EEVblog Electronics Community Forum
Products => Test Equipment => Topic started by: PoorConduct on May 19, 2016, 08:17:47 pm
-
Since coming across the eevblog website I have developed an unhealthy interest in multimeter circuit design. What I am puzzled by at the moment is the impressive AC performance of some multimeters. If I take my trusty Fluke 287 it has an almost flat frequency response to beyond 100kHz. How is this possible with a 10M input resistance? Even a relatively small amount of capacitance after the 10M will cause roll-off.
As far as I can see the divider chain does not seem to have any compensation capacitors across it, so how is the frequency response so flat?
Looking at the circuit diagrams of various other Fluke meters I see that in older models they did use some compensating caps and even some manually adjusted trimmers to improve the frequency response. This is clearly not how it's done now. Just running a simple simulation with a 10M + 1M divider and 50pF of stray or input amplifier capacitance and it is starting to roll-off at only 1kHz.
So my question is how is this dealt with in modern multimeters as I assume it is neither magic or voodoo?
-
What's the 50pF from? That seems an order of magnitude high...
-
What's the 50pF from? That seems an order of magnitude high...
I'm guessing he invented it for his simulation.
I haven't looked into this at all but it's perfectly possible there's an op-amp to buffer the signal and that the 10M input impedance of the meter's front end is simply the op-amp's input.
-
I just guessed what I thought might be a sensible figure but even if we go for 5 puff we are still rolling-off by 10kHz, so my question still stands.
In answer to Fungus, the 10M is usually in the form of a precision laser trimmed divider network specifically design for use in meters. This is the sort of thing used:
http://www.caddock.com/Online_catalog/networks/networks.html (http://www.caddock.com/Online_catalog/networks/networks.html)
Here's another question: If there is no introduction of any deliberate capacitance then how does the multimeter have some immunity to RF EMI? Once it has got into the first active stage it will be demodulated and cause DC errors, even worse it could get into the digital stages and cause chaos. So how do they make it EMC compliant?
-
That's a good question. Without a schematic for the meters you are looking at, I can't really comment on them. I have looked at some of the meters I have tested and in one case I noted a ferrite in series with five Meg resistors. In parallel with the five resistors was a single cap. Sketch out the meter you are interested in and lets have a look or point me to a schematic.
f I take my trusty Fluke 287 it has an almost flat frequency response to beyond 100kHz. How is this possible with a 10M input resistance? Even a relatively small amount of capacitance after the 10M will cause roll-off.
...
As far as I can see the divider chain does not seem to have any compensation capacitors across it, so how is the frequency response so flat?
So my question is how is this dealt with in modern multimeters as I assume it is neither magic or voodoo?
-
You only need ~500K series resistance to withstand 400V DC without creating too much heat in the resistors.
Only in the high voltage ranges will a large amount of series resistance be switched in as part of a divider.
-
In answer to Fungus, the 10M is usually in the form of a precision laser trimmed divider network specifically design for use in meters. This is the sort of thing used:
http://www.caddock.com/Online_catalog/networks/networks.html (http://www.caddock.com/Online_catalog/networks/networks.html)
OK.
Here's another question: If there is no introduction of any deliberate capacitance then how does the multimeter have some immunity to RF EMI? Once it has got into the first active stage it will be demodulated and cause DC errors, even worse it could get into the digital stages and cause chaos. So how do they make it EMC compliant?
But there could still be a voltage follower/buffer after that to boost the current and allow for a bit of capacitance.
-
How do passive oscilloscope probes work with 10 MOhm input impedance? Understanding that will take you a long way.
Also look carefully at the specifications for the meter you are looking at, many have 1 MOhm input impedance for AC, not 10 MOhm. Your figure of 50 pF is quite realistic though, and maybe too low, despite suggestions by others here to the contrary. Consider where that capacitance actually lies though... is it all at the bottom of the resistive divider?
100 kHz is nothing, my Keithley 2001 (bench meter) is dead flat to over 6 MHz, and -3 dB around 8 MHz. Its AC input impedance is speced at 1 Mohm // <140 pF.
-
100 kHz is nothing, my Keithley 2001 (bench meter) is dead flat to over 6 MHz, and -3 dB around 8 MHz. Its AC input impedance is speced at 1 Mohm // <140 pF.
That's impressive. I measured the -3dB point of the surviving handheld meters I have using a 1Vrms sine wave signal. Generator to 58U to 50ohm terminator to BNC/banana adapter. From lowest to highest:
UNI-T 90A, 2.3KHz
UNI-T 210E, 2.94KHz
Hioki 4253 (RMS), 3.16KHz
AMPROBE 530 (RMS), 4.4KHz
AMPROBE 510, 6..240KHz
Fluke 17B+, 6.7KHz
Mastech 8229, 7.3KHz
Fluke 115 (RMS), 7.77KHz
Fluke 107, 8.09KHz
Fluke 101, 10KHz
Brymen 869s (RMS), 474KHz
UNI-T 181A, 1.35MHz
All of these meters are in cal.
:-// Then again, I was thinking for over a KHz the scopes, SA and VNAs would be my tool of choice....
-
The 10Mohm is not in series with the input capacitance, it is in parallel with it, so, it has no rolloff effect. The driving impedance of the source of the voltage being measured is the thing that drives the input capacitance, and that creates a lowpass filter, not the load resistance.
-
You need some series resistance so you don't fry it when you put mains into it while it's on a low voltage setting.
-
The AC voltage range on many modern multimeters has a 10M input resistance. Sometimes the AC mV range may only have 1M. If you look at the attached diagram of the Fluke 17B front end you will see this. I have highlighted the main ACV path in blue, the range dividers in red and the protection components (for the ACV range) in yellow.
As far as I can see there are no EMC filtering components and no frequency compensating caps. I do not know the bandwidth of the 17B, I don't have one to test. However the spec of the ACV range in the manual is only to 500Hz and the input impedance is 10M//<100pF so maybe this particular model does not have a good frequency response.
I cannot find a circuit diagram for the Fluke 287 so I have sketched out the front end by taking mine apart. I think that it is very similar to the Fluke 77 that I have found a schematic for and attached. Certainly the Fluke ASIC appears to have the same part number looking at photos on the web.
I have no idea of the value of the Fluke ASIC input capacitance but I suspect that they will have done their best to minimise it. However the divider switches will all have some capacitance even if the charge injection is minimised at the cost of series resistance in the FETs. The buffer amp and the lead frame will all have capacitance and this will all be presented AFTER the 10M resistor.
I would love to spend a week in the Fluke design office although I fear that they would throw me out before then with all my incessant questions ;-)
-
I wonder how close the 17B is compared to the 17B+ at 6.7KHz. I have no idea about the 77. Only two of the meters I looked were above 10KHz.
I am surprised, 2Kish through a cap, right into the IC? :phew:
-
My apologies, the cap does not connect directly to the IC. I haven't got the meter in bits any more so I cannot see where it goes. I am guessing that it provides an AC path to the 10M divider chain. I am guessing that it is more like the attached diagram where the chip selects either the AC or DC path.