BLITTER RE-CREATION THOUGHTS

[exxos]

yay!

IMG_3713.JPG (51.48 KiB) Viewed 4833 times

[exxos]

IMG_3714.JPG (131.55 KiB) Viewed 4823 times

So VCC = 3.3V.

Left side is 5V, right is 3.3V.

Input on the 5V side is tied to 5V via a 10K resistor. Voltage on the left side is 1.30V. With a pull up on the 3.3V side, it is now 3.3V.

Now when I GND either side, the other side doesn't change. So its like the IOs are isolated. I have tied the OE to 0V and 5V and 3.3V, doesn't make any difference.

So this chip is just acting like its not even on the board. I've double checked voltages on the pins on the IC itself, so its not a bad connection anywhere... So I really don't get why these chips are not working.

[exxos]

I've plugged back in the discrete sparkfun board, and this works perfectly fine...

The IC's should behave the same way, but none of them do

I think for a initial prototype, I will just create a monster blitter board with like 60 mosfets , at least I can get something I can actually try out the blitter core with then.

IMG_3715.JPG (149.53 KiB) Viewed 4810 times

[exxos]

I connected up the IDT buffer again for a final test...

That one only seems to work with a 2.2K pulldown on the 3.3V side.. When either side goes low, the other side goes low , so it does actually work.. sort of...

I only get 3.3V on the 5V side..Which is kind of expected anyway. But what worries me actually, is if I force the 5V side directly to 5V, the 3.3V side jumps to 5V as well. It might actually work setup like this in a real circuit, but if something on the bus actually drove a logic high to 5V, it would blow up the 3.3V side. So I am not going to use a chip which easily has the potential to kill the 3.3V side up.

With the sparkfun board, I get 5V and 3.3V just fine, either side goes low, the other goes low, thats it, it just works and does what its supposed to do.

So all these "fet bus switches" are going on my "never use again list".

[ijor]

I connected up the IDT buffer again for a final test...

That one only seems to work with a 2.2K pulldown on the 3.3V side.. When either side goes low, the other side goes low , so it does actually work.. sort of...

I only get 3.3V on the 5V side..Which is kind of expected anyway. But what worries me actually, is if I force the 5V side directly to 5V, the 3.3V side jumps to 5V as well. It might actually work setup like this in a real circuit, but if something on the bus actually drove a logic high to 5V, it would blow up the 3.3V side. So I am not going to use a chip which easily has the potential to kill the 3.3V side up.

How strange! At what voltage was VCC connected at the IDT chip? Did you add a diode between the chip and vcc?

[exxos]

How strange! At what voltage was VCC connected at the IDT chip? Did you add a diode between the chip and vcc?

I have tried VCC at 5V, 3.3V, with and without diode. I have tried pretty much every possible combination of everything with all the chips I have tested.

I did have a thought just though, that is the single mosfet idea works at say 1.8V.. then it opens the door up to more FPGA devices.. So will try that now.. I think the mosfet just connects to GND either side, so the voltages either side hopefully won't matter...

EDIT:

I tried a 2,2K and 1K in series to give me a 1.55V rail, so used that on the LV side.. seems to work fine as well. I have 5V on the output, 1.5V on the input. If either side goes low, the other side goes low. So could pretty much forget the MAX10 and just go with a cyclone really... but then would need a boot ROM.. but of course for initial testing for the blitter, I will still use the MAX10 anyway just to keep things simple as possible.

[ijor]

I have tried VCC at 5V, 3.3V, with and without diode. I have tried pretty much every possible combination of everything with all the chips I have tested.

Have you considered the possibility that the chip is defective, or got damaged, or you got the wrong part, or something wrong with the board?

The IDT switches are known to work. E.g., Mark (Foft at Atariage) uses IDT switches on his 8-bit FPGA boards. In one case it's a Cyclone V connected to 5V peripherals. In other case it's a Max 10 connected directly as a replacement chip in an Atari 8-bit motherboard.

I found a dev board with a Max 10 and bus switches, TI parts in this case. Unfortunately it's not cheap:
https://www2.hdl.co.jp/en/index.php?id=437

I tried a 2,2K and 1K in series to give me a 1.55V rail, so used that on the LV side.. seems to work fine as well. I have 5V on the output, 1.5V on the input. If either side goes low, the other side goes low. So could pretty much forget the MAX10 and just go with a cyclone really... but then would need a boot ROM.. but of course for initial testing for the blitter, I will still use the MAX10 anyway just to keep things simple as possible.

All the Cyclone families can do 3.3 volt as good as the Max 10. There is no difference at the I/O in this regard. The only difference is that the Cyclone parts need 1.2V for the core voltage and then need an extra regulator. While the MAX 10 has an internal regulator and can use a single 3.3V supply. Again, this is just for powering the core without much relation with the I/O.

Anyway, why you would prefer a Cyclone over a Max 10? For the usage we are talking about, it won't have much advantages, perhaps the contrary.

[exxos]

Didn't know the Cyclone IO still ran at 3.3! Thanks. I was just thinking ahead that ultimately GLUE , MMU, BLITTER could all go in a single FPGA, Kinda is like that on the STE already.

I do have another sn74cb3t3245 to try, but don't think its worth bothering. I tried 2 IDT chips when I was doing testing, both behaved the same, I then tried a TI part, which behaved oddly as well, and then the sn74cb3t3245 which seems they forgot to put any silicon inside it at all. I may try the second chip on Monday if I get chance.

But 3 different chips from 3 different manufactures all doing all manner of various crazy.. Obviously I am doing something wrong or killing them somehow.

TF uses them on his latest board and they seem to be working fine for him. Maybe I got unlucky and all this chips are faulty. But if they are that temperamental I don't want to use them.

[ijor]

I was just thinking ahead that ultimately GLUE , MMU, BLITTER could all go in a single FPGA, Kinda is like that on the STE already.

You can fit a whole ST system on a Max 10. The whole ST chipset plus CPU, FDC, MFP, Sound chip, etc, they all fit on a single Max 10, and you could still have plenty of space left.

Yeah, I know that's not what you want. Just noting how big a Max 10 is.

[exxos]

Max10 it is then