There are many advantages of correcting an IC design with FIB Nano-Surgery, but sometimes there are also problems, and many designers have been put off the technique. Did your last FIB work as you expected? Or was the yield low? or you got results you couldn’t explain?
A quick recap of the reasons to do a FIB edit and how you would go about it, might be helpful.
SCENE: So you’ve taped out your big new design, and 4 months and ₤400,000 /$/€ later, your 1st Silicon devices have arrived for tests to start.
And immediately, there’s a problem.
If you are lucky, then a normal functional test has failed and after a few days of head scratching it becomes clear what the cause is, and what the most likely metal fix should be. If you are unlucky and it’s more subtle, but still a show stopper for customer acceptance, then the fix strategy may be less obvious.
Either way, prompt action is required and there are some difficult decisions to make. Your colleagues, customers and suppliers are all waiting to help get your product to market. Then there is the expense and the delays: to testing, to qualification, to sales and to revenue – which all add up to a serious commercial inconvenience. What ARE your options here?
- Risk a design change because there is a high confidence that the fix is understood? Do you roll the dice (the expensive and time consuming ones) and trust that the fix will work?
- Or spend a week trying to get a few chips fixed using FIB nano-surgery and give yourself some certainty that what you have modeled, will work.
But is it so straightforward? Let’s just Pro/Con this FIB thing for a moment.
Pro – If you can get ‘a few’ working devices within 3 or 4 days, then the pressure may be lifted. A new tape-out can be done with much greater certainty, and a known set of delays can be explained and mitigated for in the production and qualification process. At the same time, the clock can be ticking down, testing can start, demos can be done, boards can be designed and things keep moving. Reliability testing is still delayed, so volume shipments and bookable revenue will be late, but these issues can all be managed.
So it’s obvious – FIB can help you get it 1st time right. It’s fast and inexpensive, it de-risks the whole process and give you working chips now – you just can’t choose any other approach – that’s clear.
Con – BUT it might not work. Then you have lost 2 more weeks, you have an extra bill, more questions to answer, and everyone is still waiting. This is a significant risk and now is the most difficult time to take it.
So how do you decide?
It often comes down to your personal experience, or the advice of the colleagues around you. Did it work last time? What was your confidence in the results? Where did you have it done last time? How many worked? Did any work?
When there was a problem, did you get a clear answer from your FIB partner about why that was? Did the yield improve when they did more?
Unfortunately in a lot of cases, the answers to these questions are NOT positive. The risks are not explained, the yield can be highly variable, and problems are often not understood and certainly not corrected for. There are many examples of designers getting a bad experience and losing confidence in FIB nano-surgery to solve their ‘1st time wrong’ issues.
The fact is that when FIBs started being used to correct design errors, it was easy (1991). IC technology was simple and large, so anything with an electrostatic focusing column and an LMIS (liquid metal ion source) had a great chance of getting a fix done. Based on a high success rate, the expectation developed that getting a chip fixed was a low risk commodity – an ‘off the shelf’ service with consistent quality, based on the technique or the instrument, not the operator (OR so FIB vendors would have you believe).
This meant that as FIB became more available, there was little differentiation to help you choose between providers. And if that was true, then surely it is obvious that you should go to the cheapest provider with the newest machine? These are unfortunately STILL the most commonly asked questions at the start of ANY Nano-Surgery project.
So why are designers being disappointed? Well for the simple reason that these assumptions are no longer true (For about 15 years actually). When you actually sit down and think about it, these expectations are not just obviously wrong or hopelessly out of date, but almost entirely misleading.
The more advanced the device and the package, the more specialised the knowledge you must need to access a node, and the more experienced the operator must be. A highly yielding fix now requires a detailed understanding of how to manipulate FIB technology too. To be able to routinely manage parameters like beam tails, gas flux, end-point detection and preferential grain milling effects requires great skill and an understanding of Physics and Chemistry. You also need to know the IC construction processes, and how each material layer reacts to FIB milling with different current densities and pre-cursors. Finally you need to know how to deliver the right electrical result (not necessarily the ultimate fix, but how to prove it is correct). Every fix now needs a node access strategy, and that doesn’t exist in software.
And this can only happen AFTER you have got into the device, and kept it functional, and preserved it enough to be put into a test socket AFTER it has been modified.
There is yet another dimension to add to this mix, the University or Test House your colleague went to before, probably has the wrong sort of FIB instrument.
There is a big difference between a FIB-SEM for analytical applications which has the SEM Column mounted vertically and a FIB on the side (optimised for digging big holes), and a Nano-Surgery instrument, with the detectors, gas injectors and FIB column (optimised for nano-scale surgical operations) all in top down positions. If you want to know if you have hit the right node, it helps if your surgeon is looking directly into the hole he has cut, and not trying to peer inside over the edge of the chip.
So the experience of your colleagues, (and the reliability of their opinion) might be mostly based on WHO they used for FIB last time (not the process), and that choice may have been based on out-of-date assumptions. Getting the best results from FIB nano-surgery, still requires finding an instrument with good capabilities, but now it has to be the RIGHT instrument, and more importantly, it has to be coupled with the much rarer commodity of the right PERSON – with the right knowledge and experience based skill-set.
In conclusion – high yields and quick answers and all of the other benefits of FIB for IC Nano-surgery are still available, but the question has changed slightly.
Instead of ‘Should we do a FIB to help us out of a very difficult design flaw problem?’, it becomes ‘WHERE shall we do a FIB to help us out of a very difficult design flaw problem?‘
I will be addressing this question in my next article ‘How to choose the right FIB IC Nano-Surgery partner’
Or to put it another way, “Better call NanoScope”
Lloyd Peto – Commercial Director and FIB Technologist