Transcript The Solutions : Calculable AC Voltage Reference

16th July 2013
NCSLI 2013, Nashville
1
The Design of a Calculable
AC voltage reference using
digital waveform generation
16th July 2013
NCSLI 2013, Nashville
2
Abstract
The design of an AC voltage reference source
using a digital to analogue converter controlled
by a microcontroller to produce a calculable
RMS AC voltage reference with accuracy
suitable for calibrating high performance Digital
multimeters.
16th July 2013
NCSLI 2013, Nashville
3
Technical Objectives
• To provide a method for secondary and below
laboratories with the ability to generate high
accuracy AC Voltages
• Investigate errors associated with digitally generated
AC Voltages, as well as the practical application of
digitally generated waveforms in commercial
calibration
16th July 2013
NCSLI 2013, Nashville
4
Learning Objectives
• To investigate alternatives to traditional AC voltage
verification methods
• To enhance calibration laboratories understanding of
verification of high performance modern DMM’s
16th July 2013
NCSLI 2013, Nashville
5
The Requirement
The calibration of Today’s Modern high performance
Digital Meters which offer ACV Accuracy in the order
of 80-100 ppm
To achieve a stand off ratio of just 4 to 1
requires an accuracy of around 20 ppm. This is not
possible with a multi-product calibrator
16th July 2013
NCSLI 2013, Nashville
6
The Solutions
1) Multi–Junction Thermal Transfer Standard
2) Calculable AC voltage reference
16th July 2013
NCSLI 2013, Nashville
7
The Solutions : Multi-Junction
Thermal Transfer Standard
Advantages
Disadvantages
Wide Frequency Range
Requires both a stable DC and
AC voltage source
Proven Accuracy
Equipment is expensive
Calibration is costly, especially
when a wide range of points is
required
Even with protection, older
Thermal Transfer devices are
easy to damage by accidently
over-ranging
16th July 2013
NCSLI 2013, Nashville
8
The Solutions : Calculable AC
Voltage Reference
Advantages
Disadvantages
Due to the nature of the
device, does not require an AC
or DC source
Limited frequency range
Instrument can be internally
verified with a DC volt meter
Output voltages are limited
Due to solid state design,
instrument is hard to damage
with incorrect connections
Low cost due to ‘simple’
hardware
16th July 2013
NCSLI 2013, Nashville
9
The Concept of the
Calculable waveform
• Digitally create an AC waveform using a digital to
analogue (DAC) converter. In our experiments we
chose a waveform based on 256 steps.
• Single step the waveform to allow the level of each
step to be measured as a DC level.
16th July 2013
NCSLI 2013, Nashville
10
The Concept of the
Calculable waveform
The RMS value of the waveform is then calculated using
the following formula
1 n 2
(V12 ) + (V22 ) +........ + (Vn2 )
Vrms =
(Vi ) =
å
n i=1
n
Where :
Vrms = RMS Ac voltage output
V = measured DC voltage of ‘step’
n = number of steps
16th July 2013
NCSLI 2013, Nashville
11
History
This is not a new idea.
This technique was pioneered as
long ago as 1988 where results against
Thermal converters gave uncertainties at
the 7 volt level of just 5ppm.
16th July 2013
NCSLI 2013, Nashville
12
What has changed
There is now a much greater need for high
accuracy ACV. In 1988 we were only seeing the
first developments in High accuracy AC
DMM’s. Now almost every laboratory has a
high performance DMM
16th July 2013
NCSLI 2013, Nashville
13
What has changed
Digital Electronics along with the rapid
development of digital to analogue converters
for high quality audio has now made it much
easier to implement this concept.
16th July 2013
NCSLI 2013, Nashville
14
The Circuit
The design uses a PIC micro controller, programmed with
the waveform to directly drive a DAC. The PIC is directly
controlled from an RS232 interface.
Clock
Control
Interface
10V Reference Input
Micro Controller
PIC 16F84
DAC
AC Voltage Output
16th July 2013
NCSLI 2013, Nashville
15
The DC Reference
The D to A converter needs a short term stable
10Volt DC reference.
The Linear Technology LTZ100 reference can
easily provide better than 1ppm stability over the
short term 2/3 hours required.
As the LTZ1000 is temperature stabilised
temperature variations over the measurement period
will not add any significant contributions to an
uncertainty budget
16th July 2013
NCSLI 2013, Nashville
16
Improvements Vs
Performance
The circuit used is extremely simple. Early design’s used
two converters, with the output being switched between them
to remove glitches.
With modern converters this is not necessary.
The PIC micro controller directly drives the converter,
loading the code into the converter directly from it’s
memory, also clocking the converter after each data load.
This very simple approach requires some machine code for
the PIC to make it run as fast as possible. All timing comes
from the crystal oscillator driving the PIC.
16th July 2013
NCSLI 2013, Nashville
17
The Software
The AC reference is very simple to control, with just
a few commands needed.
S-
115Hz
260Hz
3200Hz
41000Hz
Stop waveform and set to zero
XAdvance 1 step
YAdvance 10 steps
Z-
16th July 2013
Advance 64 steps
NCSLI 2013, Nashville
18
Taking the DCV Measurements
The DCV measurements can be taken with a high
performance 8 digit DMM.
This will typically give full scale & linearity uncertainties of
less than 5ppm.
As there are over 256 measurements to be made it is
recommend that the process be automated by using a PC &
software.
16th July 2013
NCSLI 2013, Nashville
19
Measuring Software
A procedure to measure each step was written using
Procal. As each step is measured it is squared and
added to a running total.
The procedure being 256 steps long, takes about 40
minutes to run automatically.
Although the measurement procedure could be
written in any language Procal has the advantage to
work with any DMM without changing code.
16th July 2013
NCSLI 2013, Nashville
20
Basic Measurement Diagram
16th July 2013
NCSLI 2013, Nashville
21
Practical Set up
16th July 2013
NCSLI 2013, Nashville
22
Voltage Range
With a 10V input the peak output
of the converter will be 10Volts;
giving an RMS voltage of 7.07V
To get a 1V output the 10Volt output
can be resistively divided down allowing the output to
still be single stepped and measured on DC.
Note an IVD cannot be used to divide down the DC
output. However it can be used on the AC output
16th July 2013
NCSLI 2013, Nashville
23
Errors Switching Glitches and transition
This error is most difficult to evaluate. It is dependent on the
DAC used. The DAC we have chosen is typically used for audio
will be very ‘clean’. The approach used by Transmille has been
to look at an individual step on a scope and estimate an error
based on the size and duration of the transitions.
16th July 2013
NCSLI 2013, Nashville
24
Errors Switching Glitches and transition
Typical figures for this, at a frequency of 50Hz, where there is a
step every 80uS, worst case measurements give glitch times of
around 8nS, with a glitch size of around 10% of the step. This
gives transition errors around the 10ppm level. In practice as some
glitches will add and some will subtract the real error will be much
less.
16th July 2013
NCSLI 2013, Nashville
25
Errors
from Rise / fall times & DC offsets
Rise and Fall time
Providing the rise and fall times are equal they do not contribute
to errors. The error caused by a difference in rise/fall times can
be mathematically calculated
DC offsets
Any DC offsets in the DAC will be measured and added to the
calculated RMS figure. However if the device being calibrated is
AC coupled DC offsets in the AC output will give an error and it
may be best to trim any offsets out to avoid this problem.
16th July 2013
NCSLI 2013, Nashville
26
Errors
from Loading effects
Output Loading
When measuring the output with a thermal converter the load
effect on the 2 wire output will need to be considered.
When using measuring the output with a typical DMM loading
effects will be negligible
16th July 2013
NCSLI 2013, Nashville
27
Errors from drift in DC Reference Source
& DAC
As the AC source is a transfer device it is only the short term
drift between the DC measurement of the system and the
subsequent use on AC that contributes to the error.
Long term changes in the reference or the DAC linearity or gain
do not contribute as the system is calibrated before use.
16th July 2013
NCSLI 2013, Nashville
28
Uncertainties
Typical laboratory figures for uncertainty
Imported DC Voltage Measurement
Resolution of DC voltage Measurement
Switching transition errors
Short term stability of DC reference and DAC*
5ppm
0.1ppm
10ppm
2ppm
Combining for 95% (K=2) gives
13ppm
* Includes effects of temperature for +/- 2’C from Tcal
16th July 2013
NCSLI 2013, Nashville
29
Verifying the accuracy
Measurements made by Transmille
Transmille best AC measurement capability uses
our Fluke 540b thermal transfer, the accuracy of this
Instrument is enhanced by measuring the output of the
thermal converter directly. Loading effects on both AC and
DC measurements were compensated for. Tests
performed against a Wavetek 4920, Transmille 8081,
Agilent 3458A and Fluke 8508A provided similar results.
Typical readings obtained
Single step calculation =
Measurement by thermal Transfer 540B =
Error = 14 ppm
16th July 2013
NCSLI 2013, Nashville
7.055681
7.055779
30
Frequency Range
To date Transmille’s interest has been
In the frequency range 40Hz to 1kHz.
The design could generate frequencies
below 1Hz with no additional errors.
Frequencies up to 10kHz could also be
generated with increased error from switching.
16th July 2013
NCSLI 2013, Nashville
31
Improvements……
Extending the Voltage range.
A simple resistive divider
to provide 1V and 100mV outputs.
A DC voltage amplifier to provide a 100V output.
Both could be again ‘calibrated’ at DC using a known DMM
to provide the calculated AC RMS value
16th July 2013
NCSLI 2013, Nashville
32
Other Functions……
Non sine wave outputs
With this method it would be very easy to generate
other wave shapes to evaluate performance of
converters including crest factor.
16th July 2013
NCSLI 2013, Nashville
33
Power and Phase…..
Future development of this technique could be to
add additional converters to allow for several
outputs. Up to 6 outputs could easily be provided for
3 phase simulation of power, with an accurate
phase, or delay between the outputs.
If all the outputs were at the same 7V level this
would provide a very affordable solution for a phase
and power reference.
16th July 2013
NCSLI 2013, Nashville
34
References
[1] R. Kinard, and L. A. Harris “ A digitally synthesized sine wave
source with +/- 1 ppm amplitude stability. “ in Euromeas ’77 Conf
Digest. Institute Elect. Eng. Conf. Pub no 152, Sept. 1977
[2] Oldham N., Hetrick P., Zeng X., “A Calculable, Transportable Audio
Frequency AC Reference Standard”. IEEE Transactions on I & M, April
1989, Vol. 38.
[3] Oldham N., Bruce W., FU C., Cohee A., Smith A., “An
Intercomparison of AC Voltage Using A Digitally Synthesized Source”.
IEEE Transactions on I & M Vol. 39 No. 1, February 1990
16th July 2013
NCSLI 2013, Nashville
35
Conclusions
Digital synthesized waveforms can
provide a low cost and easy to use solution
for DC to AC voltage transfer.
Copies of our paper and presentation are
available on our booth on USB memory
sticks
16th July 2013
NCSLI 2013, Nashville
36