Invitation to Chat

https://discord.gg/p7UsfHD Channel #centurion-minicomputer

Notable Comments

Kromaine13 - March 20, 2022 9:36 PM

I am total new to this type of info exchange so here goes.... A few of the Centurion opcodes are No Op = 01, Delay 4.5ms = 0E Ok has been over 40 years so I hope this is correct ( should be close) Load A-Reg = 90 (16 bit value), Write A-Reg = B1 (16 bit memory address), Jump to = 71 (16 bit mem address) Other key date points Memory Mapped I/O address F000 - FFFF. 4-Port MUX @ F200 - F20F, Hawk Disk Cont @ F140 - F14F Disk Drive Unit number 8 bits at F140, Sectot address reg 16 bits at F141-F141( bit position format is: (bit 16) Cyl 1024 Cyl512 Cyl256 Cyl128 Cyl64 Cyl 32 Cyl16 Cyl8 Cyl4 Cyl2 Cyl1 Head 0/1 Sec8 Sec4 Sec2 Sec1 (bit 0) Max Cyl on Hawk 405 & Max Sectors 16. F142 Status Reg F143 Com Reg, write 0 = read sector write 1 = write sector write 2 = seek write 3 = RTZ write 4 = Format Sector ( I think ). A write to F148 resets the disk control to sector 0 Cyl 0 and does a disk read all in one command. Good luck.... Ken Romaine Discord Link

Kromaine13 - March 20, 2022 10:06 PM

Other data point that may help disk sectors are 400 byte DEC or 190 Byte HEX. Keep DMA setting are FFFF (-) 0190 = FEBF ( I think ) so when disk read of a sector is complete the DMA count is FFFF and the I/O is complete. The CPU has 8 bit ALU & 16 bit PC ( Program Counter ) CPU has 16 Interrupt level and the CPU reg are in the CPU card but can be access at real memory address 0000 - 00FF. The first program memory address is 0100 to EFFF. All I/O is by memory mapped I/O F000- FFFF. The CPU6 and map memory pages of 60K Bytes in and out but the 0000 - 00FF and F000 - FFFF are the same functions / data in all the memory pages. Regards for tonight... Ken Romaine Discord Link

Kromaine13 - March 21, 2022 9:16 PM

Team, Regarding the Diag. PROM Bd. The all the bus address should decode to a 16 bit address in the F000 - FFFF range ( memory mapped I/O space ) and not in the 0000 - EFFF system memory area address space. The 8 position dip switch, two 7-Seg. Displays and the scrap pad SRAM on the PROM Bd all should decode in the F000 - FFFF space. Think of the four PROMs on the card as being a extend Bootstrap PROM that has test programs loaded into Diag. PROMs. The boot flow should be CPU6 runs the backplane boot ROM, code test switch on CPU6 to see if Diag PROM Bd is installed and test need to be ran, CPU6 starts running code from the Diag. PROMs and looks at the cards 8 position dip switch for test number to be ran, start tesing when the cards test button is pushed, the test number running and test status is written into the 7-Seg Display ( Test # in lower display & test status into the upper display). Hope this helps...! Regards, Ken R. Discord Link

Kromaine13 - March 21, 2022 9:39 PM

Team & phire, If the PROM Bd. is decoding in the 0x8000-0x87FF range I did miss this point in the fine details of the cards design. The CPU & I/O sub system does have a Memory Card Read / Write override that can let the CPU's Registers 0000 - 00FF be in real DRAM ( memory card ) of the CPU's own onboard static RAM Registers again 0000 - 00FF ( This was a carry over from the old CPU4 and the CPU4 registers really being in the Core Memory card. My guess is the Diag PROM Bd is moving data / code In & out of the card in the F000 - FFFF I/O spaces so can have access to the full DRAM space to run the memory tests. ( The person that wrote the code in the PROM Bd, is Mr. Terry Little. Terry was about 25 years old an a very smart programmer, who knows...! It was over 40 years ago. ) Regards, Ken R. Discord Link

Kromaine13 - March 21, 2022 9:58 PM

Each of the 16 Interrupt level have a full set of like 8 bit register. I only used A & B in my simple test code to flip bits on I/O cards or make a disk drive do a seek. I do recall other registers like C & D. Each Int. Level had a "Context Register" that stored the current ALU flags (4 bits) and the Interrupt Vector (4 bits) for doing a return from Interrupt opcode to return to a lower Interrupt level. The only register that was total 16 bit was the PC as I recall. A trick that Terry did in the PROM Bd code to test Interrupt CPU logic (was my idea that worked) was to store a interrupt number in a memory address of the ConText Register but when not at the interrupt level, by pre-loading fake 4 bit vector values we to do Return From Interrupt opcode to jump up or down & to any level to any interrupt level. This may look a bit odd, its like the 2nd or 3rd Diag PROM test. Did this info help? Regards, Ken R. Discord Link

Kromaine13 - March 21, 2022 10:00 PM

By design the Interrupt comes from a I/O card hardware request, The Interrupt jumping from any level to any level was just a testing trick I came up with. Regards, Ken R. Discord Link

Kromaine13 - March 21, 2022 10:03 PM

Many of the I/O card did not use Interrupts The F140-F14F Hawk disk controller did use Interrupt and maybe the MUX @ F200-F20F. Regards, Ken R. Discord Link

Kromaine13 - March 21, 2022 10:09 PM

Any address of 0xF000 - 0xFFFF are in the 4KB I/O address space. The MMIO is name I do not recall. The PROM Bd. does have a little SRAM on it. Could F8000 & F8001 be the I/O address to write the values into the two7-Seg LED Display of test status? Regards, Ken Discord Link

Kromaine13 - March 24, 2022 8:30 PM

The Centurion OPSYS called @OSN did a round robin task selection for all system tasks. By default the task selection had 128 time slots. At system boot time a system user had one tick of the 128 ticks / time slots. If a key high need task was given 2 to 10 extra ticks that task did run faster. problem was system admins started increasing the ticks of large numbers of users which caused user task to run in bursts and others had long periods waiting for a slot of time. For 99% of the users and software apps leaving the time slots for all user set to one was best. As for hardware Interrupts the only cards I recall using interrupts are the disk controls during disk drive seek & RTZ commands and the CPU's DMA channel disk drive reads & writes. The printer interface & 4 Port MUX did not use interrupts. The later AMD2901 16-Port "Smart" MUX did use interrupts ( not part of David's system ). Max terminals on the 4-Port MUX was around about 16. The max terminals when the 16-Port Smart MUX was used was about 32 before the CPU6 & OPSYS ran out of gas. Regards, Ken R. Discord Link

Kromaine13 - March 24, 2022 8:36 PM

Do not recall the 4-Port MUX using interrupt in the operating system. Interrupts may be tested in the PROM Bd. code to be very possible. The 16-Port MUX was never in the PROM Bd best I recall. Discord Link

Kromaine13 - March 24, 2022 8:47 PM

Seem reasonable & F200 is the first 4-Port MUX. Not say the MUX could not do interrupt in hardware but I think once the @OSN operation system is booted from the disk drive & do not recall MUX interrupts being used. We had an issues in the MUX I/O driver in the OPSYS with the Hawk disk controller after a disk seek was completed, may have been corrected at some point. FYI - You are doing a great job rev. engineering the system. Ken R. Discord Link

Kromaine13 - March 24, 2022 9:01 PM

CPU6 registers are in SRAM on the CPU6 card. ( On the CPU4 the registers are in the main core memory or new 32K DRAM card. ) The CPU register layout are common all models CPU4, 5 & 6. They can be ref. by memory read write @ addressed 0000 - 00FF or normal opcodes running at the current interrupt level 0 - F. The hardware just added the interrupt request to reach the registers for that interrupt level. Which switching to a higher interrupt the hardware stored the ALU & status flag bits and current interrupt 4 bit value in the [ S reg ] so the hardware would know what to do when the higher interrupt level did a [ Return from Interrupt ] command once the interrupt request was completed. FYI - each level 0 - F had a full set of 8 bit registers and a 16 bit Program counter ( PC ). Does this help? Ken R. Discord Link

Kromaine13 - March 24, 2022 9:07 PM

It was a normal register that was set by the hardware interrupt at a given level or by a memory read / write command. Do not recall the adderss but it would have been something like 000A for interrupt 0 001A for interrupt 1 ..... 00F0 for interrupt F. I think the PC was always 000E-000F 001E-001F ..... 00FE-00FF. Ken R. Discord Link

Kromaine13 - March 24, 2022 9:15 PM

Your are correct. Did do some engineering tech work and wrote a few diag program to wiggle bits on cards to make the disk drive seek or print some text on the printer or CRT. A few bytes in machine code or assm. laug. Ken R. Discord Link

Kromaine13 - March 24, 2022 9:36 PM

Learn something new tonight about the backplane for the CPU5 & CPU6. Best I can figure this is a One-Shot timer for holding the system & bus in RESET till the DC power supplies are stable. If the power On Reset was release before the DC voltage are stable the cards with AMD2901 & microcode had a habit of screwing over the disk drive controller for the CMD and Winchester drive. Ken R. Discord Link

Kromaine13 - March 24, 2022 9:39 PM

FYI - Team, The CPU4 backplane had zero logic on the backplane. Only had bus terminator resistors on the right end of the bus like on the CPU5 / 6 backplane. Ken R. Discord Link

Kromaine13 - March 24, 2022 9:46 PM

The very old CPU4 system had a front panel that had a single step button. It was dropped from the CPU4 after the first few systems, no value to the customer. Very saw a single step function used on the CPU5 or CPU6. By the time of the CPU5 & 6 we had custom hardware in mfg and the repair dept. to step the microcode as needed using the two 50 connectors on the AMD2901 based cards. Ken R. Discord Link

Kromaine13 - March 24, 2022 10:03 PM

Be careful handling the CPU6 board. The muilt-wire was not the best if the card was flexed. Most CPU6 card problems are bad socket pin contact with the chips and open wires between points on the circuit board. The less the CPU6 card is handled at this point the better. Ken R. Discord Link

Kromaine13 - March 25, 2022 6:56 PM

Good guess for (O, I, H, R, 1, 2, 3, M) would be a a carryover from the old CPU4 front panel (no switches on the CPU4 cards) 1, 2, 3 are very likely the other 3 of the 4 Sense switches. The CPU 5 & CPU6 front panel (R/F) is 100% Sense Switch #4 a CPU opcode does a {Jump / Branch} to memory address XXXX if true to boot from disk drive #1 or no {Jump / Branch} if false boot from disk drive #0 ( could have this back words). R/F was used to select booting from Removable Disk or Fixed Disk then running the BootStrap ROM on the backplane. As for O, I, H, R & M just a guess would be Interrupt, Halt, Run. No idea for O & M, sorry. Discord Link

Kromaine13 - March 25, 2022 6:59 PM

It is the Bottom or Top dip switch on the CPU6 that tells the BootStrap ROM to jump to the Diag.PROM Bd. starting address to run the PROM Bd. test code. Regards, Ken R, Discord Link

Kromaine13 - March 25, 2022 8:06 PM

If we only had a working Hawk drive an a good system disk pack. We would press SELECT to boot stand-a-lone from a good "systems disk" pack on disk 0, run Format 1 on a new Hawk fixed platter #1, run DDump 0 to 1 to copy the "system pack" 0 to the Hawk's fixed platter 1. Spin down the Hawk swap out the Hawk pack 0 for a apps. disk pack & spin the Hawk back up. Last step would be press the OPSYS on the front panel, OPSYS would boot from platter 1 and the Centurion would be ready to start running a companies Payroll apps or the accounting G/L apps. Good job team. Looking forward to the machine code programs the team writes using the Diag. PROM Bd. as a starting point tool and the new I/O devices you come up with. Regards, Ken R. Discord Link

Kromaine13 - March 25, 2022 8:54 PM

Well the CPU4 had a 5 MHz system clock & bus clock. The CPU5 or 6 may have clocked the AMD2901 faster but the timing for the I/O bus still needs to be 5 MHz for all the old & new I/O card to keep working. Regards, Ken R. Discord Link

Kromaine13 - March 25, 2022 9:06 PM

The core cards we had did a auto re-write after read. This re-write was 100% handled in the core memory card, no extra load on the system CPU. Discord Link

Kromaine13 - March 26, 2022 9:41 AM

phire, The Centurion I/O cards @ F000-FFFF and the CPU @ 0000-00FF can source a TTL logic level signal on the backplane called {address preempt}. The {address preempt} blocks any Read / Write cycle on the DRAM memory cards and lets the CPU registers or I/O respond to the memory Read / Write cycle. Regards, Ken R. Discord Link

Kromaine13 - March 26, 2022 10:02 AM

I used the 0E delay opcode a lot when I wrote simple hardware logic test programs. The 0E caused the CPU system and DMA channel Read / Write to "Stop", yes stop for 4.5 ms. It was a easy quick way to intro a little delay without doing a loop counting NoOps ( 01 ) for thousand of cycles ( my test code was running standalone and was maybe 5 to 20 bytes long with 32K of memory so 0E was just easy. ( The programers laughed at my simple wasteful code but it worked for my testing needs. ) The problem with 0E in production code is that if a disk drive Read / Write is happening an the CPU runs a 0E the disk platter does not stop spinning and the DMA is waiting for the rest of the 4.5 ms to complete before it can complete the disk R/W. Bottom line is the read data is from the wrong sector or the data written was in the wrong sector on the the disk platter and a trashed file system...! We looked for the root cause of random trashed file for years before we figured out the 0E delay was overlooked in a part of a typewriter I/O subroutine used in the operating system for the MUX card. To much history,,,,, Regards, Ken R. Discord Link

Kromaine13 - March 26, 2022 11:25 AM

Old Centurion story.... Had a programmer working from home at night that from time to time. The programmer would crash the OPSYS and I had to drive into the office to push the front panel Load OPSYS button to reboot the system type a few commands on CRT-0 on MUX port 0 to setup system MUX port 3 to run at 300 baud for the 300 baud modem. Then the programmer could dial back into the Centurion system and go back to work. I had asked the systems programmer to write a Watch Dog Timer into the OPSYS and I would build a 10 second timer on to a old printer interface I/O card to reset the system if the OPSYS crashed and could not keep the Watch Dog Time alive by {poking} the WDT each few seconds. Only other thing I needed the system programmer to do for me was have the OPSYS code set MUX Port-0 to 300 baud not 9600 by default ( OPSYS had to be started from Port-0 and the MODEMs of the day are only 300 baud). part 1 of 2 part 2 of 2 The system programmer would not help with my request.... so the hardware engineer fix was started..! First hard wire the UART baud rate on MUX Port-0 to 300 baud with mod-wires on the MUX card. Move the dial-in MODEM from Port-3 to Port-0 now at 300 baud. Next place a micro-switch on the MODEM's analog phone's Bell Clapper arm so when the phone rings the switch toggles. Run a wire from the switch on the phone ringer arm to the Centurion from panel ( Select ) button. Now when the MODEM phone number is dialed by the programmer working at home the micro-switch toggles with each ring till the MODEM picks up the line. The Centurion system thinks a person has pushong the front panel ( Select ) button. MUX Port-0 is active at 300 baud not the default of 9600 baud and now the remote dial-in user receives boot prompts at home at 300 baud and can run standalone programs or enter @OSN to load the OPSYS and get back to work. I did not have to drive into the office at night any longer to push a button. As for the systems programmer he said I was nuts...! After this he was open to more productive ideas. Just a old story of getting around a problem. Regards, Ken R. Discord Link

Kromaine13 - March 26, 2022 11:59 AM

No the DMA channel does not share the memory cycles with anyone. In between the DMA's need to R/W memory the CPU could exec. opcodes but the DMA could take over ( all but be the 0E delay opcode of 4.5 ms, bad news for Disk DMA I/O). When the disk heads are over a sector to Read / Write its DMA time no waiting the disk is spinning at 2400 RPM on the Hawk or 3600 RPM on the CMD. If the DMA count register is not back to FFFF by the end of the sector R / W the disk control just adds one to the sector addess to be the target of the R / W operation and keeps moving data to or from the disk sector(s), The Centurion system could do 1 or 16 sector R / W operations. Each sector was 400 bytes long 0x0190 bytes. The DMA control logic needs just a few items; starting memory address for R /W, number of bytes to move and the timing control signals. A 400 byte sector read or write setup would load the DMA count register with FFFF less 0x0190 (one sector). When the DMA cycled it added one to the count, after the count was again FFFF the DMA operations are completed. Reguard, Ken R. Discord Link

Kromaine13 - March 26, 2022 12:09 PM

On the CPU4 the DMA was not part of the "CPU" is was a F000-FFFF I/O card. On the CPU5 & CPU6 the DMA function became part of the CPU 5 & ^ card but had to look like the DMA card to all the old I/O cards like the MUX, Printer interface & Hawk controller. Regards Ken R. Discord Link

Kromaine13 - March 26, 2022 12:15 PM

The Centurion system does only have a single DMA channel. The single DMA was a point of many hallway engineering debates on way performance issues the 20 to 32 terminals systems. Was not the single DMA channel but the CRT terminal software driver lookup control table. Fixed the Smart 16-Port AMD2901 MUX. Regards, Ken R. Discord Link

Kromaine13 - March 26, 2022 12:23 PM

Some one had to setup the DMA channel to get the data moved byte by byte across the disk controller into the DRAM for the CPU to complete the reading of track zero from the selected drive based on the ( R/F ) sense SW#4 status. The backplane boot ROM should have some DMA setup code ( It did in the 1720 EEPROM for the CPU4, maybe the CPU5 & 6 had a micro-code trick that on pushing a button it just setup the DMA logic in the CPU5 & 6 [ auto magically 🙂 ]. Regards, Ken R. Discord Link

Kromaine13 - March 26, 2022 12:26 PM

Was the first engineering tech at Warrex Corp & Field Engineer then later the director of customer service and director of mfg. Regards Ken R. Discord Link

Kromaine13 - March 26, 2022 12:33 PM

If any thing it would have had to be the I/O card position in the backplane. Closes to the CPU first then cards to the right last. In some of Davids YouTube videos you can see two jumpers between the CPU connector and the first I/O backplane slot. I know one was for serial interrupt chain for cards had shared the same interrupt level, maybe other was DMA priority. Regards, Ken R. Discord Link

Kromaine13 - March 26, 2022 1:07 PM

Seem to now remember the CPU-5 & CPU-6 having a DMA opcode unlike the CPU-4 have DMA registers in MMIO. A CPU-6 opcode should be something like this ( DMA Opcode, @memory start, &byte count ) Key byte count numbers to be looking in the code are FFFF less 0x0190 ( fe6f one sector disk I/O ) or FFFF less { F } times 0x0190 ( e88f 16 sector disk I/O ). Then the DMA reached FFFF the DMA I/O was completed. Please double check my hex numbers are correct. Regards,Ken R. Discord Link

Kromaine13 - March 26, 2022 1:24 PM

Odd that is not a even number or 400 bytes sectors. Oh.... The CMD drive had 512 byte sectors on the drive but Centurion I/O records had to stay at 400 byte so the sector was padded from 400 to 512 bytes so this could be the odd DMA byte count? The Hawk sector was 400 with 400 the CMD sector was 400 padded to 512 to make the CMD soft sector size. Regards, Ken R. Discord Link

Kromaine13 - March 26, 2022 6:54 PM

The Diag PROM bd. should only have to receive data or provide data to the data bus as requested by the CPU Read / Write request at a given memory address or MMIO address. Can not see why the PROM bd. ever needs to set a address on the backplane address bus. Regards, Ken R. Discord Link

Kromaine13 - March 26, 2022 7:48 PM

The Centurion models like [ Warrex Centurion IV ] really spec-ed the cabinet and type of drives in the system not so much the CPU type like CPU5 or CPU6. Based in the max memory (256KB) the picture called out it had to be a CPU-5 or CPU-6. Max CPU-4 memory was 64K - the MMIO. Regards, Ken R. Discord Link

Kromaine13 - March 26, 2022 7:54 PM

CPU-5 was around 1979-1980 and the CPU-6 was around 1981-1982 (+/-). Discord Link

Kromaine13 - March 26, 2022 7:59 PM

The CPU-6 clock was a bit dirty. We had clock fan-out driver issues and had to hand pick sometime I.C.s in key logic functions. I agree the back plane signal look a bit trashy, was the scope probe grounded to the DC backplane gound stud? Not what I call on my old Tek 465B scope. Regards, Ken R. Discord Link

Kromaine13 - March 26, 2022 8:12 PM

The adds from the old CPU-4 days. Most of the first Centurion dealers had been selling Lition Data System paper tape / card systems before they started selling Centurion. This was good for Centurion because the new dealer network already had sales, programmers and support staff ready to support the local end users. Regards, Ken R. Discord Link

Kromaine13 - March 27, 2022 1:22 PM

This Trick may help with Pi and terminal @ same time. When I would teach Centurion classes to field engineers on how to service the system I used this trick RS-232 terminal cable between MUX Port-0 and up to 5 to 10 ADDS terminals. MUX Port-0 XM Data pin was connected to all the terminals RC Data pin. Then only the XM Data pin on my terminal was connected back to the MUX Port-0 RC Data pin. Now all the students could see what I entered on my terminal and how the CPU responded but the could not enter commands to the CPU from the students terminal. Why was this needed? We did not have projectors to display the CRT-0 / Terminal-0 screen on the wall for all the students to view in the 1970's. All the system diags. had be ran from CRT-0 on MUX Port-0. With the trick RS-232 cable each student had a read-only-display terminal in front of them. Regards, Ken R, Discord Link

Kromaine13 - March 27, 2022 1:32 PM

Saw a request on seeing a CPU4. I only kept a very few pictures over the years. Regards, Ken R. Discord Link

Kromaine13 - March 27, 2022 8:47 PM

The single vert. 8" floppy was in the of the rack. Below the floppy was CDC Falcon dual platter 5MB + 5MB (same controller as the CDC Hawk but less cost). The data was copied into and backed up using the 8" 160KB floppy to the 10 MB Falcon ( real pain in the neck an slow ). Many customers lost data because they did not do customer data backups. We never built a all floppy system because the OPSYS files would not fit on a single 160KB disk. The OPSYS had to be on the same disk volume. A later rev. of the floppy controller (AMD2901 based) controlled 5-1/4 floppy (720KB) and CDC 5-1/4 Finch 24MB or 32MB hard drive, letting us build a desktop Centurion Micro Plus system with CPU6, 128KB DRAM & 4 port MUX. Picture of CPU4 was upside down to sit on a table for a lab test system, non production. Regards, Ken R. Discord Link

Kromaine13 - March 27, 2022 8:55 PM

Diff. backplane (4 slot) and lower wattage power supply, same cards & software as the big Centurion systems. We sold hundreds of Micro Plus systems to EDS and CUNDA Credit Unions for small branch banks & credit unions. Regards, Ken R. Discord Link

Kromaine13 - March 28, 2022 7:51 PM

The Hawk drive logic cards have small hole top - center in the cards. CDC had a tool with a hook & lever to pull the Hawk cards out. I personal used a #1 Phillips screw driver tip pushed into the hole then and the screw-drive shaft resting on the next Hawk as a pivot point ( watch out not to cut any traces above the hole in the card you are pulling out ). Regards, Ken R. Discord Link

Kromaine13 - March 29, 2022 7:31 PM

The ceramic resistor packs on the Hawk I/O card are the 50 pin ribbon cable terminators for the I/O cable to the Centurion disk controller. You can daisy chain up to (4) Hawk drives to the controller. The last Hawk on the 50 pin cable has the 110 & 330 ohm terminator packs and the middle drives on cable have the terminators removed. The 50 pin ribbon cable must be a shielded cable & grounded to the Hawk's cable clamp on the Hawk I/O card to prevent read / write data errors. ( You may have also noticed a terminator card plugged into the CMD Phoenix I/O card for the daisy chain 60 pin non-shielded "A" cable. The CMD also has a point-to-point 26 pin data cable to the Centurion CMD controller for each of the 3 max drives. ) Regards, Ken R. Discord Link

Kromaine13 - April 01, 2022 6:49 PM

The Centurion printer interface card will have a very simple design that dates back to the CPU4 days and still works without any revs. for the CPU5 or CPU6 systems. Will notice a number of small wire jumpers on the printer I/O card that let us adjust logic levels & signal timing facing the attached printer model. The printer interface card logic facing the Centurion backplane should be the simplest MMIO (just a few registers and no RAM or DMA). In this setup with the Data 100 line printer the printer interface signals are the standard Centronics Model 101 parallel interface. This simple Centronics parallel interface is what IBM used in the first IBM PC's. By changing the jumpers on the Cemturion printer card, printer cable and the OPSYS printer driver this same card supported parallel printers from Centronics, Control Data, Potter and Data Products. Any serial attached printers like the Texas Instruments TI-810 connected to the 4-Port MUX just like the ADDS terminal. I do not recall any printer drivers or printer tests being in the Diag PROM bd. Regards, Ken R. Discord Link

Kromaine13 - April 01, 2022 7:05 PM

The printer card will have the F000-FFFF MMIO address decode a 8 bit register to send data to the printer, think one 4 bit ( maybe 8 bit ) register to read printer status like Printer Ready, Out Of Paper & Printer Fault. A data strobe line to clock 7 bits of data out of the card & into the printer. Do not recall any "state machine logic" on the Centurion printer I/O card. Great card to learn the basic Centurion bus logic & timing. Regards, Ken R. Discord Link

Kromaine13 - April 02, 2022 3:02 PM

In ref. to the age of the Centurion system and the circuit boards. CPU4 systems only used a black cabinet half the high [single drive cabinet] of the CPU-5 & CPU-6 [dual disk drive cabinets]. Overall colors are still tan, blue and rosewood trim. Older CPU-5 & CPU-6 systems are black square tubing cabinets and much later production 1982-1985 likely had tan color folded tubes. As for the circuit the Warrex-Centurion part numders started at P/N 1001 CPU4 Disk-Autoload,{Hawk Drive], P/N 1002 Disk-II [Hawk Drive], P/N 1003 32K DRAM (4K x 1 22 pin chips - shown in old picture with eng. tech. looking at 32K card on an I/O extender card), P/N 1004 4-Port MUX, P/N 1005 Printer Interface. After most all of the old CD200 & Fedder computers area Texas had been upgraded from Sykes tape, 4K core memory and Friden Flexowriter terminals Centurion started building our own CPU1, CPU2, CPU3 and DMA cards. ( We purchased CPU1, CPU2, CPU3 and DMA cards (CPU4 model) from Fedder for a short time before we did our own re-design. The Fedder Hawk controller had major design problems and the Warrex Hawk controller was 100% new design board set. ) {part 1 of 2} Discord Link

Kromaine13 - April 02, 2022 3:02 PM

{part 2 of 2} As for serial numbers on cards & mfg dates. Early 1974-1976 cards are hand solder (not wave solder boards ) with S/N like 20030175 meaning 20th card built on March 1st 1975. The picture of the DSK/Auto card p/n 1001 with the 1702 EPROM s/n 67 was likely built in 1976-1978. The newest p/n 1001 (now called the DSKI) dropped the 1702 EPROM (Hawk & Sykes boot) and started using two small 16 pin by 32 byte ROMs (Hawk only boot) around the late 1979-1980. FYI - I recall the first CD200 - Fedder system boot strap was 8 X 12 diode matrix for the ROM code. I had to move one diode to change the boot device. An the last bit of age data is the Warrex or Centurion stickers on the cards. We put stickers on new systems and high dollar items to track warranty start date from our factory or after a circuit board was repaired [MMYY}. We started using stickers in the later 1970s. Regards, Ken R. Discord Link

Kromaine13 - April 02, 2022 8:36 PM

I met David Fedder one time in Richardson, TX, recall Fedder was from the DC - Baltimore area. Never heard of him selling new systems, we just purchased CPU4 cards ( CPU-1, 2, 3 & DMA card for a short time (less than 6 to 8 months). Warrex spent its early days upgrading the old 4K core Fedder systems till we started building the Warrex CPU1, 2, 3 & DMA boards. Fedder's early Hawk controller we called the "A" controller really had problems. The Warrex design we called the "B" Hawk controller ( DSK/Auto & DSKII) worked great. I have looked for Fedder Systems or the CD200 CPU / system with no luck finding any ref. or docs. Did find this [ William Fetter, also known as William Alan Fetter or Bill Fetter, was an American graphic designer and pioneer in the field of computer graphics. He explored the perspective fundamentals of computer animation of a human figure from 1960 on and was the first to create a human figure as a 3D model. Wikipedia ] but I remember his name as Fedder not Fetter, could be 100% wrong but it does look like him. Someone on the ( I Bought a Minicomputer Part - ? ) made a comment that he worked on Fedder Systems as a programmer in the past. Regards, Ken R. Discord Link

Kromaine13 - April 03, 2022 12:55 PM

When tracing out the circuit paths on the 128KB memory card(s) I think you should remove the 16K DRAM chips from the sockets. Not shur the DRAMs would like the VOM current flow when ohming our the circuit path. Just an idea to be safe with all the DRAM chips. Regards, Ken R. Discord Link

Kromaine13 - April 03, 2022 8:46 PM

Agree the heads look like CDC Phoenix heads. The Phoenix (3) types of heads [Servo Read Head], [Upper R/W Data Head], & [Lower R/W Data Head]. The picture show Upper & Lower R/W Data Heads. The Servo Head has two of the white plugs on the cable. One plug holds the servo head cable in place like the R/W heads but it has 4" jumper over to the servo read preamp card. Top surface of the Phoenix disk pack and the top surface of the 3 platter fixed disk assembly is all servo track di-bits for servo positioning (no data). Regards, Ken R. Discord Link

Kromaine13 - April 03, 2022 9:32 PM

Your top surface of the Phoenix is crashed and the servo will also be trashed one it mark the disk surface. Bad thing about the 3 platter fixed disk stack is its a one time install and can not be moved from a parts drive into a 2nd drive as a service replacement. Then CDC ships a replacement Phoenix fixed disk assembly it has an alignment tooling mount to the top of the fixed disk stack. One the fixed disk stack is installed the tooling is removed and that stacks life time is 100% on that spindle spinning @ 3,600 RPM. Regards, Ken R. Discord Link

Kromaine13 - April 03, 2022 9:40 PM

Can not move Phoenix fix disk platter assm. from one drive to a 2nd drive. The servo track runout will never be in spec for the servo read heads to track the center of the cylinder. Had a customer engineer remove a couple of Phoenix fixed disk stacks to clean the platters (CDC manual states never remove) and trashed two Phoenix drives (replace cost to replace fix disk stack is about $20K each in the 1980s). Regards, Ken R. Discord Link

Kromaine13 - April 03, 2022 9:48 PM

Phoenix used the servo head to local & track the cylinder, sector position and also as the read / write data clock source. Finding a working 9427H and disk pack is the best hope to find a Centurion operating system. Regards, Ken R, Discord Link

Kromaine13 - April 06, 2022 5:56 PM

The way the card cage is shown in the picture @ 4:43 PM today is normal layout all but the printer card. Normal the printer interface is not installed between the memory cards but in the gap to the left of the memory cards. We just always seemed to install the memory card to the right of all the other cards so they can get good airflow & stay cool. Yes the 26 pin DMA cable crossing over the MUX cards is odd but thats how we did it at Centurion. Recall the two bus jumper wires installed in place of the missing backplane MMIO card connector. they for the interrupt daisy chain logic. If a MMIO card that uses interrupts has a open slot to its left the interrupt daisy chain will not work for cards using the same interrupt level. So we did not leave empty slots between the CPU card an any MMIO that could use interrupts ( the printer MMIO card should pass interrupt singles from right to left on the bus ). And the CPU4 system also had the same bus term. resistor at the right side end of the bus. The CPU4 backplane had totally diff. connectors for the slots 1, 2 & 3 (the CPU4 card slots) and no logic or PROMs on the backplane. Regards, Ken R. Discord Link

Kromaine13 - April 06, 2022 6:00 PM

phire you are correct on suppressing interrupts on the same interrupt level for any MMIO card to the right of the MMIO card currently being serviced at that same level, Regards, Ken R. Discord Link

Kromaine13 - April 06, 2022 6:08 PM

The CMD controller for the Phoenix (and many other CDC drives in the CMD lineup) uses the Centurion CPU5 & CPU6 DMA logic to move the CMD command string & status in & out of the CMD controller it also uses the DMA logic to move the disk I/O read & write data from disk to main system DRAM. Regards, Ken R. Discord Link

Kromaine13 - April 06, 2022 6:15 PM

Centronics, CDC, DataProducts, Data-100 and Potter are all models we sold over the years. In most cases we had to mod a current printer I/O driver in the OPSYS, adjust the printer card jumpers (as needed) and build a new printer cable to match the I/O connector that printer vendor used on the printer end of the cable. Only parallel printers. All the serial printers like the TI-810 was driven form the 4-Port MUX card. Regards, Ken R. Discord Link

Kromaine13 - April 06, 2022 6:24 PM

I even used the printer interface card with some RC circuits connected to each of the data bits leaving the card. The RC time constant (resistor & cap) was Poked by a custom OPSYS so we could really tell how long each major OPSYS or I/O task was really taking the the OPSYS Task Selection. Just monitor the DC voltage on the RC circuit compaired to the other RC circuit DC voltage level to gain insite into offen the OPSYS got around to Poking bit for that RC circuit. This is how we found the 4-Port MUX card / ADDS driver problem.... Many of hardware & software engineers feel we had a single DMA channel issue slowing down the Centurion systemm not the DMA it was the MUX driver. Old History but fun. Regards, Ken R. Discord Link

Kromaine13 - April 06, 2022 6:38 PM

We supported the complete CDC CMD line & SMD drives with the Centurion CMD (AMD2901) controller. If it was a CDC drive with (1) 60 pin A cable and (1) 26 pin B cable we likely had a I/O driver for it. Regards, Ken R. Discord Link

Kromaine13 - April 07, 2022 6:27 PM

phire, Best that I call is the 128KB memory cards did cause and interrupt if the 128KB memory parity detect logic was triggered because of a single bit error when a byte was read out of DRAM. Having a open backplane slot to the left of a card was only a issue it two or more cards using the same interrupt level requested an interrupt at the sametime. So in the case of the 128KB DRAM cards you would never have a Data Read on two diffrent memory cards on the same bus read cycle. We did have a few CPU6 system with dual 10MB Hawk disk controllers Hawks 1, 2, 3 & 4 at MMIO 0xF140 and Hawks 5, 6, 7 & 8 at MMIO 0xF?40. Why I am bring this up is we had a CPU6 micro-code bug with the AMD2901 logic that cause intermittent lost of [disk seek complete] "interrupt" from Hawk drives 5 - 8 only if pending seeks are waiting to complete on Hawk 1 - 4 & Hawk 5 - 8 at the sametime. If Hawk drive 1 - 4 completed the seek before Hawk drive 5 -8 complete its seek the CPU6 / OPSYS lost trace of the pending seek command issued to Hawks 5 - 8. After many seconds and OPSYS clean section of code would display a message on the System Console (Port-0 on MUX-0). Both sets of the Hawk disk controllers shared the same interrupt level (I think level - 2). Regards, Ken R. Discord Link

Kromaine13 - April 07, 2022 6:42 PM

To understand the printer interface a little better take a look at the signals from the Data-100 (Centronics interface) point of view. Pin-1 on the 34 pin end of the printer cable (you had a printer cable in your box of cables) is the data strobe from the Centurion printer card to clock the 7 bits of ASCII data into the printer. I have a feeling that the 8th bit on the printer card and some of the odd logic counter string is what starts the small delay after the 7 bits of data is sent to the printer than "strobe" into the printer a ms or so later. This URL seems to cover the Centronics interface https://computer.howstuffworks.com/parallel-port1.htm Regards, Ken R. Discord Link

Kromaine13 - April 07, 2022 7:10 PM

The Hawk will not come online till the drive is at 2,400 RPM, Heads load & RTZ to Cyl-0 and the sector card in the Hawk sees sector pulses from the fix disk sector ring and the removeable disk pack sector ring. If the Hawk was working the PROM Bd. test are ( RTZ, Seek to Cyl=0, Seek to Cyl=1........ Seek to Cyl =405, Seek to Cyl=404, Seek to Cyl....... Seek to Cyl=0 then start over ). ( Hawk Seek to random Cyl 0 - 405 & I think Seek to Cyl 0, 2, 4, 8, 16, 32, 64, 128, 256 then 405 next rev direction back to Cyl=0 & repeat. ) The 3rd last test may just be a Read of the Sector Address 0 - F, head = 0 , Cyl = 0 ( Sector Address values would be set at MMIO 0xF141 & 0xF142 ). Sectors 0-F are the SELECT / OPSYS bootstrap and the disk platter's bad sector map table. Regards, Ken R. Discord Link

Kromaine13 - April 07, 2022 8:09 PM

My indepth time with the CPU was 90% CPU4 (Eng Tech) and 10% CPU5/6 ( Mgr / Dir. position = brain rot ). The Hawk DSK/Auto - DSKII, 4 Port MUX and Printer Interface did not change from CPU 4, 5 & 6. All the AMD2901 based controllers and the 128KB parity memory cards (old 4K, 16K & 32K did not have parity) only worked on the CPU5/6 systems. Wilth you digging into the opcodes I would go with what you are finding. I am 100% on the 16 interrupt levels and each interrupt level having a full set of registers that are both CPU registers at a given interrupt level or memory location 0x0000 - 0x00FF. Regards, Ken R. Discord Link

Kromaine13 - April 07, 2022 8:16 PM

Re-reading your "interrupt theory" all the MMIO cards have to know the current CPU interrupt level before the MMIO card can request an interrupt on that MMIO cards to be serviced by the CPU and the cards to the left on the same level are serviced first. Regards, Ken R. Discord Link

Kromaine13 - April 07, 2022 8:41 PM

The big change from CPU4 to CPU5/6 is the max memory supported (CPU4 65KB - MMIO & CPU5/6 256KB - MMIO per 65K blank) and the CPU4 MMIO DMA card -vs- the CPU5/6 opcode based DMA logic and DMA modes in the CPU5/6. Regards, Ken Romaine Discord Link

Kromaine13 - April 08, 2022 7:06 PM

The Hawk controller has very simple Unit Select, Sector Address & Status registers. The Command register for Seeks & RTZ is simple but Read, Write & Format starts a complex TTL state machine that counts every bit time for a disk Read, Write & Format. The R/W/F commands must handle sector address preambles & data field preambles, data, plus 16-bit CRC and the DMA timing. Repairing a bad DSK/Auto or DSKII with a bad TTL chip in the state-machine was a pain, one bit out of sequ. and the state-machine resets the R/W/F operation. Regards, Ken R. Discord Link

Kromaine13 - April 09, 2022 9:24 AM

The Centronics interface used on the Data-100 printer is the standard 34 pin shown in your 3rd picture just below the power supply. As for testing the Data-100 printer it was a Red Switch on one of the Data-100 logic boards poking out from the black electronics cover. The test switch will cause the printer to print a rolling ASCII barber-pole. Its true the Data-100 self test does not test 100% of the printers electronics but I never recall self-test running and still having a electronics issue. Regards, Ken R. Discord Link

Kromaine13 - April 10, 2022 10:52 AM

That was a Centurion dealers office. Warrex - Centurion was designed & mfg in the Richardson, TX area. Over a period of 10-15 years Warrex - Centurion mfg had it mfg site in Richardson, North Dallas, Richardson and Allen, TX. Our largest and best years are when Centurion had a 40,000 sq. ft. mfg building at 1780 JeyEll drive in Richardson, TX. during 1979-1983 (+/-). Regards, Ken R. Discord Link

kromaine13 - April 12, 2022 6:44 PM

Cincinnati Data System one of the first few Centurion dealers. I personal install Howell & Chuck's first system around 1976 (+/-), a CPU4 with 32KB DRAM and 10M Hawk drive. Howell & Chuck are the two men in the sales add picture looking at a book (one sitting is Howell, standing is Chuck). In fact C.D.S. was a dealers years before Cruze Computer became a Centurion customer then later a Centurion dealer. Buddy Cruze was a good man & good business personal also. [part 1 of 2] Discord Link

kromaine13 - April 12, 2022 6:44 PM

[prt 2 of 2 ] Buddy purchased the remains of Centurion Computer (after EDS the Centurion the mfg. was called Z-Tron Allen, TX). After Buddy purchased the remains from the bankkruptcy court I left the company at started at Proteon (the first routers p4200 and 10 & 80 Mb Token Ring years before Cisco routers and IBM 4Mb Token Ring). Regards, Ken R. Discord Link

kromaine13 - April 12, 2022 7:10 PM

Centurion was unsucessful in the word processor market. Products like Lexatron and a PC with WordStar killed any Centurion W.P. offering. Centurion was centered around accounting software for CPA, Hospital, Banking, Inventory and Oil & Gas Production. Regards, Ken R. Discord Link

kromaine13 - April 12, 2022 8:06 PM

Just download Mr. Paul Love's book on to my wife Kindle. Reading the section called [ Enter the Centurion ] lines up well with what an early Centurion Dealer like Automated Business Systems ( ABS ) in Tulsa, OK. would think about the Centurion products. I personal do not recall Mr. Love as a programer at ABS but his recall of John Warren death in a car wreck is correct. Zoe Warren (wife) reveived more that a few $$$ from GM & Firestone over John's death in her brand new Corvette. Regards, Ken R. Discord Link

kromaine13 - April 12, 2022 8:17 PM

Mr. Love did call out a few very old customers like Cains Coffee. The one bank in Texas he ref. was the Bank of Deport, TX. Mr Jeffery the bank ower would not let me drive to his bank with the Warrex logo on the company car. He did wish for his customers to know he was using a computer at the bank..! It was a book keeping machine not a computer. At this point the City of Deport, TX still had a wood board walk in front of the bank on main street. Regards, Ken R. Discord Link

Kromaine13 - April 14, 2022 5:34 PM

Max, The ref to Centurion and software contract for 3270 terminal support does not seem to be the same Centurion Computer that built the Warrex - Centurion line of business computers. For the short period of time in the 1981-1983 that EDS ( Electronic Data Systems - Ross Perot's company ) owned Centurion we did build a few hundred IBM 3270 terminal clones (Centurion 3270 terminals) for EDS internal mainframe use. We even build a few thousand Centurion XT class PCs for EDS internal use. Regards, Ken R. Discord Link

Kromaine13 - April 14, 2022 10:52 PM

1780 JayEll is a Pilgrim's Pride chicken data center for the past 10 years (+/-). When Centurion moved to 1780 JayEll the building was new. next was Centurion-EDS, EDS Tech Products (after Centurion was sold to Z-Tron). Then EDS sold the building to Perot Systems then last its been Pilgrim's Pride chicken data center. Regards, Ken R. Discord Link

Kromaine13 - April 14, 2022 10:55 PM

Draco, Because I know Buddy Cruze (now passed away) I called the phone # for Byron Williamson, not in service. Months ago I also emailed and filled out the online contact request and no reply also, Regards, Ken R. Discord Link

kromaine13 - April 15, 2022 3:45 AM

Max, 1002N Cent. Expy. Richardson was the location of the Warrex Computer office back in 1975. We have one suite in the office park then later added a 2nd suite. The 2nd suite later burned during a fire started in the wood shop which made the rosewood trim panels and top for the Centurion Computer. Around a year later we moved the company to 12505 Cent. Expy. Dallas, TX (the old Dal-Tile building). A few years later we moved to 1780 JayEll in Richardson. FYI - If you look at this URL and replace the [ CIGAR ] sign with a [ Warrex Computer ] sign that how the first Warrex site looked ( less the man in the google pic ). https://www.google.com/maps/@32.9627584,-96.7246078,3a,69.5y,133.96h,83.75t/data=!3m6!1e1!3m4!1sLQ7WHQhZyrGqBXYNvGsbZQ!2e0!7i16384!8i8192 I recall the address as 1201 not 1002, but its been over 45 years. Regards, Ken R. Discord Link

kromaine13 - April 15, 2022 4:18 AM

Chris, The main type of CRT / terminal Centurion used was the ADDS family. Only issues that came up using other vendors terminals like DEC VT100 or Hazeltine are the cursor position control codes each terminal vendor used ( no standard ). The Centurion application software packages used a "standard ADDS cursor control codes" then a OPSYS terminal I/O driver mapped the ( non - ADDS ) cursor control codes needed based in the terminal type / vendor configured for that MUX port number when the OPSYS was booted per a terminal config statement the operate could run. When the system supported 1 to 10 terminals all being ADDS the system performance hit was not an issue. But when we start having a large number of non ADDS terminals or a 10 - 16 plus terminal system the terminal cursor control codes mapping per terminal became a big performance issue. Centurion then designed a 16 port MUX card using AMD2901s and we uploaded a terminal cursor / control I/O pre-driver into the 16 port MUX card removing the load from the CPU6 & OPSYS. Regards, Ken R. Discord Link

kromaine13 - April 15, 2022 4:36 AM

Max, Your finding the old Warrex CPA is nuts and bring back some very old memories. I was dating my future wife ( now of 45 years ) when she called the office an asked for "Ken" the receptionist passed the phone call to Ken Schroeder. After my future wife talked to Ken for more than a few seconds she figured out she was talking to the wrong "Ken". She was a little embarrassed when Ken S. said oh... you need Ken R. Regards, Ken R. Discord Link

Kromaine13 - April 15, 2022 3:07 PM

With as prone to head crashes as the Phoenix drives are I would bet the Ebay drive missing all the logic cards had a fixed disk assy. head crash and the only value left was selling off the logic cards. FYI - Once the (3) platter fixed disk assy. is installed on the Phoenix drive spindle you cannot remove the disk assy from that spindle without destroying run-out alignment and causing vibration. CDC manual states once fix disk assy. is installed & the CDC factory alignment fixture is removed thats it, the disk assy life is on that one 3,600 RPM spindle. The CDC Phoenix drive really pushed the edge of disk drive technology in the late 1970's and the customer paid the price. Regards, Ken R. Discord Link

Kromaine13 - April 15, 2022 5:06 PM

Does not seem to be Warrex - Centurion. We never built a S100 bus computer. Regards, Ken R. Discord Link

Kromaine13 - April 15, 2022 10:32 PM

phire, You have the CPU4 overall 99.9% on target. The CPU4 was made up of 4 cards CPU1, CPU2 & CPU3 having a CPU bus between the 3 cards and the MMIO bus going to the cards like DMA (4th card of the CPU4) 4KB Core or 32KB DRAM cards, Printer Interface, Hawk DSK/Auto & DSKII and the 4-Port MUX. The first revs of the CPU4 did have the CPU registers in the first 0xFF bytes for the 4KB Core memory or the 32KB DRAM memory. The later revs of the CPU4 we moved the registers from normal memory cards into the CPU4's own high speed 256 byte static RAM on the CPU3 cards, still addressed at 0x000-0x00FF. Thank you for all your hard work. Regards, Ken R. Discord Link

Kromaine13 - April 15, 2022 10:39 PM

Max, I have tried to reach Bobby W. & Mel J. the main people at the Centurion Dealership office in San Ant. TX main times over the past few years. Both Bobby & Mel would be in there 80's now. Have had no reply to phone calls or email. The San Ant. dealers was one of the largest and first Warrex dealer. It seems like only the young kids @ Warrex back in the 1970s are still around. Regards, Ken R. Discord Link

Kromaine13 - April 15, 2022 10:42 PM

phire, The CPU4 system CPU & I/O cards are all TTL with a few ECL chips. Only non TTL was the 4K 22 pin DRAM chips. Regards, Ken R. Discord Link

Kromaine13 - April 15, 2022 10:49 PM

Max, John Rutland now listed as a VP in your 2021 Texas Tax pdf was a Sr. programmer at the San Ant. TX dealership. Regards, Ken R. Discord Link

Kromaine13 - April 18, 2022 6:41 PM

CDC 9427H (Hawk) had cylinder 0 about a 3/4 - 1 inch from the outer edge. The area with all the brown oxide scraped off is from the disk R/W heads sitting over cylinder zero. The scraped area is the width of the heads that should be flying over the disk. Centurion wrote the IPL disk bootstrap and bad sector map table all in cylinder 0 head 0 (upper head) the disk directory started in cylinder 0 head 1 (lower head). The disk servo had 405 cylinders x two heads = 810 tracks of 16 "hard: sectors with each sector had 400 bytes of data. ( 400 sector / 400 record length was a carrier over from old Sykes tape max record length and kept when we switched to a hard drive so not to have to rewrite the app.software. We really paid the price of the 400 byte records with the started using the CDC 9448 Phoenix which was formatted with 512 byte soft-sectors. We wrote 400 bytes of data and 112 bytes of pad to fill the soft-sector. We still kept the old app. software running without having to rewrite the software. Regards, Ken R. Discord Link

Kromaine13 - April 18, 2022 6:52 PM

Max & Team, This is the same CPU5 or CPU6 system in a different cabinet. The same card cage & circuit cards are under the Hawk disk drive area. The disk options are (3) 8 inch floppy or (1) 8 inch floppy & (1) 8 inch CDC Finch HD (20MB, 24MB or 32MB) or a single 10MB CDC Hawk as in the picture. How are you finding all the pictures? You just need to find a Warrex CPU4 system, Centurion MicroPlus (desk top CPU5 or CPU6 with 4 slot backplane) , Centurion XT class PC and the Centurion 3270 (IBM 3270 terminal clone). Regards, Ken R. Discord Link

Kromaine13 - April 18, 2022 7:01 PM

The 4-Port MUX control register is likely figured out by looking up the data sheet on the 4 UART chips used on the MUX. The baud rate, start-stop and parity could be set by the value written to the "per port control reg." by the port command in the OPSYS. I did from time to time cut the etch(s) & jumper the baud rate value into the UART I needed to override the OPSYS settings. I was one of the bad hardware guys, if the programmer would not help I used low level etch & jumpers. Sorry I do not recall the fine details after 40 years 😀 Guess looking at the PROM Bd. MUX tests may help with the status details ( default was 9600 8 1 N ) Regards, Ken R. Discord Link

Kromaine13 - April 18, 2022 7:23 PM

Agree my error, guest I was jumping the bits setting up the counter feeding the UART. If Mr Usagi could setup the ADDS terminal at a lower baud rate then you could write test values to the control reg. and figure out the needed values for a given terminal baud rate setting. Just an idea. Regards, Ken R, Discord Link

Kromaine13 - April 18, 2022 7:38 PM

0xF200 Write - data out Read - data from terminal 0xF201 Write - baud rate, stop-start bit, parity bit E/O. Read - Error staus like overrun, parity, framing.... Could have the data & status 0XF20x backwords, For other ports on the MUX just add 0xF20 +2. Regards, Ken R. Discord Link

Kromaine13 - April 19, 2022 5:30 PM

That office desk model is a CPU5 with (3) 8 inch CDC floppy drives, ADDS 520 terminal and Texas Inst. TI-810 serial printer that connected to the 4-Port MUX just like the terminal. [ Notice the terminal fits into a cutout in the desktop. CPA's like the office look. ] The all floppy drive system did not sell well to limited by single sided floppy (think it was just 80 tracks). A later model of the system had a single 8 inch floppy and one or two 8 inch CDC Finch hard drives which sold well. Customers did not back up the 20MB, 24MB or 32MB hard drives offen to the low cap floppy because it could take hours. Many customers lost end-user data files because they did not do backups as needed daily. Also one floppy was the OPSYS, next drive was the App. Software & 3rd drive was the end-users data files. Regards, Ken R. Discord Link

Kromaine13 - April 19, 2022 5:38 PM

The Centurion customers normal power OFF the system each night then rebooted the system the next work day. The time to cold boot a system to the full system ready to run app software was 2-4 minutes. MTBF was 8 to 12 months with MTTR being 10-30 minutes once on site. Regards, Ken R. Discord Link

Kromaine13 - April 19, 2022 5:46 PM

No the Hawk, Falcon & Finch would run for years with only replacing the HEPA Air filter each year on standard PM. The floppy drives just needed head cleaning from time to time 6 to 12 months. The 96MD CDC Phoenix was the problem if the room was clean and the customer did not smoke in the room the 9448 could & did last for years...! If the room was dusty or people smoked near the 9448 or used cheap disk packs head crashes became a big problem and costs $10K to $20K to repair. Regards, Ken R, Discord Link

Kromaine13 - April 19, 2022 5:50 PM

Biggest system failure was the RS232 chips on each end for the terminal cable ( MUX & ADDS terminal ). A lighting strike in a few miles of a 100ft - 200ft terminal cable would blow the SN75148 / SN75189 ( or 1488 / 1489 ) chips. This is why we started switching over from RS232 to 20ma Current Loop for serial interface, lighting did not effect as much. Regards, Ken R. Discord Link

Kromaine13 - April 19, 2022 9:17 PM

The 8" CDC Finch did have a translucent cover. The Finch was used in systems like the Centurion MicroPlus ( two small oval desktop cabinets top was CPU5 or 6 and lower was Finch, Floppy and power supply ). A few tall chassis Centurion used the Finch also. Regards, Ken R. Discord Link

Kromaine13 - April 19, 2022 9:27 PM

Max, One of my jobs at Centurion was manager / director of customer service. We repaired the board Centurion made and a large number of the boards in the OEM printers & disk drivers. The board repair turnaround time was less than 5 days in my depot ( average was 3 days ). My group also was in charge of final systems test on the mfg. floor, we call it QC Audit. Once the system shipped to the customer my dept had to cover the warranty repair costs so we also got the final say on of the system was 100% and meet the customer order, each system was a custom order as to number of terminal, DRAM ans printer. Regards, Ken R. Discord Link

Kromaine13 - April 19, 2022 9:33 PM

The blank PCBs we designed but the PCB was etched by a vendor. All the IC installed inhouse. wave solder & all mfg tech was inhouse. Cini. Data System was founded by Howel & Chuck Dewitt, I installed the first system in C.D.S. office ( a CPU4 ) . One of the Warrex - Centurion adds that you posted with two men looking a book is Howel & Chuck. Regards, Ken R. Discord Link

Kromaine13 - April 19, 2022 9:39 PM

Howel, Chuck and a few of Warrex exc's were at the MGM Grand Hotel in Vegas in the late 1970's when it burned. Really bothered Howel. After the hotel fire we never let Warrex exc's fly on the same plane of stay in the same hotel during bus. trips. Regards, Ken R, Discord Link

Kromaine13 - April 19, 2022 9:44 PM

The line was (John Warren >) Warrex > Centurion > EDS/Centurion > Z-TRON > Z-TRON bankrupt > bankrupt court to Cruze Computer, then Cruze stated buying up the remaining dealers that had sold Centurion over the years. Buddy Cruze died around few years ago but I think Cruze closed down in the early 1990's. Regards, Ken R. Discord Link

Kromaine13 - April 19, 2022 9:47 PM

Cruze purchased the Centurion system to help run his furniture business, then he like the system and became a dealer of Centurion systems. Regards, Ken R. Discord Link

Kromaine13 - April 19, 2022 9:55 PM

FYI - I reached the last design engineer to work at Centurion > Z-TRON > Cruze this afternoon, David W. told me he designed a CPU8 (AMD2901 - Mulitbus 8 bit) CPU9 (AMD2901 - Mulitbus 16 bit with math-cop) and even a CPU10 before the closed down operations. The CPU10 was a port of the Centurion OPSYS and app software to a 386 PC that supported four ASCII terminals that ran the Centurion business programs over an Ethernet LAN...! ( no Internet ). Regards, Ken R. Discord Link

Kromaine13 - April 19, 2022 9:59 PM

All of the keys on the terminal send the Centurion software a single ASCII character (full duplex) then the system sends the terminal 99% of the time the same character back, But in the case of the 10Key Pad on the terminal the (+), (-) & (Enter) offen sent a number of characters back to the terminal. Regards, Ken R. Discord Link

Kromaine13 - April 19, 2022 10:02 PM

Centurion in San Ant, TX was the first Centurion dealer and not the mfg. All the Warrex / Centurion design & MFG was done in Richardson, TX (Richardson, Dallas, Allen) till the court sold the remains to Cruze Computer in K. TN. Regards, Ken R. Discord Link

Kromaine13 - April 19, 2022 10:04 PM

Sarasota, FL is likely Scott Springer that had a dealership in Tampa / Lakekand, FL. Regards, Ken R. Discord Link

Kromaine13 - April 19, 2022 10:20 PM

Team on the way the Centurion system sees the terminal key strokes the keys on the 10 key pad like (+), (-) and (right arrow) are not the same ASCII code as the like keys on the larger alpha side of the keyboard. Guess this does not really matter till we are running Centurion OPSYS / Apps. Regards, Ken R. Discord Link

Kromaine13 - April 20, 2022 5:27 PM

Bitman, To be correct we did have a single user CPU4 with two 4K Core Memory cards. Once we had the 32K DRAM card (in the picture with a tech & scope probe) we had the CPU4 running the OPSYS with JCL using CPL, 4 terminals and a print spooler so the 4 users could share the printer. ( all in 1975 ). Regards, Ken R. Discord Link

Kromaine13 - April 20, 2022 5:42 PM

Max & Usagi, We had a Centurion dealer in Smithtown, NY and one in Cherry Hill, NJ. Do not recall one in up state NY. My best guess on total Warrex - Centurion systems sold is around 2,000 over 10 years plus a few thousand XT PCs and a few hundred Centurion 3270 (IBM clone 3270 mainfame terminals. Regards, Ken R. Discord Link

Kromaine13 - April 20, 2022 5:50 PM

Max, No steak just a lot of hamburgers as we hand solder the circuit boards for the first few years of production. My pay as Director of Customer Service was about $30K in the early 1980s not bad when a new car was $4K and a starter house was $35K on Dallas, TX. Regards, Ken R. Discord Link

Kromaine13 - April 20, 2022 6:00 PM

Correct the IBM XT & AT PC really did hurt. The killer was Lotus-123 & WordStar. A CPA could by a PC with simple accounting software for $3K-$5K then by the middle to later 1985-1987 buy a Novell server and build a local area network to share the app software. When I left Centurion I we over to Proteon ( the company that invented 10Mb & 80Mb Token Ring and the p4200 Router to tie Token Ring to Ethernet and serial WAN lines at 56 Kb & T1 rates of 1.54Mb. ). Proteon did this before IBM "invented 4Mb Token Ring"and Cisco nearly copied the p4200 and called it the Cisco AGS Router. Good engineering does not equal good marketing. Regards, Ken R. Discord Link

Kromaine13 - April 20, 2022 6:09 PM

sjsoftware, Correct once the Centurion Dealer sold & had service contracts on the end users system in the dealerships area a dealer of 4 to 6 employees could live on the service income. This was a big problem for the Centurion (the mfg.) we only made money when a dealership purchased a new systems. The dealership only needed to added 2 or 3 new customers per year to have a strong business. Now add the PC to the computer mix and Centurion was in trouble. The same thing happen to all the minicomputer companies over the next 10+ years. Regards, Ken R. Discord Link

Kromaine13 - April 20, 2022 6:19 PM

draco & Team, I alway reminber the serial settings to the ADDS terminals being 9600, 8, N, 1 and to the 300 baud modem on port-3 of the MUX being 300, 8, N, 1. Never recall the OPSYS checking parity it just checked the MUX board UART status bits for parity, timing, framing & overrun errors. Regards, Ken R. Discord Link

Kromaine13 - April 20, 2022 6:29 PM

Draco, Once we had the teleco offer 1,200 baud services we did have a customers that connected the remote ADDS terminal to the teleco 1200 modem the connected a TI-810 serial printer to the AUX port on the ADDS terminal. Now the remote Centurion user had local printer support. Back in the 1970s and part of the 1980s the phone company (telco) owned the modems. Regards, Ken R. Discord Link

Kromaine13 - April 20, 2022 6:44 PM

The exec's at Centurion did not agree the board repaired & spares should be a profit center (I disagreed). The only parts the dealer had to buy from Centurion was the boards we built. All the disk drive, terminals and printers the dealer could buy parts from the OEM if they could wait 30-60 days for the part. They did buy some of the parts for OEM devices from Centurion because my team would ship overnight. The kept complete drives, terminals & printers in the service spares inventory to pull parts from when possible, it was cheaper de-mfg a OEM device than stock line items from the OEM. Regards, Ken R. Discord Link

Kromaine13 - April 22, 2022 6:13 PM

The first 0x000-0x00FF should be many times faster than a memory R / W cycle from the systems DRAM like the 128KB memory card. The CPU6 spent a lot of money on the High Speed SRAM chips used on the CPU6 board. Back in the early CPU4 the CPU4 registers are the same as the system DRAM on the registers are in first 32KB DRAM card ( did not have the 128KB memory card and the PC was only 16 bits with memory mapping address logic ). Sorry if I am off base on my reply, you guys are above me on this level de-bugging opcodes. Regards, Ken R. Discord Link

Kromaine13 - April 23, 2022 4:18 PM

Chris, The Hawk disk controller design (DSK/Auto & DSKII) was 6 years before the AMD2901 was available. The (DSK/Auto & DSKII) is a simple MMIO with a TTL state machine which runs in sync with the Hawks Read / Write data stream to advance the state machine. The CMD controller and Floppy Controller are AMD2901 based controllers that move a command string into the controller under DMA control then the AMD2901 runs the command string then moves the Read / Write data in or out of the controller under DMA control. Regards, Ken R. Discord Link

Kromaine13 - April 23, 2022 4:31 PM

A basic sector read from the Hawk (DSK/Auto & DSKII) is.... Drive select Reg. 0xF140, Sector address Reg. 0xF141-F142 Status Reg ( I think ) 0xF143 and Command Reg 0xF148 ( 00 = read, 01 = write, 02 = seek, 03 = RTZ Return Track Zero. The (DSK/Auto & DSKII) Hawk sectors are 400 bytes ( 0x190) long and R / W can be 1 to 16 sector operations based on the total bytes the DMA was setup to move in /out of DRAM. Regards, Ken R. Discord Link

Kromaine13 - April 23, 2022 4:54 PM

Think one the older posted MMIO drawings showing some of the Diag PROM Bd. address had the Hawk (DSK/Auto & DSKII) addresses also on it from a few weeks ago. To much to keep track of. If you see status bit at 0xF144 I would go with that, its been over 40 years for me, the last time I did any machine code to toggle the controller bits & position the Hawks heads to do a disk alignment. Regards, Ken R. Discord Link

Kromaine13 - April 23, 2022 5:04 PM

Cylinder 0 -405 MSB Cyl 256, Cyl 128, Cyl 64, Cyl 32, Cyl 16, Cyl 8, Cyl 4, Cyl 2, Cyl 1, Head 0/1, Sector 8, Sector 4, Sector 2, Sector 1. LSB Max Cyl - 405, Max Heads 0 /1 , Max Sectors 0-F . So Max Tracks are 810 = 405 Cyl x two Heads with 16 sectors of 400 bytes per track. Regards Ken R, Discord Link

Kromaine13 - April 23, 2022 6:45 PM

Yes, The Printer Interface for the Data-100 printer with "this" CPU6 system is Centronics. We did offer non-Centronic printer options like CDC 9322, 9316, 9317 and DataProducts by changing the jumpers on the printer interface card, printer cable and changing the printer driver in the OPSYS. Sorry to say the Diag PROM Bd. does not have printer driver code to review, just have to trace out the printer MMIO logic on the I/O card. Regards, Ken R. Discord Link

Kromaine13 - April 23, 2022 6:49 PM

Yes the same printer MMIO card for all the parallel interface printer options. Just changed the jumpers on the printer card, matched up the cable to fit the printer connector type and set the printer type to match in the OPSYS to select the correct printer software driver. Regards, Ken R. Discord Link

Kromaine13 - April 24, 2022 10:37 AM

Chris & Team, If you are thinking about any custom ROM - PROM or some type of SerialDisk storage why not program PROM to sit in an empty socket on the Diag. PROM Bd. and use some of the TOS subroutines to handle MUX I/O and other bits of code as needed. Just an idea could a simple serial interface be designed & added on to the MUX DB9 that could stream data into an out of a USB thumb drive with a customer USB to MUX driver? The MUX's UART can run at a max rate of 19.200 baud. Regards, Ken R. Discord Link

Kromaine13 - April 24, 2022 10:49 AM

The old CPU4 had a DMA card in the MMIO space. When the CPU5 & CPU6 was designed the DMA card functions ( no more DMA card) was moved into the CPU5 & CPU6 with the addition of new DMA opcodes for DMA mode, DMA memory address and DMA byte count to be moved. The CMD controller and the Floppy-Finch(HD) controller received its command string for the controllers own AMD2901 and sent status out of the controller with CPU5/6 DMA. The controllers also moved the Disk R/W data to main system DRAM using the CPU5/6's DMA functions. Regards. Ken R. Discord Link

Kromaine13 - April 24, 2022 11:08 AM

...0x0E ought to be a long (but not infinite) loop. The delay opcode 0x0E should 4.5ms long. I do not know how the CPU5 & CPU6 handles the 0E but in the CPU4 it stopped the system clock for the full 4.5ms, it even stopped the DMA channel for the 4.5ms which caused problems if you happened to be in the middle of a disk R/W operation because the disk drive did not stop spinning so disk data got screwed up. Regards, Ken R. Discord Link

Kromaine13 - April 24, 2022 11:20 AM

You are welcome to use all of my posts (past & future) as you wish. Regards, Ken R. Discord Link