Caution: Non registered users only see threads and messages in the currently selected language, which is determined by their browser. Please create an account and log in to see all content by default. This is a limitation of the forum software.
Also users that are not logged in can not create new threads. This is a, unfortunately needed, counter measure against spam. Please create an account and log in to start new threads.
IDE access speed does change with CPU speed when the IDE speeder is on, but I haven't observed any problems with any CF card. Granted, I never tried with the Amigakit buffered interface, but given that AmigaKit just knocked off my old IDE-fix express design, not really knowing what they're copying, it may well be that they don't enable the data bus drivers fast enough, or they are not using the right kind of bus drivers.
Note that the old LS-type drivers from IDE-fix express may not be fast enough. The data bus drivers on our buffered CF card interface are about five times as fast as the ones we've used on IDE-fix express.
Further, data bus capacitance plays a huge role. Our buffered CF interface uses no cable for a reason: It helps keeping bus capacity low. So far, I haven't had a single CF card that doesn't work with the ACA1221lc at 40MHz, but that's not a guarantee. You're mentioning Sandisk, which is a high-quality and high-price brand, often target of counterfeit products.
Jens
Sorry, it's the A1200/600 CF-IDE Adapter (buffered) from amigastore.eu I have.
Haven't had that im my hands, but the mere fact that it appeared only a few weeks after our product has hit the market, has the exact same chip count and the same feature-description as mine (minus the mechanical features), I have to guess that it's the same circuit, aka "copy". Can't tell for sure unless I have a high-res photo of both circuit board sides. If you can post these here, I can give a more precise opinion.
Maybe you could edit the amigakit reference above in case someone gets offended..
I stand by what I write and say. If someone gets offended by the truth, it's his/her problem, not mine.
It's absolutely fine but 40 Mhz is noticeably snappier than 26.67 with workbench on the ACA1221LC I found.
So the system is stable at 40MHz, but only locks up if you activate the IDE speeder?
They (Thomas over in eab) was saying somewhere I think, idefix no longer required for 3.2.
Ok, fair enough. But I've never gotten over 2.6 MB/sec with the same setup in 3.2.2. So it does *seem* it might still be required maybe in some cases.
IDE-fix really digs deep in the OS structures, including the non-documented ones. I have my doubts that it's even compatible with 3.2 and higher. Further, I believe that many of the errors that the old IDE device had were ironed out with 3.2. It's not programmed to gain top speed, but to be 100% reliable and compatible with as many devices as possible. Personally, I'd always prefer the latest development instead of a 20-year old hack that happened to get best speed.
However, the 2.6MB/s do sound pretty slow for an activated IDE-speeder. You might want to contact the OS3.2 team and ask if the inner transfer-loop really uses longword-transfers.
Jens
So the system is stable at 40MHz, but only locks up if you activate the IDE speeder?
Yes.
Personally, I'd always prefer the latest development instead of a 20-year old hack that happened to get best speed.
Definitely.
For your reference attached is the Amigastore.eu A1200/600 CF-IDE Adapter (buffered).
The pass through is a good idea at first glance. Although maybe that also introduces other 'issues'.
The pass through is a good idea at first glance. Although maybe that also introduces other 'issues'.
I believe the pass through is also buffered - by the same buffers, and in an attempt to make it better, Matthias actually made it worse: Adding source termination is a good idea for fast point-to-point connections, but here, each connection forks to CF and to the pass-through, so you don't even have a point-to-point connection where source termination would actually work.
Further, the 16 data lines aren't single-direction, so you don't even have a "source" like you have for address/control lines. You'd have to modify each inserted CF card to make it work like this product pretends with it's series termination. I've made the same mistake on IDE-fix express: The only reason it always worked fine was that speeds were never high enough to make it a problem. I just didn't have the high-speed digital design knowledge back then. Could have saved me some money...
I have to revise my judgement about this thing - it's not a straight copy of my adapter, but merely an attempt to reach the same goal with fewer, less expensive components. Of course, the attempt fails. I won't go into detail (as it would reveal detailed findings of my development), but I dare to say this product does not deliver exactly what they are promising. Maybe I should say "false advertising" to finally get AmigaStore's attention? In any case, they copied my product description, but not the actual product. Judge for yourself what's worse.
Thanks - this makes me more confident that the real thing will make things work for you.
Jens
Maybe we should take a closer look at your PSU. If the CPU runs at 40MHz without IDE speeder, it *should* also work with IDE speeder. However, if power supply is on the edge (either voltage or ripple), it may already explain the behaviour.
The A1200 itself won't make a difference. It will only see 14MHz-accesses from the accelerator, no matter how fast it really runs.
MaxTransfer should not make a difference; it's a software setting and will either work or it won't, but it won't change with speed.
Can you elaborate on the two PSUs that you have avalable? Maybe take pictures?
Jens
Maybe we should take a closer look at your PSU
The power supplies are:
312503-02
391020
391020-02
Obviously all of them are v old now and haven't tested actual output/wattage whatever. Don't even know how to anyway..
The 312503-02 says 4.3A @5v on the back compared to 3.0A for the other two. So am just assuming here it has a bit more in the tank.
Would the little 020 @ 40 mhz really have much more power draw than @ 26 mhz ?
Just the ACA1211LC, the CF adapter and mouse are installed.
Tested the Sandisk Ultra CF card also with h2testw_1.4 in windows.
Seems good - well over 30 MB/sec and in line with the claimed specs
Would the little 020 @ 40 mhz really have much more power draw than @ 26 mhz ?
It's a CMOS device, so power consumption goes up linear with frequency. However, ripple requirements also go up with overclocking; if your PSUs have never been re-capped and adjusted in the past 1-2 years, I'd suggest to do just that: Exchange the output caps and adjust the voltage to be 5.00V when measured inside the A1200 while running.
Tested the Sandisk Ultra CF card also with h2testw_1.4 in windows.
Seems good - well over 30 MB/sec and in line with the claimed specs
Those modern computers don't use the IDE mode that you're using when it's in the A1200. You can't compare that - and I wouldn't expect the card to be the cause of the crash anyway. I'd expect it to be too-low voltage and transfer fails because the 68ec020 gets slightly too-low voltage for the speed you're asking it to run.
At least one of hte PSUs you have has a small adjustment potentiometer inside, so you should be able to make that adjustment. Careful, though: You have to operate both the Amiga and the PSU in an open state: The A1200 to measure the 5V rail under load, and the PSU to be able to get to the adjustment potentiometer. This should be done by someone with sufficient education/experience. I don't want anyone to die in the process!
Jens
Providing two digits after the comma would indicate that your multimeter is calibrated - if you call it "dodgy" at the same time, this indicates that you have some non-brand unit there that may depend on the battery voltage to be good to show halfway accurate readings for the first digit after the comma.
I don't expect you to buy a Fluke meter, but just to understand that measuring is a complex field, and tools must be right for the job. A not-so-expensive option if the meter that Dave Jones from EEVBlog sells. Or you search for BM235 and have it without the EEVBlog markup in a different colour. That's a 6000-count unit for well under 100,- EUR.
The floppy power connector is ideal for measuring the voltage. It's behind all spots where voltage drops (cable, connector, input filter) and perfectly represents what other chips and expansions are getting.
Assuming the readings are accurate (cough..), the A600 PSU (b) is not suitable for any Amiga with modern expansions, as the threshold of 4.85V for the local voltage regulators is not reached. The 3.3V regulators may oscillate, destroying the target parts (expecially Xilinx CPLDs).
The difference between PSU a and c is likely a measuring error, as most cheap multimeters are 2000-count only. Again, assuming that this is accurate, you should crank them up by 0.1V to have enough headroom for 40MHz overclocking. If you're at opening the PSU anyway, you should replace the output caps with fresh ones: Choose low-esr, high-temperature types, make sure that the voltage rating is sufficient and feel free to increase capacity as space allows (modern caps are higher capacity per volume these days).
If you start servicing the units, I'd go with the A600 unit first, as that's close to being unusable anyway. Gather some experience and confidence before opening the units that work OK at the moment.
Jens
Of course, no real space there to fit a fan on the heatsink without raising the 1200 off the desk. Not too keen on that (or even having a fan) so will leave it at the middle settings. Glad to resolve it though in any case.
Look for a radial fan - if you don't need it to be temp-controlled with RPM sensing and PWM control (like I use them on the ACA1240/1260), you should find something that fits.
Jens
