e011026.PDF

BEGINNERS COURSE

PC Serial Peripheral

Design (5)

Analogue measurements

By B. Kainka

So far in this series we have used the PC only with digital signals: switching,

monitoring and counting. Now we turn to the analogue domain: our programs

will understand not just ‘yes’ and ‘no’, but ‘larger’ and ‘smaller’.

nected between an output that has

been set high and an input, there are

only two possible results: either the

resistor is sufficiently small that a

clear logic one is read at the input, or

it is not. If more precision is required,

a new approach is needed.

For a 100 µF capacitor and a 1 k

Ω

resistor we have:

T = 1000

Ω ×

0.0001 F = 0.1 s =

100 ms

A doubling of the resistance leads to

a doubling of the charging time. The

same goes for a doubling of the

capacitance. So, we can measure the

charging time and deduce the value

of either one of the resistor or the

capacitor, if the value of the other is

known. All that is needed is a pro-

gram that replaces the switch in Fig-

ure 1. Figure 3 shows a simple circuit

where the capacitor is connected not

to ground but to the TXD output.

There is a good reason for this: if an

electrolytic capacitor is used, it must

never be charged with the wrong

polarity, and this is guaranteed as

long as TXD remains at –10 V.

The circuit charges and dis-

charges the capacitor via DTR and

uses the DSR signal as an input to

determine when the set voltage is

reached. The threshold voltage will

be around 1.5 V. Comparing this with

the overall voltage range of –10 V to

+10 V, we see that the threshold is

about 11.5 V/20 V = 0.575 = 57.5 %

of the ﬁnal voltage. This is not too far

from our value of 63.2 % given above

for the time constant. In any case,

the error factor introduced is con-

stant and can be compensated for

later. There are other sources of error,

which we discuss below.

000074 - 5 - 11

Charging and discharging

Figure 1. Charging and discharging a capacitor

If computers are good at one thing,

it is counting. We can use this to

measure time: a program simply

counts the seconds (or milliseconds)

that pass until a certain event hap-

pens, for example when an input

changes state. If we can convert an

analogue quantity such as a resis-

tance into a time period, then it will

be easy to measure it with a com-

puter. In this example, we can use

an RC network. Figure 1 shows a

capacitor C charged and discharged

through a resistor R. A time constant

T is associated with an RC network:

If a voltage is applied to a serial port input, it

will be read as either a logic Low (0) or a logic

High (1). The PC cannot determine the actual

voltage present. Likewise, if a resistor is con-

T = RC.

Here T is the time taken for the volt-

age across the capacitor to reach

63.2 % (=1-1/ e ) of its ﬁnal value. This

can be derived from the exponential

charging characteristic, shown in

Figure 2 . We shall not go into the

details of the physics here: it is

enough to know that the time taken

to charge to a given voltage is

directly proportional to the capaci-

tance and to the value of the resistor.

000074 - 5 - 12

Figure 2. Charge/discharge curves and

deﬁnition of the time constant T

Elektor Electronics

1/2001

BEGINNERS COURSE

Listing 1. Measuring the time constant in milliseconds

Private Sub Form_Load()

i = OPENCOM(“COM2,1200,N,8,1”)

If i = 0 Then

i = OPENCOM(“COM1,1200,N,8,1”)

Option1.Value = True

End If

If i = 0 Then MsgBox (“COM Interface Error”)

TXD 0

RTS 0

DTR 0

Counter1 = 0

Timer1.Interval = 2000

End Sub

DTR

DSR

100µ

–10V

TXD

000074 - 5 - 13

Figure 3. Automatic charging from the PC

Private Sub Form_Unload(Cancel As Integer)

CLOSECOM

End Sub

Private Sub Timer1_Timer()

DTR 1

TIMEINIT

While (DSR() = 0) And (TIMEREAD() < 1501)

DoEvents

Wend

Label1.Caption = Str$(TIMEREAD()) + “ ms”

DTR 0

End Sub

Counting time

The time measurement takes place

in a ‘while’ loop as shown in List-

ing 1 . The loop runs until either DSR

becomes set or the count reaches

1.5 seconds (timeout). The measure-

ment loop includes the command

DoEvents: this allows Windows to

process other events that may be

occurring in the system. During mea-

surements the user can still use the

mouse and other applications, and

indeed stop the program itself. This

is reassuring for the user, especially

when bugs in the program cause it

to function incorrectly. In general it

is always necessary, when program-

ming loops, to consider how the loop

can be forced to terminate. Other-

wise if the program gets into an inﬁ-

nite loop the PC will need to be

reset, either by switching it off and

on again, or by using the familiar

Ctrl-Alt-Del chord. Adding a

DoEvents call makes the loop safe;

but this has a cost in timing accu-

racy, adding an uncertainty of

around 1 to 3 milliseconds to the

measured time.

one. As we said above, a 100 µF

capacitor and a 1 kΩ resistor give a

time constant of 100 ms. Similarly,

with a 10 µF capacitor we have a

time constant of 10 ms. This can be

tested by trying various capacitors

from the junk box. Often there will be

quite a large discrepancy between

the value measured and the value

printed on the capacitor: this is

down to the large tolerance (as much

as 50 %) quoted for electrolytics. The

capacitance of an electrolytic often

changes if it is stored for a long time.

The measurements are less accu-

rate for very large electrolytic capac-

itors, with values of say around

1000 µF. The indicated value will be

too small. The reason for this lies in

the program: the charged capacitor

must be discharged, which also

requires time. Our program uses a

timer with a period of two seconds:

one second is used for charging the

capacitor up to the input threshold

voltage; the remaining second, how-

ever, is not enough to discharge the

capacitor completely. The next

charge therefore takes less time.

There is a simple solution: a diode

can be used to accelerate the dis-

charge cycle. Figure 5 shows the cir-

cuit diagram, and Figure 6 shows

how it can be constructed. With this

modification to the circuit we can

Figure 4. Measurement with 100 µF and 1 k

Ω

DTR

1N4148

DSR

1...100µ

–10V

TXD

000074 - 5 - 15

Figure 5. Improved capacitance measurement

circuit

000074-1

GND

CTS

RTS

DSR

47 µ

GND

DTR

µF not ms

TXD

RXD

DCD

1N4148

If the units ‘ms’ in the window in

Figure 4 are replaced by ‘µF’, the

value shown is not too far off the true

000074 - 5 - 16

Figure 6. Construction details.

1/2001

Elektor Electronics

BEGINNERS COURSE

Listing 2. Modiﬁcations for program Cmeas2.frm

Private Sub Timer1_Timer()

F = 1.19

DTR 1

REALTIME (True)

TIMEINITUS

While (DSR() = 0) And (TIMEREADUS() < 1500000)

Wend

T = TIMEREADUS() * F

REALTIME (False)

T = Int(T)

Label1.Caption = Str$(T) + “ nF”

DTR 0

End Sub

Figure 7. Capacitance display in nF

Listing 3. Measuring the time constant in microseconds

Private Sub Timer1_Timer()

DTR 1

REALTIME (True)

TIMEINITUS

While (DSR() = 0) And (TIMEREADUS() < 1500000)

Wend

t = TIMEREADUS()

REALTIME (False)

Label1.Caption = Str$(t) + “ us”

DTR 0

End Sub

DTR

1N4148

1...22k

DSR

47µ

–10V

TXD

000074 - 5 - 18

Figure 8. Circuit 1 with 10 k

Ω

potentiometer.

Listing 4. Determining the charging resistance

Private Sub Timer1_Timer()

DTR 1

REALTIME (True)

TIMEINITUS

While (DSR() = 0) And (TIMEREADUS() < 1500000)

Wend

T = TIMEREADUS()

T = T * 1.0000000001

R = 2200 + 7800 * (T - 76300) / (294600 - 76300)

REALTIME (False)

R = Int(R)

Label1.Caption = Str$(R) + “ Ohm”

DTR 0

End Sub

Figure 9. Measuring the charging time to

microsecond resolution

reliably measure capacitors up to

about 1500 µF. Larger values are pos-

sible if more time is allowed by mak-

ing the appropriate changes to the

program. The timeout value in the

measurement routine and the overall

timer period must both be increased,

so that the program waits long

enough for the measurement to be

completed.

What about capacitors smaller

than 1 µF? In principle the charging

resistor could be increased. How-

ever, a problem then arises: the

impedance of the DSP input (around

increased. This problem could for

example be overcome using an oper-

ational amplifier with a high input

impedance, but this is outside the

scope of this series.

700

600

Software enhancements

(ms)

500

400

It is much more interesting to try to

get around these limitations in soft-

ware. In particular it is possible to

use a timer with a resolution of one

microsecond. This increases the res-

olution of the capacitance measure-

ment one thousandfold, allowing

measurements in nanofarads ( Fig-

ure 7 ). Library PORT.DLL provides

functions

300

200

100

R (k

Ω

)

000074 - 5 - 20

introduces measurement errors

that get more and more severe as the

value of the charging resistor is

Ω)

Figure 10. Charging time plotted against

charging resistance

TIMEINITUS

and

Elektor Electronics

1/2001

BEGINNERS COURSE

TIMEREADUS for this purpose,

where ‘US’ stands for microseconds

(µs). These functions are used in

Listing 2 .

When we consider making mea-

surements in microseconds, we must

look at the effect of Windows on the

timing accuracy. In principle other

process running in parallel can inter-

rupt the measurement program and

cause large inaccuracies. This can be

prevented by raising the priority of

the measurement task, for which a

special function is provided in

PORT.DLL. Using REALTIME (True)

we can obtain greater reliability. It is

essential to set REALTIME (False)

after the measurement has been

taken. The exact accuracy achieved

depends on the PC: with a 200 MHz

Pentium MMX we measured varia-

tions of about 50 µs in the measured

value; with a faster PC this may be

reduced. If the same experiment is

carried out in Delphi, the timing

accuracy is about 20 times better.

The method is described in the Elek-

tor Electronics book ‘PC Interfaces

under Windows’, to be published

soon). It is nonetheless impressive

that an interpreted language such as

Visual Basic can achieve such timing

accuracy.

Alongside these changes to the

program, we can also improve the

basic accuracy of the measurements.

We have already seen two sources of

error in the simple equation t/ms =

C/µF, namely the threshold voltage

being slightly too low and the input

impedance of the DSR signal. There

is a third source of error: the DTR out-

put does not switch exactly between

-10 V and +10 V, but has an internal

resistance of roughly 430 Ω. The

charging resistance is therefore in

effect about 1430 Ω. Further, this

internal resistance is non-linearly

dependent on the current. The TXD

output also has an internal resis-

tance, making the voltage on the

capacitor slightly higher than

expected. The effects are too compli-

cated to be analysed mathematically,

and so we take the course preferred

by experienced engineers in the face

of a complicated calculation: test;

measure; calibrate. All the errors can

be condensed into a single correction

factor F which can be determined

with a calibration measurement. For

this we require a capacitor whose

value is accurately known (or which

can be accurately measured). Then

the correction factor can be adjusted

until the reading is correct. On the

author’s PC the value of F was found

to be 1.19; this value can of course be

used on any PC, but there will be

individual variations from machine to

machine which can only be compen-

sated for by determining the correct

value of F in each case.

Resistance measurement

Figure 11. Measurement using a 15 k

Ω

resistor.

We can measure resistance using

the same method; this is similar to

the way that potentiometers are

read by PC games cards. The circuit

for measuring resistance is shown in

Figure 8 and is essentially the same

as the capacitance-measuring cir-

cuit. Here, however, we use a fixed

capacitor and work with various

resistor values.

To test this circuit we use the pro-

gram ( Listing 3 ) which measures the

charging time to the highest accu-

racy. We use again the technique

described above for measuring small

capacitors: REALTIME (True) gives

us good timing accuracy. Figure 9

shows how the results appear on the

screen.

The circuit can be tested using

metal ﬁlm resistors with a tolerance

of 1 %. Measurements with an accu-

rate ohmmeter indicate that in gen-

eral the tolerance of such resistors is

rather better. A set of tests using

C=47 µF gave the following results:

The reasons for deviation from the ideal

linear curve are the same as were found in

measuring capacitance. With very small

charging resistances we get an error due to

the non-linear internal resistance of the DTR

output driver; with large resistances (say

around 22 kΩ) the low impedance of the DSR

input causes error.

The data obtained can be converted into

an explicit equation which will let us calcu-

late a resistance with high accuracy. Listing 4

shows the measurement routine for resis-

tance. The calculation in question reads:

R = 2200 + 7800 * (T - 76300) / (294600 -

76300)

Using this formula we can obtain a measure-

ment accuracy of around 1 % in the range

1.5 k

. It should be borne in mind,

however, that this function, speciﬁc to a par-

ticular PC, will not give the same accuracy on

a different machine. The various measure-

ments and calculations must be carried out

afresh for each PC. It is sufficient to measure

the charging times for two resistors, say

2.2 kΩ and 10 kΩ and substitute these in the

formula. Failing this, it is possible simply to

consider where the greatest sources of error

lie: in this case our attention turns immedi-

ately to the electrolytic capacitor. Electrolyt-

ics often have a capacitance very different

from the value printed on them, which has a

significant effect on the charging time for a

given resistance. If a suitable correction fac-

tor is inserted in the line

Ω

to 15 k

Ω

R/k Ω

T/ms

33.7

0.1

0.22

34.5

0.56

37.9

45.5

2.2

76.3

4.7

147.5

6.8

204.9

8.2

245.9

294.6

433.7

661.2

T = T * 1.0000000001

From the numbers alone a strong lin-

ear dependency can be seen. An

exact determination of the linearity

is possible using a graph: an analy-

sis using Excel produced the curve

shown in Figure 10 . It can be seen

that linearity is good between

around 2.2 k

then this error will be compensated for, and

reasonably usable results (see Figure 11 ) can

be obtained. A correction factor of greater

than or less than one will be required accord-

ing to whether the capacitor has a value

lower or higher than nominal.

Ω

and 10 k

Ω

(00074-5)

1/2001

Elektor Electronics

Plik z chomika:

Inne pliki z tego folderu:

Inne foldery tego chomika: