FPGA Based Videogame System

[Sneakyturtleegg]

The good news: I have fixed the bugs! The bad news: I cannot ftp into my host to post the new update. blah. it's 8 megs so a bit big for an attachment I think. I will have it up when possible.

Can you post just the files or folders that need to be updated?

[Andromeda Stardust]

The good news: I have fixed the bugs! The bad news: I cannot ftp into my host to post the new update. blah. it's 8 megs so a bit big for an attachment I think. I will have it up when possible.

AtariAge allows up to 10mb for subscribers. Ask Albert nicely and he might host the tiles for you.

 

Oh, and +1 pledge my $5 per month for Kevtris Patreon. I can spare a Cappuccino for the cause!

[cacophony]

Not sure it actually helps but I did my setup with a class 10 SD card and I haven't seen any core load take more than ~20 seconds (eg. SMS core load). Interestingly, loading the NES core (for example after loading the SMS core) only takes about 3 seconds! Why so much quicker?

 

Also, entering directories with ~200 roms only takes a second or two.

 

Kevtris, you are amazing!

Edited March 12, 2017 by cacophony

[cacophony]

That sounds exactly like my problem I didn't modify the cable, but I did try 75ohm terminators but they didn't seem to be necessary for the pvm I tried and didn't make a difference, unfortunately.

 

My console just got to analogue on Friday so hopefully on Monday I'll get some news back! I know of 3 people with the same issue so far, I'm hoping analogue will offer easy replacements/repair to those who need it. I can imagine that shipping back from Sweden would be costly and frustrating..

 

It does seem to be a simple issue, maybe one of the pins just doesn't have a good connection. I thought about opening mine up to try to take a look but given the price, I just mailed it out. Maybe there's a simple soldering fix? It's hard for me to say.

 

Would the issue you had be detectable via a composite connection to an HD TV? I'd like to verify my unit doesn't have the problem but only have access to a 1080p TV at the moment (It has a composite connection that I used to use for my toaster NES)

Edited March 12, 2017 by cacophony

[Eicar]

@kevtris It looks like something is wrong with the vertical scaling on the NES core as well. If I select 4,5x then it looks like the last two nes scanlines are missing (goes below the screen) I also get two black NES scanlines on the top.

The vertical position slider also seems to no longer work. Nothing happens when I change it. I'm running HDMI @ 1080p in NTSC mode. Tested with Super Mario Bros 1, you can easily see the brick on the buttom (title screen) is chopped off.

 

BTW: Do you have any plans for supporting FDS titles without using the Everdrive (or the real FDS hardware) ?

[Schizophretard]

Do you think that the Zimba 3000 might be useful for serious hardware preservation? Maybe some tech museums would be interested in your projects (a grant $$$)? Ask around. I think that in the future you could also try to reproduce analog outputs of other consoles with a similar level of precision as you did with the NES. Is there a better way to convince people to FPGA consoles than connecting an original console and an FPGA one to the same CRT TV model and showing that the output looks identical? Analogue should create such a comparison video for the Nt mini... This and a lag test...

 

This makes me wonder if someday when CRT's become extremely rare and need hardware replacements and/or repairs for preservation similar to how consoles do if they will also get the FPGA treatment. I'm thinking something alone the lines of the tube and case of let's say a Sony Trinitron staying original but if you open it up instead of a big mess of wiring there is an FPGA board in there leaving most of the case empty. And the FPGA wouldn't just replicate the internals of a Sony Trinitron but the internals of every CRT with all of their features. By features I mean the specs we used to look at when choosing a CRT like does it have picture and picture, does it have a 3D comb filter, is it NTSC or PAL, does it have a degausser, etc. In other words, it wouldn't lack any features that every CRT had and therefore it would be like the ultimate universal CRT FPGA board replacement that could be fitted into any make and model of CRT. The cases could even be modified for more video inputs and maybe internally have bigger and better speakers since there would be more room for them. What displays on the screen when accessing the menus with the remote would be like playing the ROM's and/or cores of all CRT's.

[atmn]

That sounds exactly like my problem I didn't modify the cable, but I did try 75ohm terminators but they didn't seem to be necessary for the pvm I tried and didn't make a difference, unfortunately.

 

My console just got to analogue on Friday so hopefully on Monday I'll get some news back! I know of 3 people with the same issue so far, I'm hoping analogue will offer easy replacements/repair to those who need it. I can imagine that shipping back from Sweden would be costly and frustrating..

 

It does seem to be a simple issue, maybe one of the pins just doesn't have a good connection. I thought about opening mine up to try to take a look but given the price, I just mailed it out. Maybe there's a simple soldering fix? It's hard for me to say..

 

Thanks for the information.

Now Im trying to make Analogue to take my issue seriously. Still referring to an unsupported cable. Really hope they take this matter seriously.

 

Im not a total novice when it comes to RGB, modded a few PCE to output euroscart-RGB, thinking the same as you did, probably some bad connection to pin 1. Maybe a super easy fix. Not afraid to do the fix myself if guided to the source of the problem.

BUT for 500USD I would expect some warranty..

 

Its a really awesome piece of hardware otherwise.

Edited March 12, 2017 by atmn

[atmn]

Anyone more having issue with the japanese Holy Diver, playing with the NES/Famicom core?

 

The graphics is messed up just right away when start playing, title screen seems OK.

 

Guess it's the namco Irem only mapper that is using that.

Working with the actual cart.

Edited March 12, 2017 by atmn

[kevtris]

@kevtris It looks like something is wrong with the vertical scaling on the NES core as well. If I select 4,5x then it looks like the last two nes scanlines are missing (goes below the screen) I also get two black NES scanlines on the top.

The vertical position slider also seems to no longer work. Nothing happens when I change it. I'm running HDMI @ 1080p in NTSC mode. Tested with Super Mario Bros 1, you can easily see the brick on the buttom (title screen) is chopped off.

 

BTW: Do you have any plans for supporting FDS titles without using the Everdrive (or the real FDS hardware) ?

Yes I want to support FDS natively, I just haven't gotten around to it yet. At the time of the design, FDS information wasn't available so I never added it. I have the scaling bug fixed, but unfortunately my webhost is not letting me upload files. It just times out trying to connect and I don't know why. I tried it again and it's still dead. I guess I will have to contact support if it isn't working tomorrow/later today.

 

Not sure it actually helps but I did my setup with a class 10 SD card and I haven't seen any core load take more than ~20 seconds (eg. SMS core load). Interestingly, loading the NES core (for example after loading the SMS core) only takes about 3 seconds! Why so much quicker?

 

Also, entering directories with ~200 roms only takes a second or two.

 

Kevtris, you are amazing!

Yeah I am only reading the card at around 140K/second right now so a faster/slower card doesn't matter yet. I hope to vastly improve this by at least a factor of 5.

 

Can you post just the files or folders that need to be updated?

I don't think I can.

 

Anyone more having issue with the japanese Holy Diver, playing with the NES/Famicom core?

 

The graphics is messed up just right away when start playing, title screen seems OK.

 

Guess it's the namco Irem only mapper that is using that.

Working with the actual cart.

This is because you need the proper nes 2.0 submapper set on the ROM header to make it work. I believe this is in the nes readme.

[kevtris]

My host is still having problems, but increasing the timeout from 20 seconds to 120 for connecting worked, took about a minute to connect and another to upload. Anyways, I hope this fixes the problems. Let me know if it doesn't.

 

http://blog.kevtris.org/blogfiles/ntm_firmware_verJB1.6.zip

[Eicar]

@kevtris The scaling seems to work better now, however there's another small snag. If I go into the nes core menu when I'm running SMB1 and toggle back and forth between 'none' and 'hq2x' then the screen will be misalligned on the buttom. This means that going back and forth several times will change the position. (not beween two postitions, but many..)

It's a bit hard to explain what's going on. If you're not able to see what I mean then I can take some images. However, it's quite easy to see when running SMB1 in 4,5x vert scale mode (hdmi@1080p) and looking at the bottom of the screen when you change back and forth between 'none' and 'hq2x'.

 

Sorry to throw all this at you as you probably have better things to do. I just wanted you to know..

[cacophony]

@kevtris The scaling seems to work better now, however there's another small snag. If I go into the nes core menu when I'm running SMB1 and toggle back and forth between 'none' and 'hq2x' then the screen will be misalligned on the buttom. This means that going back and forth several times will change the position. (not beween two postitions, but many..)

It's a bit hard to explain what's going on. If you're not able to see what I mean then I can take some images. However, it's quite easy to see when running SMB1 in 4,5x vert scale mode (hdmi@1080p) and looking at the bottom of the screen when you change back and forth between 'none' and 'hq2x'.

 

Sorry to throw all this at you as you probably have better things to do. I just wanted you to know..

 

I actually recall seeing this continuous position shift while toggling behavior on the factory firmware as well, but I can't remember the specifics. Are you sure it's a recent regression?

 

FYI, that vertical position slider is only for 5x height scaling.

Edited March 12, 2017 by cacophony

[rezb1t]

 

Would the issue you had be detectable via a composite connection to an HD TV? I'd like to verify my unit doesn't have the problem but only have access to a 1080p TV at the moment (It has a composite connection that I used to use for my toaster NES)

Yep! I first noticed it with composite and component. With component, the colors were all messed up and I noticed unplugged the red cable(R-Y) did nothing to the colorspace

 

 

Thanks for the information.

Now Im trying to make Analogue to take my issue seriously. Still referring to an unsupported cable. Really hope they take this matter seriously.

 

Im not a total novice when it comes to RGB, modded a few PCE to output euroscart-RGB, thinking the same as you did, probably some bad connection to pin 1. Maybe a super easy fix. Not afraid to do the fix myself if guided to the source of the problem.

BUT for 500USD I would expect some warranty..

 

Its a really awesome piece of hardware otherwise.

Yeah getting them to take it seriously in my case was difficult too :/ I ended up testing 4 tvs, 3 different capture devices and had to order a few extra cables from Monoprice. I love the device too but this is a defect in a few units that needs addressing

Edited March 12, 2017 by rezb1t

[Eicar]

 

I actually recall seeing this continuous position shift while toggling behavior on the factory firmware as well, but I can't remember the specifics. Are you sure it's a recent regression?

 

FYI, that vertical position slider is only for 5x height scaling.

I'm testing firmware 1.6 now. Yes, the 5x height scaling slider works as it should now. The problem I'm talking about now is about the fancy upscalers, hq2x, hq3x, x-ray etc. The image will shift slightly if you go back and forth between 'none' and hq2x.

I'm not sure if this was a problem before. I havn't seen it before...

Edited March 12, 2017 by Eicar

[Eicar]

@kevtris Is there any reason why I get the "load core yes/No" prompt everytime I go from any NES game and try to load Castlevania III ?

Does it require a "special" NES core? However, I don't get the "load core" when I go to another game afterwards...

[Dweezicus]

Anyone more having issue with the japanese Holy Diver, playing with the NES/Famicom core?

 

The graphics is messed up just right away when start playing, title screen seems OK.

 

Guess it's the namco Irem only mapper that is using that.

Working with the actual cart.

Yeah, I am having the same issue with Holy Diver playing directly off of the SD card. I haven't messed around with the rom headers yet, just play it off my everdrive instead (having save states for this game makes it a little less frustrating anyway). On a side note, that game is seriously like 10x more amazing if you play it while listening to the Dio album of the same name. Honestly, I wish someone would create a rom hack with chip tune versions of Holy Diver instead of the original soundtrack.

[retro_fan]

Honestly, I wish someone would create a rom hack with chip tune versions of Holy Diver instead of the original soundtrack.

Maybe instead of the chip tune version you would like to listen to the real deal? It seems that the NES can actually play high-quality audio (it starts at 9:57):

 

https://www.youtube.com/watch?v=tA9nlUtepR4

Edited March 12, 2017 by retro_fan

[Kismet]

 

This makes me wonder if someday when CRT's become extremely rare and need hardware replacements and/or repairs for preservation similar to how consoles do if they will also get the FPGA treatment. I'm thinking something alone the lines of the tube and case of let's say a Sony Trinitron staying original but if you open it up instead of a big mess of wiring there is an FPGA board in there leaving most of the case empty. And the FPGA wouldn't just replicate the internals of a Sony Trinitron but the internals of every CRT with all of their features. By features I mean the specs we used to look at when choosing a CRT like does it have picture and picture, does it have a 3D comb filter, is it NTSC or PAL, does it have a degausser, etc. In other words, it wouldn't lack any features that every CRT had and therefore it would be like the ultimate universal CRT FPGA board replacement that could be fitted into any make and model of CRT. The cases could even be modified for more video inputs and maybe internally have bigger and better speakers since there would be more room for them. What displays on the screen when accessing the menus with the remote would be like playing the ROM's and/or cores of all CRT's.

 

It's not going to be possible. The reason CRT's die has more to do with the phosphors wearing out (which are toxic) and the CRT's leaded glass being toxic, to reduce x-ray emissions. To say nothing of the amount of voltage in the power supply that can kill someone. The problem is not the board, it's the CRT tube itself. You have about 50,000 hours max on a CRT or Plasma, and that translates into either 12 years of heavy usage, 6 years of 24 hour usage, or 24 years of typical usage. Usually the power supply goes BANG (and maybe lets out some smoke) before the CRT becomes too dim to use, but we're talking about the difference of a PSU failure versus the CRT tube itself inevitable becoming useless.

 

What is likely going to replace CRT's for arcades are curved "trinitron" large-pitch OLED style screens with a specialized anti-glare filter. But until someone comes up with that, the more likely things are just conventional TN-based LCD screens, with the mirror being removed from the cabinet to ensure the viewing angle is useful.

 

What would be useful, albeit limited audience would be to design a LCD/OLED driver that connects directly to the LVDS that uses the kind of line-by-line scaling found in the OSSC. That would eliminate all the latency, but the tradeoff is that it likely would only be useful for making arcade monitor replacements, since building your own LCD monitors would still involve buying the panels from LG/Samsung/Japan Display/etc

[atmn]

This is because you need the proper nes 2.0 submapper set on the ROM header to make it work. I believe this is in the nes readme.

Ops, sorry. Should have checked the documentation before. Think I'll go for the MMC1 patch guess that will do it.

 

 

 

Yep! I first noticed it with composite and component. With component, the colors were all messed up and I noticed unplugged the red cable(R-Y) did nothing to the colorspace

 

Yeah getting them to take it seriously in my case was difficult too :/ I ended up testing 4 tvs, 3 different capture devices and had to order a few extra cables from Monoprice. I love the device too but this is a defect in a few units that needs addressing

 

You have right, pin 1 are active using all the analogue outputs, shared the same pin using the detection pins in various ways. If pin1 fails all analogue signals are broken..

 

When I start mine console I get the text red and then white (all signals are there) and then it goes greenish. The transformation from red-normal-greenish are like in a split second.

 

 

 

Since you have sent your in they will have a confirmed case and also they are aware of the problem on certain units... I hope.. Guess there is some flaws with testing the units, bad connection is not totally uncommon, but analogue signals are more and more uncommon. (Harder to test)

 

Yeah, I am having the same issue with Holy Diver playing directly off of the SD card. I haven't messed around with the rom headers yet, just play it off my everdrive instead (having save states for this game makes it a little less frustrating anyway). On a side note, that game is seriously like 10x more amazing if you play it while listening to the Dio album of the same name. Honestly, I wish someone would create a rom hack with chip tune versions of Holy Diver instead of the original soundtrack.

Hehe they most likely are unofficial ment to be. :

Totally stuck on level 4, so save states are very handy.

[kevtris]

I'm testing firmware 1.6 now. Yes, the 5x height scaling slider works as it should now. The problem I'm talking about now is about the fancy upscalers, hq2x, hq3x, x-ray etc. The image will shift slightly if you go back and forth between 'none' and hq2x.

I'm not sure if this was a problem before. I havn't seen it before...

 

that's normal, it has always done that. It has to do with how the scaling maps input pixels to output pixels.

 

@kevtris Is there any reason why I get the "load core yes/No" prompt everytime I go from any NES game and try to load Castlevania III ?

Does it require a "special" NES core? However, I don't get the "load core" when I go to another game afterwards...

yes, because there's a separate core for MMC5 (castlevania 3) and a couple other high complexity mappers vs. all the other mappers. So when you play CV3 it has to load the secondary core, then when you play something not MMC3 or mapper 90 (or one of about 4 mappers) it has to load the "stock" NES core. once it's loaded it doesn't need to load it again. This is a limitation of the size of FPGA in use. I was hoping to squeeze it all into a single FPGA but ended up not being able to.

[kevtris]

 

I actually recall seeing this continuous position shift while toggling behavior on the factory firmware as well, but I can't remember the specifics. Are you sure it's a recent regression?

 

FYI, that vertical position slider is only for 5x height scaling.

 

@kevtris The scaling seems to work better now, however there's another small snag. If I go into the nes core menu when I'm running SMB1 and toggle back and forth between 'none' and 'hq2x' then the screen will be misalligned on the buttom. This means that going back and forth several times will change the position. (not beween two postitions, but many..)

It's a bit hard to explain what's going on. If you're not able to see what I mean then I can take some images. However, it's quite easy to see when running SMB1 in 4,5x vert scale mode (hdmi@1080p) and looking at the bottom of the screen when you change back and forth between 'none' and 'hq2x'.

 

Sorry to throw all this at you as you probably have better things to do. I just wanted you to know..

I don't think I have seen it change between different positions. you mean if you change between one and hqx, it will be in different positions each time? or you mean changing the width doesn't seem to change behavior? I'm sorta confused. I need to be able to reproduce this so I can fix 'er.

[Eicar]

 

that's normal, it has always done that. It has to do with how the scaling maps input pixels to output pixels.

 

yes, because there's a separate core for MMC5 (castlevania 3) and a couple other high complexity mappers vs. all the other mappers. So when you play CV3 it has to load the secondary core, then when you play something not MMC3 or mapper 90 (or one of about 4 mappers) it has to load the "stock" NES core. once it's loaded it doesn't need to load it again. This is a limitation of the size of FPGA in use. I was hoping to squeeze it all into a single FPGA but ended up not being able to.

I understand that the mapping of input and output pixels might cause a shift. However, if I start out with 'none', then go to 'hq2x' and then go back to 'none' then I would expect to have the same image alignment which I started with. However, this is not the case. I need to go back and forth between 'none' and 'hq2x' several times in order to get back to the correct starting alignment.

Just try Super Mario Bros 1 @ 4,5x (vert scale) (1080p) and look at the bottom of the screen. Then go back and forth between 'none' and 'hq2x'. You can see that the bottom of the two bricks are sometimes cut off, and sometimes it's not. If you are still unable to reproduce this I can make a video of it.

[cacophony]

I understand that the mapping of input and output pixels might cause a shift. However, if I start out with 'none', then go to 'hq2x' and then go back to 'none' then I would expect to have the same image alignment which I started with. However, this is not the case. I need to go back and forth between 'none' and 'hq2x' several times in order to get back to the correct starting alignment.

Just try Super Mario Bros 1 @ 4,5x (vert scale) (1080p) and look at the bottom of the screen. Then go back and forth between 'none' and 'hq2x'. You can see that the bottom of the two bricks are sometimes cut off, and sometimes it's not. If you are still unable to reproduce this I can make a video of it.

 

 

I don't think I have seen it change between different positions. you mean if you change between one and hqx, it will be in different positions each time? or you mean changing the width doesn't seem to change behavior? I'm sorta confused. I need to be able to reproduce this so I can fix 'er.

I can confirm that it does this too. It doesn't require 4.5x height (I just did it with a 5x height) and it does it for all the scalers (hq3x, hq4x, etc).

 

Just switch between hq3x, none, hq3x, none, hq3x, none, etc

or hq4x, none, hq4x, none, hq4x, none, etc

 

And as you're doing that watch how the image sometimes sometimes shifts left, sometimes right, sometimes not at all. Kinda odd, but I don't use scalers so never really noticed it before.

Edited March 12, 2017 by cacophony

[kevtris]

OK thanks for the info on how to repro it. I will see what's going on tonight.

[coderkind]

Has anyone tested RGB on a Nt mini since the last restock? I received my unit last week and it's performed great over HDMI, but Retro Gaming Cables have been out-of-stock on their SCART cables for the past few weeks as far as I can tell and it'd be good to verify if my console has the analog fault. Postage of the various recommended Monoprice cables to the UK is extremely expensive, but I might have to spend the money it seems.

 

I'm surprised the analog issue hasn't been picked up on by any professional reviews (or maybe it was a very small batch of the consoles that were affected?).