Printable View
Several people have made it known that they want a new bba adapter that doesn't cost 180$ or so.
I was hoping to start this topic where we can post any and all info on the subject and maybe eventually work towards building a new one. Anyone with information on the parallel port, programming for it, the current BBAs, hardware that might be usable, etc; please post it here.
This topic branched off the DC Cluster topic so you can also look there for more info.
There's an error in the pictures I gave you, the guy who scanned them scanned his modem in place of hi BBA by mistake :-S
a homebrew BBA would be a massive step for the scene :)
good luck
Guess I can start off with what we know already.
The BBA, or at least one model, uses the realtec 8139 ethernet contoller IC. This lets the dreamcast connect to an ethernet network.
The dreamcast parallel port has 16 data pins (pinout can be found on the internet) and operates at 25mhz. Allowing 50MB/s data transfer.
Most ethernet contollers (pci) operate with 32 pins at 33mhz.
I can think of 3 ways we may be able to connect the two together:
-Find a chip that can operate at 12.5mhz (half the parallel port) and send two sets of 16bits for each clock.
-Find an ethernet controller that uses 16bits already.
-Overclock the maple bus to be able to operate at twice the slowest pci clock. I've seen 16.75 so that means overclocking the bus from 25 to 35mhz. Then doing the pairs of 16bits idea.
Any ideas on feasability or know of something we can use?
Should be easier to have a chip working at 25Mhz just translating 16bits in 32 with some buffers, the realtek must be working at 25Mhz (as one can change pci speed in one's pc bios and it won't affect ethernet cards).
Finding an ethernet chip working at 16 bits should lead us to obsolete parts, but can be done.
Overclocking the G2 bus (maple is for joysticks) doesn't seems a good idea for me, it could alter the standard behaviour of the DC since the sound chip is connected on it.
But wouldn't taking the 16bit to 32bit mean halving the mhz to 12.5? I haven't been able to find any controllers that will run at this speed. Or is there another way to do this? I'm taking classes on electronics but haven't gotten very far yet, maybe in the following years...
It would mean buffering pci strobes of data.
Like i said, pci devices sould be able to run at 25Mhz.
If for example the pci device sends 16 times 32 bits (at 25Mhz), one keeps it in a memory and then (or even in the same time) one sends it on the g2 bus 32 times 16 bits.
In the other direction it would just be wait for two times 16 bits before writing 32 bits on the pci device.
That seems to be another job for SUPER CPLD defender of the logic and numerical data.
Sounds like an amibitious project ;D Semicolo's DC modem disassembly tutorial and extender board is a great start though http://semicolo.free.fr/Dreamcast/. Thanks for that semicolo!
Ah, I understand now. I thought it was something that was always streaming data. Not sure if that makes things easier or harder...
Other than telling the hardware when to wait and storing the seperate 16bit sections both ways, anything else that might cause problems between the two interfaces?
why not use an isa ethernet card as your basis? that way you dont need to do any weird things since it would be a 16 bit bus anyways...
From what I saw, the ISA only runs at 8mhz or so. That might make it a little hard too, and I bet that would only allow a base10 ethernet and waste a lot of what the port could handle. Of course I'm not an expert like I've said so if anyone knows otherwise feel free to correct me.
Because even if you could get a full speed ISA bus, it's not very fast?
Still significantly faster than the serial port?
Yeah, full speed ISA would be some 50x faster than the serial port but we could still get like 5x more than that if we can make use of a base100 ethernet. If we can find a 16bit one that runs at a higher mhz that would be really helpful.
I'm sure that everyone has seen this http://www.realtek.com.tw/products/p...aspx?modelid=6
It supports PCI clock speeds from 16.75 to 40MHz and can be interfaced to PCI, mini-PCI or Cardbus...
Couldn't someone have a look inside their own BBA and let us know whats inside? Wouldn't that be a easy route to develop a "replacement"?
The fact you have to consider though is that anything that you send over a bba would be well under 16mb as it would have to fit into the DC's ram AND have space to execute. Therefore 16mbps would pretty much be the maximum needed for anything. 16bits x 8mz =~ 16mpbs that means that in under a second you could send all the data you want to. IMO that is an acceptable speed =P.
The BBA uses some proprietary logic chips to do all the work and so far nobody has let us know how they do it. So we're stuck trying to find our own way. Still hoping someone has a BBA and can give us more information to go on. Thanks. ;D
Edit: Wow, we posted at the same time...
Well I wanted to get it as fast as possible in order to work as a cluster. Even though it can send all the data it could hold in memory it would in this case be processing it and sending it back just as much. Was hoping to kill two birds with one stone. ^^
Again the pictures :
http://perso.wanadoo.fr/pryonic/photos/BBAcote.JPG
http://perso.wanadoo.fr/pryonic/photos/BBArecto.JPG
http://perso.wanadoo.fr/pryonic/photos/BBAverso.JPG
There's really a Realteak 8139 in it, the trouble comes from the Sega chip
Maintaining compatibility with the official BBA would be nice, there are good informations in kos kernel on how the BBA talks to the G2 bus (DMA etc ...)
Quzar, some thoughts:
1) If you're going to build a custom piece of hardware anyway, why would you purposely use an older, slower interface if you're still going to have to fudge the thing just as much.
2) We're not currently talking about just sending data to store in memory. If you're designing a cluster, you give it a piece. It chugs it, sends results, gets new piece.
3) Assuming you WERE just using it for dev purposes (or otherwise just filling the DC's RAM for the hell of it) the serial port would suffice and indeed be a whole lot easier/cheaper to work with.
Yes, that would be very nice to keep it backwards compatible with the previous BBA(s?). Is there any information on the glue logic chips that sega used or can anyone maybe probe it to find out for us. (doubting, but would be really helpful) Also, if anyone has the BBA, how fast does it currently send/recv data?
I was looking at making this replacement because it would be helpful in so many different things. Possible cluster or online vs dev games, development of software, all kinds of stuff. Hope we're up to the challenge.
I don't know if anyone has tried to break the sega chip. I cannot help on this, I don't own a BBA nor a logic analyser.
Does anyone know where we can get the realtec chip to use for prototyping? I've seen places to order it but that's only in bulk (orders over 500$). I've thought about contacting realtec themselves but wanted to ask those who know more about ICs first.
Still haven't found any more information on the BBA.
I was more thinking of using an ethernet pci card with a rt8139 on it than building one. (and say the pci interface could work for other pci cards)
I see the product webpage is little more than marketing hype:
http://www.realtek.com.tw/products/p...aspx?modelid=7
I think unless you are a motherboard or high volume PCI card mfr you are out of luck. If I were you I would get a logic analyzer, study a BBA in operation and then use a glue chip with someone elses chip to make a functional clone.
Stock BBA is 10 mbits/sec. But the problem is that you're going to have a very hard time obtaining what is needed to build a drop-in replacement that is BBA-compatible. Whatever you build is probably going to end up being a custom solution that only works for devving and homebrew that supports it.
Which is why I recommend USB <-> DC Serial if your goal is just good speed for a dev cable. If your goal is a BBA replacement, good luck.
[quote author=Alexvrb link=board=hardware;num=1101681619;start=15#24 date=12/04/04 at 17:54:16]...
Which is why I recommend USB <-> DC Serial if your goal is just good speed for a dev cable. If your goal is a BBA replacement, good luck.[/quote]
;D
I guess then the best way to describe what I hope to eventually build is the highest speed connection I can to a network. Though it may be useful for dev; I mainly want it as something that can be used in homebrew games and software to communicate with either other dreamcasts (for multiplayer) or a computer (clusters or streaming programs like video players).
I had hoped that it wouldn't be too much harder to keep it compatible with the previous BBA it doesn't sound too likely so the best we can hope for is a good replacement.
Still looking for anyone with a bba to give us some information on it's workings though.
BBA is only 10mbits, and that's plenty fast for everything except maybe a DC cluster, in which case 100mbits would be plenty (for this hardware, at any rate). So really, you don't need to worry about maxing out the expansion port unless you're building a custom non-ISA ATA interface or some other fast storage device. Further, BBA compatibility would be awesome, but is unlikely. I'd be happy with a homebrew-use-only 10 or 100 mbit ethernet adapter of any sort... but that's still a whole lot of work.
For dev and streaming video/audio, Axlen's cable should suffice (given appropriate software support in the latter case).
Holy crap! I think I have an old NIC card with that RealTek chip in it! I'll have to double check to see.
BTW, I am behind deving a new BBA 100%, and would help as much as possible.
I've been looking into a lot of options lately but sadly, haven't been able to find any new information. We really need someone to look at their bba and give us some more info on how it works.
If we can find another ethernet controller ic that we can buy online for fairly cheap; we should start studying to see if we can't get it to work. How much data would be able to be sent at a time if it must be stored on the ethernet ic until it has had time to be sent to the dc, which recieves it at half the speed. Hopefully the ethernet ic can handle this dynamically so it doesn't overrun it's memory because it can't get the data out as fast as it's coming it.
Hrm... well just a bunch of questions to be asked for those that understand electronics, hope I made some kind of sense. O_<
Well the transfer speed shouldn't be a problem since the dc G2 bus is 400 Mbps when ethernet is 100 Mbps, it should be ok with some buffers.
I've got 2 or 3 ethernet cards with a realtek 8139 too.
OK, I checked the NIC card I have, and unfortaunately, it is not the same chip. However, when I get the opportunity, I will take a picture of it, and post it.
one could use the realtek 8100 too, it's the same as 8139 but has physical layer integrated, with the same problem of interfacing it onto the G2 bus and obtaining the chip in small quantities.
If the homebrew BBA is feasible, I think that we would instead have to obtain the chip in huge quantities, since every single DC owner would probably ask for one.
Any idea of how much a homebrew BBA would cost (supposing the G2 / Realtek interfacing coud be easily solved) ?
I also wondered, since the modem extension is included by default in each off-the-shelf DC package, if it could be possible to re-use the modem extension, but twist it to interface itself with a realtek system to go ethernet.
What do you think about this?
I find it very unlikely that the masses would be interested, especially because it could never be cheap, affordable maybe, but not cheap enough to buy on a whim. Anyway, you still might need to buy in fairly large quantities, depending on the chip. A lot of times there's a relatively large minimum order, you know? So unless you can scavenge from cheap PC networking hardware, no individual could really get it rolling anyway. But that's assuming someone had a design for a fast, simple ethernet interface. ;)
Supposing the G2 / pci can be done by a CPLD and we make about 100 boards, these two pieces of hardware would cost about 15 US$ (and a realtek ethernet pci card can be found for 10 US$ I think).
Well not a bad idea for the modem thingy, we'll have to figure if the chip used are pci ones (like in winmodems) and if it's the case reuse the bridge. Darn I think I dumped mine already...
so could we load on homebrew using this?
I'm still here, just been out of town snowboarding for a few weeks. Also have several other projects on my hands so any work I've done has been really slow.
I still haven't been able to find any other networking chip for sale that doesn't have a minimum quantity of like 500+ so if anyone knows of one please let us know. I guess the other way to go is find a common ethernet card and use it (or pieces of it)?
Just a passing thought, anyone think it would be worthwhile to just build a g2 to pci adapter and use an ethernet card to do the rest of the work? I don't mean trying to build one where you could put any pci card into it, just an ethernet card or perhaps several different ones unless they are very different hardware wise. No idea as to the difficulties in this idea. :-/
I went through all of my spare hardware, and in about 8 ethernet cards I didn't have any realtek ones. x.x
Crikey... never in my right mind did I imagine anyone else would have thought of actually trying to build a replacement BBA! But that's mainly because up until now I've only been visiting the B00B! DC Research forums which are sadly rather idle :P
Not that I'm trying to build one, but I have thought about it on occasion, including the idea Chaotis mentioned, though it would probably be quite hard to do. Having said that, getting the pinouts for a PCI port would be handy to start with. I just did a Google Search and came up with this.
(EDIT: And I suppose a pinout for the BBA connector would be very useful as well, though I couldn't find that on Google. A bit of DC reverse-engineering would probably be needed to produce details like that.)
Well I started thinking of using a cpld for bridging the G2 to pci
I found some verilog source on the Internet (copied there : ftp://semicolo.homelinux.org/dreamcast/verilog.php.html)
Credits go to Ben Jackson.
In early thinking for now (well in fact I've got all the needeed hardware but the pci connector)