grenley@nsc.nsc.com (George Grenley) (11/14/87)
Greetings, O gang of loyal 32kniks. Recently, I posted a follow-up to a gentlemen inquiring about the 32203 DMA chip. I have stirred up quite a lot of discussion. Good. In the posting, I said the part wasn't really recommended for new design. That is not quite accurate. The part works just fine, is not about to be obsoleted, and is recommended for new design, particularly with the 32016 and 32032 CPUs. The '203 is a 24 bit address, 16 bit data DMA controller. It is best suited to use w/ 32016, but will also work with 32032 if you add a couple of extra buffers. The part works at 10 mhz. While it is possible to use it with newer processors such as 32332 and 32532, I think you will find in most cases the CPU can do the transfer faster in s/w, since they are 32 bit, and also higher clock rate. You would want to check your particular application to be sure. That is what I meant when I said not recommended, it isn't that it wouldn't work, it is just that it wouldn't be the best way. (I'm a 532 designer, so I am biased toward horsepower, lots of it!!!). There are some drawbacks to the '203. It is not easy to do scatter gather with it, because the command for the next block has to be at the end of the data block of the last xfer. Also, the part can't DMA to/from 8 bit devices and combine/split 16 bit memory. In other words, when servicing an 8 bit device, memory references will be 8 bit. It does do 16 bit to 16 bit xfers. Call your local sales office. If they don't have a data sheet, call Tony Radi, our apps guy, at (408) 721-6773. If you are having trouble gatting one of our sales offices to respond (see following posting), please call our 32000 series marketing director, Anne Wagner. Her number is (408) 721-5955. Please DO NOT call her for routine stuff, as she has a lot to do. But, if one of our sales offices is not up to snuff, we want to hear about it. In article <317@gen1.UUCP> peter@gen1.UUCP (Peter CAMILLERI) writes: > > I am currently in the proccess of designing a system based around >the 32016 CPU core. As the system is intended for use in an embedded >application, the use of a 32 bit bus is not appropriate. The configuration >we are considering is a CPU, TCU, ICU trio. However the problem is that >this set can draw close to 1 amp, according to published specs. ( One >point here, I admit they are TI specs ) As you can imagine, cost, both >in $ and ma, is important. > > Unfortunatly, being in Toronto, puts us at too much more than arms >length from the source of the data, and repeated calls to the local NSC >branch have yielded nothing. > > Any help will be most appreciated Okay, here ya go: Call the people mentioned above. Tony can provide technical info on current and future (yes, future) CMOS products. If the NSC office in Toronto is being rude, let Anne know. This doesn't mean that the sales office is obligated to drop everything when you call; I've worked in the field and I know how busy it can be. But they should at least send you data sheets. Regards, George Grenley disclaimer: My name is not Charlie Spork. I don't run this place. OK?