The DLP from Texas Instruments...

For the switches that route signals in an FPGA, ideally, one wants something that does not delay signals, even if it takes longer to program than ordinary memory.

So the gate which determines where a signal should go should not be symmetric in terms of the time taken for the input to affect the output; the control input can be slow, but the signal input should not be delayed at all.

Electromechanical relays have this characteristic; the magnet takes time to move the contacts, but the signal that flows through the contacts, if they are closed, just flows through conductors and does not actuate anything - and so, is not delayed.

Of course they're not very practical on an FPGA chip. Or are they?

Texas Instruments' DLP chips, used in projection TVs, involve moving little pieces of metal around on chips. On a small scale, surface tension and adhesion are big problems, so this may be difficult... but maybe it is possible to build an FPGA whose programmable routing to the logic elements is made from a kind of electromechanical relay!

John Savard

Reply to
Loading thread data ...

Not sure where you are going with this but I can hazard a guess.

I have read of optical switches that route light beams using essentially the same idea as DLP, it all just smoke and mirrors, I mean electro static controlled mirrors or MEMs devices. Have no clue what that has to do with FPGAs, though unless you think of an optical switch as being a logic gate. Mind you DLPs are not that big I mean relatively few pixels barely even 1M. An FPGA no doubt has many more SRAM controlled switches 1 or 2 orders more. Also DLP mirrors are not analog, they use PWM IIRC to simulate analog levels so I'd guess an optical steered mirror FPGA switch would be binary and far bigger than an FPGA route switch. Also its taken TI literally 2 decades to make this thing producible.

BTW you do know something about mos switches too, at least for very small interconnects the signal can propagate as an analog signal through a mos pass gate, while the logic delay for switching proper on/off is soso, the transit time through a mos device is an order less. Problem is it must be rebuffered fairly often.

I also imagine TI has most of the patents on DLP like technology too.

johnjakson at usa dot com

Reply to


The micro-machanical structures are huge. They are hundreds, if not tens of thousands of times bigger than the transistors used in the FPGA fabric.


Reply to

A trip to Sears or other high end TV dept store you may see a DLP chip on a TI info card. Probably the only chip most ordinary folks may see.

Its about 1cm by 2cm, The cells are about 14u by 14u for 1M pixels. Usually RGB done by color wheel from 1 array. Cells toggle at 5KHz enough to fake 10b gray level.

Old news, SRAM cells at about 1u sq so about 200 or more smaller, sram to sram.

Quite a few papers at (TI site) and applications beside usual projectors but very little on optical networking use of DLP from TI

Looks like TI has a 2M pixel device for 1080p HDTV coming in July05 products, I want one.



Reply to

One of the reasons of a lot of research into photonic crystals as a replacement for mems(micro electro mechanical) devices.

Supposed to be the next step towards optical processors as well as replacement for current optic fibre materials / do away with non optical amplifiers in present optical fibre systems

formatting link

Reply to
Alex Gibson

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.