• Please review our updated Terms and Rules here

XT-FDC project level of interest

Also, have we given up on the DP8473? I'd like to at least give it a try. The main advantage is it supports some unusual modes but is also a hedge against parts obsolescence. Unfortunately, not a great one since it is probably more rare than the PC8477B.
The parts list quotes "DP8473V". Out on the Internet is both DP8473V and DP8473AV, with the DP8473AV appearing (on the surface) to be readily available. I have looked but have yet to establish the difference between the V and AV versions (i.e. if the AV is a substitute for the V version).
 
The parts list quotes "DP8473V". Out on the Internet is both DP8473V and DP8473AV, with the DP8473AV appearing (on the surface) to be readily available. I have looked but have yet to establish the difference between the V and AV versions (i.e. if the AV is a substitute for the V version).

Hi
I don't have a datasheet but will check it out later today. Do the pinouts match? If they do its probably worth a try at least. Some issues like this can really only be settled by trying it in build and test. If it is simply dead or severely brain damaged then so be it and we can drop it. However if it can be made to work we should at least consider it.

Thanks and have a nice day!

Andrew Lynch
 
Hi Chuck! Thanks! What's the best way to do that? I am thinking three 74LS07s.

Also, have we given up on the DP8473? I'd like to at least give it a try. The main advantage is it supports some unusual modes but is also a hedge against parts obsolescence. Unfortunately, not a great one since it is probably more rare than the PC8477B. Are there any other options?

I am feeling good about the prototype board. The community investment in these four PCBs is paying off big time.h

7407s will work just fine--or you could assert /INVERT and use 7406s, but then you'd have to invert the input signals as well (74LS14?).

The DP8473 is a great controller, but it's very scarce and, AFAIK, there are no pin-compatible substitutes. Future Domain, Ultrastor and Adaptec (and probably DTC) used them at some point, but I don't know if they ever made it onto many motherboards, which might account for their scarcity. Oddly, the DIP (2 drive) version is more common than the PLCC. About the only novel use is in writing 128-byte MFM sectored data. I have more fingers on one hand than the number of formats that I've seen that way.

At least that's my take.
 
Andrew, could you adjust your copper pour or traces routing so there are no ground fingers between signal lines?
These are bad at higher frequency because they act as antenna.
Currently checking signal lines if they are centered on the Via's without any offset, this is something that is known to cause problems too.
 
Andrew, could you adjust your copper pour or traces routing so there are no ground fingers between signal lines?
These are bad at higher frequency because they act as antenna.
Currently checking signal lines if they are centered on the Via's without any offset, this is something that is known to cause problems too.

I have to admit that I was a bit surprised by the routing on the PCB. In particular, note that the +5 first point of connection after it leaves the bus is a signal pin. I guess the auto-router doesn't treat power buses specially. If I were doing a hand tape-up of a board (does anyone do this any more?), I'd do the power rails first, then work the signal lines around them.

But I'm old and not up to speed on "modern" stuff. :(
 
I have to admit that I was a bit surprised by the routing on the PCB. In particular, note that the +5 first point of connection after it leaves the bus is a signal pin. I guess the auto-router doesn't treat power buses specially. If I were doing a hand tape-up of a board (does anyone do this any more?), I'd do the power rails first, then work the signal lines around them.

But I'm old and not up to speed on "modern" stuff. :(

Hi Chuck,
The VCC and GND rails are a separate net class (power) and their traces are 24 mil. The rest of the signal traces are the default 8 mil. It is a bit hard to follow the GND traces since they are combined with the GND fill but they are still there. All the traces are routed at the same time but I make sure the power traces don't use vias to the extent possible.

Thanks and have a nice day!

Andrew Lynch
 
I always hand route. I believe Sergey does too. Auto routers just don't consider things like routing through filter/prop-up caps first, reducing ground impedance, or any EMI impact. In general these retro computing projects have good immunity to these problems, but you hit limits once you start going high speed or lower signaling voltages.
 
I always hand route. I believe Sergey does too. Auto routers just don't consider things like routing through filter/prop-up caps first, reducing ground impedance, or any EMI impact. In general these retro computing projects have good immunity to these problems, but you hit limits once you start going high speed or lower signaling voltages.

Hi
Fortunately operating with VCC at 5V at 4-8 MHz and 74LSxxx technology reduces much of the effect on these small project boards. Hand routing can be good but time consuming on larger and/or multiple projects. The high frequency effects really don't kick in strong until 16 MHz or beyond.

FreeRouting.net does a pretty good job if you are willing to let it optimize the board sufficiently. Generally not as good as a hand routed board but at a fraction of the time necessary to walk it through manually. I wouldn't make absolute statements on hand routing vs auto routing though since I've seen plenty of examples good and bad of each kind.

Thanks and have a nice day!

Andrew Lynch
 
Andrew, could you adjust your copper pour or traces routing so there are no ground fingers between signal lines?
These are bad at higher frequency because they act as antenna.
Currently checking signal lines if they are centered on the Via's without any offset, this is something that is known to cause problems too.

Hi
You must be working with some pretty high frequency stuff if that's an issue. Anything in a PC/XT ISA bus will be hard pressed to go above 8 MHz except for some tiny exceptions like local oscillators. Most of the board will be *much* slower. I can adjust the KiCAD fill zone parameters easily enough but has anyone actually seen any issues with it yet?

Thanks and have a nice day!

Andrew Lynch
 
I'm very interested in these comments.

I've been using a ground fill and covering as much as possible (again 2-layer boards). I've not had any signal quality issues - because of the relatively low ISA bus speeds presumably.
 
Andrew It's something my Boss taught me he Designed some OPL 3 stuff back in the mid to late 90's.
He worked for A trend, and did some mother boards and video cards for them.(He passed away last month.)
With ISA Bus you should assume signals up to 16MHz, with the XT bus you should keep in mind some third party boards exist that run up to 12MHz.
Besides that Near the back plane there is this one trace making sharp corners where those can be eliminated.
On U16 between the Pins are copper pour where there shouldn't because they will cause signal interference especially there!
P18 the first FD connector should be rotated 180 degrees so that the ground plane can be continues without disturbing the signal wires.

Design technical we would need to bring down the number of components, since then we regain some board space for some proper routing.

And besides that Andrew in general I design for low speed applications, at least i Used to. :(

PS: Problem area's.
-GND/Copper pour, should be have a be forced to a higher width so they don't create fingers between signal lines.(The FD connectors and U2 and U16)
-Sharp corners in the design these should be evaded whenever possible.
-Not all Via's are connected correctly some even have sharp corners before their connection.(Pin 17 on U4)

Besides the copper pour the 74 logic part is done in a excellent manner you said this was not done by Hand?
 
Last edited:
I can add that the external drive connector (DC37F) works just fine with my external drives. Unless there are any additional updates, I'll start cutting some code.

-----------------------------------------------------------------------

A crazy thought just crossed my mind. One could make an interesting "combo" board by combining the XT-CF logic and the XT-FDC on the same board. The I/O range could even be continuous with the FDC occupying 3x0-3x7 and the CF occupying 3x8-3xF. The same BIOS ROM socket could support both as well. This would give the 5150/5160 crowd what they'd like to have most--a hard disk replacement and a floppy controller capable of high-density.

Opinions?
 
A crazy thought just crossed my mind. One could make an interesting "combo" board by combining the XT-CF logic and the XT-FDC on the same board. The I/O range could even be continuous with the FDC occupying 3x0-3x7 and the CF occupying 3x8-3xF. The same BIOS ROM socket could support both as well. This would give the 5150/5160 crowd what they'd like to have most--a hard disk replacement and a floppy controller capable of high-density.

Opinions?

Yes. I have an NCR with 5 slots and more than 5 cards to stuff into them. Just let me know when to send the PayPal. :)
 
Hi
Would you please refresh my memory on the updates to the board? What I have so far:

1. move the 8" floppy connector to the left
2. move the external floppy connector to the right (need some dimensions)
3. fix the FDC chip select logic bug
?4. there was some discussion about adding 74LS07s on the external floppy connector signals?
?5. eliminate the DP8473

I am not seeing anything that would require a new prototype PCB at the moment. The problems we have so far can be worked around for now and fixed in the "production" boards.

Questions? Comment? Thoughts?

Thanks and have a nice day!

Andrew Lynch
 
A crazy thought just crossed my mind. One could make an interesting "combo" board by combining the XT-CF logic and the XT-FDC on the same board. The I/O range could even be continuous with the FDC occupying 3x0-3x7 and the CF occupying 3x8-3xF. The same BIOS ROM socket could support both as well. This would give the 5150/5160 crowd what they'd like to have most--a hard disk replacement and a floppy controller capable of high-density.

Do you mean with a 40-pin header (allbeit restricted to use with a CF adapter)? I found that CF /CS1 isn't really needed (only used for drive reset on error) so possibly all that's needed is an address decoder (might be able to use some OR gates if you have any spare instead, to get the contiguous IO space mapping) and a couple of resistors. Looking through the specs again this week the CF spec data bus drive is 8mA, the same as the XC9572XL I've been using, so maybe not even a buffer either.
 
Hi Andrew,

Here's what's on my list:

  1. Fix the bit A7 decoding for the I/O port.
  2. I can give you exact positions relative to the ISA edge connector; will that do for the external connector? (I need a reference point)
  3. The second internal drive connector P2 needs to be moved a bit away from the external (DC37) position--otherwise, there's too little clearance if both are installed.
  4. The 50-pin 8" header needs to be moved over as well (too close to the bracket side of the board.
  5. Add some space between the crystal/10 pF caps (C12/13) and the PLCC pads so that there's clearance for them after a PLCC socket is installed.
  6. Put some distance between JP11 and P18.
  7. Drop the DP8473 and associated components.
  8. Adding some 74LS07 drivers (do you want to add receivers) might give a little protection.
  9. Very minor, but I hate to see Vcc on a jumper pin--too much possibility for mayhem. Can K9 be made a simple 2-pin affair that simply grounds a pulled-up pin?

Has anyone else built up their board?
 
Do you mean with a 40-pin header (allbeit restricted to use with a CF adapter)? I found that CF /CS1 isn't really needed (only used for drive reset on error) so possibly all that's needed is an address decoder (might be able to use some OR gates if you have any spare instead, to get the contiguous IO space mapping) and a couple of resistors. Looking through the specs again this week the CF spec data bus drive is 8mA, the same as the XC9572XL I've been using, so maybe not even a buffer either.

No, I was thinking about a CF socket. It shouldn't be too complex in hardware to do, given the sharing of resources.

Just thinking aloud here. You could also use an SMSC FDC37C669 which would give you the IDE, Serial, floppy and parallel ports. Since you could run the CF card in 8-bit mode, that would be enough. The 669 will support 4 floppies with a little external logic. They still seem to be available.
 
Last edited:
I was thinking about a CF socket. It shouldn't be too complex in hardware to do, given the sharing of resources...You could also use an SMSC FDC37C669 which would give you the IDE, Serial, floppy and parallel ports

The FDC37C669 looks idea - but being QFP/TQFP, I'd be nervous that it could limit the appeal of the project. Unless a short-run of 669s-on-through-hole-breakout-boards is the thought?
 
Back
Top