Stedy's fix(?) for Compact flash IDE issues

[Stedy]

Hi,

Let me start by apologising for creating yet another thread on Compact Flash IDE cards when there are many on this forum, I have my reason. Please read on and respond to my queries at the end.

I believe I know why some cards/adaptors work with the Amiga and some don't, though I need some information from other users. I have 3 systems with CF/IDE drives, an A600 with 128MB Sandisk CF, , A1200+IDEFIX with Hamex 4GB CF and CD32/SX32 with a 512MB Sandisk CF.

I was testing something in my A600 earlier today and after 10 minutes, the 'HDD' stopped responding. A reboot and the drive was not found. Power cycled, still no drive. My A1200 was in pieces, so I whipped the IDEFIX-97 (IDE buffer) out and 2 minutes later, my A600 had a buffered IDE interface, nothing else changed and the system worked, reliably.

My oscilloscope was nearby so I measured a few IDE signals with and without the IDEFIX:

Pin   Signal    Rise time Amiga    Fall time Amiga   Rise Time IDEFIX  Fall time IDEFIX
3/7  Data 3/5   74ns                   10ns                   70ns              10ns
25   IOWR       29ns                   10ns                   11ns              9ns
35   DA0/(A2)   25ns                   10ns                   11ns              9ns
37   CS0/(CS1)  12ns                   10ns                   14ns              12ns

Note names in brackets are as shown on the A600/A1200 schematics, source of signal names is the ATA specification.

The main difference between the IDEFIX buffered and normal Amiga was related to IOWR and IORD for Read/Writes and the address signals, DA0-DA2.

The ATA specification recommends a maximum rise time for signals of 12.5ns, we violate this on the data but for other signals, when buffered, they are met. Modern logic TTL devices, CPLDS, FPGA and IDE interfaces, typically expect signals that have rise/fall times of 10ns or less. They may tolerate longer rise/fall times providing the signal is noise free. This varies from part to part.

I don't think we have a gross timing violation, though I need to hook the logic analyser up soon.

The other potential area is relating to pull-up resistors on the IDE/ATA bus. The IORDY signal should have a 1K pull-up resistor, to ensure that when the last devices drives it, it rapidly does to the in-active state. Those of you following the Terriblefire accelerator project should be aware the pull-up on the board was changed from 47K to 1K, this is the reason why. This issue also affects the A600/A1200. On both machines, the _WAIT signal (pin 27 of IDE) has a 10K resistor, the IDE/ATA specification requires 1K. IDE hard drives do not need seem as bothered by this but compact flash cards are.

For reference, on the A600, R715 (a resistor pack) and on the A1200 R715G should be changed. Rather than remove resistors, place a 1K resistor in parallel with the existing part. This might fix some issues, I will test this soon.

The A600/A1200 were designed before the IDE (ATA) specification was formally ratified so minor infractions are understood.

Enough technobabble, what do I need from you?

If you have a working IDE to CF adaptor and card, post the model numbers and sizes here. Importantly, do you have a buffered IDE interface?

More importantly, those of you who do not have a working card/adapter combo, also post manufacturer, card size and if you use an IDE buffer?

I will measure the IDE port timing and try changing the IORDY/_WAIT resistr on my A600, to see if it fixes my system, if it does not, an IDE buffer may be required. I could create a IDE buffer and compact flash in 1 PCB but only if required, I need feedback from my fellow Amiga users.

I fixed SCART cable issues back in 2009 by making seemingly minor changes, I want to do the same for IDE/compact flash cards in 2017!

[idrougge]

Working here:
A600: Sandisk 1 GB (old type, no Ultra or anything else)
A600: Sandisk Ultra 8 GB
A1200: Transcend 4 GB Industrial

No buffers.

[th4t1guy]

Adapter: SYBA SY-IDE2CF-NB25
Card: Kingston 8GB Elite Pro

Working on cd32 SX-1 IDE

[Daedalus]

Excellent investigative work as always Stedy! I have a Kingston CF/4GB, no other designations other than 3.3V/5V, and it's a 2008 model. It works in a generic eBay adaptor on at least three different A600s, and probably a dozen or more A1200s (I use it for testing after repairs). None of them have buffered interfaces. I haven't found any machine it hasn't worked on.

[meynaf]

On my A1200 I have used so far :
- Sandisk Ultra II 2GB
- Lexar pro 800x 16GB (this one needs to lower MaxTransfer, while the other doesn't)
No part number available for the adapter itself.
No buffers, working fine except for the fact only 1 CF is seen at bootup when there are 2 on the adapter.