SNES rev 1 black screen at startup

[Poot36]

Well I can confirm that the glitching graphics that I was seeing with the first test program that ran was due to a bad cart connector connection. Also the new test works as well. I guess this is good but it is weird that the PLB instruction would cause it to crash. I have attached a LA trace of this working startup. Do you think running the SNES at a lower (1.79Mhz) or higher (3.58Mhz) speed would change its behavior?

[tepples]

I think that program is already a slow ROM (mode $20: high nibble means slow ROM, low nibble means LoROM). Someone could patch the init sequence to make it a fast ROM (mode $30) and see what breaks.

[lidnariq]

You actually already thought of that, tepples; it inspects its own header and adjusts the execution speed register according to the $10s bit at memory location $00FFD5.

To change any of the spadtest things I've sent to take advantage of this, use a hex editor to change the byte at address 0x1FD5 from 0x20 (" ") to 0x30 ("0").

----

On the other point: If it's just the PLB instruction, then I'd be surprised if there were anything you can fix externally.

The silver lining is that modifying other existing ROMs (to further validate that the PPUs work) becomes feasible. So ... here's a copy of nocash's Magic Floor where I removed the PLB instruction. It's still only mode 0.

magicsns-no-plb.zip

(3.19 KiB) Downloaded 136 times

—also tests the APU (yeah, I know you already removed the module).

Here's a modified copy of a, uh, certain group's intro. PPU mode 1.

Anthrox - 3d Demo (PD)-no-plb.zip

(12.57 KiB) Downloaded 124 times

And here's a modified random trippy mode 7 something:

Mode 7 Demo (PD)-no-plb.zip

(1.77 KiB) Downloaded 120 times

[Poot36]

Well this is interesting, the Magic Floor appears to try to run but never does (screen flickers slightly over and over and will sometimes get graphical garbage on screen), the Anthrox demo is glitching video: https://youtu.be/HAledVusSGI it shows black and wight most of the time but color bars are present and when I press the Start button to view the credits they are shimmering as well (also have a problem having this one start up continually on power up), the mode 7 demo works great. All of these demos work properly on my working SNES so I think I can rule out the homebrew cart and rom file multiplying. Could I still have bad solder connections on some of the chips or cartridge connector or does the mode 7 demo rule that out? I will get a LA trace from the Maic Floor one tomorrow. I have also found out that the previous owner of this SNES was trying to see if it would work and tried to use a NES power adapter on the SNES using jumper wires (I assume they thought it would work due to the label on the bottom of the SNES mentioning a AC adapter). Would having AC from the NES adapter have damaged the SNES? In the schematic I see a diode as well as a 5V regulator that should prevent the AC from reaching the chips. I also know that the audio amp still works because I was originally getting some random audio out.

[lidnariq]

Do you have an oscilloscope as well as that LA? Could you check how clean the 21.5MHz clock output is?

It'd be nice to check both medium-term (ms range) stability as well as cycle-to-cycle jitter.
----

9VAC before the fuse and regulator shouldn't have harmed anything, although it might cause the filter caps to get hot and shorten their lives.
9VAC after the diode and fuse but before the regulator probably permanently damaged the electrolytics on the input stage. ... and possibly could have damaged Q18 (the NPN on the input stage for the semi-regulated higher-voltage supply)... maybe?

[Poot36]

I can check the 21.5Mhz clock but you won't see a square wave due to my scope only having a 100Mhz bandwidth (a Tektronix 465).

[lidnariq]

Fortunately, sharpness of the transition doesn't really matter here.

[Poot36]

The NES power adapter was connected to the barrel jack on the outside of the SNES so I am going to assume that it would have been ok. I will take a look at the XIN signal with my scope and see if there is any of the problems you mentioned. Got any ideas why the Anthrix intro demo would corrupt like that? It looks like it is trying but failing on correctly displaying the colored ball text (not sure you can make this out in the video, if you want I can try to use my Huppage 1600 TV tuner capture card but I have no clue as how to reduce the bit rate so I do not wind up with a very large file to upload). Stupid 512kbits/s upload speed should be 1mbits/s.

[lidnariq]

NTSC console, right?

If the 21.5MHz master crystal is significantly changing frequency, the TV's chroma killer will activate, producing a monochrome signal. This is consistent with the scanlines being very jittery like what I saw, also.

The easiest ways to check for jitter or drift is to set an edge trigger and set the horizontal position a few cycles later (for jitter) or many cycles later (ten thousand or so, about 1ms, for drift).

If it is significantly changing frequency, I'd strongly suspect misregulation of the +5V rail; try putting another 'scope probe on +5V and looking for noise there.

You could also check PPU2 pin 3, which is master clock ÷ 6. (The division will mask jitter some, but it might be helpful anyway)

I was able to see what you were talking about in the youtube video, so I don't think you need any more sophisticated capture stuff yet. But I'd assume a single field (or maybe a few sequential ones) would be sufficient, should the need arise.

Stupid 512Kbps upload speed should be 1Mbps.

My downstream is like 10mbit/s. My upstream is like 200kbit/sec.

[Poot36]

I don't think that I can do all those things that you just mentioned on my 1970s purely analog scope. The mode 7 and controlle test colors are sold but maybe they work differently then this test. I am using the S-Video input on my Sony PVM series studio broadcast monitor so I would like to think it would have a more robust color sync circuit. I think I screwed up the units, my download speed is around 6mbits/s so I guess that is the trade off (it is a ADSL line and according to the modem the maximum upload speed that the line can support is a few kbits/s above 1mbits/s so I think that if I removed the dialup modem (used for scan to fax on the computer) that is in parallel with it the speed may go back up (using the telephone out on the dialup modem to go to the ADSL modem).

[lidnariq]

I could have sworn I could check for jitter on the tube-based Tek 547 ...

Anyway, in the manual for the 465, it talks about "Delayed Sweep Displays", which should be the right thing for this.

[Poot36]

Well I just looked at both signals on my scope and they look solid (used the fastest timebase and the 10X mag button and scrolled around the waveform with the horizontal position knob). There was a bit of ringing on pin 3 of PPU2 but that was probably due to my long ground lead. I don't see the advantage to the delayed sweep to magnify the trace as selecting the fastest timebase and the 10X mag button will show me the same thing as far as I can tell.

[lidnariq]

How about checking while the Anthrox Intro is running?

[Poot36]

I will do that. Will have to reprogram the eeprom though. Because my SNES in a rev 1 it has some differences in the clock circuitry from the schematic that I have, my X1 crystal is the 21.4Mhz one and I have a X2 crystal that has this text on it ([M]4.00 f) I think this is a 4Mhz crystal attached to U9 (74hcu04A). I have the LA trace from the Magic Carpet demo attached to this post.

[lidnariq]

Magic Floor is hanging on APU communication:

Code: Select all

sub_0A2B: ; $00EA2B / $80EA2B
        lda     $2140                   ; $0A2B | $2140 -> $002140
        cmp     #$AA                    ; $0A2E
        bne     sub_0A2B                ; $0A30

(This was generated using https://github.com/andlabs/65816disasm )

Nocash's corresponding faux-x86-for-65816 code:

Code: Select all

apu_up_prepare_first_block:
@@wait_for_0bbaah:              ;\
 mov  a,[2140h]                 ;
 cmp  a,0aah                    ;
 jne  @@wait_for_0bbaah         ; Wait until Word[2140h]=BBAAh

The LA trace indicates that it's always reading $00 from the APU, not $AA.


OH. And the 4MHz ceramic resonator is the CIC's, not the CPU's.