Boot loop after update_all

[HenrikErlandsson]

History: My hardware, and my Mister and menu.rbf files were dated 2312xx, I had run update_all 2401xx.

Ran it again today and was a success! It updated to latest Mister and menu.rbf.

But on my hardware, all recent such file versions result in a "sometimes" boot, average 1 of 8 boot attempts (power cycle).

I reverted to menu_20240408.rbf and MiSTer_20240505, and now it boots every time.

I used the latest Mister.ini, and this older version of Mister complained (while working correctly) about the subcarrier option, so I've commented it out for now.

This means that I'm happy for now because I figured out what caused the bootloop, and I have the latest arcade cores in my cab (Thanks for e.g. Salamander and Galaga '88!)

But obviously I would like to continue to run update_all without having to get the SD card out of the tricky spot inside the cab and replace those 2 files manually.

There's nothing wrong with the hardware, SD card etc. So what changed that could break the boot so hard? Can I run the latest Mister/menu.rbf and provide you a bootlog?

[HenrikErlandsson]

Changing this position of SW2 (Note the strange numbering) on the DE10 nano board will let a user of MisterFPGA update_all past that fateful date of 202403xx. AFAIK this has nothing to do with Jammix?

[sorgelig]

SW2 is used with secondary SD card. It's needed for cores where secondary SD card is supported but I/O board has missing a pin. I don't understand how it affects your case.

[HenrikErlandsson]

I'm trying to understand the text in the image. As best I can tell, a change in this file with this date is what forced me to "turn off the 24-bit DAC". Due to interference. I don't use a secondary SD card.

Does it mean I don't have 24-bit video color anymore? I'm guessing JAMMA RGB is affected if so, and not HDMI?

[sorgelig]

It's only affects analog output and only on systems where resistive 24bits are implemented (i know only misteraddons used that).

[HenrikErlandsson]

I tested the bpp on CRT and certainly no MisterFPGA can do 24-bit yet. It's more like 18-bit, and unbalanced so that if you make a 24bpp grayscale gradient in menu.jpg and display it, the gradient is steppy, and some grayscale colors come in the wrong order, i.e. a darker one in between two increasingly bright steps in the gradient.

So the R vs G vs B intensity for CRT gamma is wrong. Can I calculate and replace the resistors? (I'm assuming there's no digital recalculation of weights, if so the software must know the gamma of the display connected in order to calculate correctly.)

[sorgelig]

Unbalanced is because resistors have a tolerance and it may produce such gradient mismatch. So even with standard MiSTer 18bit analog video, resistors must be very precise. That's why 24bits with resistors is a bad idea.
Of course you can try to replace resistors with more precise versions. With correct resistors you should get a correct gradient.