• Please review our updated Terms and Rules here

ISA USB adapter

Had anyone ran across this yet?

Had anyone ran across this yet?

I've been looking for a USB to ISA adapter and came across this thread on Google. I wanted to know if anyone had come across this site yet and checked it out? it looks as though it could use some serious improvement as most of the cards it lists are out of stock. Curious what a bunch of guru's such as yoruselves think of the products listed. Thanks.

http://www.arstech.com/
 
ARSTech makes ISA (and PCI) bus expanders that run as a USB slave device for machines that don't have either type slot. Not much help. What people want is the other way around.

I'm actually starting to tackle this project. However it's a spare time thing and that may not be much as I'm probably switching jobs in 2-3 weeks. My original scope was pretty pie in the sky - embedded MCU running Linux with a full USB stack supporting various USB classes - audio, HID, MSI, 802.11 network - being emulated as popular legacy equivalents - SB/GUS, game port/serial mouse, IDE disk/cdrom, NE2K/3C309. And the original scope also included adding a Super I/O chip for native floppy, 2S, 1P (bi, ECP/EPP), 1G, MPU-401, etc, an AC'97 codec, plus lower and XMS memory fill-in to 16MB of zero wait state RAM, and a real-time clock. (see what I mean about feature creep?)

However I've recently scaled back the scope significantly for round 1 to just USB MSI to IDE disk emulation and possibly CD-ROM support; along with serving up a controller BIOS from an SPI flash chip. It should allow for a 2 layer half height half length 16 bit ISA board (8 bit slot compatible) that supports emulation of basic IDE PIO and DMA modes. And the Vinculum-2 SDK seems robust enough that I can get away with not having to do a generic end-point queuing engine plus add a full USB software stack on the MCU. I can just handle everything at a block level over a simple fifo. My price point on components is now sitting at $~45 in unit quantity 1 (<35 for 25+), the PCBs are at $5-35 depending on final process requirements and quantity, back-plates are still up in the air, I can solder everything with equipment I already have (for a minimal assembly comp)... with the ultimate goal of a sub-100 dollar price point.

I've designed an ISA carrier board for a Lattice CPLD breakout board, FPGA eval board, and a Vinculum-II eval module and ordered the full parts kit. Should be here Wednesday so I can fit parts to to-scale pads before I order PCBs on Sat/Sun. The CPLD primarily is for a ISA to LPC bridge and has all 80+ ISA signals buffered into it with FETs (the idea was to play around as a busmaster eventually). It also has an additional 512 SRAM and 4MB SPI flash for an optional 8-bit soft-core MCU and 32MB SDR DRAM to experiment with XMS/EMS fill-in. The FPGA eval board has similar SRAM/Flash on-board plus additional flash on the carrier board, EEPROM, RTC, Temp sensor, and Super-I/O chip for evaluation. The the Vinculum 2 eval module with a high speed FIFO interface to the FPGA eval, a SPI interface to the CPLD, JTAG and tri-state IOB control to both for turn-key USB stick firmware flashing, and a MAX232 pinned to a header for debug.

I'll see how far this goes. I'm still a pretty inexperienced HDL coder but basic ISA decoding is pretty straight forward; even DMA. The plan was to build the eval platform versatile enough that I can accomplished my scaled down goals while still opening the door to running Linux on a 32-bit MCU core in the FPGA, writing a USB host driver for the Vinculum-2 over fifo, and going full bore with part or all of the original plans eventually. It's more of a 'is this component of the final design feasible and how much impact is it.' There isn't enough gate capacity to run everything at once.

There's a 4-week turn on the PCBs for cost mitigation so it will be a month before anything is running for real. Till then it's writing HDL and firmware code and simulation. So there won't me much in the way of project movement for at least a month (and that's all dependent on new employment at what is honestly my 'dream job').
 
One thing I forgot to mention is all hardware design, HDL, firmware and host code associated with my work will eventually be GPLv2'd as soon as it is mature enough to put out there with some prototype boards. At least for the simple disk only emulator, the plan is to make it appear 100% compatible to the traditional primary IDE/ATA/ST-506 register interface - locking the I/O addresses to 1f0/3f0, the IRQ to 14, and DMA to 2. Thus (hopefully) eliminating the need for special drivers. I planned on using the ISA PnP enumeration protocol with vendor specific extensions to setup a config register to allow the jumperless configuration of everything else - either using the on-board BIOS ROM image or a DOS config utility.

In all honesty, it will probably be late summer / early fall time frame before the first real prototype design is in copper. And that will be a small run of a half dozen boards or so free to other developers / testers willing to contribute.

The size/family of CPLD I'm currently targeting isn't even shipping yet from Lattice.
 
The more I work through this, the clearer it's becoming the first production design just needs to be a 2"x3" module board with a 40-pin right angle IDE header on one side and a USB type A connector on the other. A USB upgrade mode jumper, CPLD, small SRAM, VC2 USB controller, TTL buffers, linear regulator, USB power supply, XO, and the rest passives in the middle. I made a few other simplifications to the design tonight that altered my BOM significantly so the $100 target price point is looking extremely high. For a stand-alone module, the parts, PCB and assembly cost is now putting it in the $50-$75 range. I'll do my best to get it towards the lower figure.

I'm still going to go forward with the integrated ISA card, but it's looking more like I will have enough I/Os now for a dynamic IRQ and DMA config setting (as well as I/O ports). And the ISA card still adds ROM BIOS via on-board flash.
 
Back
Top