Minimig v1.1 ARM Hardfile Demonstration

[yaqube]

The Minimig in the video is equipped with 20 MHz 68SEC000 (the price difference between 16 and 20 MHz version was marginal). The speed gain of the CPU run @ 28 MHz over standard A600 is only 2.91 due to memory running at 7 MHz and being shared with custom chips. I have also made some tests with memory run @ 14 MHz (requires faster memory chips) and achieved speed gain of 5.05.

I have made some tests to find out what the maximum operational frequency of the 68SEC000 rated at 20MHz is and mine is able to run at 39 MHz. But without equally fast memory access the speed gain is very small (SysInfo reports 3.03 times the A600 speed).

I do not expect the 16 MHz rated parts to be much slower.

[yaqube]

Darrin wrote:

So you don't foresee any problems with the 16MHz chip being run at 28 Mhz?

That's my expectation. Actual tests will reveal if it runs or not.

Are you planning a new core release for the unmodified v1.1 anytime soon (bug fixes, df1:, etc)?

Yep, I have already rewritten some parts to make the Minimig more cycle exact (it's not 100% exact and will never be) and now some games and demos work in the same way as run on a real Amiga.
But there are still some others that do not run correctly and I would like to find out why and fix the problems if feasible.

After implementing upgrades from SD-Card for the ARM controller board I will port the firmware to the PIC (most of it has been already done). The FPGA core will be the same for the ARM and the PIC (only the former will allow to have 4 floppy drives and hard disk support).

[yaqube]

Everblue wrote:
1. 28MHz CPU - with the current CPU on board?

Yep.

2. To replace PIC with armboard (for HD Support), it wont involve soldering right?

Basically no soldering. The ARM board is a plug-in replacement for the PIC. But unfortunately the original Minimig board design imposes a limitation on the maximum communication speed with an SD/MMC card. To overcome this problem you need to short two pads of two resistors (very simple soldering task). Otherwise the emulated hard disk transfer speed is limited to ca 200 KB/s.

3. I would like to be able to install run WHDLoad - will that need extra RAM to be installed for all (working games) to er, work?

I can't answer this question because I haven't done any such tests. I expect that with the base memory of 1.5 MB chip RAM those games which require 1MB of any or chip memory should run. But this is not confirmed.

[yaqube]

Darrin wrote:

yaqube wrote:
But unfortunately the original Minimig board design imposes a limitation on the maximum communication speed with an SD/MMC card. To overcome this problem you need to short two pads of two resistors (very simple soldering task). Otherwise the emulated hard disk transfer speed is limited to ca 200 KB/s.

Does that mean that with the ARM and the pad shorted you could implement a "turbo floppy mode" where the read/write speed to the ADF images is accelerated?

Nope. The "turbo floppy mode" doesn't need the ARM board at all. But with the ARM the turbo mode is faster. Now I can't tell how fast the floppy turbo mode is when using the PIC but with the ARM basic transfer speed is ca 30 KB/s and in turbo mode ca 60 KB/s. No need to short the pads but it's advised since the overall system speed is higher.

[yaqube]

mikej wrote:
Yaqube, could you clarify which pads needs to be shorted and why?

Dennis had problems with his MMC card because it was still active even when its _CS line was held high. So he gated the MMC clock line with very simple solution: he put the R51 (1K) in series with the R50 (270R) and shorted the connection between them to VCC (with the PIC RA0 pin) whenever the MMC card was inactive.

The drawback of this solution is that there is a huge impedance between the source of the clock (the PIC) and the MMC card effectively limiting rising/falling edge speed of the CLK signal. I have modified the program not to disable the MMC clock and it works with all mine MMC and SD cards (I have only tested about ten different types). Bypassing these two resistors allows to run an SD card with 24 MHz clock. This solution isn't perfect but relatively simple.

The best solution would be to put a series source-terminating resistor to match the PCB trace impedance (60-90 Ohm) and have proper clock routing topology.

[yaqube]

The LPC2388 is planned to be used in the next generation of Minimig. The RTC with battery back-up is on the list.

The PIC replacement version is based on the AT91SAM7S256 and has no RTC. But can be used in any Minimig V1.1 board (designed by Dennis).

I have talked to Peter about a possibility of including the ARM controller on the Minimig-ITX board. It's seems to be feasible.

[yaqube]

@Loriano

I will talk to you before the final board design will be completed. We should design it in an enclosure-friendly way.