Finally, selling my old Xilinx/Viewlogic software package
in [Embedded]
|
From: Petter Gustad on 22 Mar 2010 17:40 glen herrmannsfeldt <gah(a)ugcs.caltech.edu> writes: > Some time ago, I was wondering about using Verilog for PC board design. I'm pretty bad when it comes to drawing schematics for PCB's as that was never my primary job. I have also played with the idea of using an HDL or even EDIF as I can write some neat Common Lisp software to throw s-expressions around. The big problem is not to create the netlist, but to interface it to the parts database and the back-end tools which is usually proprietary. One possibility would be to use gEDA or similar open source PCB tools for the back-end work. Petter -- ..sig removed by request.
From: Jonathan Bromley on 22 Mar 2010 19:27 On Mon, 22 Mar 2010 20:38:42 +0000, Peter wrote: >Towards the end of my era of doing complicated logic designs, a very >nice product was from somebody like Altera. It was a FREE VHDL >compiler, crippled to work with just a few low end devices e.g. a >22V10. Cypress WARP, maybe? I designed a good few PAL/GAL devices with it. There was a cut-off-at-the-knees version of the Veribest VHDL simulator, too - can't remember who shipped that. I still have a copy on my machine, but I can't get it to run under XP. -- Jonathan Bromley
From: D Yuniskis on 23 Mar 2010 05:03 Hi Peter, Peter wrote: >> Why? Won't you need to deal with supporting older products? > > I thought about this very hard. > > Of the FPGA designs I did, for those which might come back, I would > never have time to set up all the software (on a PC with a DOS > partition, which I don't have anymore), go up the not insignificant > learning curve on the software, and then be able to make a commercial > project out of it. Understood. I try to keep "live" copies of all my tools so the first issue isn't a problem. But, I will concede that the second issue can be daunting -- especially if it has been any serious amount of time since last used (I find moving between different OS's to be a chore even after just a few days! :< Different applications are much worse!) > Two of the designs were perhaps more significant. > > One was a 32-channel sound generator (basically 32 programmable pulse > generators feeding a 32-input OR gate) which was originally done in > 1992, on a full size ISA card, no kidding. About 8 years ago the > customer wanted a PCI version of this card. I said to them I will give > them the whole design free of charge but they will need to find > somebody else to do it. I posted around Usenet looking for an FPGA > programmer in SE UK and one or two turned up and I passed them on but > never heard anything else. The *huge* issue is the commercial risk: > one has to quote a fixed price, which would make it very profitable if > the old tools worked fine, nothing went wrong, and I could still get > the XC3030/3090 TQFP chips (which I probably could). But if something > goes wrong, and e.g. I have to buy new tools, I would lose some high > 4- or low 5-digit sum on it. Yup. I won't even quote upgrades that look like they will venture into The Great Unknown. Experience says that even T&M can be a losing prospect :< > The other was basically a complicated custom UART supporting some > weird characters (122 bits long) with CRCs and all kinds of stuff. > This project was done in 1993 and was hugely profitable, helping me to > establish my manufacturing business (the kind of thing every > consultant should be doing before his hair goes grey and his sandals I don;t recall yours as showing much age! :> Though it *has* been a few years (and I am always amazed at just how *quick* things like this can change ;-) > fall apart :)). To my astonishment the project came up 12 years later, > via a completely different customer (but the same end user) and I > simply bought more of the XC3090 devices (PLCC, seemed to be available Wow! I wouldn't have thought them to still be available! I've an XC4xxx in a design that I had abandoned figuring it would be impossible to get... > OK) and to my amazement it worked. I first sent a few devices to the > customer (who was pretty smart) and he put them into his old boards, > and it all worked, so evidently Xilinx did not change the silicon to > make it much faster (and break my timing, which was done according to > the *then* Xilinx guidelines which said it was OK to use local > interconnect (with some 'max skew = 2ns or whatever' parameter on the > wire) for clocks because the interconnects were much faster than the D Ah! > to Q propagation delays). The customer may buy again but I just need > to source some more of the old devices. He won't need a logic > redesign. > > The main project I did with the tools was maybe 1 man year spent on > prototyping various evolutions of a very low power ASIC, with about > 5000 gate equivalent of logic, a real time clock, etc. This one is not > coming back. So, your attitude is "move on to new tools *if* the need arises"? > These tools will be very valuable to somebody who did designs in those > days, which might come back for modifications. The devices are for the > most part available; I checked the XC3090 etc about 2 years ago and > Xiling were still selling them, and there are bound to be plenty on > the surplus market which these days is huge.
From: Gerhard Hoffmann on 23 Mar 2010 19:47 rickman wrote: > On Mar 22, 1:46 pm, Jim Stewart <jstew...(a)jkmicro.com> wrote: >> rickman wrote: >> >> snip.. >> >>> I wonder how long it will be before we give up on schematics >>> altogether and just write pin lists or net lists? >> After I'm retired I hope. A well drawn schematic is a >> thing of beauty, helping techs to troubleshoot and >> customers to understand a product. >> >> You could also ask when mechanical engineers will stop >> using fab drawings and just send the data as G codes. > > I guess if you have an analog design with a lot of small components a > schematic is good, but for many digital designs there is almost no > point. Look inside any number of modern products and you see one, two > or three large IC packages with lots of traces between them and few > smaller components. The schematics for these products are horrendous, > often much less clear than a simple pin list. Its not that they were > drawn badly, but that it is impossible to draw a 300+ pin part with > much utility. Even by breaking the part into sections it still ends > up being a pin list with a box around it. > > I'll grant that analog designs can gain from schematic, but many of > the digital ones are pointless when drawn. I recently played with the Altium Designer, seems to be quite OK, and I exported a low noise amplifier in VHDL for the fun of it: entity MAT02 is port ( base: inout something; emitter: inout something; collector: inout something ); end entity MAT02; ............ q1: mat02 port map ( base => input, emitter => tail_current_source, collector => left_c ) was quite pointless for an analog design. But then you can spice that amplifier, at least from the original circuit. (cited from memory under influence of some nice Rioja :-) ) OTOH you can easily bridge the gap between a Xilinx user constraint file and the things on the board that are connected to the FPGA. I like it. Maybe I'll buy it. Gerhard
From: Gerhard Hoffmann on 24 Mar 2010 17:34
Peter wrote: > That was another unfortunate product - their windoze version could not > properly import DOS Orcad schematics (really clever). DOS Orcad was a great improvement on its successors. (with kind regards to Tony Hoare) >> We ran 4 Compaq '286s overnight to possibly get a working 30[249]0 the next morning. > > What is that? XC-3020, 3040, 3090. (regular expressions) Ooops, wasn't that a 3042, really? Gerhard |