DMM GENERAL: bar graph filtering/responsiveness

[-KK-]

Hello everyone,
 
I recently bought a Brymen BM319s, my first multimeter with a bar graph display. I tested the function on the turn indicators of a car (i.e. 0-12V square wave @ approx. 1 Hz). After manual range setting, I expected to see the bars turning on and off all at the same time, consistently with the clear cut between levels of the square wave.
 
Conversely, what I see is pretty slow and smooth transitions, which almost take the complete half-period to stabilise on the 0 or 12 V value, resembling much more a sine than a square.
 
The bar graph seems to follow the slowness of the main display which, due to the usual averaging, displays 2-3 intermediate values even if the measured signal is a very clean square wave (scope-verified). I know that this behaviour is found in very low-end DMMs, which use for the bar graph the same 4-5 Hz refresh rate as for the rest of the screen. I do not expect it from a DMM over €100, and which explicitly states the refresh rate specific for the bar graph (the usual 40 Hz, BTW).
 
Is a filter applied to the bar graph display? If so, why is it set so much lower than that needed for basic antialiasing?
 
What's your experience in similar situations with DMMs from this and from other brands (e.g. the BM235 and 121GW)?

Thank you all for your help, today as well as in the past 

[rsjsouza]

A DMM bargraph is the equivalent of the needle in analog VOMs (multimeters). In other words, it replicates the movement of the needle when faced with a varying voltage on the DMM's inputs. In order to show a smooth movement, the bargraph needs to operate at a higher update rate when compared to the measurement itself (the numbers) - the ADC must run faster than the BM319S's 5Hz (5 updates/sec).

One caveat is that, for the bargraph to smoothly display a wide voltage variation that goes across several DMM ranges (0~20V on a 6000 count meter, for example), the DMM autorange must be disabled - otherwise the autorange and the bargraph will try to catch up with the input voltage and the reading will be very confusing.   

Another detail is that many cheap DMMs update the entire display (including the bargraph) at the same slow speed of the digits (3~5Hz typically), rendering a varying input voltage into a rough movement of the bargraph segments. This turns them useless for their original purpose.

The BM319S manual shows the 24 segment bargraph operates at 40 updates per second, which is quite suitable for this operation. 

[-KK-]

Hi rsjsouza, thank you for your answer.

One caveat is that, for the bargraph to smoothly display a wide voltage variation that goes across several DMM ranges (0~20V on a 6000 count meter, for example), the DMM autorange must be disabled - otherwise the autorange and the bargraph will try to catch up with the input voltage and the reading will be very confusing.   

As I said in the beginning, this behaviour happens with autoranging disabled.

Another detail is that many cheap DMMs update the entire display (including the bargraph) at the same slow speed of the digits (3~5Hz typically), rendering a varying input voltage into a rough movement of the bargraph segments. This turns them useless for their original purpose.

The BM319S manual shows the 24 segment bargraph operates at 40 updates per second, which is quite suitable for this operation. 

Indeed, as I mentioned I do not expect the BM319s (or any other DMM from Brymen) to be that type of cheap DMM. The smoothness of the bargraph motion makes it visually clear that the refresh rate is no doubt well above 20 Hz.


However, it is still not clear to me how the bar graph works.
Does the ADC actually work at 40 Hz, and feed all of the resulting data to the bargraph control circuitry?
Or does the ADC work at the usual 5 Hz, and for most of the 40-hertz "screen updates" the bargraph does not display actual data, but simply an exponential interpolation between consecutive samples (giving only the appearance of VOM-like mechanical smoothness)?

Are both options in use, with the first one being limited to DMMs of higher price?

I see how the manual mentions an "Update rate" of 5 per second, but that sounds more like the screen refresh rate (for the digits area) than the sampling rate the ADC is capable of. The two might be equal of course, I do not know the model and specs of the ADC in the BM319s.

[rsjsouza]

Hi rsjsouza, thank you for your answer.

One caveat is that, for the bargraph to smoothly display a wide voltage variation that goes across several DMM ranges (0~20V on a 6000 count meter, for example), the DMM autorange must be disabled - otherwise the autorange and the bargraph will try to catch up with the input voltage and the reading will be very confusing.   

As I said in the beginning, this behaviour happens with autoranging disabled.

Sorry, I missed that.

Another detail is that many cheap DMMs update the entire display (including the bargraph) at the same slow speed of the digits (3~5Hz typically), rendering a varying input voltage into a rough movement of the bargraph segments. This turns them useless for their original purpose.

The BM319S manual shows the 24 segment bargraph operates at 40 updates per second, which is quite suitable for this operation. 

Indeed, as I mentioned I do not expect the BM319s (or any other DMM from Brymen) to be that type of cheap DMM. The smoothness of the bargraph motion makes it visually clear that the refresh rate is no doubt well above 20 Hz.


However, it is still not clear to me how the bar graph works.
Does the ADC actually work at 40 Hz, and feed all of the resulting data to the bargraph control circuitry?
Or does the ADC work at the usual 5 Hz, and for most of the 40-hertz "screen updates" the bargraph does not display actual data, but simply an exponential interpolation between consecutive samples (giving only the appearance of VOM-like mechanical smoothness)?

Are both options in use, with the first one being limited to DMMs of higher price?

I see how the manual mentions an "Update rate" of 5 per second, but that sounds more like the screen refresh rate (for the digits area) than the sampling rate the ADC is capable of. The two might be equal of course, I do not know the model and specs of the ADC in the BM319s.

In the designs I have seen where a bargraph is involved, the ADC operates at a higher sampling frequency and the actual measurement (digits) is averaged, which guarantees stability by dampening noise.

However, I can see how a manufacturer could add some minimal averaging to the bargraph as well, so it gives that smooth feel to it - the issue then seems that Brymen may be adding too much averaging to your taste.

I don't have a Brymen, but the clip below shows the bargraph dampening between Keysight and Uni-T

https://youtu.be/Kreo_Ohq4YQ?t=1190

[joeqsmith]

I am not sure if I ever compared the bargraph on this meter or not.  The video is on-line if you are interested.  Yes, I would expect it to switch like you suggest in your first post.  If the video doesn't show it, it's certainly an easy test for me to setup if you want some idea how it compares with others I have looked at.



https://youtu.be/uY6ZcINF7L0

[joeqsmith]

I tried it on my BM319s using a 0-10V DC signal switching at 1Hz 50% DC (I suspect you meant 2Hz with a 50% DC, 1 second on 1 second off)  and as I suspected,  the segments turn on at the same time.    If I set the range to the lowest voltage setting (to force an over range), all of the segments turn on and off together.    Any filter is higher than 1Hz.   

Maybe your signal is actually floating but if you are looking right across the directional lamp, I would expect your results to be the same.    Post a video clip of it if you can. 

[David Hess]

As far as update speed, good meters update the bar graph much more often than the main display.  Both of my multimeters which have bargraphs flicker the display when measuring 60 or 120 Hz AC or pulsed DC.

One caveat is that, for the bargraph to smoothly display a wide voltage variation that goes across several DMM ranges (0~20V on a 6000 count meter, for example), the DMM autorange must be disabled - otherwise the autorange and the bargraph will try to catch up with the input voltage and the reading will be very confusing.

As I said in the beginning, this behaviour happens with autoranging disabled.

My Beckman multimeter which has autoranging and a bar graph does *not* allow autoranging to be disabled.  The manual range can be set all you want but it only affects the display and autoranging  of the input is always occurring.

I only discovered this when trying to measure the input resistance on different ranges and the manual ranging had no effect.  My autoranging Tektronix multimeter does obey the manual range setting.

[-KK-]

My Beckman multimeter which has autoranging and a bar graph does *not* allow autoranging to be disabled.  The manual range can be set all you want but it only affects the display and autoranging  of the input is always occurring.

I only discovered this when trying to measure the input resistance on different ranges and the manual ranging had no effect.  My autoranging Tektronix multimeter does obey the manual range setting.

Beckman behaviour sounds strange, but still, interesting. In a scenario like this (slow on-off signal), I guess it would show an instantaneous drop to 0 volts, but a slight delay when the signal goes back to 12 volts, as the autorange needs some time to move from the mV range to the 20 V one.

Regardless of this specific case, I cannot see why they would want to do something like this. Kind of defeats the purpose of manual ranging, which should emulate the behaviour of a traditional, manual-ranging VOM or DMM.

[-KK-]

Minor update after some research about ADCs in DMM chipsets.

Unfortunately, the BM319s chipset (BTC0207) looks custom-made, so no data sheet is available to find out its ADC specs. Kudos to joeqsmith for showing the chipset in the video, so I did not have to take the time to fully dismantle my BM319s. Wonderful and very thorough review, by the way.

I checked the datasheets for two other DMM chipsets, the Hycon HY3131 (used in the 121GW) and the Cyrustech ES51990 (attached). Both chipsets have multiple ADCs, one is slow (5 Hz) and high resolution (50k or 6k counts, for the digits display) and the other(s) are fast (20-50Hz) but not as precise (feeding the bargraph). HY3131 also has a 3rd ADC for simultaneous measures (required for power measurement).

Note that, just like the 319s, even the 121GW manual doesn’t go into detail about the chipset, stating simply “nominal sample rate = 5 Hz”. Quite a waste not being able to perform 40-Hz acquisitions, trading a little accuracy for sample rate.

In the designs I have seen where a bargraph is involved, the ADC operates at a higher sampling frequency and the actual measurement (digits) is averaged, which guarantees stability by dampening noise.

Given the wide difference in sensitivity (6000 vs. 600 count for the ES51990), I do not think it makes sense (metrologically) to average the fast ADC readings to achieve a more stable reading in the digits portion, as the data sources differ. A single ADC having both high-sensitivity and high-speed  (+ software averaging) would appear to be the best option, but clearly there’s some technical difficulty in achieving this economically.

I’m no EE, maybe someone more expert on this matter can shed some light on the speed/accuracy/cost balance in ADCs? Oscilloscopes seem able to achieve decent accuracies at sample rates orders of magnitude larger.

[joeqsmith]

Minor update after some research about ADCs in DMM chipsets.

Unfortunately, the BM319s chipset (BTC0207) looks custom-made, so no data sheet is available to find out its ADC specs. Kudos to joeqsmith for showing the chipset in the video, so I did not have to take the time to fully dismantle my BM319s. Wonderful and very thorough review, by the way.

I checked the datasheets for two other DMM chipsets, the Hycon HY3131 (used in the 121GW) and the Cyrustech ES51990 (attached). Both chipsets have multiple ADCs, one is slow (5 Hz) and high resolution (50k or 6k counts, for the digits display) and the other(s) are fast (20-50Hz) but not as precise (feeding the bargraph). HY3131 also has a 3rd ADC for simultaneous measures (required for power measurement).

Note that, just like the 319s, even the 121GW manual doesn’t go into detail about the chipset, stating simply “nominal sample rate = 5 Hz”. Quite a waste not being able to perform 40-Hz acquisitions, trading a little accuracy for sample rate.

In the designs I have seen where a bargraph is involved, the ADC operates at a higher sampling frequency and the actual measurement (digits) is averaged, which guarantees stability by dampening noise.

Given the wide difference in sensitivity (6000 vs. 600 count for the ES51990), I do not think it makes sense (metrologically) to average the fast ADC readings to achieve a more stable reading in the digits portion, as the data sources differ. A single ADC having both high-sensitivity and high-speed  (+ software averaging) would appear to be the best option, but clearly there’s some technical difficulty in achieving this economically.

I’m no EE, maybe someone more expert on this matter can shed some light on the speed/accuracy/cost balance in ADCs? Oscilloscopes seem able to achieve decent accuracies at sample rates orders of magnitude larger.

Glad you found the 319s video helpful.   I've looked at a few of their products and they always seem to hold up well to anything I expose them to. 

I was recently thinking about about how I may be able to combine the data from the two ADCs.  Not an average as you suggest but some other way.  If you are interested, I have attached a link to a video where I explain what I did and show it in operation.   

https://youtu.be/e_YzwO62feQ?t=1248

[David Hess]

My Beckman multimeter which has autoranging and a bar graph does *not* allow autoranging to be disabled.  The manual range can be set all you want but it only affects the display and autoranging  of the input is always occurring.

I only discovered this when trying to measure the input resistance on different ranges and the manual ranging had no effect.  My autoranging Tektronix multimeter does obey the manual range setting.

Beckman behaviour sounds strange, but still, interesting. In a scenario like this (slow on-off signal), I guess it would show an instantaneous drop to 0 volts, but a slight delay when the signal goes back to 12 volts, as the autorange needs some time to move from the mV range to the 20 V one.

Regardless of this specific case, I cannot see why they would want to do something like this. Kind of defeats the purpose of manual ranging, which should emulate the behaviour of a traditional, manual-ranging VOM or DMM.

I do not know why they did it like this but I brought it up as something not to be taken for granted.

Similarly, do not assume that an autoranging meter has a constant input resistance; most do not.

Given the wide difference in sensitivity (6000 vs. 600 count for the ES51990), I do not think it makes sense (metrologically) to average the fast ADC readings to achieve a more stable reading in the digits portion, as the data sources differ. A single ADC having both high-sensitivity and high-speed  (+ software averaging) would appear to be the best option, but clearly there’s some technical difficulty in achieving this economically.

I’m no EE, maybe someone more expert on this matter can shed some light on the speed/accuracy/cost balance in ADCs? Oscilloscopes seem able to achieve decent accuracies at sample rates orders of magnitude larger.

In this particular case there is no tradeoff because the meter's display rate has nothing to do with the ADC's sample rate and has everything to do with noise.

The input is integrated over a whole number of power line cycles to produce differential and common mode rejection of power line frequencies which would otherwise produce an additional error term.  In practice this means integrating over units of 1/10th of a second to cancel both 50 and 60 Hz interference but the actual total integration time can be quite long to reduce noise for higher resolution readings.

There is no difference (1) between a single long integration and multiple short integrations so if the ADC is integrating over short periods, the bar graph can be updated at a high rate with lower noise limited resolution while the display is updated at a lower rate with higher noise limited resolution by combining multiple short integrations.

(1) Of course there is a difference but it has no bearing on multimeters which have a resolution of "only" 5-1/2 digits (200,000 counts) or less.

[-KK-]

In this particular case there is no tradeoff because the meter's display rate has nothing to do with the ADC's sample rate and has everything to do with noise.

The input is integrated over a whole number of power line cycles to produce differential and common mode rejection of power line frequencies which would otherwise produce an additional error term.  In practice this means integrating over units of 1/10th of a second to cancel both 50 and 60 Hz interference but the actual total integration time can be quite long to reduce noise for higher resolution readings.

David, are you saying that the slow ADC (the only one, for DMMs w/o bargraph) actually samples at frequencies where power line noise becomes relevant (>100 Hz), then does averaging/integration/whatnot, and finally outputs to the display at just 5 Hz?

I assumed there was an anti-alias filter in place on the analog side (before the ADC), which would also explain the response slowness that affects the digits reading. Even for high-speed ADCs (i.e. bargraph, typically 40 Hz), power line frequencies falls well upwards of the fs/2 cutoff for a filter like this, and thus their effect on the reading should be completely eliminated.

From what I saw about slow (i.e. hi-res) ADCs in the chipsets datasheets, integration is not mentioned, but low-pass filters are.

Let me know if I am mixing up two things that have nothing in common with each other, unfortunately I am not familiar with power line cycle integration.

[David Hess]

David, are you saying that the slow ADC (the only one, for DMMs w/o bargraph) actually samples at frequencies where power line noise becomes relevant (>100 Hz), then does averaging/integration/whatnot, and finally outputs to the display at just 5 Hz?

I am saying that it could.  Or it could sample at 10Hz providing full 50/60Hz rejection and update the display over longer integration time.  If it is sampling faster than 10Hz for the bar graph, then the added noise from power line interference will be insignificant compared to the bar graph's resolution anyway.

I assumed there was an anti-alias filter in place on the analog side (before the ADC), which would also explain the response slowness that affects the digits reading. Even for high-speed ADCs (i.e. bargraph, typically 40 Hz), power line frequencies falls well upwards of the fs/2 cutoff for a filter like this, and thus their effect on the reading should be completely eliminated.

That is not how it works because an anti-aliasing filter which could remove power line noise is difficult to make in the analog domain and slows settling time.  There may be a filter but the cutoff frequency is much higher and it just removes excess noise and RFI.

From what I saw about slow (i.e. hi-res) ADCs in the chipsets datasheets, integration is not mentioned, but low-pass filters are.

There are exceptions but the trick is that integrating (averaging) the input over 1 or a whole number of power line cycles creates a null in the frequency response at the power line frequency and its harmonics.  The actual frequency response is the non-linear sin(x)/x function.  If 1/10th of a second is used, then integration happens over 6 cycles at 60Hz or 5 cycles at 50Hz so both 50Hz and 60Hz interference are rejected and the first notch is at 10Hz.

Some meters have an option to select 50Hz or 60Hz rejection allowing them to operate at a faster sample rate of 50 or 60 samples per second without worrying about interference.  Bench meters usually detect the line frequency automatically and may even synchronize with it.  But these days usually 1/10th of a second is used to cover both 50Hz and 60Hz rejection.