CTPCI : Project story
December 2011 to January 2012
After a long break (sorry !), hardwre tests and many hardware bugs removed in the CPLD (and the 2 CT60 CPLD for local bus arbitration again).
The PCI arbiter is re-designed and now USB board is running on PCI bus with Radeon.
The IDE port is now OK too. The Interupts run too.
PLX DMA read in SDRAM seems OK and is used by VDI
USB is not yet finished...
April 2010 to September 2010
Design and tests for the software and many hardware bugs removed in the CPLD (and the 2 CT60 CPLD for local bus arbitration).
The Radeon card is running fine but the use of other PCI baords is not possible. A problem on the CPI bus ?
March 2010, 3rd
I soldered an osc. from my stock (an old plastic DIP model) on a CTPCI board and it runs 100% well with the Radeon PCI.
Ouf, the PCB is bug free...
I will transfer the SMT osc. from the prototype to a new board and send this board to didier to let him work with the software.
The assembling compnay is now urged to find a part and financial solution with the part furnisher that did the mistake. I should get news this day or tomorrow.
March 2010, 2nd
Control of the oscillator on an other board : hum, it's strange, the pin #2 & 4 (GND & VDD) are connected together inside the osc. This explains well the smoke....
It seems so that the parts funisher of the assembling company did a mistake about the model of oscillator and did not take account about my osc. reference/manufacturer list with a standard pin-out (CTS & Fox references).
March 2010, 1st
I spent a bit time with the CPLD source to try to finish the PLX BURST Read from SDRAM.
Yes ! I get the boards and it's beautifull !
First power on test and.... it's smoking on the 33 MHz oscillator... Aie !
The 33 MHz oscillator for PCI seems to be inversed on PCB. I move it by 180°.
Smoke again !
The PCB are come back to the assembling company and I have to bring some parts (connectors)...
Visit at the assembling company and last layout recommendations to avoid an assembling mistake.
The 2 boards PCB are now in production phase and the assembling company is receiving the parts.
It seems there is no problem with the PCB files reading on the CAD system.
PCB routing modifications on the 2 boards to be compatible with 80 wires IDE flat cables.
Sent of the files to the PCB manufacturer for files control on their CAD system.
Unfortunately, they close on the last week of the month. It will be for January 2010 now.
The PLX transfers now with DMA (060 bus master) with read & write but only with single accesses; the burst read is not yet running.
September 2009, 1st
CTPCI orders are open.
Stills the PLX DMA and the IDE accesses to be tested and debugged.
August 2009, 21th
Running fine with 2 flat cables 80-wires of 45 cms long.
I secured timing on CTPCI and ABE/SDR chips.
Now, the timing limit of chips is far above 100 Mhz...
The only frontier to overclock more than 100 Mhz is the 060 and its cooling quality.
Those of you that were not able to overclock after 105 or 107 Mhz may have a very nice surprise...
The world record is 108 MHz without CTPCI . Mine is at 105 Mhz + CTPCI and with the standard CT63 low profile cooler...
Other news, the 060 BURST with MOVE16 instruction is working...
Stills to be tested :
- PLX master for DMA
- IDE port
I can launch some software in 1027*768* 256c or TC...
Good news : Cubase Audio 2.6 runs fine in these resolutions !
August 2009, 13th
Second day of running CT60 + CTPCI at 100 Mhz and what a pleasure to see the first benchs with Kronos en 640*480*256c et en 1024*768*TC and a big desktop !
Didier did a great work on VDI if you see the new VDI that is 10 times faster than with a CT60-100/25Mhz and 7 times faster than with the (now very) old Hades 060 ...
I did some successfull tests with 2 flat cables of 30 cms (40 wires) instead of the two of 10 cms I was using so far... I have to test the 80 wires flat cables that offering better performances (and more easy to find ), so I'm confident...
I have now to test the BURST of the 060 with the famous MOVE16 instruction and I just received the test software from Didier (yeah, we may increase the benchs !).
I get too the new tos from Didier that use the DMA from PLX to transfert data from SDRAM to Video DDR. I will be able to test the famous old CT60 arbiter (tested only with Blitter and SDMA chips so far...).
August 2009, 12th
A few days of brainstorming and some new tests witht he logic analyser to persuade me that the essential of the problems are coming from the clokcs and precisely from the cypress PLL chip that divide by 2 the CT clock for PLX.
I have unsolder it and connected directly the CT clock to CPLD and used CPLD to divide the CT clock. The CPLD get a delay of around 5 ns on the new PLX clock.
But the CTPCI is now running perfectly at 66 Mhz and all tests are good. The radeon is running fine ! No more crash.
I now boot at 95 Mhz but it stills some timing problems I have to correct in the CPLD.
We will nextly be able to test the PLX DMA.
July 2009, 19th
Yeah, we got it, the first screen from the Radeon board. It seems we get some problems with BIOS reading and finally the last problems were PLL initialization.
We don't know why but the VGA BIOS is not writting the PLL registers...
Didier forced the good data to PLL registers...without BIOS and it runs now...
OK we have bus error before the desk but that is not important... ;-)
July 2009, 16th
After 2 weeks of trying to speak with the PLX we successed the PLX registers reading & writing.
Actually tests show that we can run the CT60 up to 85 MHz with a good link with CTPCI.
PLX9054 is not able to run above 50 Mhz, it is why my logic is controlling a CT60 CLK/2 clock for PLX..
Sure the PCI is independant with 33 Mhz.
The final tests will be to go to 100/50 Mhz stable configuration.
BUT actually, after 10 days of stress and tests we are not able to communicate well with the Radeon board.
Oh, sure the PCI BIOS scan is working fine and find the board with it's ATI ID...in all the 4 slots and we can access to the PCI registers of the Radeon.
But when we load the BIOS ROM from ATI and lauch the X86 emulator, the things are bad...: master aborts and parity errors are coming and/or the X86 code is bad.
We reduced PCI flat cables to 10 cms and downgraded the PCI clock to 16 Mhz, with same results... what let me think this damned PCI bus is good and there is a software vicious bug or a not completed Radeon init...
A good news is that there is only 2 minors modifications on the CTPCI prototype and 1 minor modification on PCI slots board...if we have no more hardware bug discovered...
I have only to pay 2 new films for first batch production.
Next news should be the good ones...
June 2009, 19th
One week ago I get the last components and I just finished now to assemble the boards.
First good news : no mecanical errors; all is fitting perfectly on CT60.
Some pictures before I power on, begin the electrical controls and launch the software.
June 2009, 2nd
After some manufacturing problems because of wrong DRC settings in the software, here they are, the 2 PCB arrived this morning (June 2nd) : PCB-proto.jpg
New design of CTPCI :
- Add an IDE port (mode PIO4) 100% compatible & 50 % faster than the Falcon one (PIO3 with Combel) with logic registers to optimize the tranfer timings according to the CT60/63 clock speed. Sorry, one wire must be connected on falcon motherboard to use this IDE port. It will be possible to use the 2 IDE ports for 4 devices.
- New CPLD with twice logic (288 registers). The transfers of PLX reads (from SDRAM) are now supported with BURST. The writes of the PLX (into SDRAM) are not supported as BURST because of the clock difference between PLX & CT60/63 (PLX clock is divided by 2).
- Remove the drive 5 1/4 power connector
New design of PCI 4 Slots :
- Change dimensions for a better fitting into PC towers.
- Change routing from 4 layers to double face PCB to reduce cost.
- New atx holes drilling.
December 5th, 2008
Adding a drive 5 1/4 power connector on the main and PCI slots boards.
Adding a 5 to 3.3V 4A voltage linear regulator on the PCI Slots board.
Adding the fitting holes on the PCI slots board.
October 28th, 2008
Routing is finished. Verifications are OK. Next step : prototype.
Autorouter try : 97.4% of 1138 nets. Stills some settings to be done, some manual PCB work, some parts to be validated and tuning of the clocks traces.
September 2007 to March 2008
The hierarchical schematics are finished. This design method allows the reuse of some blocks in some others designs... I will so continue to layout of the boards with need to create some parts in the library... I present you the 'TOP' sheet of the CTPCI project : the blocks point to schematics sheet. Here is the PLX sheet.
Happy getting a bit money to buy some parts to assemble a powerfull PC in my week appartement to run Dx Designer & Pads Layout + Router software flow from Mentor Graphics. I get a licence that costs 6000...
January to June 2007
The project is going slowly because of my job far from my house and so some heavy days and shorts week ends... Finally I finished the design of the CPLD and I will go back to schematics.
After a long period very busy, I go back on CTPCI design...
I try to find a better CAD software to boost the PCB design.
Thought about a new IDE port : negative (no place & may be available on PCI).
Deepened study of the PLX 9054 to determine the logic conception with the CT bus.
Study the PCI-->Local (PCI master & DMA) data transfert rates according to the technical choices for the 'glue' logic : CPLD or FPGA.
Project on the rails again.
PCI knowledges refresh.
Start the writing of the developers documentation (CTPCI Hardware Guide).
Confirmation the project can be developed.
Announcement of the project and opinion testing among the CT users.