FPGA Based Videogame System

[Andromeda Stardust]

You bring up an excellent point Kimset. The 8-bitdo is very dependant upon the wireless environment. An apartment with crouded networks, forget about it. A suburban home with a single Wifi connection and hardly anything using it? Clear sailing...

[leods]

Just a couple problems with the above posts:

 

The older 8bit do Bluetooth controllers were showing no extra lag. The new one, that is 2.4 ghz tech but not Bluetooth might be showing the problem. It could be a thing of the dongle. As I said, I don't actually know what's going on, I'd just take a look if at all possible to make sure everything's allright, or make necessary corrections to the marketing material.

 

Testing seems to have shown modern console wireless controllers have no added latency compared to wired controllers for the same platform. I think I specifically saw in the Shoryuken forums a comparison between the best xbox 360 wired arcade sticks and the 360 wireless controller, and if it showed any lag it was under one frame of extra lag as compared to the arcade stick. If I remember correctly though, it was basically no lag. The reason tournaments use wired are two: first, no pairing required. You plug your controller in and you play. Second: by the time you have 20 controllers trying to work wirelessly at the same time you'll run into trouble.

 

I also prefer wired when it's practical. It's more relyable and is a no frills long lasting solution. But the input lag generated by the wireless ignal will be dependent on implementation. Even multiple conversion don't need to take that long. Even if you need somehting like a ms to convert one way and a ms to convert back, that much input lag just doesn't matter. Many wired controllers actually show more input lag than some wireless solutions. So again. It just depends on implementation.

[Kismet]

 

I also prefer wired when it's practical. It's more relyable and is a no frills long lasting solution. But the input lag generated by the wireless ignal will be dependent on implementation. Even multiple conversion don't need to take that long. Even if you need somehting like a ms to convert one way and a ms to convert back, that much input lag just doesn't matter. Many wired controllers actually show more input lag than some wireless solutions. So again. It just depends on implementation.

 

That's the thing though, the build quality is not a consistent thing from vendor to vendor, or even product to product, even the software driver can be poor. People have been trained to buy products based on price, not the underlying quality. As a general consideration, a BT device can not have less latency than the same device wired, because the wired device is not supposed go through the radio. However it's entirely possible that simply "plugging in" a wireless device, only activates the charging circuit, and the wireless device continues to communicate over bluetooth.

[F34R]

Anecdotal. The PCB's are all wrong, period.

What's wrong with it?

[Great Hierophant]

 

USB introduces lag when it goes through the ASIC. It's usually not a measurable amount of latency, but depends on the chip used. USB operates, or can operate faster than the SNES controller.

 

A NES or SNES controller can be read extremely quickly on real hardware, it only takes a couple of hundred machine cycles to get the state of the buttons. 1 machine cycle = ~560ns. So it may take 200us to get the state of the buttons. USB polls at a maximum of 1000Hz, or 1ms and you have to deal with the protocol, the OS driver and the translation if using an emulator.

[Guspaz]

What's wrong with it?

 

The contact pattern for the SNES30 controllers is different than original controllers, which affects how likely they are to detect diagonal inputs. This may or may not apply to the new SN30 controllers (the one with the horizontal start/select button layout), AFAIK nobody has done a teardown of those yet.

 

People who deride the 8bitdo controllers as useless because of the PCB layout tend to be a very small minority of hardcore enthusiasts who don't reflect the vast majority of users. Most people are quite happy with them. I can notice a difference from a real SNES controller, but it doesn't bother me much in practice.

Edited December 8, 2017 by Guspaz

[cfillak]

I know this isn't NT Mini, Super NT, Analogue or Kevtris related but I think it's interesting:

 

https://www.neosdstore.com/shop/index.php?id_product=12&controller=product&id_lang=1

 

 

It's basically the final solution for PC Engine CD. It's a board that plugs into the expansion port of any PC-Engine (including CoreGrafx and SuperGrafx I believe). It provides RGB out, an SD card slot for loading HuCard ROMs and ALL CD-ROM images including CD-ROM2, Super CD-ROM2, and Arcade CD-ROM2, all without needing the system cards. So it acts as an RGB solution, an Everdrive, and an optical drive emulator all in one. I believe the ODE functionality is all implemented on an FPGA.

 

This makes things interesting for Kevtris's planned TG-16/PCE core. I wonder if Analogue thinks the TG-16/PCE has enough popularity to carry its own product. If they do ever release a Turbo NT, and the jailbreak doesn't have CD-ROM functionality built in, I hope the system would at least have the expansion port to allow compatibility with the Super SD System 3.

[F34R]

 

The contact pattern for the SNES30 controllers is different than original controllers, which affects how likely they are to detect diagonal inputs. This may or may not apply to the new SN30 controllers (the one with the horizontal start/select button layout), AFAIK nobody has done a teardown of those yet.

 

People who deride the 8bitdo controllers as useless because of the PCB layout tend to be a very small minority of hardcore enthusiasts who don't reflect the vast majority of users. Most people are quite happy with them. I can notice a difference from a real SNES controller, but it doesn't bother me much in practice.

What game(s) can I test this on? Mine is the NES30.

[kevtris]

 

A NES or SNES controller can be read extremely quickly on real hardware, it only takes a couple of hundred machine cycles to get the state of the buttons. 1 machine cycle = ~560ns. So it may take 200us to get the state of the buttons. USB polls at a maximum of 1000Hz, or 1ms and you have to deal with the protocol, the OS driver and the translation if using an emulator.

Most games only poll at 60Hz and use the hardware controller reading which is quite slow. It takes around 200uS. it's quite a bit faster to read it in code actually. If you're using slowrom (2.6xMHz, 8 clocks/cycle) this is 32 cpu cycles/button. You can do the read/shift function in around 18 or so (LDA 4016, AND #01, LSR, ROL buttonreg, DEX, BNE loop) cycles. Of course, this means you have to do it in code. The anti-lag methodology that the nt mini uses is to read the controllers about 12.5% before vblank, when most games read them so that "fresh" data is always available. This way it doesn't have to be polled thousands of times a second to reduce the lag by increasing sample rate. The various cores use this "smart" polling to reduce latency.

[Tusecsy]

What's wrong with it?

 

The 4 directions have split pads on the original PCBs, so when you're hitting diagonals its say half up and half right. On all the 8bitdo controllers its just right on/off, up on/off. It creates a totally different feel, completely obvious to anyone with even basic experience of the original controllers. You'll hear a lot of complaints about the D-Pad for this reason. It's just lazy/cheap manufacturing, which runs through everything 8bitdo does.

 

But hey, form over function is the new way to go when it comes to electronics ::cough:: cell phones ::cough::, so maybe I'm just a relic.

[Tusecsy]

I know this isn't NT Mini, Super NT, Analogue or Kevtris related but I think it's interesting:

 

https://www.neosdstore.com/shop/index.php?id_product=12&controller=product&id_lang=1

 

 

It's basically the final solution for PC Engine CD. It's a board that plugs into the expansion port of any PC-Engine (including CoreGrafx and SuperGrafx I believe). It provides RGB out, an SD card slot for loading HuCard ROMs and ALL CD-ROM images including CD-ROM2, Super CD-ROM2, and Arcade CD-ROM2, all without needing the system cards. So it acts as an RGB solution, an Everdrive, and an optical drive emulator all in one. I believe the ODE functionality is all implemented on an FPGA.

 

This makes things interesting for Kevtris's planned TG-16/PCE core. I wonder if Analogue thinks the TG-16/PCE has enough popularity to carry its own product. If they do ever release a Turbo NT, and the jailbreak doesn't have CD-ROM functionality built in, I hope the system would at least have the expansion port to allow compatibility with the Super SD System 3.

 

Thanks for the post, heard about this thing, now the wait for a cheaper solution. Is there a review anywhere?

 

Analogue isn't gonna waste their time on repackaged Super NT FPGA's. That's something garbage companies like Retron do. I doubt programming the CD-Rom core on top of the regular TG16 one is too much of a leap for someone on Kevtris's level but I'm not that well versed on the difference.

Edited December 8, 2017 by Tusecsy

[cacophony]

What game(s) can I test this on? Mine is the NES30.

Contra is a good one because "down" causes your character to duck and "down/left" or "down/right" causes you to stand with a diagonal downward firing. My NES30 interprets even the slightest hint of left/right while pressing down as a down/left or down/right, which make Contra really frustrating to play. My original NES controls do not do this.

[leods]

Well. I took a look at the controller testings for different consoles, and interestingly enough the PS4 wireless controller is 2ms slower than the best possible board for the PS4. This testing is made by wiring controllers to the same button and comparing in game lag. 2ms means that most of the time your actions are going to happen in the same frame. There is no licensed controller that is more than 0.3ms faster than the DS4 in wireless mode. It's most likely your lowest lag option.

 

Links to some testing:

 

http://www.teyah.net/sticklag/overview.html

 

http://www.teyah.net/sticklag/results.html

 

Then a guy tested the Nintendo Switch controller with cable and wireless. Guess what? The way it's implemented the wireless is faster:

 

 

So. I am not here to claim that Wireless is somehow faster. I'm also not here to deny wired should in theory be faster. But I am here to say that 1: actual real world implementation is so varied that you can't just assume wired is faster. 2: wireless will add negligible lag if well implemented. So again, don't just assume that wireless is going to be bad, because even if it actually does add a couple miliseconds to your latency, that is something not even a world champoin in any game would actually be able to notice.

 

I don't want to really extend the discussion much, I think this is loosely related to the topic at hand, but not that much that it should occupy various pages. I do think though, that actual research and real information (tests with accurate results achieved with good testing) might inform people on what actually goes on.

 

And to finish: Modern gaming seems to involve a lot more latency than classic gaming. Even modern fighting games often suffer from added frames of lag, and gaming in general seems to happen at higher lag overall than classic gaming did. For purists the whole lag consideration should be taken seriously. Chose carefuly what monitor/TV you plug your Analogue NT / SNT / Zimba 3000, cause that will have huge impact on your experience. LED TV and monitor latency vary wildly between 10ms (less than one frame) to upwards of 100ms (around 6 entire frames).

 

I'm still curious about more specifics on the 8bitdo controllers.

 

And now I'm actually also curious about SNES controller implementation, and wether different models have different latency etc...

 

 

edit:

 

From this topic:

http://forums.shoryuken.com/discussion/49295/do-wireless-360-and-ps3-pads-lag-any-reason-to-stick-to-a-wired-pcb

 

the results:

 

"In short, both the Cthulhu and Sixaxis are, in my opinion based on these tests, neck and neck. I'll be using the 1ms updating code in all of the boards I send from here on out; no one will notice the change, I dont care how hardcore you are, but it will make me feel better sending them out with that extra 0.001% advantage. What I can say for sure is that anyone using a Sixaxis as the core of their board has nothing to fear from latency. I am seriously impressed with Sony that they were able to pull that off; I did not expect it to perform that well at all. I don't know whether to feel sad that my board didn't wipe the floor with the sixaxis, or proud that my board goes toe to toe with Sony engineering. *shrug*"

Edited December 8, 2017 by leods

[F34R]

 

The 4 directions have split pads on the original PCBs, so when you're hitting diagonals its say half up and half right. On all the 8bitdo controllers its just right on/off, up on/off. It creates a totally different feel, completely obvious to anyone with even basic experience of the original controllers. You'll hear a lot of complaints about the D-Pad for this reason. It's just lazy/cheap manufacturing, which runs through everything 8bitdo does.

 

But hey, form over function is the new way to go when it comes to electronics ::cough:: cell phones ::cough::, so maybe I'm just a relic.

I'll pull out my originals and plug them back in. Haven't used it since getting the NT Mini...

[F34R]

Contra is a good one because "down" causes your character to duck and "down/left" or "down/right" causes you to stand with a diagonal downward firing. My NES30 interprets even the slightest hint of left/right while pressing down as a down/left or down/right, which make Contra really frustrating to play. My original NES controls do not do this.

Ahhh.. I'll check it out. Thanks.

[Guspaz]

 

The 4 directions have split pads on the original PCBs, so when you're hitting diagonals its say half up and half right. On all the 8bitdo controllers its just right on/off, up on/off. It creates a totally different feel, completely obvious to anyone with even basic experience of the original controllers. You'll hear a lot of complaints about the D-Pad for this reason. It's just lazy/cheap manufacturing, which runs through everything 8bitdo does.

 

But hey, form over function is the new way to go when it comes to electronics ::cough:: cell phones ::cough::, so maybe I'm just a relic.

 

The 8bitdo controllers and the SNES controllers work identically for detecting direction inputs from an electrical standpoint. Both controllers have four binary contact switches that bridge to ground. The only difference is the physical pattern of the printed carbon on the PCB, which is a different shape. The SNES basically has a circle with a gap between the signal and ground, while the 8bitdo has silkscreened on a sort of an X pattern with ground on top/bottom and signal on left/right, or the opposite.

 

The difference in the silkscreened pattern changes how likely the conductive rubber pad under the d-pad is to close the switch when pushed in a certain way. This changes the "feel" of how the input translates into the game, but it is absolutely NOT a difference in how it works.

 

The difference in how it works in game is definitely not totally different, and it's actually a more complex pattern than the original controllers, hardly "lazy/cheap manufacturing".

 

Original SNES:

 

 

8bitdo:

 

Edited December 8, 2017 by Guspaz

[Great Hierophant]

Contra is a good one because "down" causes your character to duck and "down/left" or "down/right" causes you to stand with a diagonal downward firing. My NES30 interprets even the slightest hint of left/right while pressing down as a down/left or down/right, which make Contra really frustrating to play. My original NES controls do not do this.

In short down/left or down right = death, down = a chance to survive.

[cfillak]

 

Thanks for the post, heard about this thing, now the wait for a cheaper solution. Is there a review anywhere?

 

Analogue isn't gonna waste their time on repackaged Super NT FPGA's. That's something garbage companies like Retron do. I doubt programming the CD-Rom core on top of the regular TG16 one is too much of a leap for someone on Kevtris's level but I'm not that well versed on the difference.

 

That's a pretty long video review of a prototype board (but near final I believe). It's got a really cool UI that puts the Everdrive UIs to shame.

[Andromeda Stardust]

I know this isn't NT Mini, Super NT, Analogue or Kevtris related but I think it's interesting:

 

https://www.neosdstore.com/shop/index.php?id_product=12&controller=product&id_lang=1

 

 

It's basically the final solution for PC Engine CD. It's a board that plugs into the expansion port of any PC-Engine (including CoreGrafx and SuperGrafx I believe). It provides RGB out, an SD card slot for loading HuCard ROMs and ALL CD-ROM images including CD-ROM2, Super CD-ROM2, and Arcade CD-ROM2, all without needing the system cards. So it acts as an RGB solution, an Everdrive, and an optical drive emulator all in one. I believe the ODE functionality is all implemented on an FPGA.

 

This makes things interesting for Kevtris's planned TG-16/PCE core. I wonder if Analogue thinks the TG-16/PCE has enough popularity to carry its own product. If they do ever release a Turbo NT, and the jailbreak doesn't have CD-ROM functionality built in, I hope the system would at least have the expansion port to allow compatibility with the Super SD System 3.

Holy shit this is awesome. Is there some reason this won't work on Turbografx16?

 

Compatible with : Pcengine, Coregrafx, Coregrafx II and Supergrafx.

AFAIK, the only difference between a PCe and TG-16 expansion bus is the shape of the sheath. Both are 23x3 pins so should work with a simple rectangular adapter plug.

 

I currently have a US Turbo and a PC Henshin adapter for imports, as well as a TED. I would love to give CD games a spin on native hardware, but the flakiness and expense imo of CD drives / duos isn't worth it.

 

EDIT: watched the video from the above post. What happens if you plug a Hucard into the system with a super SD3 in the back?

[cfillak]

Holy shit this is awesome. Is there some reason this won't work on Turbografx16?

AFAIK, the only difference between a PCe and TG-16 expansion bus is the shape of the sheath. Both are 23x3 pins so should work with a simple rectangular adapter plug.

 

I currently have a US Turbo and a PC Henshin adapter for imports, as well as a TED. I would love to give CD games a spin on native hardware, but the flakiness and expense imo of CD drives / duos isn't worth it.

I don't know why it isn't listed as compatible with TG-16. I've never owned anything in the PC-Engine family so I'm not super familiar with the hardware differences. PC-Engine consoles seem pretty cheap though so I might jump in. What's cool is that all you need once you have the Super SD System is a PC-Engine, even an RF-only model, because the board handles everything including RGB out. With the Super SD System for PCE, the Saturn Satiator, and the PS-IO making strides, it seems like we're finally getting Everdrive tier solutions for CD based systems. It seems like Sega CD is the only major CD based system left without a method of loading backups that doesn't involve burning discs. Let's hope the Super SD System inspires Kevtris to implement CD support in his TG-16/PCE core

[ApolloBoy]

I don't know why it isn't listed as compatible with TG-16.

My guess is because the TG-16's expansion port has a slightly different shape compared to the PCE. It's the exact same pinout on both systems but the TG-16 has a trapezoidal connector whereas the PCE is rectangular.

[hyrulebr]

Holy shit this is awesome. Is there some reason this won't work on Turbografx16?

 

You can see in this part of video that TG16 is working (but not PAL). Only the current shell will be incompatible.

 

https://youtu.be/cyEJk0LmM8I?t=1664

 

EDIT: watched the video from the above post. What happens if you plug a Hucard into the system with a super SD3 in the back?

 

According to dev you can use yours hucards and system cards:

 

- you don't need a hucard. Everything is emulated in the sd system 3: hucards, system cards, arcade card... . You can however use your own hucards and that part will be disabled on the sd system 3, so you can still benefit, for example, from the backup ram or use your own system card to boot cds. Some limitations apply though, so it's recommened to use the internal emulation and run without a hucard.

 

 

Other information:

 

- you don't need a new power supply, we have tested with the 650mA psu of pcengines and it works fine.

- csync is TTL but can be easily feed to a 75ohm sync inserting a 470ohm resistor in the cart. The sd system 3 has access to the original pce csync, that is 75ohm sync levels, but it's not selected, currently a TTL csync is generated from the h & v sync signals.

[Tusecsy]

 

The 8bitdo controllers and the SNES controllers work identically for detecting direction inputs from an electrical standpoint. Both controllers have four binary contact switches that bridge to ground. The only difference is the physical pattern of the printed carbon on the PCB, which is a different shape. The SNES basically has a circle with a gap between the signal and ground, while the 8bitdo has silkscreened on a sort of an X pattern with ground on top/bottom and signal on left/right, or the opposite.

 

The difference in the silkscreened pattern changes how likely the conductive rubber pad under the d-pad is to close the switch when pushed in a certain way. This changes the "feel" of how the input translates into the game, but it is absolutely NOT a difference in how it works.

 

The difference in how it works in game is definitely not totally different, and it's actually a more complex pattern than the original controllers, hardly "lazy/cheap manufacturing".

 

Original SNES:

 

 

8bitdo:

 

Just look at the pictures. It's obvious the 8bitdo version is a garbage copy of the original. The contra test also proves it. I didn't realize how bad the actual button contacts were too.

 

You can't say it's both identical and "different but better" at the same time. It's lazy/bad manufacturing, and in no way an improvement, period.

Edited December 9, 2017 by Tusecsy

[F34R]

NES30

 

 

NES o/g

 

 

NES Classic

 

 

Just had a small run through the first area of Contra with the original controller and the NES30 from 8bitdo.

Using the NES30, I did see a couple times when I pressed down, and the dpad was slightly pressed with a little left/right pressure on the DOWN part of the dpad, that it registered as diagonal down.

[cfillak]

Back on the topic of a possible official Analogue DAC for the Super NT. I like being able to have my NT Mini plugged into both an analog and digital display at the same time. I get why the two displays can't both be used at the same time, with the downclocking etc., but I like being able to switch between the two outputs without messing with cables. So I hope any official Analogue DAC could replicate this feature by having a VGA style analog out port like the NT Mini, maybe with a switch to choose between Composite, S-Video, YPbPr, and RGB. but ALSO an HDMI passthrough, so you can switch between analog and digital with a controller input like on the NT Mini.