PCB Drilling Machine 5.PDF

GENERAL INTEREST

Drilling Machine (5)

part 5 (ﬁnal): ready to go!

Design by T. Müller (Radix GmbH)

www.radixgmbh.de

In this final part on the construction of the PCB drilling machine, all its

functions are tested and the machine is calibrated. The machine is then

ready for its ﬁrst job.

By the time you read this, the first

PCB drilling machines have already

been delivered — and perhaps also

assembled. It is therefore high time

that our description of the machine

be completed. All that remains is the

calibration of the machine. The

machine must be calibrated before

the driver software is able to deter-

mine the reference point, calculate

drilling coordinates correctly, posi-

tion the arm(s) and turntable and

raise and lower the drilling head. A

number of position sensing switches

play a key role in initialising the

machine. In this final part we

describe a complete calibration pro-

cedure for the arms and turntable

with the aid of the TanBoTest soft-

ware, and the use of the TanBoDrive

driver software for drilling PCBs.

Please note that the German wor-

king title of the project, 'Tanbo' (for

Tangential Bohrer) is retained in the

names of the files and programs

developed for the project.

Position sensing swiches

In the machine enclosure there is an

optical switch situated just bet-

ween the two drive shafts for the

tool arms, which operates on the

reflective principle. The position of

the turntable and of tool arms 1

Elektor Electronics

9/2001

PCB

GENERAL INTEREST

and 2 can be determined using this

switch. The sensor consists of a

light-emitting diode (LED) which

emits infrared light vertically

upwards, and a phototransistor,

also pointing upwards, which

detects when a reflective object

passes in front of the LED. For this

reason there is a reflector made of

metallised foil on the pointer atta-

ched to the tool arm drive shaft, as

well as on the underside of the turn-

table.

This reﬂective light switch opera-

tes using unmodulated infrared light

and is therefore highly prone to inter-

ference from other IR sources such

as daylight, artiﬁcial light, and even

the light from cigarette lighters!

While under the circular part of the

turntable the switch is completely

covered, but the turntable’s two ﬂat-

tened sides, which are opposite one

another, can let stray light in. And

why is the turntable not a perfect

circle? It saves a little material, but

more importantly this is the only

way that access can be provided to

the sensor to allow for inspection

and cleaning.

If stray light should interfere with

the sensor, the turntable need only

turn until the sensor is again in the

dark, and then continue to turn until

the sensor sees light again but only

over a narrow angle . In this way,

the position of the reﬂective foil can

be found. Upon subsequent rotation

in the opposite direction the transi-

tion from light to dark must be found

at the same point. Once the system

has determined the position of the

turntable, it then immediately knows

where the flattened sides are that

give rise to the risk of light interfe-

rence: the software takes these into

account.

The process for initialising the

tool arms is simpler, since in each

case the sensor is covered by the

turntable and is therefore protected

from stray light. The system looks for

(and ﬁnds!) the reﬂecting position of

the arm.

If two arms are in use simulta-

neously, there is a problem that must

not be overlooked. If both arms

attempt to travel to their end positi-

ons at the same time, there is a risk

of collision. This difficulty is over-

come elegantly as follows: the Win-

dows software notes the position of

Figure 1. The TanBoTest utility tests the various microswitches and determines the operating

parameters to be stored in the conﬁguration ﬁle.

the two arms at the end of a run and

stores the relevant information in a

temporary ﬁle. When the program is

next run, the ﬁle is loaded, and so a

collision can be avoided. If the pro-

gram terminates abnormally (which

can be the rule rather than the

exception under Windows!), this ﬁle

will be found to be missing the next

time the program is run. In this case

the program enters manual control

mode. A graphical representation of

the drilling machine appears on the

monitor: the tool arms can be drag-

ged into position on the screen using

the mouse so as to represent the

actual state of the machine. This

need not be absolutely accurate, alt-

hough the machine will rely on your

input.

Two microswitches are provided

in the tool arms to determine which

of three possible states the drill head

lift mechanism is in.

mally-closed contact). If the arm is not ﬁtted,

there is no BO switch, and the corresponding

switch inputs (BO1-BO4) on the circuit board

are open. The software can check each arm

individually to determine whether the corre-

sponding drill lift mechanism is in the up

position.

State 2: Drill head lift mechanism is lower

position, determined via BU. All BU contacts

of all arms are normally open and are connec-

ted in parallel. This can work because only

one head is allowed to be in the down posi-

tion at a time. Safety logic in the GAL pre-

vents multiple drill head lift mechanisms from

being actuated at once. And if (because of a

mechanical failure) more than one drill head

should be down at the same time, this can be

detected via the BO contacts.

State 3: This state is between the other two:

neither up nor down. Assuming nothing has

gone wrong, this can only happen during

motion of the drill head. If this state occurs

while the head is not in motion, the drill lift

guide is probably jammed.

State 1: Drill head lift mechanism at

the top of its travel, in rest position.

If no current flows in the solenoid,

the force of the spring built into the

lower part of the arm must push the

drill head to the top of its travel and

thus open switch BO (BO is a nor-

During drilling the Windows software can

measure the time between the closing of the

upper switch and the closing of the lower

switch, which gives the time taken to drill the

hole. A broken drill can be detected by this

9/2001

Elektor Electronics

GENERAL INTEREST

time being too short, and an overly worn drill

can be detected by this time being too long.

If it takes longer than five seconds to drill a

hole, the controller automatically switches off

the drill lift mechanism and the drill motor.

The ﬁrst test

The TanBoDrive driver software requires that

the drilling machine be fully functional, both

electrically and mechanically. The input data

— in this case drilling data in Excellon format

— are specially processed and transformed

into the required commands, in the form of

motion commands and drilling commands, to

be sent to the microcontroller. More than this

TanBoDrive cannot do. Linear interpolation,

as required for interpreting HPGL-format ﬁles,

is included (otherwise the machine could not

even move from hole to hole), but HPGL

import is disabled and will only be enabled

when the previously-mentioned milling arm

has been thoroughly tested. TanBoDrive

should be compared to a printer driver: if the

controller board or the mechanics are not in

order, the program simply gives up.

First, then, servicing and diagnostic soft-

ware is required to test all the functions of the

controller board and the connected units. The

various switches are particularly important,

since they must be correctly wired as well as

correctly ﬁxed. All components must ﬁrst be

made to work as intended and all tests must

be passed before we can move on to drilling a

circuit board.

This requirement is met by the TanBoTest test

software (Figure 1) . When the software is

started a dialogue box appears showing the

states of all the switches, updated every 300

ms. The optical switches can be tested with a

ﬁnger or any reﬂective object. The drill motors

can also be turned on and off, and the lift

solenoids can be supplied with a selectable

current, in order to help check (and, if neces-

sary, improve) the smooth running of the gui-

des. The program, freely available from this

url

Figure 2. Initialisation program for calibrating the arms and turntable.

ger side and above, the left-hand

arm is tool arm 1 and the right-hand

arm is tool arm 2.

The switches are so vital to the

operation of the machine that the

switch inputs to the controller board

are not ﬁtted with a socket, but rat-

her soldered directly. Although a

little more tedious to assemble, the

connection is more secure. A length

of black PVC insulating tape should

be used as a strain relief on the cable

which comes from the drive shafts,

so that the connecting pins are not

under mechanical stress. Ensure also

that the cable is free to move.

Here is a summary of all the func-

tions of the test software:

Cycle button initiates a drill cycle

with the selected parameters.

The Press box sets the force with

which the drill is pushed down on

the circuit board. This value — to a

certain extent — affects the speed of

drilling. If the value is too small, the

solenoid will not exert enough force

to drive the drill down. If it is too big,

the drill can in extreme circum-

stances be damaged when it hits the

circuit board. The value should be

set to suit the drill: a value between

24 and 40 is normal. If satisfactory

operation can be obtained with a

value below 25, it is a credit to the

conscientiousness and care with

which the assembly has been built:

congratulations!

The Target box selects between out-

put drive stages 0-3 for drill and coil.

The Brake box sets the braking value

needed to cause the drilling head to

come to a gentle stop at the end of

its upwards travel. If the value is too

high, the head guide will rebound

down before ﬁnally coming to a halt;

if on the other hand the value is too

low, the upper switch will be

(ab)used as a mechanical endstop,

which is definitely something to be

avoided. Find the setting where the

head does not rebound and then add

one or two to the value.

The Coil box can be set from 0-65 to

control the solenoid current from 0

(off) to 65 (100 %) for the output

stage selected under Target . Beware

that the original solenoids are over-

loaded at the 100 % setting and can-

not run continuously at this current.

A duty cycle of about 30 % is possi-

ble. Continuous operation is possible

at a setting of 26. The software swit-

ches the coil off if the setting is left

unchanged for 5 s.

www.radixgmbh.de/deutsch/

menu_bestellung.html

is self-explanatory and shows the states of all

the switches. It is essential to check the mes-

sages in the dialogue box and to test the

three positions of the drill head lift mecha-

nism. Test the optical switches with a reﬂec-

tive surface and with stray light. Set Target

to 1 and push the drill head down: you should

see activity on BO1 and BU (but not BO2);

likewise for the other target. Using the box

marked Coil , adjust the solenoid current to

ﬁnd the value required to drive the drill head

down. The numbering of the targets is clearly

deﬁned: looking at the machine from the lon-

The Drill On button switches the out-

put drive for the drill motor of the sel-

ected target on and off.

The Hold box sets the time (in units

of 52 ms) for which the drill remains

down after the circuit board is dril-

led through. Too short a time leads to

swarf not being fully cleared from

around the hole on the underside of

the board, while if the time is set too

long, it is simply time wasted. The

default value of 10 (i.e. 0.52 s) is a

The next four boxes are to do with

the drill cycle. Each target has its

own press and brake values, while

the shake and hold values apply to

all four targets. Each click on the

Elektor Electronics

9/2001

GENERAL INTEREST

good typical value for clean holes of

all sizes.

can be made to move simulta-

neously. When trying the speed test,

the stepper motors should not be

connected!

The Shake box is aimed at the auto-

matic clearing of faults. If for exam-

ple the guides are very dirty and not

running smoothly, the force of the

spring may not be enough to lift the

head fully up and operate the upper

switch. This fault can be simulated

by pushing the guide down a couple

of millimetres so that the switch BO

indicates ‘Head not up’ . If the head

stays in this position when it is

carefully released, then the fault con-

dition has been replicated. The

shake value controls how the head is

driven down a small amount and

then released, without braking, so

that it springs back. The force with

which this is done must of course

not be so great that the drill hits the

circuit board when it is not running.

The EventsOn check box, when

checked, gives up some processor

time during the speed test to the

operating system, and hence to

other processes. This is done to

allow other tasks to make progress.

The TanBoDrive control program

uses a dynamic algorithm to deter-

mine how much processor time to

allow the operating system, ensu-

ring as necessary that the system

does not come to a complete halt.

At start-up, the software must be

informed of the address of the LPT

port to which the controller is atta-

ched. This setting can be obtained

from the Windows system informa-

tion, and is generally 0378-037F HEX .

Open an ASCII text editor with an

empty ﬁle and enter on the ﬁrst line

the first of these numbers (here

0378). Now press Return twice and

save the file as lptaddr.txt in the

same directory as TanBoTest . This

file is processed by all the drilling

machine programs and must be pla-

ced in the same directory as all those

programs.

An important part of the test pro-

gram is the measurement of the

transfer rate for data streams over

the Centronics interface. See the text

box for further details.

The Cycle button initiates a com-

plete drilling cycle on the selected

target. The feed time is measured

and displayed (minus the Hold

value). If the cycle completes with-

out error, OK is displayed; other-

wise, Error is displayed. If there is

an error with a feed time of around

4.5 to 5 s, this indicates that the

head never reached the lower posi-

tion: this condition can be simulated

by setting Press=0. With a feed time

of around 0.5 to 1 s, the fault is that

the upper end position was not rea-

ched: this condition can be simula-

ted as described above with

Shake=0.

Adjusting the arms

The values established for the drill

cycle using the TanBoTest program

are recorded in the conﬁguration ﬁle.

After you have checked the opera-

tion of the hardware — and, if neces-

sary, corrected it — using the Ta n -

BoTest program, the machine must

be adjusted and calibrated. Calibra-

tion — also known as setting the

zero point — presents no difficulty

with a linear machine, but how can

we adjust something that is circular

and does not have a start or an end?

Further, several of the components of

the drilling machine can be assemb-

led in different ways: for example,

the drilling axis can be moved to and

fro by several millimetres before

being ﬁrmly screwed down. In order

to calculate the X/Y coordinates

exactly, the length of the arm must

be known absolutely precisely. It is

all done, as we said in the first

Figure 3. The arm positions for calibrating the

turntable: home/wait (a), wait/home(b) and

wait/wait (c).

The Speed button sends data at a

steadily increasing rate to the con-

troller until the PC’s processor can

no longer keep up with the data

stream and the controller’s FIFO

empties.

article in this series, without precision mea-

suring instruments of any kind.

All that is required for exact adjustment is

the TanBoInit software (Figure 2) , which also

uses the lptaddr.txt file. Note: if at any time

during the adjustment procedure anything

untoward should happen, or if you think from

the way the arms are moving that a collision

is imminent, click on the ‘emergency brake’

ALL MOVING STOP: the machine will come

to an immediate halt.

First we calibrate the arms, and then the

turntable. The other way round is not possi-

ble, because the turntable can only be set up

The Speed indication in mm/s refers

to the speed of interpolation on two

axes, for example to simultaneous

motion of the turntable and the tool

arm at the given speed. The mecha-

nical maximum speed limit is around

80 mm/s. If the speed test gives bet-

ter results then there is a surplus of

processing power, and further axes

9/2001

Elektor Electronics

GENERAL INTEREST

with properly calibrated arms. Each arm can

be moved at will using the buttons CW

(clockwise) and CCW (counterclockwise). In

TanBoInit , as before, which of the arms 1-4 is

moved can be set via the Target box .

First check that all the tool arms turn

appropriately in response to the CW and

CCW buttons. Select also the targets for

which an arm is not ﬁtted, and press CW and

CCW. In these cases nothing should move.

The ordering of the target numbers has

already been checked in the test program.

The four BO switches are shown in a column

above one another. If you push down

the head on arm 1, you should see

activity on BO1 and not on BO2.

Check this very carefully, since

otherwise the collision detection will

not work. The drives are extremely

powerful and are easily capable of

destroying one another!

The turntable has its own CW

and CCW buttons, while the Target

box remains always assigned to a

tool arm. Before an arm can be cali-

brated, the turntable must be rota-

ted so that the reﬂective light switch

is off.

Begin with the calibration of arm 1

(target 1). Move arm 2 (if fitted) to

the WAIT position, as shown in

Figure 3a . In the kit of parts which

contained the brass chuck, you will

find a short polished metal pin, 20

mm long, left over. This is the only

calibration tool required. Push this

pin as far as possible, without using

excessive force, into the hole in the

aluminium shaft at the centre of the

Buffer Underrun?

The drilling machine controller software forces Windows to ope-

rate in real time. Although Windows can output data very quickly,

it cannot coordinate the timing of the output exactly, and for this

reason the controller includes a FIFO (ﬁrst in, ﬁrst out) memory

which allows the transmitted data items to be brought back into

step with one another.

This works perfectly as long as we can ensure that the data

stream from Windows never stalls for so long that the FIFO beco-

mes completely empty. The FIFO has 35 slots, from which we can

make the following example calculation.

Let us suppose that a stepper motor is to be turned at the rate of

500 steps per second. The motors used in the drilling machine

have an angular resolution of 1.8°, and so this corresponds to a

rotation rate of 2.5 rotations per second. Since the following gear-

box has a ratio of 200:1, the output shaft turns through an angle of

4.5° per second. The circumference of the circle swept out by the

machine is 1510.6 mm, and so the head moves at a speed of 18.88

mm/s.

What data rate is required for this? The speciﬁcation of the Cen-

tronics interface allows for at least 20000 bytes/s. On an older PC

(150 MHz Pentium) the bandwidth was measured at around

46000 bytes/s.

Each motor step requires exactly one transfer, and so the requi-

red 500 transfers per second occupy a mere 2.5 % of the availa-

ble bandwidth. Bandwidth is therefore not a problem.

At 500 steps per seconds each step lasts 2 ms. The FIFO, with its

35 slots, can hold 70 ms of data. In these 70 ms the processor we

used runs for 10.5 million cycles. If these cycles are consumed by

Windows and the FIFO cannot be reﬁlled with data in time, then

there will be a break in the data stream.

For this reason it is not possible to run programs in parallel with

the drilling machine that load the system heavily or which con-

sume practically all its processing power.

How can we check that the data stream is in fact continuous?

A control signal on the controller board indicates when the FIFO

is empty.

This signal is taken to pin 12 of the Centronics interface (paper

empty) and is asserted when all the bytes in the FIFO have been

processed.

So why is it so important to have a continuous data stream? Step-

per motors that are to be used at a high stepping rate must be

brought gradually up to the desired target speed using a specially-

designed ramp function. The moment of inertia of the rotor is so

great that it cannot follow a rapid speed change and drops steps.

This effect is much more noticeable when starting than when

stopping, and depends on the individual motor and its driver

stage. So, if the pulses are suddenly stopped while the motor is

still turning, the rotor will not come to an immediate halt, but rat-

her jump on one or two positions because of its moment of iner-

tia. That does not matter, or course, if the system has feedback,

for example in the form of an encoder on the drive shaft.

And the solution? The best answer is not to allow interruptions in

the data stream. Before issuing each motor pulse command, we

can check whether the controller FIFO signal is asserted, indica-

ting that the FIFO has emptied. Then — if we know the characte-

ristics of the motor — we can decide on the basis of the pre-

viously-set command rate whether the motor might have got out

of step (this can only happen above a certain, relatively high, rota-

tion speed), and whether the drive in question should return to

the limit switch position to recalibrate itself. If this behaviour

recurs—that is, if you notice that the machine is returning to its

reference points too frequently—then either you will need to use

a faster computer, reduce the load on your system, or reduce the

pulse rate for the motors.

None of this is really a problem, but the unnecessary recalibration

movements do affect the average speed of the machine and ham-

per the system somewhat. A reduced motor step rate can

increase the overall speed of the system, if it reduces the number

of recalibration movements.

And there is an even better solution: the FIFO memory can be

expanded, as used to be done with print spoolers. Exactly the

same principle can be applied to the drilling machine. With a small

static RAM of for example 8 kbyte the capacity of the FIFO can be

increased to 8.25 seconds of data at the frequency of 500/s men-

tioned above.

This extra FIFO would take the form of an additional unit; it is not

possible simply to change the microcontroller. Microcontrollers

with a large amount of on-chip RAM are signiﬁcantly more expen-

sive. The additional module could be supplied with power via the

OptSpare pin (5 V DC). Another possibility would be a small

module on a printed circuit board to ﬁt in the microcontroller

socket which includes the extra buffering.

If such an idea is implemented, you will be able to read about it

either in Elektor Electronics or on the drilling machine website.

Elektor Electronics

9/2001

Plik z chomika:

Inne pliki z tego folderu:

Inne foldery tego chomika: