• Please review our updated Terms and Rules here

ATX Turbo XT mainboard project revision 3 - now released on GitHub

R

rodney

Experienced Member
Joined
Jul 14, 2022
Messages
268
Hello everyone,

A while ago I designed and built Revision 3 of my ATX Turbo XT mainboard.
If anyone is interested, you can check out my designs on GitHub.
Please read all the information I have provided carefully first if interested, before deciding to build this project for yourself.

I made a page about the project on my website.
You also can see some photos of the finished mainboard.

Regarding future projects I am planning, my next mainboard project is in preparation and will be a 286 ATX mainboard.
I will again not be using any chipset, but instead do a similar design as the 5170 and derived systems.
Just like with the XT, I will be removing the parity checking and DRAM logic and design the system based on SRAM memory.
And I will be adding common AT interfaces like COM, LPT, IDE, FDC, LAN, SCSI etc onboard as far as space permits.

Right now the problem is to find a suitable and stable AT 286 mainboard without a chipset.
I bought two such mainboards so far, however both had malfunctions and stability issues and are not ideal as a reference board for my project.

Anyway, I hope perhaps some people can enjoy to check out my designs.
Personally, I find this XT a pleasure to use and right now I am not experiencing any issues or problems so far.

Kind regards,

Rodney
 
rodney said:
Also I noticed that the chips are very sensitive, when you lay your finger on them, the PC may freeze or not release the DMA cycle if active
Click to expand...
This to me sounds very much like the result of a floating signal somewhere.

Are you using any CMOS (HC or HCT) chips in your design? I'd be looking at them because TTL chips kind of have a natural pull-up on their inputs so shouldnt be as susceptible to that kind of thing.
 
Hi dhau, thanks for your reply. The turbo speed (default) is 8Mhz.

tom.storey, thanks for checking out my project and for your suggestions. Though I would not immediately suspect it to be because of HC or HCT logic because it only happens when I go near the DMA chip, no other chip displays this sensitivity. I think it also could be some chip design issue. I have built various computers and mainboards with loads of HC and HCT chips and I never had this experience before like I did when touching the 8237. Maybe indeed it would decrease when using LS logic, but I still prefer HC and HCT because they are more power efficient and also provide better amplitude control. Just not every area is a good candidate for HC or HCT. The XT is a really sensitive computer in various areas. I could experiment with some resistor networks on the DMA controller, this could perhaps solve the issue. Thanks for your suggestions!

Kind regards,

Rodney
 
Glad to see that you feel the project is finished. I personally think it's quite an accomplishment.

Did you ever have a chance to try out @MicroCoreLabs drop-in 8088 replacement with your MB?
 
Hi acgs,

Thank you for your reply. I didn't check this thread for a while and found some replies here, which is always a pleasure to see some interest for my work.
I am always hoping that there are other enthousiasts out there who get excited by these kinds of projects.

When I did my XT project, it was mostly to prove something to myself but also soon after that I also understood the potential that perhaps others could find it interesting as well.
It all depends on what you find attractive and how you feel about the system.
After building it and playing around with the XT PC, I really started to like it more and more, it kind of grew on me.

Doing the mainboard like this, there is not much more required to make a full system.
I feel it really makes the system more attractive to use a real mechanical harddisk and some floppy drives.
I love the harddisks which make some real noise when the heads are writing data to the platters.
But unfortunately the nice noisy ones are mostly smaller capacity.
I remember when I was in school that I used some PCs which had really cool harddisk activity noise.
When using a more modern larger capacity harddisk, I feel it's pretty fast and I don't really feel like I want to use a CF card.

I didn't have a chance yet to try the MicroCoreLabs drop-in 8088, but this is definately very high on my wishlist!
I already made a folder in my project directory to remind myself to try this.
When I get around to it, I will definately update with my findings to this thread.

In fact, when I was trying out the V20 for the first time instead of the 8088, I was already imagining how cool it would be to emulate a CPU, this can elevate the system to a much higher level.
And I also really like newer logic which is generally more low power as well.
I wouldn't be surprised if the bus activity looks much better with this CPU solution, I will definately check it.

When I try this solution, I will definately have a good look around as well to see if there are other similar solutions to be found which I could try out.
Perhaps simulating the 8087 would also be an option, I am curious to see if there are some speed benefits for example with games.
I think emulation has huge potential because theoretically perhaps higher level processors could be emulated and connected to the XT system.
If software can be made to think that a higher level CPU is present, perhaps it could become possible to run even more software on this system.

It's really attractive to me to explore the limits of an 8 bit system, which is another reason for me to build this XT, since an 8 bit IBM compatible PC is really one of the most complete and elaborate systems possible with 8 bit technology. To expand that even further would be really cool!

Kind regards,

Rodney
 
HI Sergey,

It's great to read your reply here, thanks!
I also love your projects, really cool stuff, I have to say it again, I am grateful for your inspirational work being out there.
When I see this kind of projects, I love to look at the schematics and photos to see the hard work that went into them.
And I feel can't lag behind and should follow your example, and stuff by others whose work I have seen too.

Definately it has been a threshold for me to get myself into open source, but now I finally did, the path is open and I will publish much more stuff in the future.

If anyone is succesful in building my design, I would love to hear about it some day.
I know it's a big project to do, anyway, time will tell.
But at least I am already thankful for any interest.

Kind regards,

Rodney
 
Hi Chris527,

Thanks for your reply. Definately you are right, it has been a huge project to do.

When I made the schematics for the first prototype, I have additionally used about 40 sheets of paper notes, logic tables, diagrams, other notes.
And then all the documents, various lists in text form I assembled, cross referencing several schematics several times and measuring stuff on my clone turbo XT PCB to figure out how they did the turbo circuits.

Finally I stepped away from all of the examples and took a chance to do my own version of the system regarding clock generation and switching.
And in the end I had to convince myself to stop checking and commit to the PCB.
That first POST screen is a bit like magic happening!
I was pleasantly surprised to notice that I was not able to crash the PC by switching between fast and slow CPU speed.

Maybe when I find the time, I will add some of my notes to GitHub as well, just for fun to let visitors on GitHub see all the work.
In reality it was of course much more work than you can see just from the end results.

Kind regards,

Rodney
 
Hello everyone,

I have a small update on my XT project, if anyone read my pages about it, I mentioned that there were some delicate timing issues in the DMA circuits.
So these circuits have the most critical timing of the whole XT which especially becomes apparent when trying to format a floppy disk, which is very telling about how things are regarding the DMA timing.

I noticed on my Taiwanese clone turbo XT mainboard that they included a few 74ALS series ICs, which made me remember that I wanted to test with some ALS or F logic. Recently while working on my discrete chipset 286 project which is based on the 5170 as the starting point I was reading various TTL logic datasheets and noticed that the F and ALS logic types are especially fast, where I saw really short propagation delays on some 74F type logic ICs especially. Since I have a few parts mainboards which don't work reliably enough I decided to replace the following ICs on my XT Rev 3:
U24: 74ALS573
U55: 74F08
U58: 74F74
U59: 74F00
U60: 74F74
U24 is responsible for latching A8-A15 from XD0-XD7 through DMA_ADSTB from the 8237 and the address is output by /DMAAEN onto the address bus.
U55 - U60 are involved in detection of the HRQ_DMA from the 8237 and signalling back to the 8237 that the CPU has released the bus and it can proceed to do the DMA cycle.

After replacing the above faster ICs in those positions I have done more testing with the DMA controllers I bought from Aliexpress.
Now all of the NEC types D8237AC-5 are working 100% in my XT mainboard, even those where you almost can't read the printing on them.
I tested some older DMA controllers however those seem to not want to work properly.
Probably the DMA clock speed is too high for those older ones, which is not really worth it to me to look into further.
If I wanted to, I could turn off turbo speed and change the DMA CLK circuit to run at 3Mhz or something like that.
That should be sufficient to let even the slower DMA controllers be able to keep up with the system I believe.

I have a suspicion that all the more recent DMA controllers from Turbo XTs and early ATs will now function fine just like all of the NEC branded ones I have tested.
I have some Mitsubishi ones on an AT mainboard which I will desolder later and test them out.
I am pretty sure they will be working too. If not, I will report it here.

So indeed using faster logic was the solution to getting DMA 100% stable in my design. Thankfully no layout changes are needed.

Kind regards,

Rodney.
 
You must log in or register to reply here.
Back
Top