sd cartridge for 7800?

[+batari]

On 2/25/2020 at 6:14 AM, RevEng said:

I really love that you're scaling it back, and being very wary of feature-creep.

 

That said, I think support of the $450 version is important for homebrew. The $450 location is a lot more useful than $4000, because you don't waste 16k worth of address space for access to 16 registers. The XM, for example, only supports pokey at $450, and not at $4000. While both $450 and $4000 have been used in homebrew titles - with $4000 sticking around due to CC2 being the most common flash cart - any dev with an XM in their hands will very likely be using $450 from here on out.

 

Of course, you can always say "get an xm" to anybody looking for pokey@450 support. Same with the HSC, and same with some of the on-cart RAM titles - it shouldn't be terribly difficult to hack them to turn on the XM RAM instead, and I'd be willing to take that on. I just wanted to share my perspective on the relative importance of the pokey locations.

It’s not exactly that POKEY @ $0450 won’t work but that I’m not sure the hardware will support it. It might work, but I can’t make any promises at this stage. 
 

edit: part of the reason I’m not sure is the complex address decoding for $045x vs allowing $4000-$7FFF. If the address decoding can be simplified (e.g. allow the entire page at $04xx (Or even more lax decoding, such as $400-$7ff) without breaking anything then that would be simpler to do. 

[RevEng]

That's fair. And if it doesn't work out, I totally agree with your decisions to focus on affordability and simplicity.

[RevEng]

The only devices in 400-47f would be XM, Xboard, and other pokey carts. XM users will likely have a way of disabling the pokey and other registers (for those homebrew carts using $450), or XM users can use your cart with a78 files that don't enable pokey. Xboard never hit critical mass, and can be disregarded, but any Xboard users could also use a78 files without pokey enabled.

 

There's RIOT RAM from 480-4FF, but no games should be writing to it, as it's unreliable in 7800 mode. The range up to 7FF looks to be clear and/or RIOT RAM mirrors. The next thing up from there is the HSC, at $1000. (info not at the link, but just from memory)

[+batari]

2 hours ago, RevEng said:

The only devices in 400-47f would be XM, Xboard, and other pokey carts. XM users will likely have a way of disabling the pokey and other registers (for those homebrew carts using $450), or XM users can use your cart with a78 files that don't enable pokey. Xboard never hit critical mass, and can be disregarded, but any Xboard users could also use a78 files without pokey enabled.

 

There's RIOT RAM from 480-4FF, but no games should be writing to it, as it's unreliable in 7800 mode. The range up to 7FF looks to be clear and/or RIOT RAM mirrors. The next thing up from there is the HSC, at $1000. (info not at the link, but just from memory)

I didn't know that about RIOT RAM. I have to check if Concerto BIOS uses it, as it might, and if it's unreliable, it might be a possible source of issues some people have having.

 

But either way, I will go forward making an assumption that I can use the whole area from $400-$47F: It reduces the number of address bits to decode from 12 to 9 and that does help a little.

[TailChao]

12 hours ago, batari said:

Thanks, I’ll look into that. The Maria RAM access is so fast it’s hard to keep up with it. If it can be detected ahead of time that would help a lot.

She's faster than Sally, but not too fast. Ditching the 6502 style R/W for a PHI2 gated OEn and WRn already helps. Whenever Sally is driving you only want to perform a read or write to cartridge memory when PHI2 is high - but Maria only reads and you have to ignore PHI2 when she's taken the bus.

[tep392]

10 hours ago, batari said:

I didn't know that about RIOT RAM. I have to check if Concerto BIOS uses it, as it might, and if it's unreliable, it might be a possible source of issues some people have having.

 

But either way, I will go forward making an assumption that I can use the whole area from $400-$47F: It reduces the number of address bits to decode from 12 to 9 and that does help a little.

The XM control registers reside in the area from $470-$47F, so you should avoid that area if possible. You don't want writes to those registers to enable the POKEY.

[+batari]

1 hour ago, TailChao said:

She's faster than Sally, but not too fast. Ditching the 6502 style R/W for a PHI2 gated OEn and WRn already helps. Whenever Sally is driving you only want to perform a read or write to cartridge memory when PHI2 is high - but Maria only reads and you have to ignore PHI2 when she's taken the bus.

Maria in fast DMA mode apparently needs 165 ns access time or less, but on a cartridge (due to gate delays) that is more like 100 ns, which is pretty fast considering the date of the technology. One report even said 100 ns might be too slow. So that is what I am dealing with, I think.

[+batari]

29 minutes ago, tep392 said:

The XM control registers reside in the area from $470-$47F, so you should avoid that area if possible. You don't want writes to those registers to enable the POKEY.

If the XM is disabled from getting a magic number from Concerto, does it still write to these areas?

[tep392]

The XM BIOS will disable the XM POKEY if it see's the Concerto's magic number or detects CPUWIZ's POKEY board and they will be locked out until power cycle.  But the XM isn't totally disabled, and games can still utilize features like HSC, YM2151, PIA and XM RAM by writing to those XM registers.  If your cart's POKEY is enabled by writes to those registers, it may start making unwanted sounds.  I haven't seen how it is going to work yet, but there will be a way for the game code to write to registers and prevent those from going to the POKEY, if your decoding overlaps the XM registers.  Future games will need to be written to do this and there will be routines in the developers kit that can be used that will take care of this automatically.

[RALPH!]

On 10/14/2014 at 8:37 AM, Rybags said:

In theory an adaptor could be made to allow using Atarimax flashcarts from the computer on the 7800.

Problem is though, they only use 8K address range which would limit what games could be run off them. Plus the banking is different - I imagine no existing 7800 games would be 8K visible with the entire range bankswitched.

Will need to change the addresses to work from $4000 to $7FFF and have $8000 to $FFFF always mapped. AtariMax board connector pins address different with control logic chip. May be better to use similar chip for XEGS that can have only part switching window. The address and window size set by pins on board connected to chip, not board itself.

[Curt Vendel]

On 2/24/2020 at 11:23 PM, Mitch said:

If we are counting 7800 consoles, I currently own 26 of them. 4 of them are PAL and 3 are protos.

 

Can anyone top that?

 

Mitch

Here, hold my beer...

 

 

[Rybags]

PMSL.

I can just imaging Curt opening a triple garage door and revealing stacked too the roof and to the back, the entire Atari surplus of 7800s that they couldn't sell.

[Curt Vendel]

16 hours ago, Rybags said:

PMSL.

I can just imaging Curt opening a triple garage door and revealing stacked too the roof and to the back, the entire Atari surplus of 7800s that they couldn't sell.

I have a lot but almost all are May-June 84’s and a couple of April 84 protos

[+batari]

Finally, I am getting back to sorting out the issues with Concerto and working toward a re-release.

 

I built two Concertos the other day and I am putting together about 20 total for now. More will be assembled down the road if the problems can be fixed (I think many can be.) It took a while to find all of the parts, and especially, to get my device programmer going again (it's a parallel port model and requires an older version of Windows, so I had to get an old laptop put together for it, but finally it's going!)

 

I got my 12 consoles out of storage, am ordering a new POKEY chip and some power supplies from Best (currently using one that has been repaired too many times.) None of my consoles have a video mod so I am going to get one (prefferrably the same kind that others reporting problems have, to rule that out.)

 

I do apologize for the delay, but I always knew I would get back to this eventually when my life got back to something resembling normal. Without getting into boring detail, there was a lot to unravel in my life to make this possible. Part of it was that in order for me to be able to do difficult tasks (like, debug flashcart BIOS!) I need to be in a certain state of mind to be able to think deeply enough and process. I got pretty overwhelmed years ago when people reported some issues and at the time wasn't in the right state of mind to solve the problem. But now I feel I am back and ready to tackle the issues once again with a fresh set of eyes. And, in particular, the integrity test results from RevEng's Utility Cart reveals some useful information.

 

Yesterday, I tested all 12 of my consoles and ALL of them are 100% green on the integrity test, every time. So I need help from the community from someone reporting issues:

 

I see a lot of posts where people reported integrity errors at 80, 84, 90, 94, C0, C4, etc. There is some variation but there is also a pattern here that might help reveal the exact cause. Problem is, each square represents 256 bytes, and just knowing the 256-byte range isn't enough to reveal a cause. But, the utility cart can show any area of memory in Memory explorer in 128 byte chunks, so:

 

Run the integrity test, and recall the red areas. Then, open the Memory Explorer in the Utility cart, then scroll down to problem areas. Say, if you are getting 80, this means locations $8000-$80FF, so scroll down to 8000, take a picture, then to 8080 and take another, and post a picture of each screen here. This may pinpoint the exact byte(s) not being loaded correctly and what values are loaded in their place, and that could very well reveal a probable source if loading issues.

 

Once I have this information, I will hopefully be able to solve this problem and post up a test BIOS for you all to try!

 

Thanks!

[+-^CrossBow^-]

I'm just sorry I didn't jump on the initial release for these as since then I've had likely every variant of the 7800 known come past my hands I could have tested with using the Rev's Utility. In fact I use that utility now and the integrity test for burn in testing on systems both before and after video upgrades.

 

Speaking on this, the 7800 stock configuration with a Pokey cart like Commando running will draw about 720ma to power the entire system. When you add in a UAV that might add take it to about 730ma. I'm sure other mods etc would be similar. Point is, some of the issues might be resolved by replacing out the factory voltage regulator and replacing it with a new 1amp or better, 2amp output variant to ensure there is enough power to the system at all times. 

 

@batari do you know how much current draw the current Concerto requires on a stock system? Again, without having one of these myself for testing I'm just thinking out loud. I've also noticed that many revisions of the 7800 use different speed RAM in them as well to say nothing about that extra timing circuit in the mix.

 

[RevEng]

In case he's not watching the thread - could @Jinks lend a hand? I recall that your unit was one that turned up with red blocks, until it gets toasty anyway.

[+batari]

49 minutes ago, -^CrossBow^- said:

I'm just sorry I didn't jump on the initial release for these as since then I've had likely every variant of the 7800 known come past my hands I could have tested with using the Rev's Utility. In fact I use that utility now and the integrity test for burn in testing on systems both before and after video upgrades.

 

Speaking on this, the 7800 stock configuration with a Pokey cart like Commando running will draw about 720ma to power the entire system. When you add in a UAV that might add take it to about 730ma. I'm sure other mods etc would be similar. Point is, some of the issues might be resolved by replacing out the factory voltage regulator and replacing it with a new 1amp or better, 2amp output variant to ensure there is enough power to the system at all times. 

 

@batari do you know how much current draw the current Concerto requires on a stock system? Again, without having one of these myself for testing I'm just thinking out loud. I've also noticed that many revisions of the 7800 use different speed RAM in them as well to say nothing about that extra timing circuit in the mix.

 

Why, yes I do! This is a stock RF 7800 with a Concerto, no POKEY, Concerto running the menu:

 

 

I can't find my POKEY so I am trying to get one from Best Electronics along with some 7800 PSUs.

 

I am glad to hear the UAV only adds 10 ma. I plan to get one. From experience, the Longhorn mod sucks up like 75 ma, or at least the one I have for a 2600 does, and I think his 7800 one is based on that design. So maybe I also need to install a Longhorn on one of mine as it seems more likely to cause power issues.

[+-^CrossBow^-]

You know I actually don't run my bench at 9v when testing 7800s. I simulate the same input voltge that original does so I tend to have mine set around 12v on the input. That might account for the higher amperage reading since the VR having to waste more out as heat and looses further efficiency in the process. My bench supply might be a bit off kilter too but it should be pretty close.

 

You should take those same readings from a 7800 with a DC-DC buck converter installed...hehe. 

 

Old pic from a few months ago but here is what my bench supply reads from a stock 7800 running the integrity test through Mateos with Pokey installed. As you can see on my supply is it just over 700ma and I was guessing about 720ma or so. On the pic below, the current is in High mode so the top scale is what you use as my supply goes up to 3A output.

 

 

[+batari]

I just tried my PSU at 12v, and got 682 ma, so 12 ma more than at 9v.

[+DrVenkman]

1 hour ago, batari said:

I built two Concertos the other day and I am putting together about 20 total for now.

I have a spare POKEY just waiting for this, along with two 7800s I can test with: a totally-stock, unmodded AT-84 with a fully socketed motherboard including the SRAM chips; and a UAV-modded A1 serial number unit.  I also have a scope, a bench power supply and a decades out-of-date and vastly-underutilized engineering degree that likes to chew on annoying technical problems.  

[+DrVenkman]

1 minute ago, batari said:

I just tried my PSU at 12v, and got 682 ma, so 12 ma more than at 9v.

That's probably just excess current turned into heat by the 7805, assuming the Vcc at each chip remains ~5.0V.

[bszarek001]

15 minutes ago, DrVenkman said:

I have a spare POKEY just waiting for this, along with two 7800s I can test with: a totally-stock, unmodded AT-84 with a fully socketed motherboard including the SRAM chips; and a UAV-modded A1 serial number unit.  I also have a scope, a bench power supply and a decades out-of-date and vastly-underutilized engineering degree that likes to chew on annoying technical problems.  

I’m also dying to get my hands on one and test. @batari, if you need beta testers, I’d love to be a part of your project.  My unit is also UAV modded, rev D.  

[+batari]

There will be enough for everyone eventually, assuming the most egregious issues can be solved. But for now, priority for the remaining 20 is for game developers first, as there are several who are making 7800 games but have no way to test on hardware, so I'd like all game developers who want to buy a Concerto to get one.

 

I am sure there are less than 20 game developers out there, so if you want to help test and have a number of consoles to test on, I am also willing to sell you one of the 20.

 

The price for Concerto is $99. I would knock off $10 if you just want a bare board, not in a cart shell. There isn't a label for Concerto yet. I will of course have a label eventually.

 

I am also still working on Concertino, by the way, if $99 is too much and you are willing to wait a bit!

[+-^CrossBow^-]

If anyone is curious...here is what the current draw from a 7800 with a UAV running the same integrity test through my Mateos cart with a Traco 2A rated 2-2450 installed.. 

 

 

 

Yeap, that is reading right at 300ma as opposed to the linear VR. 

 

 

[+Karl G]

5 minutes ago, batari said:

There will be enough for everyone eventually, assuming the most egregious issues can be solved. But for now, priority for the remaining 20 is for game developers first, as there are several who are making 7800 games but have no way to test on hardware, so I'd like all game developers who want to buy a Concerto to get one.

 

I am sure there are less than 20 game developers out there, so if you want to help test and have a number of consoles to test on, I am also willing to sell you one of the 20.

I'm interested in having one as a fledgling 7800 developer if there are enough to go around to developers who need them. I wouldn't want to take one from someone who is developing more actively than me at the moment. I've been working on Cosmic Cabbie off and on, and plan on devoting more dedicated time to it after Penult is finished.

 

I have two 7800 consoles I could test on - one at my house, and one at my girlfriend's house.