1
\$\begingroup\$

This question already has an answer here:

I need a chip like the esp32 with an output channel thats at least 48kHz. Also, it needs to be able to pair and connect to another esp32 over bluetooth, kind of like Airpods.

Basically, if I want to get a custom IC made, what so i do? What kind of company do I contact or who do i talk to, assuming I have the money (I dont but there's a reasonable chance I convince someone it's a worthy investment)? Any clue?

\$\endgroup\$

marked as duplicate by Tom Carpenter, DerStrom8, Transistor, Marcus Müller, Ale..chenski Jun 24 '18 at 0:58

This question has been asked before and already has an answer. If those answers do not fully address your question, please ask a new question.

  • 3
    \$\begingroup\$ Unless you have >£100k to play with, it's a non-starter. \$\endgroup\$ – Tom Carpenter Jun 23 '18 at 22:29
  • 2
    \$\begingroup\$ On top of that, there would be more than a year development time and costs. \$\endgroup\$ – Tom Carpenter Jun 23 '18 at 22:30
  • 1
    \$\begingroup\$ Googling "custom ic" finds a few companies straight up (no idea if any of them are any good). \$\endgroup\$ – Tom Carpenter Jun 23 '18 at 22:32
  • 5
    \$\begingroup\$ I'm not sure what "an output channel thats at least 48kHz" means, but it seems to me that there are plenty of programmable chips already on the market that can be made to do what you want. You should build your prototype using one of them, and only if the product is successful should you think about custom silicon. \$\endgroup\$ – Dave Tweed Jun 23 '18 at 22:34
  • 2
    \$\begingroup\$ @Kevin It’s both of those ranges. It depends on many variables. \$\endgroup\$ – Blair Fonville Jun 23 '18 at 22:35
5
\$\begingroup\$

It would be very wise to follow the advice of one David Tweed who suggested to you to get your product first prototyped and built using off the shelf chips and programmable FPGAs. This is how to cost effectively get all your specialized IP perfected and your software developed and field tested. Along the way as you do this also make sure that you develop concepts for testing and vetting the technology.

If the market likes your product and starts to suck up more than you can manufacture then the time may be right to start considering custom silicon. There are some pretty key numbers to consider however.

  1. One of these of these is the sales volume because it is rare that custom silicon can be justified unless you are manufacturing 100's of thousands or millions of units.
  2. Another number is the product price versus the cost of goods and labor. If the profit margin in your product is especially high you may be able to justify the custom silicon at the 10's of thousands build volumes.
  3. The third important number is product longevity. How long would the product sales volume hold up once you embark on the path to get custom silicon developed? You could find that after a huge investment that some other technology comes along and wipes out your market place almost over night.

If your product ideas are incredibly innovative and everything goes well for you, and I certainly hope that it does, then you do need to be prepared for taking silicon from a foundry and testing and validating that the chip works correctly. If you developed test and qualification strategies for your technology early on this will be a huge benefit to you at the time you get ready to turn out a spin of your product with the custom chip in it.

There are multiple ASIC design houses that would be more than happy to help you spend your money doing your custom chip design. However you may also want to look directly to chip foundries that also provide a design service. They are in the best position to most effectively leverage IP blocks that they have at their disposal for use in your design.

\$\endgroup\$
1
\$\begingroup\$

If you want to duplicate and extend the ESP32 functionality then you could start with Tensilica ...they own the Xtensa processor IP and will happily help to provide you with your own SOC design.

Expect at least a 1 year cycle time, and if you have any bugs, double that.

A more cost effective way to approach your problem might be to use a programable device such as the Cypress PSOC 4 series with inbuilt BLE. These are not as cheap as the ESP32, but still reasonable.

If the ESP32 meets your requirements other than your 48kHz PWM, then consider simply using that device with a small PSOC 4 as an I2C peripheral device.

I'm assuming your 48kHz problem is with a PWM. The PSOC 4 meets this easily.

\$\endgroup\$
  • \$\begingroup\$ Thanks. the 48kHz PWM is the problem, as you correctly inferred. \$\endgroup\$ – Kevin Jun 24 '18 at 15:23

Not the answer you're looking for? Browse other questions tagged or ask your own question.