MiSTer FPGA Forum

Hi,

I observed the Neogeo has sometimes sound pitch issues in the music, like they are running too fast or too slow. it is not always happening.
observed this on two differents MiSTer, two differents microsd, two differents monitors, various controllers.

Is it a known issue, if so, is there any hope for a fix or workaround ?


Any advice welcome

Thank you

@madmax: Yes, it's a known issue (at least it's a very commented issue).
I have been seeing this years ago, on different MiSTer configurations. What I don't know is if the core author has even been informed. I guess he's aware, but I am not sure.

Yes it’s a very well known issue, documented here:

https://github.com/MiSTer-devel/NeoGeo_MiSTer/issues/60

Sadly there hasn’t been much activity or drive by core authors to fix it.

Ok. Thank you for the insight. I hope it can be fixed in the future.

Unfortunately it looks like Neo Geo core has been abandoned, pity as there are quite a few things to fix.

Insert Disk Two wrote: ↑Tue Jul 20, 2021 8:28 am Unfortunately it looks like Neo Geo core has been abandoned, pity as there are quite a few things to fix.

I am not sure if you are serious, the latest version is from last week and the latest change in the Github is from 5 days ago.

@Chris23235 All the updates over at least the past six months except one very small alignment fix are all around just keeping the framework updated. There hasn't been a meaningful functionality or bug fix for a long, long time.

What "meaningful functionality" do you miss in the core?

The remaining differences to the original system are marginal and getting these points right is most probably harder by the things getting more marginal. So the "small alignment fix" was when it comes to the differences of an original system not small at all.

All this talk about the core being "abondoned" sounds like the core was released a decade ago in half functional state and then left. But it is quite the opposite, the Neo Geo core is one of the most precise cores you can find.

The biggest reason there have been no real improvements is because of the way the core is constructed. As a 'reverse-forward-engineered' solution, it's actually more difficult to understand what the core is doing than it is to understand the original hardware. It's like copying chips by sampling every possible combinations of inputs and outputs and reducing the set...it *can* be done, but it's rarely a good idea because it's damned near impossible to ever debug.

I could not conceivably disagree more that it's one of the more 'precise' (and by that I mean accurate, and presume you do as well) cores on any FPGA. There's a reason it's got as many bugs as it does that are so hard to fix. I know where the trope began (from the author's self-aggrandizement), but I'd actually rather see it rewritten by hand from scratch and perfected over time than continue to exist in its current form. I've seen people sell off thousands of dollars of actual hardware because they read the NG core hype and drank the Kool-Aid. Not even close.

rhester72 wrote: ↑Tue Jul 20, 2021 7:22 pm I know where the trope began (from the author's self-aggrandizement), but I'd actually rather see it rewritten by hand from scratch and perfected over time than continue to exist in its current form. I've seen people sell off thousands of dollars of actual hardware because they read the NG core hype and drank the Kool-Aid. Not even close.

You are free to provide a more accurate core. What you call "many bugs" are just a small number of differences that can be noted when running the core side-by-side with a real NeoGeo system and with some of them it will remain unclear how much come from the framework of the MiSTer itself, because a real NeoGeo didn't output it's data to the monitor the way the MiSTer does.

I guess what you mean by "'reverse-forward-engineered" is that the core is built on decapping the original's chips and rebuilding them on the FPGA on the basis of scans. This is one possible way to recreate hardware in an FPGA and it is not as esoteric as you make it sound. Chip decapping and rebuilding in a hardware description language based on scans is the basis of many FPGA designs that are used in many cores. FX68K is based on scans and it is the backbone for a dozen or so cores.
If you say " I'd actually rather see it rewritten by hand from scratch and perfected over time than continue to exist in its current form" you should write your own NeoGeo core, going a different route then furrtek and proving that a more accurate core can be done.

Chris23235 wrote: ↑Tue Jul 20, 2021 5:32 pmThe remaining differences to the original system are marginal

I wouldn’t exactly label “audio that is painful to listen to” as marginal. It’s so out of pitch it triggers the same reaction in me as the worst of The Voice contestants.

rhester72 wrote: ↑Tue Jul 20, 2021 7:22 pm The biggest reason there have been no real improvements is because of the way the core is constructed. As a 'reverse-forward-engineered' solution, it's actually more difficult to understand what the core is doing than it is to understand the original hardware. It's like copying chips by sampling every possible combinations of inputs and outputs and reducing the set...it *can* be done, but it's rarely a good idea because it's damned near impossible to ever debug.

I could not conceivably disagree more that it's one of the more 'precise' (and by that I mean accurate, and presume you do as well) cores on any FPGA. There's a reason it's got as many bugs as it does that are so hard to fix. I know where the trope began (from the author's self-aggrandizement), but I'd actually rather see it rewritten by hand from scratch and perfected over time than continue to exist in its current form. I've seen people sell off thousands of dollars of actual hardware because they read the NG core hype and drank the Kool-Aid. Not even close.

Chris23235 wrote: ↑Tue Jul 20, 2021 8:06 pm You are free to provide a more accurate core. What you call "many bugs" are just a small number of differences that can be noted when running the core side-by-side with a real NeoGeo system and with some of them it will remain unclear how much come from the framework of the MiSTer itself, because a real NeoGeo didn't output it's data to the monitor the way the MiSTer does.

I guess what you mean by "'reverse-forward-engineered" is that the core is built on decapping the original's chips and rebuilding them on the FPGA on the basis of scans. This is one possible way to recreate hardware in an FPGA and it is not as esoteric as you make it sound. Chip decapping and rebuilding in a hardware description language based on scans is the basis of many FPGA designs that are used in many cores. FX68K is based on scans and it is the backbone for a dozen or so cores.
If you say " I'd actually rather see it rewritten by hand from scratch and perfected over time than continue to exist in its current form" you should write your own NeoGeo core, going a different route then furrtek and proving that a more accurate core can be done.

Since this is rapidly devolving into rampant fanboi-ism, I won't bother replying further.

Hey hey

https://github.com/MiSTer-devel/NeoGeo_ ... -897176697

ryosanisback wrote: ↑Thu Aug 12, 2021 2:32 pm Hey hey

https://github.com/MiSTer-devel/NeoGeo_ ... -897176697

There is a pull request merged 20 hours ago in the official Mister repo. Maybe it has something to do with this issue.

Hurray !

Can confirm too that the posted build https://github.com/MiSTer-unstable-nigh ... 1_ba00.rbf sounds much better indeed, tested with Pulstar and AOF3. Good news

Holy shit finally some movement on this! Can’t wait to try it out after work!

Yes this build is much improved imo, hopefully this gets included in the official main neo geo build soon, thanks for the link to this nightly release.

Yes!!! At long last, the wobbling sound is gone with this fix! I hope official inherits it soon!

Yeah, the sound seems much better on the 'unstable' build. Sadly, though, the issue with Zed Blade's title music bursting in at random during the stages currently persists!

It's up! The fix got merged today!

I fired Blazing Star a second after the download and I can attest it sounds perfect now!

Confirm that Viewpoint, which had great music on the original machine no longer sounds like a strangled cat! Kudos to the fixer!!!!

Anyone check the issue with Zed Blade yet?

Sarge wrote: ↑Thu Sep 09, 2021 11:01 pm Anyone check the issue with Zed Blade yet?

It's still not fixed.

Okay, thanks. I meant to check when I got home yesterday, but got sucked into other stuff.