CARC DVM Evening

 On 26th April CARC met for an evening on Digital Voltmeters. This was part of an occasional series that we are running on test and measurement. Mike G0KAD have a short talk on DVM and how they measure voltage using the dual ramp technique. A number of members bought along their voltmeter from their shack, mostly digital and a few analogue. Stewart G3YSX provided a number of "standards" that he built a while ago based on band gap reference ICs that provided a 5V output. We measured these voltage sources and recorded the results.

The measured results are tabulated below. However the first problem is to determine how good the references are. I  approached this by assuming that although out of calibration the four five-digit DVMs were likely to the most accurate. The average voltage reading for these DVMs is shown in the Average 4 (decimal places) row. I then had Excel calculate the average and the standard deviation. For two sources the standard deviation was 800uV and the third one (V2) had a standard deviation of 1,1mV. The standard deviation is a statistical method of calculating the range of values in which 2/3 of measurements are expected. Thus in the case of V1 there is a 66% probability that V1 was between 4.9989 and 5.0005v. Whilst I did not calculate it below there is a 95% chance a value being within two standard deviations of the mean, i.e. in the case of V1 between the values 4.9981v and 5.0013v.

In the table I have highlighted in green measured values that lie within one standard deviation of the particular  source.

Owner	Model	V1	V2	V3
M0IJP	Q1442	5.00	4.99	5.00
G4PEO	EEBLOG BM235	4.992	4.992	4.992
G4PEO	AVO8 MK3	4.910	4.925	4.925
G4PEO	FLUKE 179	4.999	5.000	5.000
G4PEO	UNIT UT61E	4.988	4.988	4.988
M0HQM	MAPLIN WG020	4.99	5.00
G3ZIY	7150	5.0005	5.0003	5.0007
G3YSX	ISO TECH IDM505	4.999	5.000	5.000
G3YSX	SDM3055	4.996	5.000	5.001
G3YSX	Sanwa U50D	5.0	5.0
G6RC	Sinometer	4.98	4.98	4.98
G0KAD	F77	4.98	4.98	4.98
G0KAD	F189	5.0003	5.0007	5.0011
G6RC	HP34401A	4.9994	4.9980	5.0000
M0WID	WG022	4.99	4.99	4.99
M0WID	FLUKE75	5.00	5.00	5.00
M0WID	AVO	4.95	4.95	4.95
G4ANN	F189	4.9985	4.9989	4.9991
Average 4		4.9997	4.9995	5.0002
Std Dev 4		0.0008	0.0011	0.0008
Average - Std dev		4.9989	4.9984	4.9995
Average + Std dev		5.0005	5.0006	5.0010

However it is difficult to draw significant conclusions from the above beyond noting that for  most practical benchwork all the meters tested were good enough, and only specialist  applications need anything better that a £20 eBay/Amazon/Bangood model. To understand things better I looked at the measurements of source V1 and looked at the  errors in a little more detail. I used a reference voltage of 4.9997 as determined above. I then sorted the table on the square of the percentage error.
Owner Model V1 Error (%) Error mv G6RC HP34401A 4.9994 -0.01 0 M0IJP Q1442 5.00 0.01 0 G3YSX Sanwa U50D 5.0 0.01 0 M0WID FLUKE75 5.00 0.01 0 G0KAD F189 5.0003 0.01 1 G4PEO FLUKE 179 4.999 -0.01 -1 G3YSX ISO TECH IDM505 4.999 -0.01 -1 G3ZIY 7150 5.0005 0.02 1 G4ANN F189 4.9985 -0.02 -1 G3YSX SDM3055 4.996 -0.07 -4 G4PEO EEBLOG BM235 4.992 -0.15 -8 M0HQM MAPLIN WG020 4.99 -0.19 -10 M0WID WG022 4.99 -0.19 -10 G4PEO UNIT UT61E 4.988 -0.23 -12 G6RC Sinometer 4.98 -0.39 -20 G0KAD F77 4.98 -0.39 -20 M0WID AVO  4.95 -0.99 -50 G4PEO AVO8 MK3 4.910 -1.79 -90 It is not surprising that the AVOs were at the bottom of the table. They are old and a  long time out of calibration, They are sensitive to their environment, their pre-measurement calibration  their exact orientation and to how well the mirror scale is used. Some of the meters at the top of the table are there because they did not "try too hard" that is the results  were reported to a lesser degree of precision.  There is a certain measure of you get what you may for in the results. All the meters except one AVO were better that 1% accuracy and the majority better than 0,2%/19mV in 5V. Stewart/G3YSX