Forums

In the form of a Google document, attached here for posterity.

https://docs.google.com/document/d/1MrPrSDpp6PmHPx45LzxTVJt2uCfl0MajA8GVtpWcois/...

(Google's PDF export seems to be broken currently)

Highlights:

- Red, green, blue levels measured using a video capture card, and checked for any odd biases produced by the card and the slightly funny video signals
- Red and green levels are pretty much linear
- On an SMS1, blue isn't, in particular the lowest intensity blue is 22-23% above where it ought to be
- Confirmed with an oscilloscope
- On an SMS2, the blue is more "normal" and the overall intensity is a bit higher (higher range) than SMS1/JSMS
- On a Genesis with PBC the colours are non-linear in a different way

SMS.zip (2.14 MB)

HTML + images

I was wondering if all SMS 1 have the same Sony video encoder chip...

I believe not. I guess it may be VDP related, perhaps unbalanced voltages at the encoder are the root cause. The doc does not list VDP or encoder versions.

so if we have both first revision SMS and second revision SMS using the same video encoder chip we could rule out that...

That's my Google Doc which I do hope to update over time. I didn't even know we were going to find this, so slowly I am getting more details about the captures others submitted. It seems quite clear at this time that this is the VDP, the 315-5124 and not the encoder as it shows in various encoders with that VDP as well as RGB bypasses with that VDP. And it also seems the 315-5246 is always linear.

If you’re going into this rabbit hole, I guess you can scope the voltages from the VDP? How do the RGB out voltages correspond to perceived brightness on a typical TV of the era, compared to a typical PC screen now, especially regarding gamma? Does the progressive scan change that (as the electron beam double-scans the same rows)?

Those are all great ideas. I've written them down for when I magically find time :-) Many thanks.

I did some measurements myself a while back ( https://www.smspower.org/forums/15700-SMSVideoOutputLevels ), somehow I didn't notice this blue difference you did. This is cool info none the less ~

That is INTERESTING, TimEE, and god knows I am happy to be wrong somehow. I hope more people test and maybe we all learn something. From personal measures and submissions it appears that consoles that have the 315-5124 have this blue push.

It is fascinating to see a more or less linear result with yours.

Always willing to learn more!

This got me curious. @Micro_Repairs sent me a dead SMS1 PCB to pick up the 315-5124 and decap it.
Here's where the blue LSB offset comes from:

The black "forks" are the gates of the transistors for each level of each channel, only a single one for a given channel is enabled at a time.
In pink are resistors between GND and +5V, the longer the strip, the higher the resistance.
The transistors connect different taps to select different output voltages.

Red and green have taps at x0: 0/10, x1: 1/10, x2: 2/10 and x3: 3/10.
Blue however has its B1 tap at approx 1.2/10 instead (circled in red).

I don't know what are the odds of this being a mistake. The IC design tools at the time probably had copy and paste functions so it's hard to imagine how that particular point could have moved.
Or maybe those were parametrized structures and a value was accidentally changed ?

249DFF48-074E-4EB0-86BF-97DE79D74E44.png (1.01 MB)

Attachment fairy

AMAZING work Furrtek. Phenomenal approach and exactly what was needed here. Thank you.

If the assumption is that it was done purposely, then what's the advantage? I don't see any immediate advantage here. Does this help create a more noticeable difference on some of the colors in the 6bit RGB range? I know some colors are so close that they almost useless (this happens in any RGB space though that's 6bits and above).

if that was done on purpose (and I still suspect it's not the case) the only reason I could see is that the darkest blue is very very dark, so really close to black that they might have decided to lighten that a little bit, screwing all the 16 colors that use the lowest blue level as a result.

Can we see the rest of the decapped die?

That's interesting, thanks for confirming this by decapping the VDP chip (I don't think anybody did it before you) and sharing picture of the DACs.

Out of curiosity, I looked more closely to the available 315-5313A (Mega Drive VDP chip) high-res die picture that was shared some years ago and it appears the tap for SMS mode (Mode 4) B1 is also shifted compared to Red and Green channels DACs (see attached picture below, ignoring other taps that are the thirteen MD Mode 5 specific color levels and are unchanged between blue and green DACs).

So it looks like it is not a mistake but intentional design from either Sega or Yamaha engineers. Not sure what the ground reason is as the colors using B1 does not look much different with non-linear value (dark/brown colors look slightly more blueish but that's not outstanding in RGB, maybe it is different with composite)

315-5313A_analog_DAC.png (3.83 MB)

315-5313A analog green & blue DACs

SMS_blue_diff.gif (16.98 KB)

SMS color differences

this is very interesting, and matches with the suspect that the MD VDP is an evolution of the first revision SMS VDP rather than an evolution of the second revision VDP (zoomed sprites were bugged in that, so they got removed - and extra height video modes weren't present either, so they weren't added...)

My guess is that the development forked from VDP1 to VDP2 to Game Gear, and from VDP1 to Mega Drive VDPs. Thus we see why the Mega Drive lacks 12-bit colour support as well as the extra height modes.

Perhaps this also aligned with the design, switching to NEC for the former and staying with Yamaha for the latter?

I think the best way to find out is to decap more chips. I’m sure we can get some to Furrtek or anyone else who is able to help...

Nice find Eke ! I didn't pay enough attention to the 315-5313A die shot, I was convinced that the taps were identical across the 3 DACs.

My microscopy setup is fully manual for now, so I can't provide entire die shots without a lot of work (hundreds of pictures). If someone has the right setup for this, I can send the die in.

What magnification level are you using for the shots? I’ve got a rig on a metallurgical microscope that I was planning to automate the xy stage and have about 600x zoom into the camera.

That was x200 magnification from a biological microscope with hacked top illumination and a HD w€bcam. You'll certainly get better results with your equipment.

I'm trying to get an openflexure delta stage attached to mine, but focus and vibration are still an issue.

Decapping and cleaning is something I can do fairly quickly and with a good success rate however. The dies can then be shipped back or sent to someone else for imaging if needed.
Faulty chips are generally fine if they didn't crack, smoke or caught on fire.

The possibility of seeing the SMS VDP die is awesome.

Just being able to confirm the full function of all the VDP registers bits makes it worthwhile IMO.

If I ever get a chance to finish motorizing mine, I’d be willing to take photos. I have a couple genesis chips to decap but don’t have the skill or safe environment to do that myself.

I have a couple VDP IC's I wanted to get professionally decapped and high resolution imaged but I struggled to find a company that specializes in this.

furrtek wrote

This got me curious. @Micro_Repairs sent me a dead SMS1 PCB to pick up the 315-5124 and decap it.
Here's where the blue LSB offset comes from:

The black "forks" are the gates of the transistors for each level of each channel, only a single one for a given channel is enabled at a time.
In pink are resistors between GND and +5V, the longer the strip, the higher the resistance.
The transistors connect different taps to select different output voltages.

Red and green have taps at x0: 0/10, x1: 1/10, x2: 2/10 and x3: 3/10.
Blue however has its B1 tap at approx 1.2/10 instead (circled in red).

I don't know what are the odds of this being a mistake. The IC design tools at the time probably had copy and paste functions so it's hard to imagine how that particular point could have moved.
Or maybe those were parametrized structures and a value was accidentally changed ?

SUPERB WORK MY FRIEND !

I use a biology microscope to take pictures and use an semi-automated software called ImageJ to stitch the pictures together. Or also PanaVue ImageAssembler.

The silicon electrical analyst in me looks at this and thinks you can put this die in a focused-ion beam electrical microscope and edit the circuit so that B1 is tapped into the resistor ladder like G1 and R1. This would depend on what kind of resistor we're talking about, not sure if this is a polysilicon resistor or a diffused resistor. The procedure would involve cutting the current metal line tapping into the resistor and adding another tap in the "correct" place. There are challenges, but I think there would be a decent chance of success.

Author	Message
Maxim Site Admin Joined: 19 Oct 1999 Posts: 14998 Location: London	Notes & Measures: Nonlinear Blue on Sega Master System 1 & Other Findings by bfbiii Posted: Sat Dec 19, 2020 8:36 pm
	In the form of a Google document, attached here for posterity. https://docs.google.com/document/d/1MrPrSDpp6PmHPx45LzxTVJt2uCfl0MajA8GVtpWcois/... (Google's PDF export seems to be broken currently) Highlights: - Red, green, blue levels measured using a video capture card, and checked for any odd biases produced by the card and the slightly funny video signals - Red and green levels are pretty much linear - On an SMS1, blue isn't, in particular the lowest intensity blue is 22-23% above where it ought to be - Confirmed with an oscilloscope - On an SMS2, the blue is more "normal" and the overall intensity is a bit higher (higher range) than SMS1/JSMS - On a Genesis with PBC the colours are non-linear in a different way SMS.zip (2.14 MB) HTML + images

sverx Joined: 05 Sep 2013 Posts: 3968 Location: Stockholm, Sweden	Posted: Mon Dec 21, 2020 5:30 pm
	I was wondering if all SMS 1 have the same Sony video encoder chip...

Maxim Site Admin Joined: 19 Oct 1999 Posts: 14998 Location: London	Posted: Tue Dec 22, 2020 12:55 am
	I believe not. I guess it may be VDP related, perhaps unbalanced voltages at the encoder are the root cause. The doc does not list VDP or encoder versions.

sverx Joined: 05 Sep 2013 Posts: 3968 Location: Stockholm, Sweden	Posted: Tue Dec 22, 2020 11:49 am
	so if we have both first revision SMS and second revision SMS using the same video encoder chip we could rule out that...

bfbiii Joined: 22 Dec 2020 Posts: 4	Hi there Posted: Tue Dec 22, 2020 9:17 pm
bfbiii Joined: 22 Dec 2020 Posts: 4	That's my Google Doc which I do hope to update over time. I didn't even know we were going to find this, so slowly I am getting more details about the captures others submitted. It seems quite clear at this time that this is the VDP, the 315-5124 and not the encoder as it shows in various encoders with that VDP as well as RGB bypasses with that VDP. And it also seems the 315-5246 is always linear.

Maxim Site Admin Joined: 19 Oct 1999 Posts: 14998 Location: London	Posted: Wed Dec 23, 2020 12:04 am
	If you’re going into this rabbit hole, I guess you can scope the voltages from the VDP? How do the RGB out voltages correspond to perceived brightness on a typical TV of the era, compared to a typical PC screen now, especially regarding gamma? Does the progressive scan change that (as the electron beam double-scans the same rows)?

bfbiii Joined: 22 Dec 2020 Posts: 4	Posted: Wed Dec 23, 2020 12:39 pm
bfbiii Joined: 22 Dec 2020 Posts: 4	Those are all great ideas. I've written them down for when I magically find time :-) Many thanks.

TmEE Joined: 31 Oct 2007 Posts: 855 Location: Norway (used to be Estonia)	Posted: Wed Dec 30, 2020 7:41 am
	I did some measurements myself a while back ( https://www.smspower.org/forums/15700-SMSVideoOutputLevels ), somehow I didn't notice this blue difference you did. This is cool info none the less ~

bfbiii Joined: 22 Dec 2020 Posts: 4	Posted: Sat Jan 02, 2021 3:16 pm
bfbiii Joined: 22 Dec 2020 Posts: 4	That is INTERESTING, TimEE, and god knows I am happy to be wrong somehow. I hope more people test and maybe we all learn something. From personal measures and submissions it appears that consoles that have the 315-5124 have this blue push. It is fascinating to see a more or less linear result with yours. Always willing to learn more!

furrtek Joined: 05 Mar 2006 Posts: 54 Location: France	Posted: Mon Mar 08, 2021 1:05 pm Last edited by furrtek on Thu Mar 11, 2021 5:28 am; edited 1 time in total
furrtek Joined: 05 Mar 2006 Posts: 54 Location: France	This got me curious. @Micro_Repairs sent me a dead SMS1 PCB to pick up the 315-5124 and decap it. Here's where the blue LSB offset comes from: The black "forks" are the gates of the transistors for each level of each channel, only a single one for a given channel is enabled at a time. In pink are resistors between GND and +5V, the longer the strip, the higher the resistance. The transistors connect different taps to select different output voltages. Red and green have taps at x0: 0/10, x1: 1/10, x2: 2/10 and x3: 3/10. Blue however has its B1 tap at approx 1.2/10 instead (circled in red). I don't know what are the odds of this being a mistake. The IC design tools at the time probably had copy and paste functions so it's hard to imagine how that particular point could have moved. Or maybe those were parametrized structures and a value was accidentally changed ? 249DFF48-074E-4EB0-86BF-97DE79D74E44.png (1.01 MB) Attachment fairy

bfbiii Joined: 22 Dec 2020 Posts: 4	Posted: Tue Mar 09, 2021 12:05 pm
bfbiii Joined: 22 Dec 2020 Posts: 4	AMAZING work Furrtek. Phenomenal approach and exactly what was needed here. Thank you.

turboxray Joined: 09 Mar 2021 Posts: 11	Posted: Tue Mar 09, 2021 5:35 pm
turboxray Joined: 09 Mar 2021 Posts: 11	If the assumption is that it was done purposely, then what's the advantage? I don't see any immediate advantage here. Does this help create a more noticeable difference on some of the colors in the 6bit RGB range? I know some colors are so close that they almost useless (this happens in any RGB space though that's 6bits and above).

sverx Joined: 05 Sep 2013 Posts: 3968 Location: Stockholm, Sweden	Posted: Wed Mar 10, 2021 10:51 am
	if that was done on purpose (and I still suspect it's not the case) the only reason I could see is that the darkest blue is very very dark, so really close to black that they might have decided to lighten that a little bit, screwing all the 16 colors that use the lowest blue level as a result.

asynchronous Joined: 14 Aug 2000 Posts: 764 Location: Adelaide, Australia	Posted: Wed Mar 10, 2021 11:00 am
	Can we see the rest of the decapped die?

Eke Joined: 06 Feb 2009 Posts: 110 Location: Toulouse, France	Posted: Sat Mar 13, 2021 2:01 pm
	That's interesting, thanks for confirming this by decapping the VDP chip (I don't think anybody did it before you) and sharing picture of the DACs. Out of curiosity, I looked more closely to the available 315-5313A (Mega Drive VDP chip) high-res die picture that was shared some years ago and it appears the tap for SMS mode (Mode 4) B1 is also shifted compared to Red and Green channels DACs (see attached picture below, ignoring other taps that are the thirteen MD Mode 5 specific color levels and are unchanged between blue and green DACs). So it looks like it is not a mistake but intentional design from either Sega or Yamaha engineers. Not sure what the ground reason is as the colors using B1 does not look much different with non-linear value (dark/brown colors look slightly more blueish but that's not outstanding in RGB, maybe it is different with composite) 315-5313A_analog_DAC.png (3.83 MB) 315-5313A analog green & blue DACs SMS_blue_diff.gif (16.98 KB) SMS color differences

sverx Joined: 05 Sep 2013 Posts: 3968 Location: Stockholm, Sweden	Posted: Mon Mar 15, 2021 5:28 pm
	this is very interesting, and matches with the suspect that the MD VDP is an evolution of the first revision SMS VDP rather than an evolution of the second revision VDP (zoomed sprites were bugged in that, so they got removed - and extra height video modes weren't present either, so they weren't added...)

Maxim Site Admin Joined: 19 Oct 1999 Posts: 14998 Location: London	Posted: Mon Mar 15, 2021 9:04 pm Last edited by Maxim on Mon Mar 15, 2021 9:39 pm; edited 1 time in total
	My guess is that the development forked from VDP1 to VDP2 to Game Gear, and from VDP1 to Mega Drive VDPs. Thus we see why the Mega Drive lacks 12-bit colour support as well as the extra height modes. Perhaps this also aligned with the design, switching to NEC for the former and staying with Yamaha for the latter? I think the best way to find out is to decap more chips. I’m sure we can get some to Furrtek or anyone else who is able to help...

furrtek Joined: 05 Mar 2006 Posts: 54 Location: France	Posted: Mon Mar 15, 2021 9:20 pm
furrtek Joined: 05 Mar 2006 Posts: 54 Location: France	Nice find Eke ! I didn't pay enough attention to the 315-5313A die shot, I was convinced that the taps were identical across the 3 DACs. My microscopy setup is fully manual for now, so I can't provide entire die shots without a lot of work (hundreds of pictures). If someone has the right setup for this, I can send the die in.

Enforcer Joined: 16 Mar 2018 Posts: 29 Location: Indiana	Posted: Mon Mar 15, 2021 11:08 pm
Enforcer Joined: 16 Mar 2018 Posts: 29 Location: Indiana	What magnification level are you using for the shots? I’ve got a rig on a metallurgical microscope that I was planning to automate the xy stage and have about 600x zoom into the camera.

furrtek Joined: 05 Mar 2006 Posts: 54 Location: France	Posted: Tue Mar 16, 2021 1:04 am
furrtek Joined: 05 Mar 2006 Posts: 54 Location: France	That was x200 magnification from a biological microscope with hacked top illumination and a HD w€bcam. You'll certainly get better results with your equipment. I'm trying to get an openflexure delta stage attached to mine, but focus and vibration are still an issue. Decapping and cleaning is something I can do fairly quickly and with a good success rate however. The dies can then be shipped back or sent to someone else for imaging if needed. Faulty chips are generally fine if they didn't crack, smoke or caught on fire.

asynchronous Joined: 14 Aug 2000 Posts: 764 Location: Adelaide, Australia	Notes & Measures: Nonlinear Blue on Sega Master System 1 & Other Findings by bfbiii Posted: Tue Mar 16, 2021 2:50 am
	The possibility of seeing the SMS VDP die is awesome. Just being able to confirm the full function of all the VDP registers bits makes it worthwhile IMO.

Enforcer Joined: 16 Mar 2018 Posts: 29 Location: Indiana	Posted: Tue Mar 16, 2021 4:00 am
Enforcer Joined: 16 Mar 2018 Posts: 29 Location: Indiana	If I ever get a chance to finish motorizing mine, I’d be willing to take photos. I have a couple genesis chips to decap but don’t have the skill or safe environment to do that myself.

thatawesomeguy Joined: 25 Jul 2007 Posts: 744 Location: Melbourne, Australia	Posted: Tue Mar 16, 2021 7:28 am
	I have a couple VDP IC's I wanted to get professionally decapped and high resolution imaged but I struggled to find a company that specializes in this.

asper Joined: 25 Nov 2015 Posts: 143	Posted: Sat Mar 20, 2021 12:57 pm
asper Joined: 25 Nov 2015 Posts: 143	furrtek wrote This got me curious. @Micro_Repairs sent me a dead SMS1 PCB to pick up the 315-5124 and decap it. Here's where the blue LSB offset comes from: The black "forks" are the gates of the transistors for each level of each channel, only a single one for a given channel is enabled at a time. In pink are resistors between GND and +5V, the longer the strip, the higher the resistance. The transistors connect different taps to select different output voltages. Red and green have taps at x0: 0/10, x1: 1/10, x2: 2/10 and x3: 3/10. Blue however has its B1 tap at approx 1.2/10 instead (circled in red). I don't know what are the odds of this being a mistake. The IC design tools at the time probably had copy and paste functions so it's hard to imagine how that particular point could have moved. Or maybe those were parametrized structures and a value was accidentally changed ? SUPERB WORK MY FRIEND ! I use a biology microscope to take pictures and use an semi-automated software called ImageJ to stitch the pictures together. Or also PanaVue ImageAssembler.

rothgar Joined: 17 Mar 2021 Posts: 1	Posted: Fri May 13, 2022 12:12 am
rothgar Joined: 17 Mar 2021 Posts: 1	The silicon electrical analyst in me looks at this and thinks you can put this die in a focused-ion beam electrical microscope and edit the circuit so that B1 is tapped into the resistor ladder like G1 and R1. This would depend on what kind of resistor we're talking about, not sure if this is a polysilicon resistor or a diffused resistor. The procedure would involve cutting the current metal line tapping into the resistor and adding another tap in the "correct" place. There are challenges, but I think there would be a decent chance of success.

Forums

View topic - Notes & Measures: Nonlinear Blue on Sega Master System 1 & Other Findings by bfbiii