elvovirto wrote on 2023-02-21, 15:44:

After looking a bit at the end of that linked thread, I figured I'd check for an E-IDE card on eBay. There are only two ISA opt […]
Show full quote

Jo22 wrote on 2023-02-21, 14:45:

I do remember these enhancers! They really do work, at least for Award/Ami BIOSes. Haven't tested them with others yet.
Some also shipped with those Year 2000 upgrade cards.

After looking a bit at the end of that linked thread, I figured I'd check for an E-IDE card on eBay. There are only two ISA options currently listed - and one of them is over 100 bucks.

When I think of all the stuff I tossed or donated over the years it's like...man...I wish I still had (Insert Hardware I used to have here).

We feel ya. We're all through this, I think. On the other hand, these things are +25 years old by now. A quater of a century.
It's a wonder these things are around, at all, still. In parts (pun intended) because some of it was being preserved by computer fans (and/or simply forgotten).
But yes, once you re-start tinkering with these things, you find yourself thrown back into the 90s, mentally.
So it somewhat surprises you (us) when we see those hefty price tags (seeing them with the eyes from back then).

douglar wrote on 2023-02-22, 01:48:

Seems unlikely that swapping CF adapters is going to make a difference in this case. They just plug the 40 pin IDE through to the CF, no chips or intelligence in the device. Seems like your issue is placing an IDE device at the higher clock speed it the issue, regardless if it is an SD bridge or a CF adapter, since things do work at the slower speed.

Yep. And that's exactly the problem, I'm afraid. CompactFlash supports three different modes of operation, and all the CF IDE adapters are just ..adapters. No converters.
If they did actual conversion, the CF card could operate in one of the non-IDE modes and let the converter do the IDE/PATA talk.
But being just a mechanical adapter also means that the CF card is directly sitting on the ISA bus, with all the weird timings elvovirto's mainboard is throwing at it.
If the motherboard BIOS isn't even ATA-2 compatible (it isn't) a lot of things must sound like gibberish to the CF card, so it reacts weird.
It's just not only an LBA issue here that EZ-Drive and friends could fix. The board needs a set of fresh HDD routines, first of all, I think.
That's why I think that those EIDE BIOSes, XUB etc. are worth a try. Sure, it won't fix the timings, but maybe the new routines are more robust.

Alterntively, let's just use SCSI. Try getting an ISA SCSI controller with a boot rom.
Then, use a Raspberry Pi to simulate an HDD. The Macintosh scene uses this approach, I belive.
https://blog.adafruit.com/2021/02/26/rascsi-a … i-raspberry_pi/

Jo22 wrote on 2023-02-22, 18:26:

If the motherboard BIOS isn't even ATA-2 compatible (it isn't) a lot of things must sound like gibberish to the CF card, so it reacts weird.

Here's the crux - Pretty sure he said that the storage and video work when he runs the mobo at 8mhz and doesn't work when he runs it at 16 mhz.

So if it does work at one clockspeed, it's not an LBA miss, a CHS mapping, an ATA compatability, or a Pin 28 issue, because changing the clock speed wouldn't fix those things.

It's an ISA issue. Seems like the bus on his computer can't handle multiple loads on the ISA bus at higher frequency.

The problem with the SD adapter sounds like the classic Sintech/FC1307 bug where the adapter freaks out if the IDE interface is reset and it can't find an active FAT16/32 partition in the card's partition table.

The instability issue with the CF based cards sounds like they're pulling too much power causing problems when the system is run at full speed - they do have an external power connector attached (with a working 5v line), right?
(It's absolutely nothing to do with how the IDE card talks to the drive/card.)

douglar wrote on 2023-02-22, 18:48:

Here's the crux - Pretty sure he said that the storage and video work when he runs the mobo at 8mhz and doesn't work when he run […]
Show full quote

Jo22 wrote on 2023-02-22, 18:26:

If the motherboard BIOS isn't even ATA-2 compatible (it isn't) a lot of things must sound like gibberish to the CF card, so it reacts weird.

Here's the crux - Pretty sure he said that the storage and video work when he runs the mobo at 8mhz and doesn't work when he runs it at 16 mhz.

So if it does work at one clockspeed, it's not an LBA miss, a CHS mapping, an ATA compatability, or a Pin 28 issue, because changing the clock speed wouldn't fix those things.

It's an ISA issue. Seems like the bus on his computer can't handle multiple loads on the ISA bus at higher frequency.

I know, that's why I added a small comment about it in my post.
This problem calls for an oscilloscope, I think. A storage scope, favorably.
To see how the signal is at low/high speed.

jmarsh wrote on 2023-02-22, 18:52:

The instability issue with the CF based cards sounds like they're pulling too much power causing problems when the system is run at full speed - they do have an external power connector attached (with a working 5v line), right?
(It's absolutely nothing to do with how the IDE card talks to the drive/card.)

But isn't the CF already running overvolted at 5v?
CF is 3v3, but it gets 5v here unless a little 5v to 3.3v converter is running on the CF adapter.

CompactFlash can work at both 3.3v and 5v. Neither is "correct." Some IDE adapters have a jumper to make a linear regulator convert 5v to 3.3v for the card's power, but that's pointless since the 5v IDE address/data lines are connected directly to the card's pins.

jmarsh wrote on 2023-02-22, 20:42:

CompactFlash can work at both 3.3v and 5v. Neither is "correct." Some IDE adapters have a jumper to make a linear regulator convert 5v to 3.3v for the card's power, but that's pointless since the 5v IDE address/data lines are connected directly to the card's pins.

I don't think the same.: 3.3v is the prefered voltage, hence the regulators. 5v is a legacy voltage, CF cards care "5v tolerant".

The same story can be said about flash chips, by the way. Some voices say the 3.3v parts shouldn't be connected directly to old 5v equipment.
There's a whole lot of discussion in the flash cart scene. Some are even more "hard core" than me and say the "diodes in series" solution is a no-go. 😁

Edit: Removed. Have to check my books..

jmarsh wrote on 2023-02-22, 20:42:

CompactFlash can work at both 3.3v and 5v. Neither is "correct." Some IDE adapters have a jumper to make a linear regulator convert 5v to 3.3v for the card's power, but that's pointless since the 5v IDE address/data lines are connected directly to the card's pins.

Apparently you need to use 3V mode for the top speeds (UDMA6+ ?) which are useless in practice anyway 😜

Jo22 wrote on 2023-02-22, 21:35:

I don't think the same.: 3.3v is the prefered voltage, hence the regulators. 5v is a legacy voltage, CF cards care "5v tolerant".

Devices that are "tolerant" of higher voltages usually only accept them as output, they don't output at those voltages. That is not the case with CompactFlash.
The regulators on those adapters only perform voltage conversion for one single pin - the other pins are still connected directly to 5v signals from the PC so they achieve pretty much nothing.

The same story can be said about flash chips, by the way. Some voices say the 3.3v parts shouldn't be connected directly to old 5v equipment.
There's a whole lot of discussion in the flash cart scene. Some are even more "hard core" than me and say the "diodes in series" solution is a no-go. 😁

That is not the same story at all. It's the same as connecting SD cards to 5v - they're not designed for it, the results are undefined and likely to lead to permanent damage.
And diodes are a terrible solution; they may work for power, but for data (depending on direction) they either introduce delays or may not work at all due to raising the digital low voltage above acceptable thresholds.

Ryccardo wrote on 2023-02-22, 21:55:

Apparently you need to use 3V mode for the top speeds (UDMA6+ ?) which are useless in practice anyway 😜

That's "sort of" true for some cards due to them not supporting those modes when powered with 5V, but what it really means is that those modes aren't usable when they're running in True IDE mode since IDE uses 5v signalling by design.

jmarsh wrote on 2023-02-22, 22:04:

Jo22 wrote on 2023-02-22, 21:35:

I don't think the same.: 3.3v is the prefered voltage, hence the regulators. 5v is a legacy voltage, CF cards care "5v tolerant".

Devices that are "tolerant" of higher voltages usually only accept them as output, they don't output at those voltages. That is not the case with CompactFlash.
The regulators on those adapters only perform voltage conversion for one single pin - the other pins are still connected directly to 5v signals from the PC so they achieve pretty much nothing.

I think this might be a misunderstanding.
You can use a single voltage as a power source for pull-up/down circuits. Which could be installed both ways in a CF card, for both input and output. It's not necessarily only used as a voltage for the internal operating voltage of an IC.

Anyway, it might be interesting to stop speculating and do perform some tests here.
Not necessarily now, but maybe in the future.

I wouldn't be surprised if the CF cards would either always output 3.3v on their data pins or
if they make it dependant on the input voltage were the voltage regulator connects to.

PS: I'm a fan of classic 5v TTL tech, too. It can be powered by 4,5v latern batteries (3R12). 😃

Proof is in the pudding as they say. Some boards just do not tolerate CF-IDE adapters either due to chipset or BIOS... Or improper ISA CF adapters. Seen many threads about this and is my conclusion.Bored with near a foot of snow in last few hours and have to give an opinion even if it irks others ;p

Jo22 wrote on 2023-02-23, 01:53:

I wouldn't be surprised if the CF cards would either always output 3.3v on their data pins or
if they make it dependant on the input voltage were the voltage regulator connects to.

Of course it's dependant on the source voltage, that's how ICs (unless explicitly stated otherwise) work. Which is yet another reason why it's a bad idea to use a 3.3v source when connected to a PC - the IDE controller expects 5V from the connected drive, worse case it connects some of those lines (IRQ in particular) directly to the bus with no buffers and the voltage isn't high enough to be recognized as high by other devices.

It is also the case that you should (again unless explicitly stated otherwise) never connect any voltage to an IO pin that is above the source voltage, which is what will be happening when you feed a compact flash card with a 3.3V supply and connect it to the 5V ISA bus.

Horun wrote on 2023-02-23, 06:03:

Proof is in the pudding as they say. Some boards just do not tolerate CF-IDE adapters either due to chipset or BIOS... Or improper ISA CF adapters. Seen many threads about this and is my conclusion.Bored with near a foot of snow in last few hours and have to give an opinion even if it irks others ;p

Pretty sure this is what we've run into, as some folks have tossed around before - the BIOS and/or IDE controller and/or motherboard chipset just can't handle things.

I got another CF adapter in the mail today (yay delays) and the end result, plugged directly into the controller, is the same. Garbled text, no POST, opens CMOS setup instantly upon power-on at 16Mhz.

I think this system is just going to get shelved for the time being. It's a decent little system and matching monitor, but until I have the time/cash to test a card with an option ROM or to swap the motherboard for something slightly newer, it'll just sit on a shelf. Once I chop the Varta battery off of it, anyway.

Thanks everyone for your experiences and your inputs.

Re: "Fixed" 386sx motherboard works but not with 16-bit VGA card

"Imo best course of action would be isolating pin 28 between CF_to_IDE converter and IDE controller. "

try that

rasz_pl wrote on 2023-03-01, 04:48:

Re: "Fixed" 386sx motherboard works but not with 16-bit VGA card

"Imo best course of action would be isolating pin 28 between CF_to_IDE converter and IDE controller. "

try that

I'll have a look.

For reference, I actually found the system board on TRW, for anyone that wanted to have a look at it.

https://theretroweb.com/motherboards/s/unknown-386sx-neatsx
https://theretroweb.com/motherboards/?showIma … ufacturerId=118