe I should contact to the fab and ask how much the chip will cost

 

[eld03]

Its Sensor , with six mask process. these are already in place at CMS-CERN experiment.