FS: Radius Rocket Clone "Prototypes"

[JDW]

@max1zzz first mentioned his work on a Radius Rocket prototype here. I was informed about his MLA thread on the subject today. Specifically, it seems he now has a working board:

I have an original Radius Rocket 33 with a full FPU 040, installed in my IIci. Here are some pics:





Max, is there are reason you chose to go with the 25MHz GALs instead of 33? Is it because you had no access to the 33MHz edition GALs?

Honestly, it would be great if the GALs could be revised to work at 40MHz, especially because some 040's can clock that high. Although even at lower clock speeds, Radius Rocket's pull a huge amount of current from the power supply...

As a result, it gets pretty hot. You can see the temps on mine, using a UTi260B thermal camera...

 

[Bolle]

I don't think a 33MHz GAL based model exists. With the speedbump they also redesigned the whole thing and funneled all the logic into those two gate arrays.

 

[Trash80toG4]

If that's the case and we have the formulas for the GALs, might someone accomplish that funneling of GAL logic intto modern gate arrays?

Thinking here would be that the logic upgrade would be more power efficient and capable of running at the full 40Mh Stage II clock while still supporting RocketWare. RocketShare limitation of the Stage II is a PITA.


edit: moving to a pair of 72pin SIMMs might be better for achieving that goal as well?

 

[phipli]

Lets take a pause and breath in how awesome it is that someone cloned a Rocket before we ask them to do an even more difficult job of redesigning it into something even the original manufacturer didn't make .

Also, apparently modern FPGAs often struggle to get the low latency achievable with that era of GALs, especially as 5v parts are rare so you frequently need level shifters to interface with period hardware. Custom silicon on the other hand would presumably be prohibitively expensive.

Board looks amazing, nice that they've even found the PDS connectors.

 

[Trash80toG4]

Lets take a pause and breath in how awesome it is that someone cloned a Rocket before we ask them to do an even more difficult job of redesigning it into something even the original manufacturer didn't make .

Indeed! Said that in the first response to his announcement over in the MLA Trading Post thread.

Board looks amazing, nice that they've even found the PDS connectors.

Amen to that, I've go the SCSI2 Daughtercard on hand.

 

[JDW]

I've go the SCSI2 Daughtercard on hand.

All this time I was thinking the DSP board (for Photoshop acceleration) was the only daughtercard. Now I see Radius also had a SCSI board too, which is nice:

Radius Rocket Daughterboards

I suppose this is what I was thinking of, from the back of the PhotoBooster retail box: I trust Radius' own official material, but that Stage II DSP board doesn't seem to have the longer top connector like I think it should (maybe just the lighting/resolution playing tricks...) Hopefully this...

68kmla.org

 

[max1zzz]

Indeed! Said that in the first response to his announcement over in the MLA Trading Post thread.

That's because it was a secret project...... Nah it's only because I find writing project threads quite tiring so tend not to do it these days, I did post some status updates both on here and 68k with pictures of the board while it was in progress

I don't think a 33MHz GAL based model exists. With the speedbump they also redesigned the whole thing and funneled all the logic into those two gate arrays.

I believe that is correct, at least I haven't seen one. I have however tested the clones at up to 32mhz with no issues. 33mhz seems to be a step too far though and causes the rocket's to be unstable

Board looks amazing, nice that they've even found the PDS connectors.

Funnily those where actually very easy to find, the are the same part as used by by SCA SCSI drives so I recognised the connector straight away. I still need to test that the headers actually work though. I do now have a SCSI II card in my hands both for testing and eventual cloning

As for the future of these I am currently working on a variant that has ram cards rather than 30 pin simms due to the difficulty of me actually getting 30 pin simm sockets in the UK (Basically I can't and have to proxy order them form the states which is a pain) this is currently in testing and proving to have some stability issues but I think this is partially down to the way I designed the RAM cards and partially down to the ram chips I chose. For the future I do also want to at least partially move to a CPLD based design to cut down on the parts count however a fully CPLD based design would not be easy as it's proving difficult to find something that can disassemble dumps from the 4 GAL26CV12's however at a minimum it should be possible to move the RAM controll logic into a pair of CPLD's replacing 10 IC's with just 2

 

[Trash80toG4]

A pair of 72pin SIMMs should be easy enough to implement with connectors available, no?

 

[max1zzz]

You would think that, and that was my first idea however the way the rocket is designed it actually isn't. 72pin simms need 2 RAS signals (one per rank) and the rocket only generates one for all 8 30 pin simms which means only single rank simms would work, this isn't the end of the world however it would mostly limit the rocket it 32mb of ram (16 per bank) however more importantly it would generate a bunch of complaints of "It only sees half my simm so must be broken" from people who haven't read (or have but haven't understood)

I understand some people won't like the idea of ram cards but it's the best solution I could come up with, the design for the ram boards will be open source and I will still make some 30pin versions as long as I can get the slots at a reasonable price

 

[Trash80toG4]

Whatever it takes will suffice! I think such will be fine with those who want a Rocket on board. You won't see it very often and anything that any reduction of power consumption and heat generated will be welcome.

Since the PDS connectors are readily available, it would be amazing to build an intermediate card to feed the Fast/Narrow SCSI II Daughtercard output to an internal IDC cable connector. Not everyone has a pair of Radius StudioArrays to hook up to the external connector.

Years back, it used to be that everyone looked at slapping RPi onto anything inside a Mac. Now I'm pushing the Fast/Narrow BlueSCSI II/WiFi onto my targets of affection (compulsions) it seems. Slapping it onto said intermediate card would handily reduce power requirements and heat generated by spinning rust.

You'd need to set the SCSI II Daughtercard back a bit and put the external connector on the intermediate board which would be a complication of course, but who'd really want to make use of the external port?