Oh Blimey I bought a CNC Lathe !!!!

[philf]

Some picture to hold your interest 

You've got my interest! 

I thought setting up Mach3 to work with with my mill was hard enough.

What are you planning to make with it when it's finally fixed?

 

Phil.

[loply]

awemason, maybe you have already addressed this in an earlier post, but I'm wondering what the advantage in retaining all of the original control circuitry is, as opposed to replacing it with a modern solution (whether LinuxCNC or some other 'drop in' replacement)?

It seems like a lot of effort and expense and I can't help but feel... what if it breaks again?

[awemawson]

PhilF : Make - good heavens man I'm not going to MAKE anything - it's the journey I enjoy  Well perhaps the odd run of a few hundred thousand widgets. 

Loply: partly answered your point in my comment to PhilF - but actually the TX 8F is a very capable controller, and on this machine the spindle amplifiers and axis servos are not interfaced in the conventional way, but are driven by a scusii like 50 pin interface proprietary to Mitsubishi - not saying it's not possible but not just 'step and direction' or '+10-0--10' that can be bought as a comodity.

Spent a bit of time this evening gluing the Optical Tool Setter box back together - looks ok in the picture but in fact obviously has been dropped - diametrically opposite corners broken.

Still not found an MC465a - ordered a few ics today and will start replacing the buffer logic that it talks to the bus with. A couple of standard through hole DIL SN74LS645-1N octal bi-directional bus drivers which I know I can cope with, but eight SN74ls244 SOIC chips as well. Never worked with tiny surface mounted components before so this will be breaking new ground for me.

This is the card I'm working on:

[awemawson]

Fortunately I think I've proved the two custom Programmable Logic Arrays to be ok as I've swapped them between the two faulty cards that exhibit distinctly differing fault symptoms, and the faults remain the same. Not absolute proof that they are ok but they probably are.

[awemawson]

Quandry Quandry:

Chips arrived this morning, and I started on the 'easy' through hole SN74LS645-1N - as there are a pair of them I am assuming that they are the 16 bit wide data bus buffers. First one wasn't too bad, but the power pins were a pain as the board has a ground plane and a power plane so it took a lot of heat to suck out the solder. Changed the first, put in a 20 pin socket, whacked in a replacement chip, tried it - unchanged symptoms  - anyway onwards and upwards, went ahead and removed and socketed the second SN74LS645-1N. Tried it - symptoms have changed to EXACTLY the same as my other board (which was already packed up to be returned as faulty!). I can only presume as both boards now give the same symptoms the fault lies elsewhere. Looking on the bright side it does look as though I have probably fixed one faulty board to chip level. 

Meanwhile they've cut our water off due to burst pipes on the main road and are bringing a bowser round for our livestock, and bottles for us, so probably little time for progress this afternoon due to water carrying.

[MrSleepy]

If you do not have a hot air smd rework station then you should get one.. or use a standard hot air gun carefully..

An increasingly common feature of soic packages is a pad actually under the chip...but I'm sure yours will not.

On dips I have used a technique of cutting the pins off with a fine dremel cutting disk then unsoldering the remains individually.. but it takes practise a great care.


HTH Rob

[Swarfing]

If you can get a thin piece of wire under the legs then just use a soldering iron, heat a leg at a time pull the two ends of the as you go lifting the legs.

[MrSleepy]

The wire underneath trick works okay on older packages , but sometimes you have to go so thin on the wire that it snaps frequently making it worse...and if theres a conformal coating then its a nogo.

Circuit Specialists do a cheap hot air desoldering gun ... £57

http://www.circuitspecialists.eu/soldering/desoldering-hot-air-rework/hot-air-gun-w-digital-display-for-smds-csi-hotgun2

Handy for occasional use,

Rob

[Swarfing]

Wire trick works fine for me. The cheap hot air guns you get from Aldi have a selection of nozzles, one of those is a small round affair which works well too.

[awemawson]

I took delivery today of a kit containing a magic low melting point high fluidity solder. The idea is you swamp the legs of the SOIC with the new solder, that dissolves the original and stays molten long enough to lift the device away from the board. Not had a chance to unpack it yet due to other issues but the video looks good:

http://uk.farnell.com/jsp/search/productdetail.jsp?SKU=1850214

All sorts of confusing issues with the controller - seems I had the wrong PLC eprom inserted which accounts for some of the problems but I don't seem to be able to clear the Z58BRAM error. I am seriously considering reverting back to the 10 slot box as that uses a different BRAM card that cured the same errors in that box. I'm told the card addressing along the backboard is different but I would have thought that the 6 slot would be a subset of the 10 slot, and I will not be using the extra four slots! I will pull them both out on the bench and buss out some pins, but it's a bit of a daunting task as each slot has two 96 way sockets so there's a bit of effort involved ! 6 x 96 x 2 = 1152 points to be correlated on two backboards 

[MrSleepy]

I took delivery today of a kit containing a magic low melting point high fluidity solder.

They do have traces of lead in them though..which can be a problem.

[awemawson]

Just received an email from Traubs agents in the UK quoting me for a replacement MC465a which is the card I 'think' I fixed today.


£2461.50 plus VAT @20% plus carriage No I don't think I'll be buying one of those - anyway hopefully I don't need one now.

[awemawson]

I took delivery today of a kit containing a magic low melting point high fluidity solder.

They do have traces of lead in them though..which can be a problem.

I've been soldering components with 60/40 tin/lead solder for the past 45 years and I expect that this series of boards being made in the early 1990's was still tin lead so I don't suppose it'll be much of a technical problem, and certainly no worry to me whatsoever.

It's interesting to note that high reliability critical kit is still being made using tin/lead solder due to the fatigue effect of lead free which over time produces dry joints.

[MrSleepy]

£2461.50 plus VAT @20% plus carriage No I don't think I'll be buying one of those - anyway hopefully I don't need one now.

Wow thats ... well it reminds me of parts prices for my German Scheaff HR16 minidigger.

[vtsteam]

I favor the cutting pins method on standard DIP packages so I don't over heat or over pry the package being removed, rather than trying to solder suck with the piece in place. In the old days we used to slip a nail file under -- similar to the wire method.

Overheating a plated through hole in a multilayer board can be big problem. Individual leads are much easier to desolder in one go than an attached package worth of leads, which often require re-heats of sticky legs. It's the re-heats with pressure that detach pads or mess up a plated through connection.

I use fingernail clippers to snip leads instead of a Dremel.  No tendency to ricochet with a hand tool as there is with a power tool.

awemawson, are you replacing the DIP pkgs with sockets or are you soldering the new chips in place? I'd probably go with sockets to make sure your new chips don't include a dud. You can then substitute around to check.

Caps tend to have limited lives more than most other components in an old board. I don't see too many on your board, though.

I'm sure I'm just repeating things you already know, but throwing stuff out there just in case some little bit is helpful.

I got my fingers crossed for you.   

Sorry to hear about the water situation, I know how that is. Makes everything difficult. 

Soldier on! 

[awemawson]

Vsteam thanks - water situation resolved late last night. Makes you appreciate it when it comes back!

Yes I always cut pins on DIL packages to remove, then take out individual pins. Although I do have a proper Weller de-soldering station with a vacuum pump my method of choice is to flow the solder - push the pin as far as it will go towards the component side with the tip of the iron, then give it a quick blast from a fine compressed air nozzle. Blows pin and solder clean away leaving a nice undamaged hole but I work on a pad of workshop tissue to contain the bits that come though. In this case the power pins on their massive power planes resisted solderwick, the vacuum sucker and the air blast - I just couldn't get the thing molten all the way through until I went from a #7 Weller bit to a #9 (which is higher temperature. I suppose this is one case where that low melting point stuff in the video above would have come in handy, but it hadn't arrived by then.

Oh and yes I always socket replacement chips !

[dsquire]

awemawson

I have been following along since the start of your project on the CNC lathe purchase and process of bringing it back to life. Slow and steady, one step at a time. Think each step through first then on to the next. When you solve the last piece of the puzzle it will be so worth it that we will probably all hear a big
 

  "YES".     

Good luck with it and I will continue to keep watching. 

Cheers 

Don

3158

[awemawson]

Don thanks for the encouragement.

I've reverted back to the 10 slot controller as I cannot overcome the Z58 BRAM failure that I get in the 6 slot. It takes MC465a BRAM cards, two of which give the same failure, whereas the ten slot takes an MC462 BRAM card which has gone in and shows no such issue. I have yet to prove that the input / output addresses appear in the same places and that is tomorrows job with a 'scope and big piece of paper. However I cannot see that the 6 slot is anything other than a subset of the ten slot but I may be proved wrong!

[vtsteam]

At the end of the day, I bet it's just one *%$#! line different!

Fingers crossed! 

[awemawson]

OK I didn't fancy buzzing out all those backboard lines to prove if the slot addressing in the 6 slot was different from the 10 slot as the experts had told me, so I went ahead and tried it anyway. There are two input output cards and a channel card to drive the additional axis:

MC301 has 64 Digit Inputs and 45 Digital Outputs
MC323 has Analogue i/o (not used in this config, and 64 Digital Input and 64 Digital Outputs
MC652 drives an additional daisy chain interface to the V axis (opposing spindle)

Now I can monitor inputs on built in displays on the controller, but I've not yet found how to drive outputs, but my reasoning goes that for these inputs and outputs to feature at the correct addresses so long as inputs on a card are proved to be in the right place, the outputs should also be ok address wise. Now there is a logistical issue testing this - the controller is on the front, and all the wiring is round the back! So I made up a long test lead consisting simply a single pole switch on one end of a wire pair, and a pair of 'bootlace ferrules' on the other end. A '1' is +24V on this system, so one wire was fixed to +24v, and I sequentially removed a selection of inputs from the termination board substituting my other wire. This way I could watch the interface monitor screen, and toggle the switch and watch bits being turned off or on. All inputs showed up in the right places so I concluded that the MC301 & MC323 backboard addressing must be what the system expected. Now for the MC652 card: Each drive (Axis and Servo) talks along its daisy chained bus and gives its status to the system the same way. Each has a rotary pcb mounted hexadecimal switch to set its 'drive number' on the bus hex 0-E are potential drive numbers and 'F' is unequipped - ie servo amp  not used. Sure enough making adjustment to the drive numbers showed up on the appropriate screen of the controller so again I conclude its address as far as the system program is concerned work just as it did before in the 6 slot backboard. So I MIGHT be on a winner.

Fortunately all the i/o wires are screwless 'Wago' types that you open and close pushing a fine screwdriver in a  hole, so connecting and disconnecting wasn't as hard as it might have been.

Talking of winners, I grabbed some 96 way din connector extender cards on ebay which arrived today. The system has single connector (half height) boards and double connector (full height) boards. In the photos that follow we have:

A/ Single connector extender - can be used unmodified
B/ Dual connector extender - needs slicing up and the connector spacing adjusting
C/ Tripple connector extender comprising three single bolted together with spacers - I may use two of the element of this rather than slice up B/
D/ a shot showing how close to the required spacing the dual extender is
E/ a shot of the now empty and forlorn 6 slot controller

[vtsteam]

Classic hacking. 

[awemawson]

Whilst it was on the bench I set too with the power supply in the 6 slot. It worked fine except that there was a faulty centre biased toggle switch that latches it on or off, which duplicates a function on the operator control panel. It would stay in the on state, so when youb turned off at the operator control panel, it would immediately turn back on! It's an obsolete micro miniature pcb mounted 90 degree bent 0.1" pin spacing with two extra mechanical pins to secure it and an extended toggle. Closest I could find had the right 90 degree bend 0.1 pin spacing but the two mechanical mounts were in the wrong place and the toggle was short . A bit of judicious pin bending got it in the pcb (desoldering with the Weller vacuum sucker worked a treat this time). I reckoned to glue a tube extension to the toggle when re-assembled but in practice it's ok.

Hopefully I can set up a test bed with this 6 slot and the extenders - anyway some pictures:

A/ Old and new switches
B/ New in situ

[awemawson]

I spent the day trying to work out why the controller isn't acting on the 'door open and 'door closed' switches. It is quite a complicated system with lots of interlocks. The machines sliding door has a switch for when it is 'open', another for 'closed', and a third for 'locked'. Locking is by an electric shot bolt with switch feedback. Then each rotary axis has a special safety monitoring relay that tells when the spindle is rotating. These three rotary axis monitors are all connected in two series chains of normally open and normally closed contacts, which in combination with the door switches operate 'open', 'closed' and 'locked' relays. Or rather they should but they don't

No complicated controller interaction here - just 'simple' relay & switch logic but I've been chasing my tail all day. I knew that one of the 'zero speed' relays is faulty but I have temporarily linked them all out. Turns out I have lots of relays with bad contacts. Measure a normally closed contact and the chances are it's open! Even after a spray of contact cleaner which de-oxidises contacts there are problems. I suspect that it's quite a bit to do with the fact I'm testing with an Avo meter on ohms that will only bias the contacts with 1.5v whereas the normal operation is 24v. I will set up a battery & bulb tester this evening and see what happens. Machine is quite clean inside so I'm surprised contacts are that  bad.

[lordedmond]

Not trying you to suck eggs,but when I first went out on the plant at the end of my apprenticeship I was put with a electricians mate who was abut to retire his two bits of advice to a new wet behind the ears sparks was 1 always test with the voltage that the job uses ( i.e. use a test lamp with on leg on the live side. and two listen to the plant operators when we go to a break down they may just have seen something that is important , brush the off and they will not help you again.

enough of my rambling as you have found out with old equipment most relays do not have much wipe as they close and need a minimum current flow though the contacts to operate correctly, get a 24 v lamp and use that to bell out the contact chain AVO are next to useless in this case as are 1000v megga's on very old VIR cables ( they fail the megga test but are fine on the voltage they used on )

another tip that has held me in good stead was if you have the drawing always start testing from the middle to each rail thus you have halved the amount you need to test, repeat with the remaining section and the fault can be quickly found


note the above does not conform to the current H&S regs of live testing  but bear in mind I was was brung up up on open 220vdc open switch gear with only a rubber mat to stand on as well as 550vac barrel type crane controllers

as a comment you have come a long way with this project and I wish you well with this endeavour and I do hope you will have a good end to your journey


Stuart

[awemawson]

Stuart,

I totally agree about listening to the operators. After years working on process control computers from biscuit factories to oil refineries through nuclear power stations what you say rings very true.

Before reading your post I'd cobbled up a 24v T5.5 'wedge bulb' on the end of a probe to do precisely what you suggest, and also ordered a Fluke T50 tester! Not convinced the T50 is the total solution as it uses leds, but in this instance better thasn the Avo, which reigns supreme in other uses.

Another extremely useful tool I'm using is a telecoms wire tracer. Hook the oscillator on a wire and you can trace with the receiver where it actually goes rather than where the circuit says!

Andrew