|
ForumsSega Master System / Mark III / Game GearSG-1000 / SC-3000 / SF-7000 / OMV |
Home - Forums - Games - Scans - Maps - Cheats - Credits Music - Videos - Development - Hacks - Translations - Homebrew |
Goto page 1, 2, 3 Next |
Author | Message |
---|---|
|
Phantasy Star on Everdrive glitches?
Posted: Fri Jan 01, 2016 6:53 pm
|
I recently got an Everdrive, and I wanted to play Phantasy Star on it. When I start the game, the graphics glitch a lot. Is it the ROM file, the Everdrive, or my Game Gear? Or something else entirely? | |
|
Posted: Wed Mar 08, 2017 8:27 pm |
I'm either very patient, or very forgetful. After I fix my umpteenth Game Gear, I plan to play this game. I would greatly appreciate an answer.
Thanks, Capt. 2110 |
|
|
Posted: Wed Mar 08, 2017 10:54 pm |
I think the answer is nobody knows. If other games work then it's even more strange. | |
|
Posted: Thu Mar 09, 2017 12:17 am |
Share the rom so others can try. | |
|
Posted: Thu Mar 09, 2017 9:49 am |
Got a chance to try on another GG?
If the EverDrive works OK with everything else, the ROM might be the problem. |
|
|
Posted: Thu Mar 09, 2017 9:21 pm |
The Everdrive works on other Game Gears, and it plays other games fine. I'll post the ROM when I get the chance. | |
|
Posted: Thu Jan 28, 2021 9:38 pm |
From what I remember at the time, the SMS adaptor for the GG didn't support 4 Mega cartridges. It might be that it wasn't the adaptor's limitation, but rather the GG's.
I remember playing Phantasy Star on my GG and having artifacts on the screen. But it didn't prevent gameplay... |
|
|
Posted: Fri Jan 29, 2021 12:01 am |
We're discussing artifacts on the screen for this game in another forum thread right now. it seems it might effect some master systems as well as certain game gear models.
I'm curious if you can see if what's seen in this video are what you remember? Also, did you play the actual cartridge in an adapter or use an everdrive? Thanks! |
|
|
Posted: Fri Jan 29, 2021 3:04 am |
Yes, that's what I remember seeing - small colored blocks out of place across the screen.
It was the Brazilian cartridge on the adaptor, translated by TecToy - so early 90s. |
|
|
Posted: Fri Jan 29, 2021 7:30 am |
This is very odd. The SMS adaptor doesn’t have any inherent cartridge size limitation. I guess we can rule out the Everdrive as this happens with a real cartridge. The thing that’s weird is that the corruption seems to happen when the game does something very simple that pretty much every other game does. | |
|
Posted: Fri Jan 29, 2021 7:54 am Last edited by bsittler on Fri Jan 29, 2021 9:49 am; edited 1 time in total |
Oh, wow! One of my Game Gears does this too and I thought it was a hardware failure of some kind. Good to know it's not just me and is likely some kind of bug in the earlier chips
edit: I sometimes also see this in the EverDrive menus on the affected Game Gear. Also, do you see this behavior regardless of power source (batteries vs. DC adapter?) I see this behavior only on my oldest Game Gear, and regardless of power source. Four newer Game Gears I tested don't seem to do this. The oldest one, however, usually glitches at least one character on the SMSTestSuite-v0.32 title screen |
|
|
Posted: Fri Jan 29, 2021 9:43 am |
Wow, very interesting information. I just modded my bro's GG, it's an older model one, and his copy of PS is doing this. I thought maybe it was the converter, or the cartridge was dirty. But after cleaning both, it persisted. I even got him a GG X7 for Christmas, with the same results. I thought it might of been something I did when I installed the McWill, or maybe when I replaced the caps. But other games seem to play fine. | |
|
Posted: Fri Jan 29, 2021 11:57 am |
so this seems to confirm after all that 26 cycles aren't always safe and that 'old' Game Gears (the two asics model?) are the ones showing glitches if we write data 'so fast'.
I'm going to check again what was written in the official documentation, and I think we should stick to that... edit: the Game Gear official docs here seems to miss one page, 'vdp-05' that is just where I think that information is. Master System docs instead say 29 cycles are needed. Also, I might be able to develop a quick test that checks if a Game Gear is a single asic or two asics one, if this proves to be consistent. |
|
|
Posted: Fri Jan 29, 2021 12:42 pm |
speaking about fixing Phantasy Star anyway, Maxim might just want to add a single NOP into the many 26-cycles OUTI / JP loops so that they become
-:
OUTI ; 16 NOP ; 4 JP - ; 10 which should slow down enough (as that would take 30 cycles) ... or if he prefers a 29-cycles loop there's even the option of using -:
OUTI ; 16 DJNZ - ; 13 but paying attention to double the loop counter beforehand since both operations decrement the B register. |
|
|
Posted: Fri Jan 29, 2021 12:53 pm |
Funny enough, I never played my GG with batteries, so I don't know if it also happened with batteries. |
|
|
Posted: Fri Jan 29, 2021 12:54 pm |
You're a good brother :) |
|
|
Posted: Fri Jan 29, 2021 12:55 pm |
There are many such loops in the game, but all of them should (?) be happening during the inactive display period. Inserting delays into them will be hard.
Could you make a tester that exercises different access speeds (in and out of active display) and verifies which produce errors? |
|
|
Posted: Fri Jan 29, 2021 2:37 pm |
I don't expect troubles with 26 cycles timing during the inactive display period, given OUTI blocks use 16 cycles timing fine, as far as we know. I suspect there's some VRAM write operation that runs during active display instead, and that uses 26 cycles timing which causes problems on some (very few? a whole class of them?) Game Gears.
OK, I started with something simple. This writes 128 16-bit words to PNT, with 26 cycles between first and second byte writes in the first test, and 27 cycles in the second test, always within active display. It's attached (with sources). If we will need more tests I'll add them later. I need you guys with 'faulty' Game Gears to test this and report the results - thanks! |
|
|
Posted: Fri Jan 29, 2021 10:17 pm |
and
Thanks for replying! so glad to get confirmation that this isn't my system or everdrive and that it's actually some sort of hardware bug! I haven't actually noticed it on my everdrive menu, but I'll keep an eye out for it. I've only seen it in master system specific games or software so far. Also, I can confirm it happens the same with batteries or AC power. So far I've found Phantasy star (and hacks) Outrun SMSTestSuite SMSMemTest I plan to look for more now that we've confirmed it's actually multiple systems. @sverx, I finally got home to test this out! 26 cycles seems to fail about 1/2 to 3/4 of the time on my system. 27 has not failed at all. Also, I figure since we're started to get confirmation maybe can get serial numbers and info from confirmed systems. Mine is a twin ASIC with a "837-7719-01 USA" board. it has a silver finish inside (I've read that means it's japanese made?) and has serial number B20007984. seems to be an earlier model from what I've read. I can definitely take picture inside or include specific numbers on chips too. EDIT: just noticed bsittler included his info with his picture. he has a japanese model, while mine is US. the confirmed serials are B0051252 B20007984 Another person on reddit has a japanese model game gear with a 837-7398 main board. his serial is B10469708 and his console does NOT have this problem. |
|
|
Posted: Sat Jan 30, 2021 2:21 am |
Thanks for the test ROM! This is really strange! I tested it with six different EverDrives (two different pre-"X" EverDrive GG's, a pre-"X" Master EverDrive, an EverDrive GG X7, a Master EverDrive X7, and a Chinese clone of the pre-"X" Master EverDrive with microSD slot), two different adapters for the SMS EverDrives/clone (Nuby Converter and Master Gear Converter), and five different Game Gears (Majesco US[?] Model 2110G serial AG91074495, Sega US[?] Model 2110 serial 031362818, McWill-modded Sega EUR Model 2110-50 serial B31243347, McWill-modded Sega EUR Model 2110-50 serial K27111288, and early Sega JP Model HGG-3210 serial B0051252.) In the one Game Gear where I ever got non-passing results (the early JP model) I get different results depending on which EverDrive I used! With the pre-"X" EverDrive GG's it seems to pass 100% of the time and I never see any corruption. With all the other flash cartridges, and regardless of how connected (in the SMS case), I see occasional-to-frequent corruption. With the flash carts that sometimes fail it is rare for the 26 cycle result to ever be other than "failed" (though it does happen sometimes), and sometimes the 27-cycle result is also "failed". Also while the visible VRAM corruption is most often dropped writes ("blanks" where a letter should be), sometimes it's more serious — either corrupted tiles, or a mostly or entirely blank screen. Sometimes both test results are shown as "passed" in spite of visible corruption on the results screen. Powering down the console for a few minutes and then powering it up again seems to result in different "typical failure rates" different times. Sometimes I see severe/obvious and frequent corruption, other times the corruption is infrequent. The console where I see failures fails both when connected to a DC adapter (with no batteries inserted), and when battery-powered. Other consoles tested work regardless of power source. Screenshots from the console that sometimes give non-passing results, taken using the Krikzz Master EverDrive connected through a Nuby converter, though other non-passing combinations behave similarly. The other consoles, in addition to always passing, never exhibit screen corruption with this test program. Likewise the two pre-"X" EverDrive GG's never exhibit screen corruption even in the console that exhibits failures with other flash cartridges. So, maybe some early VDP's are just rotten? edit: also tried another Majesco model 2110G serial AG91052431 and another Sega model 2110 serial 140021212 and they likewise always passed the tests |
|
|
Posted: Sat Jan 30, 2021 11:13 pm |
Forgot to mention: glitches appear on the early JP Game Gear using both the Master EverDrive X7 and GG EverDrive X7 regardless of whether the in-game menu feature is enabled | |
|
Posted: Sun Jan 31, 2021 6:08 am |
Perhaps when testing for these glitches, it may be useful to have a way to play the original game on the original hardware to see if the glitches occur in that context as well. If so, then it may indeed be some previously-unknown hardware issue caused by certain variables. Identify any possible variables, control for them, and then test one by one. |
|
|
Posted: Sun Jan 31, 2021 4:53 pm Last edited by sverx on Sun Jan 31, 2021 4:58 pm; edited 1 time in total |
This seems to confirm that there are systems where a 26 cycles wait between two writes can't be considered 'safe'. I wonder if there's a way to understand if this is just your specimen, say due to ageing/failure of components, or if the whole batch of Game Gears that are your very same model show the same behavior. :| @bsittler: I don't get if you have found a Game Gear where the "27 cycles wait test" fails regardless of the cartridge you're using. Also, I was quite expecting to see screen corruption (garbled text on display) on those Game Gears that fails on tests, as the code that writes text on screen is still considering 26 cycles safe. So I wouldn't expect to see much screen corruption and all passed test often (but theoretically it could sometimes happen). |
|
|
Posted: Sun Jan 31, 2021 4:57 pm |
I will run the tests some more and collect counts/statistics with different cartridge/console combinations. I don't understand why the cartridge used would matter in these tests unless the cause is something like a Vcc (power) sag. I certainly was seeing different behavior in the same console depending on which flash cartridge I used.
Agreed, it just makes accurate diagnosis more tricky! If the result display were always done more "safely" it would make more test runs usable for statistical purposes |
|
|
Posted: Sun Jan 31, 2021 5:05 pm |
I wonder, could you add test cases for 28 cycles and 29 cycles too? Given that I am sometimes (rarely) seeing failures with 27 cycles I'd like to understand what cycle count is truly "safe" with this combination of hardware | |
|
Posted: Sun Jan 31, 2021 5:10 pm |
small update - this also tests for 30 cycles wait, which should NEVER fail on a working Game Gear (if it does, I would think there's something wrong with VDP and/or VRAM). Also, 'FAILED' it's written in CAPS so that it's easier to see that even if just one char appear on screen ('passed' is still all lowercase)
edit: sorry, I didn't see your post. Let me see if I can add some 28 and 29 easily... |
|
|
Posted: Sun Jan 31, 2021 5:31 pm |
Update - it now tests for 26, 27, 28, 29, 30 cycles wait.
According to Master System official docs, 29 and 30 should *never* fail - if it does, you have an out-of-specs component somewhere. |
|
|
Posted: Sun Jan 31, 2021 5:32 pm Last edited by Maxim on Thu Dec 16, 2021 12:41 pm; edited 1 time in total |
I was thinking you could continuously test at a range of speeds and record a failure count. Many years ago I made a program that did this but relied on visual inspection to see any errors, and I think it did tend to be bad bytes rather than missing bytes at the higher speeds.
Note also that some real games exceed the 26 cycles time sometimes - presumably not in ways that are noticeable though. |
|
|
Posted: Sun Jan 31, 2021 5:41 pm |
well, I'm not really testing what kind of error is happening, just if VRAM contents are what are expected to be after writing there... | |
|
Posted: Sun Jan 31, 2021 6:40 pm |
tried out 1.2. Like before I only get fails on 26, at about the same rate. I never saw a fail on 27-30. | |
|
Posted: Sun Jan 31, 2021 6:56 pm |
Nice! I have a ton of questions, and hope you might be able to answer at least some of them to help characterize this better Which model of flash cartridge did you use, and how connected? If EverDrive, is it GG or SMS? And is it the old one, or one of the X series? Is it Krikzz or a clone? What does the EverDrive hardware/version info screen show? And if it's SMS, which converter do you use? Also, which screen do you use? And if the original LCD, is it still using the original backlight too, or a modern replacement? |
|
|
Posted: Sun Jan 31, 2021 8:13 pm |
Hey! It's a Everdrive GG X7. Only a couple months old. OS 1.04, FPGA v3, asm date 13.11.2020. The everdrive works perfect on my other 1ASIC unit. on the unit that fails, it now has a aliexpress LCD in it, but it had the same issues with the original LCD too. |
|
|
Posted: Sun Jan 31, 2021 10:10 pm |
but you're having issues with that Game Gear even using a Phantasy Star original cartridge and a Master Gear converter, correct?
Maxim: do the original ROM uses 26-cycles/write VRAM transfers? (if so, this means they violated SEGA technical specifications... and this puzzles me, honestly.) |
|
|
Posted: Sun Jan 31, 2021 11:21 pm |
This is a new approach.
I'm now writing a block of 256 bytes (a sequence of 128 pairs of two different bytes) to VRAM with *exactly* 26/27/28 cycles between all the writes and I now do this while interrupts are disabled. All this is to see if I can get less 'volatile' results (always fail a test, always pass another). Let me know the results, please. Also, I admit it wasn't easy to write two different bytes to VRAM in a tight loop of exactly 27 cycles, I came up with this: di
_loop_copy_27: ld a,l ; 4 out (0xBE),a ; 11 nop ; 4 nop ; 4 dec b ; 4 ld a,h ; 4 = 27 out (0xBE),a ; 11 jr nz, _loop_copy_27 ; 12 = 27 ei on the contrary, in the case of 26 and 28 it was straightforward (OUTI/JP loop the former, OUTI/JR loop the latter...) +++attachment removed, there was a mistake+++ |
|
|
Posted: Sun Jan 31, 2021 11:33 pm |
(disregard v1.3, there was a stupid bug)
|
|
|
Posted: Mon Feb 01, 2021 5:12 am |
This v1.4 ROM seems to be an "EverDrive detector" of sorts! The unused background of the screen is all "@" on EverDrive GG X7 or Master EverDrive X7 regardless of which Game Gear console I use. With older EverDrives (both SMS and GG) there are varying degrees of screen corruption, but at least the unused background is properly filled with spaces. Is this perhaps symptomatic of some incomplete initialization of RAM or VRAM? I'm still collecting data for both v1.2 and v1.4 ROMs, and will follow up when I have more. edit: Indeed, looks like it's showing some VRAM written by the EverDrive menu system |
|
|
Posted: Mon Feb 01, 2021 8:17 am |
Ran many more tests across five different Game Gear consoles and six different flash carts. With SMS flash carts I tried two different converters. I ran each suite around 100 times on each hardware combination, sometimes more. All tests were conducted on battery power. In-game menu was set to "off" for X7 EverDrives.
All of the newer consoles work fine and never fail the write tests. Only my old JP model Game Gear ever fails. Only other failures are that the Majesco Game Gear, the old JP Game Gear, and one of my non-modded US Game Gears won't flash ROMs on the old pre-"X" Krikzz Master EverDrive (full-sized SD slot with an adapter for microSD) though they will run ROMs flashed in another console, and the Master Gear Converter is incompatible with the Majesco Game Gear. Also the Nuby Converter won't physically fit in one of My McWill-modded European model Game Gears due to placement of an after-market VGA connector. All I can conclude is that either my old JP model Game Gear has a failing VDP, or some combinations of adapters and flash cartridges put it out-of-spec in terms of power leading to incorrect behavior. With old pre-"X" GG EverDrives I can't get it to fail, ever. With newer GG EverDrive X7, with Master EverDrive X7 in a Nuby converter or Master Gear Converter, with a Chinese clone Master EverDrive (microSD based) in a Nuby or Master Gear Converter, converter, and with an old pre-"X" Krikzz EverDrive (full-sized SD slot, though I use microSD in an adapter) in a Master Gear Converter, I get failures somewhere between 40% and 50% of the time for the "26 cycle" tests in both v1.2 and v1.4 of the test. Very rarely (less than 2% I think?) I see failures for the "27 cycle" tests in the v1.4 version too with the Master EverDrive X7. With an old pre-"X" Krikzz Master EverDrive (full-sized SD slot, though I use microSD in an adapter) in the Nuby converter, it fails somewhere around 70% of the time, and the failures also occur in 27-cycle, 28-cycle, 29-cycle and even 30-cycle write tests in v1.2! In v1.4 all tests sometimes fail. In both suites sometimes the screen is blank or almost blank, and sometimes characters are visibly corrupted. However powering off the console, waiting a few minutes, and then powering it back on returned it to more "normal" failures (26-cycle only, and around 40% to 50% of the time.) My suspicion: the different SMS converters, EverDrives, and possibly even microSD cards may have different side-effects due to power consumption somehow, OR I have a rotten VDP. Hard to figure out which with my limited test equipment, unfortunately. |
|
|
Posted: Mon Feb 01, 2021 11:32 am |
no, that's me that had removed the screen clear part by mistake... sorry, here's a fixed version. |
|
|
Posted: Tue Feb 02, 2021 6:02 am |
No worries, I was still able to read the results even in v1.4 despite the uncleared remnants. However I just ran a mini-test with v1.5, and I realized I had another way to test! I used one of the retrocircuits.com SMS flash carts from Mortoff Games (512 KB configuration, no SRAM or battery installed) flashed with v1.5 of your test. Even with the Nuby converter, this didn't fail even once in 150 attempts! In the same console and Nuby converter v1.5 of your test fails around half the time for the "26" test in a Chinese Master EverDrive clone (clone of pre-"X" Krikzz hardware but with microSD instead of full-sized SD card slot.) The other tests pass consistently. With a pre-"X" Krikzz EverDrive GG, v1.5 never fails in that same console. At this point my strong suspicion in that different cartridge power requirements may be to blame here. I guess I should install the SRAM and try flashing the Phantasy Star retranslation and get to the "problematic" area. edit: just tried the same flashed ROM cart (still flashed with v1.5 of your test) in the Master Gear Converter in that console, and it fails the "26" test around 50% of the time! The other tests pass consistently. So I guess the Master Gear Converter probably draws more power than the Nuby Converter somehow? |
|
|
Posted: Tue Feb 02, 2021 8:19 am |
I'll add a data point.
Twin Asic Game Gear with Everdrive GG X7 fails the 26 cycle test 50% of the time. |
|
|
Posted: Tue Feb 02, 2021 9:19 am |
You can play stock Phantasy Star with no SRAM, it just breaks saves. The retranslation uses the save RAM for some of the new settings and so may be broken in that scenario. | |
|
Posted: Tue Feb 02, 2021 10:08 am |
Nice! In fact I had already flashed the Retranslation to the cartridge and it does as least boot. I can't change fonts but at least music player and game play are possible |
|
|
Posted: Wed Feb 03, 2021 11:14 am |
I think this confirms that under best circumstances there should not be any VDP that fail at writing VRAM at 26 cycles/bytes. But that components age and additional hardware might drain power or anyway interfere to the point that the VDP can't keep the pace. |
|
|
Posted: Sat Feb 13, 2021 8:52 pm |
BTW I just got another Game Gear, an early no-BIOS European model that has been modded with a new screen (McWill-like but supposedly lower power consumption and with adjustable broightness), and it also behaves like my early JP model, intermittently failing the most stringent of the VDP tests when using a pre-X GG EverDrive. I haven't tried the test on the Mortoff Games flash cart with it yet since that one currently contains Phantasy Star Retranslation instead | |
|
Posted: Thu Feb 18, 2021 12:04 am |
Could have sworn I replied a while back, clearly didn't.
sverx, I don't have an adapter for trying an actual SMS cartridge yet. I hope some day, but don't currently have plans unless a good deal on a converter comes up. nyandeyanen over in the phantasy star retranslation thread mentioned he had also seen MANY twin asic models that have this issue, and also brought up a good point - apparently when the master gear converter was released, it was reported to NOT be compatible with Outrun and Phantasy Star. However, it seems like others get those games to work perfectly fine on a master gear converter. It makes me wonder if sega didn't know there was an issue on early units, since these games seem to be the one with issues? That said, bsittler being able to get one of these consoles to work with certain cartridges or adapters does make it seem like there's some other thing at play. |
|
|
Posted: Thu Feb 18, 2021 12:18 am |
FWIW the Master Gear Converter, Gear Master, and Nuby Converter seem to behave differently in the 2-ASIC Game Gears. The Nuby converter seems to work best in these, the Gear Master not at all, and the Master Gear Converter works but seems to exacerbate the VDP errors | |
|
Posted: Thu Feb 18, 2021 12:32 am |
So the standard "gear master" won't play them at all, while the "master" variant will play them (but with VDP issues)? | |
|
Posted: Thu Feb 18, 2021 1:00 am |
That was my experience, yes. I was frankly a little shocked as I had assumed those were simply rebranding of the same adapter! | |
|
Posted: Thu Feb 18, 2021 10:12 am |
what I don't know is if those Master System to Game Gear adapters are simply passive PCBs re-mapping pins from one system to another or if there's more than that.
I suspect a simple rewiring wouldn't change any timing and thus make some games incompatible but I'm absolutely no expert in this topic anyone knows details about how those adapters work? |
|
|
Posted: Thu Feb 18, 2021 1:35 pm |
They are all just physical adaptors. Some of the more exotic pins may not be the same on all. | |
Goto page 1, 2, 3 Next |