Coil Winder

[CJ]

I was thinkin, we could try all three:

Logic Chips
PIC Chips
PC

Since we have a simple pulse, why not RS232 it out to a laptop and use basic?

Hardware is more fun than software, and those logic chips are cheap, so we will try that first.

 

[AnalogPackrat]

Ceej,

Are we on for a Wed night TTL hacking session? I have a tube of assorted stuff ready to go. No 74xx161s, but I have some 7474, 7404, and several others. Got a breadboard and some 22-24ga insulated solid core wire? Do I need to make a lunch time surplus run for the missing bits? Let's get that variable speed traverse working!

A P (or is that D P?)

 

[CJ]

I went down to Anchor today and bought

10 ea. 7400
10 ea. 7404
5 ea. 74LS90
2 ea. 74HC90
4 ea. 74HC161
10 ea. CD4001
10 ea. CD4011
10 ea. 4013
5 ea. 4027


And still have the NTE crap, 74HC123 and 74HC109.


Just drop by, look at this thing, draw me up some ideas, and then I can breadboard it while your at home in front of the fire place on these 20 degree nite, drinkin wine with the wife.
Come back and trouble shoot after I blow it up.

cj

 

[gyraf]

why not RS232 it out to a laptop and use basic

If you go PC, it rather calls for a parallel (printer-) port. Easy to use with basic. Has fast inputs and individually-settable output bits. Is used for simple stepper CNC control all the time - quite a few basic examples out there iirc.

Jakob E.

 

[CJ]

Ah yes, a stream of pulses does not make a word.

I had to build my own cable to hook up the encoder. I could have soldered directly to the pins, but this is more kosh.

Weird DF-11 connector for the Renco encoder.
Did not have the $900 dollar crimper, so I soldered the ribbon cable leads directly to the connector pins, then they snap into the black housing.
I left the pic in largde rez so you can read the digi-key p/n's if you happen to get stuck with this cable problem:

http://vacuumbrain.com/The_Lab/Winder/encoder_cable.jpg

Here is the schematic I used coming out of the encoder.
Just a few changes. There was enough capacitance in the ribbon cable to where adding C according to the RC formula just messed up the symetry of the pulse. With no C, the pulse was very square.
The diagram was right on the Renco site, so a little searching would have saved me a week, but then again, I would have nbot learned anything.

Here is another wire gauge chart I made, I took about 7 differnt charts and took the best stuff from all of them, eliminating the 9 extra dims for thickness, triple build insl, etc.. and extended the gauges down to #16.
Turns per inch will be adjusted as I determine real world values during actual winding.
Every winder has their own binder full of winding data that they take to the grave. Turns per inch, DCR, all that good stuff.
I just hazzarded a guess as to turns per inch from 45 down to 58, as there is no data on the web for these sizes, I guess it varies too much for a fixed number. I just decreased the Lay% by a seat of the pants increment according to dia. So take the turns per inch from 45 on down with a grain of salt. To get turns per inch, you just divide 1 inch by wire dia, then multyply by the Lay factor. The Lay factor is due to the fact that wire varies, and so do windewrs and their machines, so a perfect Lay is not possible. (what?)
Excel:

http://vacuumbrain.com/The_Lab/Winder/awg_wire_data.xls

Now that I have an adjusted turns per inch chart, I can figure out the division constants needed for the different gauges.

The math is fairly easy, the motor puts out 5000 pulses per rev, but the Gecko divides that by ten in order to get the ten microsteps.
So we work with 500 pulses from the dc motor.

The stepper uses 0.9 degrees per pulse, so 360/0.9 = 400 pulses per rev.

So 500/400 gives 1.25 revs of the stepper per every rev of the dc motor.

Figure in the 80:1 lead screw, and were done:

1.25/80 = 0.015625 or exactly 1/16".

We could have done the math this way:

1 1/4 = (5/4) / 80 = 5/320 = 1/16 = 0.015625"

Here is another spread sheet for working out the divider constants.

You just take the turns per inch values and divide by the tuns per inch of the dc/stepper setup to get the fraction of travel that you need for the various gauges:

http://vacuumbrain.com/The_Lab/Winder/pulse_divider_constants.xls

Final results are:

Gauges 29-34 = divide by 2
Gauges 35-37 Divide by 3
Gauges 38-39 = divide by 4

Gauges 40-42 = divide by 5
Gauges 43-44 = divide by 6
Gauges 45 = divide by 7

Gauges 46 = divide by 8
Gauges 47 = divide by 9
Gauges 48 = divide by 10

.Here are some simple divider circuits out of the Mims book:

http://vacuumbrain.com/The_Lab/Winder/divide_a.jpg
http://vacuumbrain.com/The_Lab/TA/Winder/divide_b.jpg

 

[CJ]

OK, the divide by two chip works great, easy to breadboard ratrher than hardwire thanks to the AP-!
I also built the divide by 5, 6, 7, 8, and 9 and they all work great!

http://vacuumbrain.com/The_Lab/Winder/divide_2.jpg

Divide by two happens to work out almost exactly to #34 gauge, at least in theory.

So lets slap down a layer of 34 with the divide by two setting and see what happens:

Not too shabby. I was supposed to get 128 turns per inch and I got 130.

So it looks like I am heading for a bunch of fixed divider chips, a bunch a bunch of variabel divider chips, and a bucnh of comparators.
Maybe burn a board if it all works, then if someone wants to do a controller, we already have the circuit and artwork.

Party time in the ghetto region of Mt. View.
I'm out.
:thumb:

 

[Learner]

gee.....i hope I don't catch the coil winder bug :?

But I couldn't help it thinking, out of all the transformer you disected. Were all the wires wounded consecutively side by side strand after strand? Especially the UTC ones I am refering to.

 

[CJ]

Most of them were machine wound. Theree were a few hand guided, random wound affairs, like the API 2622.

Marinair inputs could be hand guided, but theyy did a good job.

 

[Learner]

So all the coils you unwound are not flat but have a bulge in the middle? Somehow it seems to get like that when you wind randomly....mm... still remember my fishing days...

dan, I feel this coil winder bug is coming to get me. arggg... :shock:

 

[CJ]

I am going to use seperate channels for each division.
The HC4017 works alright for all the divisions, but it does not put out a symetrical sqyare wave like the 7490 chips, and when you divide by two, the pulse width and frequency approach the bandwidth limit of the driver.

I figure the verniers for each gauge will have to have different values of R and C to make them work well at their assigned freqs. So better to have multi channels, besides, it looks more impressive!

Also, as you divide 1/2, 1/3, 1/4, ...as you get out to 1/9 and 1/10, there is actually very little between thw two from a mathmatical standpoint, so we can drop the divide by 9, which was acting quriky anyway.

so,

Channel 1 - no division-just a vernier and highspeed comparator
Channel 2 - Divide by 2- high speed comp - vernier - high speed comp
Channel 3 Divide by 3 - slow comparator is OK here, max freq is 250 KC/3 = 83 KC- vernier - low speed comp

Channel 4 - Divide by 4 - then same as above
Channel 5 - Divide by 5 - then same
Channel 6 - Divide by 6
Channel 7 - Divide by 7
Channel 8 - Divide by 8
Channel 9 - Divide by 10

Here is the board ready to wire, left room for futire expansion, such as divide by 12 for ultar thin wire.

I noticed that ther is not much difference in winding pitch, we are covering 1/64 th inch to 1/640th inch.
The diff is therefore: 0.015625 - 0.0015625 = 0.014 about 14 thousandths of an inch. If you divide that up by 22 gauges, not much difference, especially with the #40 and thinner.

 

[mikka]

Very sweet winding job... :thumb:

 

[burdij]

You could save some wirewrapping by implementing a programmable divider such as is shown in this circuit:

This uses a single synchronous presettable counter to divide the incoming rising edge clock. The output is a low going pulse one clock width wide. Most of the logic including the tri-state driver 74HC244 is used to interface two BCD thumbwheel switches to set the divisor ratio. There are four ICs total. Other families (LS, AC, C, etc.) should be usable, too.

There is no logic to test for a ratio of "00" so I am not sure what behavior will occur there. I also didn't test the circuit so some debugging might be necessary.

 

[CJ]

Now ya tell me!

I did try a programable, as mentioned above, but this brute force method seems to be the ticket. Plus, I get to learn all these logic chips.
Almost done wiring up that mother board seen above.

I thought it might be cool to have an LED on each frequency channel, when you slow down the winder, there would be 9 LED's all blinking at different rates.
It would eother look really cool, or put someone into a seizure. :razz:

 

[CJ]

30 days forward, ten days back, i just found out that the variable divider circuit that I was
gonna use to vary each freq channel will not work.
It only divides by integer equivalents.
So it's either software, or some type of analog sloution, like a true rms converter to sample the pulses, turn it into a dc voltagte which drives a VCO.

Either way might work.
I am worried about interupts causing lost pulses if using a PC or pic chip.

The 16C54 is a very basic chip, no interupts, been looking at that.

 

[Learner]

[quote author="CJ"]

The 16C54 is a very basic chip, no interupts, been looking at that.[/quote]

Don't bother with the old 16Cxx series OTP, OTP is a pain and a waste of time for developing/testing programs you might ended up wasting a bunch of them.

Go for the new 16Fxx series with flash and eeprom and a USB ICD2 that allows you to hook up to your circuit to test the program, so you can edit your program in MPLB then compile and see result of how the circuit behave then refine the program repeatedly.

Haven't really gone thru the thread to know why you need dividers but I assume to count the turns?

 

[CJ]

The dividers are to get the correct winding pitch for the 20 or so gauges that I want to wind.

We have the 16C71 and 16C74B as well as the 16F stuff here at work, chip burners for all of them, the configuring of addresses and other header code gets worse as you go up the PiC ladder.

All I need to do is divide the input pulses by a constant.
That's all. Nothing else. I can burn as many chips as needed.
So I do not need 8000 lines of code and a fancy chip for that. I hope.
We hire a consultamnt that is very good with the PICS.
But very expensive.
But a program to divide could not be more than an hour, @100 per, so maybe.....

So far, this thing has only cost $384, so I am still doing alright.
I do not know what a precision machine costs nowdays, but I must have it beat pretty good.

 

[Learner]

[quote author="CJ"]The dividers are to get the correct winding pitch for the 20 or so gauges that I want to wind.

[/quote]

Whats the relation between the winding pitch and the gauge of the wire? beides higher gauge requires more turns to get the same inductance which takes longer?

 

[CJ]

the smaller the wire, the more turns per inch.
the n more turns per inch, the less you want the traverse to move.
so they have to be synced in order to get a nice layer.

 

[burdij]

Is the problem caused by the need to have decimal divisor ratios? Can the tach/encoder produce pulses at 1000x the rpm of the winding motor or even 100 times more? Perhaps gearing up the tach off the motor shaft or getting a tach with more resolution. A friend of mine built an encoder using a CD and a CD pickup head to get a tach with millions of bits of resolution in each revolution. Then it would just be a matter of increasing the digits of precision in the divider chain and presetting the decimal ratios.

 

[CJ]

I thought of that.
Ther encoder puts out 5,000 per rev, but they make them up to 12,000 per rev, which might just be the ticket.
Right now, I am missing about ten gauges that fall in between the cracks.
With 12,000, I bety I could pick most of them back up.

Time to price an encoder. :thumb: