** The ICT 1301 Resurrection project **


The 2003/2004 Repair Diaries

The visit diaries are now entered here, in reverse chronological order, so the latest news is on top. ( more updates as and when we visit. Do check back for pictures, updates and news on this project )


" Link to the 2005 diaries "


Last entries in 2004 are below


32nd Episode:-
Last 2004 session:-17/11/04
General tidy up day, lots of boards wrapped in then, a quick try of the machine, we now have started to get drum errors on reading decade transfers, and store parity errors on decade writes ! Is it the cold ? or do we have a new fault ?
Worked around the drum and could see on a single decade transfer the bits being written to the drum, but cannot see stabiltiy of the data during a decade read. In fact it can be seen dithing, relative to the read clocks, which suggests digit counters. Working along the line of digit times we were getting a lot of spurious pulses, unfortunatly this is typical of 1301 logic. As we were running out of day it was decided that we would close work this year as the low temperatue will stop us doing further fault finding, and could even be giving us more problems.
We now have what we belive is a corrected copy of the initial orders, we have recoverd from the Drum, but we will not rewrite these until we are sure we are writing good parity with the data.
So it was covers on and Flossie was put to bed for the Winter. We do seem to have come a long way since the start in 2003, however we still have a long way to go !
Check Back Soon !

31st Episode:-
More work on the drum and in an attempt to rectify the Channel Write's which are both giving us store parity on write and drum parity on read, however we can write individual decades with decade transfers and read a whole channel of individual decades back ok !
Just to keep us on our toes we had a failure in the control registers, with 4 bits bieng streamed from control register 1 to control regiter 3, this was a package in 5D26.
One of our biggest problem today was the fact that the programs loops were getting corrupted with the store parity errors. Trevor Chapman was with us today, and suggested that we move the code for the program loop up the store beyond the first two working store pages. Well done Trevor we no longer have a corrupted program, which means we can now fault find in a dynamic situation, which is just what you have to with this machine!

So the state of play is that we can write and read decades, Read a whole channel (20 decades), but cannot write a Channel due to store corruption when we try.
30th Episode:-
Today started with a failure in the first page or ( barndoor ) Bit 1 in Digit 2 failing, replaced the A7 package and then down to fault finding the drum some more.
Continued work on the Drum clocks to try to get a better copy of the initial orders code off of the drum, working on the theory that we could set the clocks to read the data as it was written.
The overview is now that we can read and write to both Normal and Reserved Bands, with Decade transfers, ( block of 10 words ). We can also read 20 decades with a Channel transfer, but cannot Write a channel to the Drum
29th Episode:-

Today is designer day ! Follow this link for news of the day

Desinger DAY News and Pictures



28th Episode:-
Rogers Worked alone this time on a Sunday and isolated the cause of the Channel transfer problem, I am advised that the Channel read now works !
And yes it does !

27th Episode:-
A lot of today was spent in capturing the content of the reserved bands, this has been done by a small piece of code in the 1301 which transmits, an RS232 format signal to a PC connected via the output of a programmed indicator. A simple TTL to RS232 level conversion is applied and the PC functions in "Capture Mode"
Work was halted whist we repaired digit 9 bit 1 in store page 1, a new A7 package solved that problem
During emails which followed the day, Roger confirmed that with a bit of tinkering, we have a ( possible? ) working copy of I/O's ( At last ! )
26th Episode:-
more Drum mayhem, but we do seem to be winning, we now understand that if we choose the orders in the control registers with care we can even set up a small loop of code in the store and not get it wiped each time we try to fault find.

Using this we have isolated our problem to closeure of the Drum order and signal R6, the drum order runs into the start up of the main machine, at that point the drum is driving the control registers and the machine is trying to obey the next instructions. Nasty things result !

If we keep the following instructions limited to STOP orders we seem to have some control over the situation. Attempts to fault find Micro order by Micro order all fail due to the loss of synchronisation to the drum. Ultimately our investigation of the day took us to bay 11 and fault finding around this area was proving, eerrrrr! Touchy ! If we touched anything in the area different things happened! most of the logic in level B seemed to have corrosion around the logic boards and the connections.

So we took a dynamic decision when Roger produced from our pile of spares a complete replacement rack. During this removal and replacement the remains of a birds nest was removed from the rest of the logic, I do wonder if we will find more oddities like this as we progress through the project. The new rack gave us a more stable situation but essentialy unchanged problems, we know we are in the right area but we can only remove one problem at a time and keep going.

The upshot of today is that I have suggested that Roger makes soup this evening, YES, BIRDS NEST SOUP, of course !
25th Episode:-
For this session it was back to the drum problems which continue to dog us and we have isolated some of the issues in resolving the current problems. The machine will do different things depending on what is in the third control register when a drum order is issued. we can now predict what will happen ( mostly ) and have isolated the problem to what we suspect is " Running on ", the drum order completes, but the control is not passed back to the main machine and the control registers keep spinning and adding up the drum order until we FALL off the end of the working drum. The main signal which seems to be set for too long is R6 this should only be up for ten times one word. ie: 52 micro seconds times ten = 520 plus a few microseconds for the setup and close down of the drum order, but it is set for at least twice as long and maybe more.

We do however have an advantage, we have the storeage oscilloscope back after its repair. We have sorely missed this valuable tool, and now although we have to check signls running from one side of the machine to the other we can see even single pulses and analyse them. So isolation of signals in the function mesh, to try to trap the END micro order duration, started. Soon bay 3 looked busy with isolators in 3E6 pin 7, 3E6 pin 6, 3E12 pin 14, 3E14 pin 12, 3E16 pin 9 and 3E23 pin 4. This was to identify when the drum order which uses the T5 signal changes to T3, to signify the end of the drum order.

We closed activity today feeling somewhat defeated, but closer to the definition of just what the problem was to start with!

However a high point today was the finalisation of a simple piece of software which uses a programmed indicator to send RS232 signals out of the machine.

WHY ? to unload flossies software

HOW ? to send Store, and hopefully Drum and Tape data, to a laptop and archive it !

WHEN ? I do wish you had not asked me that ! ( for the answer is:- when we get the Drum Working )


24th Episode:-24/8/04
At start of this session Roger has cleaned up the logic gating around the gereration of " Q times " these are the main 1 mhz clock divided down to 250khz and split into 4 quarter or "Q" pulses. our slip pulse problem remains however, and was finally isolated to 11F9 , the defective Binary was replaced and now we have both an alarm and syncronisation to the drum.

The big try " Our first Drum order " the Control registers were set up as 830100-010700 and with fingers crossed start was pressed. SOMETHING HAPPENED, but what ? We did not get the result expected and no data came from the drum, but the control registers showed, that we had tried to count the tranfer through the registers. A deeper analysis showed that the machine had stared the tranfer but was unable to find the address on the drum. The drum is devided into 20 " Decades " around the circumference and this is mapped out in the digit, word and Decade counters. The coincidence with the required address was not bieng detected correctly and fault finding found 11D7, pin 13 , also 11A22 not working. Now we have a drum transfer that completes, still no valid data transferd to store, but we are getting something from the drum which is overwriting the old store contents. A further problem seems to be that we are transfering only half of the decades we request, this was tracked down to the decade counter bieng counted twice each transfer so 11A6 pin 14 fixed this.

Now somethimes we feel we have won a major battle with this project and this was one of those moments, as Roger proudly reminded me we were now further forward than we had ever been since the project started. On the other hand sometimes flossie pulls problems and tricks out of the hat which we do just not see coming ! This was also one of those days.

On setting up a drum transfer of 810000-010000 ( transfer just 10 words into store from the start of the drum ) and pressing start we were greeted with a machine that went " Very Busy !" the final end to which was that the machine stopped with 816101-010601 in the Control registers and we had transfered all of half of a drum into the store about 3 times over, and all of the store had been wiped of our software and replaced with random data from the drum.

So flossie has gone from Not-Starting to Not-Stopping in one step.

Today closed with us feeling we had just won, and then lost any advantage we had gained from our work. As we now could not put any tests into store to test the drum, as flossie would wipe any software as soon as we pressed start.
23rd Episode:-18/8/2004
The work this session started with finalising the drum setup, all of the work so far has been a prelude to this point, for without the drum we have no where to store any software. Even small test loops have to be input by hand, and are lost at power down, so drum store is vital if we are to unload the historic software from this machine. The work to complete the tape control unit and fault find the data paths will also require some simple test software so all our effort focus now on getting the drum working !

the first set up was to check the gate which is only used for the drum reserved bands writing unit, but it has a counter output with the last bit of all of the counters. So ( Bit 2860 ) on 11D4 pin 22 was used to check the alignment of the clocks and the datum on both sides of the drum logic bays 11 and 12. Some considerable time spent on these to ensure as near perfect alignment, which has paid off since, due the the fact that we could be sure that future faults were not in this area.

A condition where the machine cannot stay in step with the drum still dogs this work however, the condition and error is called ( Slip Pulse ) and although we have fault found the whole chain of clock gennerating logic, the signal is still there.

For some light relief to end todays work we foucused on another problem where the multiply order has failed. This proved to be a defective gate in 6D4, pin 7 giving only a half level out, neither a logic one or a logic nought. Multiply now works, and the simple test of running " Ghost " proves all is ok.


" Link to the Earlier 2004 diaries "


" Link to the 2003 diaries "


"(Shedland Home)"

©"Copyright Shedland Software design"