I'm getting a very small (test) specialized wafer run done (think MEMS or something along those lines). Probably on 100mm wafers, probably silicon.
Any rough rules of thumb for what the incremental process cost of running additional wafers will be? There's mask cost, which could be considered fixed, and presumably some setup for each process stage and each run, then a variable cost dependent on how many wafers are run in each batch (substrate cost would figure into the latter).
Thanks for any info.
Best regards, Spehro Pefhany
--
"it's the network..." "The Journey is the reward"
speff@interlog.com Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog Info for designers: http://www.speff.com
Quite variable.. area (and also resolution) make a huge difference. Of course you need one mask for every layer, so it's * n. Apparently for state-of-the art resolution and wafer size, it's in the $millions for a set. My test will not cost nearly that much. NRE for a mask ROM (single layer customized) used to be in the $1-2K range, IIRC.
Not sure how 'custom' those really are.
Best regards, Spehro Pefhany
--
"it's the network..." "The Journey is the reward"
speff@interlog.com Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog Info for designers: http://www.speff.com
It depends on the processes, layers, thickness and geometries. Without knowing the exact requirements, it's difficult to estimate fixed cost, let alone incremental cost. If you tell me more about your device (perhaps NDA), I can tell you more about the cost.
"Done" ?? Where? Have you looked into MOSIS? They do MPW's (multi-project wafers) for almost every foundry in the world. And they will price out all of your questions.
...Jim Thompson
--
| James E.Thompson, CTO | mens |
| Analog Innovations, Inc. | et |
| Analog/Mixed-Signal ASIC's and Discrete Systems | manus |
| Phoenix, Arizona 85048 Skype: Contacts Only | |
| Voice:(480)460-2350 Fax: Available upon request | Brass Rat |
| E-mail Icon at http://www.analog-innovations.com | 1962 |
Remember: Once you go over the hill, you pick up speed
Generally the process is set up so that you are not tweaking anything along the line for a particular product, so there isn't a setup for each stage other than loading the mask. I believe ion implanters are the exception here since the implanter can be shared between processes, or the company uses an outside service for implants. Many a water run has been screwed up by incorrectly set up implants.
At some point of scale, the process line does get tweaked for the product, but that would be a case where the process line is dedicated to one product. And not every manufacturer does this.
I have to wonder about the "maybe silicon" comment. You request is quite vague. But you would be a fool not to get the production cost before you even start the development phase, so the whole question is dubious.
There are other elements in the cost of the final product. Packaging low volume chips won't be cheap. Is your lead frame a custom? Stamped or etched? Do you need backlap?
If you really are making a MEMS, does the fab have some sort of item in the test pattern that will correlate to the viability of your structure? This probably needs some more explanation. When you process wafers, there are test patterns on the wafer. The devices on the test pattern have limits. You get the right to refuse buying the wafer if the test criteria isn't met. The goal here is to isolate whose fault is it that the product doesn't yield. Now if there is no MEMS on the test pattern, then you have less assurance that the wafers will yield, and you have no recourse in refusing them.
Yup. There's really only one or two foundries in the world wot have a handle on this stuff, and not only can't I talk about it, but it's so far off the chart in several ways that it wouldn't help.
But from a process pov it isn't much different from making a 100mm wafer full of 555s, in a run of 1-5 wafers.
So is making a 'run' of 1 wafer vs. 5 wafers (ignoring substrate cost) double the work? 25% more?
Best regards, Spehro Pefhany
--
"it's the network..." "The Journey is the reward"
speff@interlog.com Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog Info for designers: http://www.speff.com
MEMS growing/etching/trimming is anything but standard, other than the equipments and clean rooms.
Even with NDA, the fabs won't disclose what they do for other customers. But for 20K, they will guide us away from mistakes.
It depends, growing/etching may be 25% to 50% more. Trimming may be
125% more. The reason MEMS is popular with 100mm is that some of these order equipments can't handle more than 1 or 2 wafers. So, they are dirt cheap, relative to the 150mm or 200mm equipments. =A0
You realize an implanter is about $4 million. I've worked at semis that used outside implanters.
formatting link
One of a handfull that do implants.
If the original poster is going to a fab, the fab is not going to tweak the process. With a fab, what you see is what you get. If you need something custom, that is a big deal. I've worked at fabless semis where we set up our own tweaks, but we are talking serious bucks.
Trust me, you never make a one wafer run. Shit happens, especially on prototypes. Assuming you are not doing a multichip, the cost of tooling will be more significant than the wafer cost.
Well, we get some .5 uM mixed-signal chips made through MOSIS for about $12,000 for a chip of about 3 x 5 mm. This is on the old AMI CMOS process, now apparently owned by ON Semi. For that price, we get 40 pieces, and that is a mixed run, combined with other chips from other groups. They are U-test-em, just singulated and packaged.
The price goes up almost linearly to about 200 parts, then it starts to level off dramatically. We've never had more than a few hundred parts made at a time.
I guess English isn't your first language. Let me try this again. When you work at a fabless semi, you deal with an outside fab. You can, if you have the personnel in-house, set up a process in that outside fab. Of course, this takes money and the clout that goes with it. You may have to buy some gear for that outside fab. A sputterer for thin film is a prime example since very few fabs can do thin film or even require it.
The clean room is step number one. You don't have a fab without a clean room. That is money you are required to spend if you want to be in the business. However, you can have a fab that does some steps, such as the ion implantation, out of the fab. The fact these outside implant services exist is evidence some companies use them.
This is all part of the Silicon Valley culture. Most of these outside companies doing semi services are start-ups from employees in the valley. Implanters, fib, backlap, FA, etc. You simply can't thow a million here and there and control your cash flow, nor can a small company keep expensive fab items utilized 100%. Bean counters like to expense. They don't like to buy and then depreciate.
Semis startups are not what the venture market wants these days, or even for the last decade.. Venture capitalists don't dump piles of money on what they consider old technology. Cash flow is a big deal, and the less venture you take, the more the employees (well founders) make. So you start fabless, then buy an old fab or build one if you have volume, but even that has to be done on the cheap.
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.