[LegendCJS]

My lab has just dusted off an old Minitech Mini-Mill/2 purchased in 1993. It has been sitting without much use since it arrived, and we have lost all documentation for it. However, the Minitech website claims that Minitech's customization of the Mach3 software should work with every machine they have ever sold back to their founding in 1983. We downloaded the Minitech customization of the Mach3 software but can not get the motors to spin.

Two electrical engineering graduate students spent a couple hours with a oscilloscope and probe and chip data sheets and determined that the 4N28 opto-couplers that are part of the mill's electronics and isolate the parallel port are incapable of passing through a pulse any shorter than around 100 micro-seconds. However, the longest step pulse we can seem to get Mach3 to send is 15 micro seconds long. I believe that Minitech would have tested out their customization of Mach3 before making the claim that it will work on every machine they ever sold, so I come to you asking if there is anything obvious I have missed. The only place in software I saw that allowed a change in the step pulse length was in the Motor Tuning configure window. The label over the text input box for setting the pulse length is labeled for a 1-5 us range. But typing in any number larger than 1 to 15 micro seconds just makes it reset to 15 micro seconds (which is odd considering the label says it will only accept pulse widths between 1 and 5 micro seconds.) We have tried both active high and active low logic options as well.

If any one has experience with something similar, any suggestions would be greatly appreciated. Minitech tells me that they are unlikely to find documentation for this old mill.

Thanks everyone in advance.

[Torchhead]

4N28's are wretchedly slow. BUT... Are you sure you are feeding them the signal with the right polarity and the correct "common" voltage. Have your EE's pull up the spec sheet for the 4N28 and note the rise and fall times. It should turn on with a 5 usec pulse. Problem occurs when the pulses are to close together (pulse rate too high). You need to first determine that the optos will respond to SOME level of signal. The freq response (overall) of the 4N28 drops off after about 10KHZ.

On older machines without microstepping that was normally enough to get by.

The rise and fall times of the NPN in the opto is dependent on the current through the transistor. High values of collector resistors (lower collector current really slow down the max freq.

Slow your velocity way down and deal with pulses easier to see.

TOM Caudle
www.CandCNC.com

[LegendCJS]

Thanks for your response Torchhead.

The EEs directly measured the response time of the 4N28 optocouplers in place and found that they are in the 40us to 50us range, as currently installed in the driver board.

I got a reply from Jack at Minitech (great guy) who hooked me up with their current tech support guy (Robert). He gave me some old documentation that helped out. However even in the old documentation (which must be for a newer version of the mini-mill/2 than I have) the driver schematic shows that they switched to all 4N33s (a faster optocoupler.) The Sherline 1/2 stepping mode for Mach3 I've read gives a 40us pulse width, and trying that is the only way to get pulses longer than 15us in Mach3. With that mode enabled it just barely works on the x-axis only. I figure its due to the statistical tolerances between the 4N28-s and also temperature- as it only works if the machine has been off for a long period of time.

I tried EMC to see if that program would give longer pulses and it does. EMC's longest pulse size allowable is 50us, and with 50us all three axes now work well. I recommend that any Artsoft employee reading this post change their Mach software to allow a wider range of step pulse sizes. There is no reason that it shouldn't be completely variable. The hard coded 15us limit +40us for a special mode seems very arbitrary in my opinion.

I had ordered some 4N33s to swap out the 4N28s with, but I might not need them.

But now that the software issues are solved/understood I still have some questions. I'll post them here for completeness, but they don't really pertain to Mach software anymore:

What are the torque differences between full stepping and half stepping mode? My driver board has a jumper option currently soldered in place for full stepping mode, but I could change it to half stepping mode very easily.

What do people recommend for re-lubricating the slides and lead screws? This mill had been sitting unused and static gathering dust since 1993 potentially, and I can notice some stalling of the steppers due to insufficient torque. I want to tear it down and oil all the moving parts. The slides are brass bushings on ground steel rods, and the lead screw has a kind of nut that is a few inch long plastic thing wrapped in a layer of rubber and around that a wire spring for gripping the lead screw threads.

[Torchhead]

Wow. You guys read spec sheets different than I do. A 4N33 is a darlington with lots of gain, but the switching characteristics of the darlington pair are worse than the single NPN.

Perhaps you are not driving the Optos with enough current? A 4N28 needs 15 or 20 ma of drive to work. The "gain" current in VS current out, is less than 1 on a 4N28. It's much higher on a 4N33 BUT you pay for that in poorer response. There are a lot faster optos (6N136) but they don't have the same pin out so you can't just drop them into the same circuit. There are a few with the same pin out that have slightly better rise times but I still contend the poor response of the 4N28's does not agree with the spec sheets if they are run at the proper input currents and output current through the transistor. It's customary to buffer the drive pins of a parallel port (new ones are only 3.3VDC). You could also consider using a BOB that has the high speed optos on it and bypass them on the driver card. The waveform on the inputs of the Optos should be nice and crisp. If they are then the output should exhibit some roll off but if it's designed right it should have a usable waveform. Usually the turn on is a lot shorter than the turn-off. More current through the opto NPN makes it turn off faster. A lot of the spec sheets show testing setups and the wave forms you can expect.

You did not say, but a direct connection to a parallel port (especially on a new computer) is not enough to fire older optos. The parallel port pins are not designed to source current and do it poorly. They will sink current to an extent, but that means you have to turn on the OPTO (cathode side) with a LOW and source the current with an external +5 on the anode side. Poor drive current has stopped more than one CNC from working.

The reason MACH cannot pump out 50 us pulses is that it has to do other stuff in the time period and is bound by Windows and it's timer period. MACH is the only control software running under windows that can get up to 100, 000 PPS on 6 axis through a parallel port and still service the other inputs at high speed. EMC does not have to deal with the Windows OS.

There won't be a lot of torque difference between full and half step but you will have to provide double the PPS to go the same speed. You will have slightly less torque the more fractional steps you use. 1/2 step helps a LITTLE with resonance problems but only microstepping (8:1 or 10:1 will reduce it significantly)

Tom Caudle
www.CandCNC.com
Totally Modular CNC electronics

[ger21]

Mach3's Sherline Mode will get you 40us pulses.

[LegendCJS]

Thanks ger21 and Tom.

The computer I am using is brand new, so that, combined with the 1kohm resistors between the parallel port pins and the optos (at the 3.3 volts you said that would only give a current of 3.3mA through the LED in the opto, if the port can supply it in the first place) probably explains the slow response times of the 4N28 via low current. I'm not seeing a 5-volt supply anywhere on the parallel port side of the optos. The collector in the opto is hooked to +5 volts, and the emitter has a 470 ohm path to ground (I think). Maybe that explains the fact that only EMC's 50us pulses are able to get the steppers to turn.

The Sherline 1/2 step mode does give 40us pulses, but those didn't work reliably.

And I'm taking the word of my lab mates when they tell me that the 4N33 will work faster (ME here, not a EE).

But thanks again for all the help.

[jalessi]

LegendCJS,

Have you thought about tossing the original controller in the garbage.

Upgrading to something like a Gecko G540 may be less painful.

http://www.geckodrive.com/product.aspx?c=3&i=14469

Welcome to the Zone,


Jeff...

[Torchhead]

Thanks ger21 and Tom.

The computer I am using is brand new, so that, combined with the 1kohm resistors between the parallel port pins and the optos (at the 3.3 volts you said that would only give a current of 3.3mA through the LED in the opto, if the port can supply it in the first place) probably explains the slow response times of the 4N28 via low current. I'm not seeing a 5-volt supply anywhere on the parallel port side of the optos. The collector in the opto is hooked to +5 volts, and the emitter has a 470 ohm path to ground (I think). Maybe that explains the fact that only EMC's 50us pulses are able to get the steppers to turn.

The Sherline 1/2 step mode does give 40us pulses, but those didn't work reliably.

And I'm taking the word of my lab mates when they tell me that the 4N33 will work faster (ME here, not a EE).

But thanks again for all the help.

Total current through the input LED is the applied voltage less the LED forward voltage (1.2V) so you have 3.3 - 1.2 (about 2.1v). Current is volts divided by resistance or about 2.1 ma. No mystery why the opto won't fire or does so only with a wide pulse.

Since the current through the transistor on the opto is a function of the current gain of the opto if you use 2.1 ma you will only get that or less in the collector. even at 470 ohms you have less than 2 ma so the switching characteristics will be poor. At some point it will just cease to work. Raising the Gain ratio (with a darlington) will help BUT you pay a price. The core rise and fall times of the package get a lot slower.

It this point the problem is a lack of drive current. Once that is resolved the next hurdle is the switching times of the opto and that is where the darlington will limit you.


Drive the opto with a full 15ma with a high speed source and you will get the thing to work.

Have your EE's google 74HCT541. You will need to have a +5 source that is referenced to the same ground as the signal source (PC GND) and can supply about 60ma or more. A wall wart regulated + 5 will work. It is an octal buffer chip that has a schmitt input and will take low level (down to 2V) signals and put out nice buffered (up to 50ma per channel) 5 V pulses. It works in either source or sink modes.

You will still be limited as to the upper pulse rate but it will let you get movement.

I agree that in the long run just using a more modern controller (like the G540) would probabaly be cheaper. I don't know what you pay the EE Twins but the G540 is only 300 bucks and has the motor drivers and basic BOB in one tiny package. Drivers are limited to 3.5 per phase and 50VDC max. We use the same drives (G250) in our BladeRunner Series but they are coupled with more modern hybrid steppers that provide over 600 oz-in of torque at only 3.5A. There have been lots of improvements to the drivers and the motors since '93. Actually a lot since even '03 (:-)


Tom Caudle
www.CandCNC.com