Idea: open source DMM/audio oscilloscope

[NiHaoMike]

Even the expensive multimeters seem to lack features that are easily feasible today. Even stuff as basic as graphic LCDs and display of multiple values at once. So how about design an open source multimeter?

A preliminary list of features:
*DC-20kHz bandwidth (That means a sampling rate of about 200kHz, very easily attainable.)
*+-32768 counts (16 bit) or greater at DC
*LCR measurement
*1kV CAT III rating
*separate voltage, current, and impedance connectors, with connector checking
*"soft" pushbutton controls (That is done to allow remote operation if equipped.)
*Transflective color LCD with LED backlight (backlight off by default, maybe with ambient light sensor to automatically set backlight)
*Bluetooth or optoisolated USB (optional)
*Lithium ion or NiMH battery with integrated charging (NiMH is more common in stores, but Lithium ion offers superior energy density)
*target price $50-$150 (parts only) depending on features

And here's a list of how some things could be implemented:
*ARM CPU, perhaps a Cortex M3. Firmware and user data is stored on a SD card in the battery compartment, allowing the user to easily swap it.
*The input voltage divider could consist of a precision high voltage resistor from the input terminal to the buffer, with precision resistors to ground switched in and out by electronic switches and a pair of large zener diodes for transient protection.
*Hall effect current sensing for extremely low insertion resistance on high mA and A ranges, active I-V converter for uA and low mA. A relay switches between the two, with a pair of shunt diodes to ground and a HRC fuse for protection.
*24 bit ADC, so calibration can be handled digitally (Dual channel 24 bit, 192kHz ADCs are common for audio. One channel could be connected to the voltage divider circuit and the other to the current sense circuit, allowing for simultaneous voltage and current measurement.)
*The impedance input protection could consist of a HRC fuse, a pair of doubled shunt diodes (or other overvoltage protector) to ground, and fusible resistors to the current source and voltage sense. The doubled (or tripled, etc.) shunt diodes allow for enough output voltage for semiconductor testing.
*The power supply could be integrated, accepting 100-240v AC or 140-340v DC. Alternatively, an isolated DC/DC converter could be integrated, accepting 12v from an external power supply.
*The software, of course, would be open source so any feature possible with the hardware could be implemented.

[EEVblog]

As always, the devil is in the detail, and there are an awful lot of details in a project like that. That's why no one has ever really bothered.
But don't let me stop anyone!

Dave.

[NiHaoMike]

And i'm pretty close to coding USB Communication via FTDI FT2322 too. hahah But only if you find it valuable.

I'm thinking Bluetooth since it is inherently isolated.

-Color LCD (poor battery performance though).

My iPod has a 2.5" color LCD and its battery life is good (with backlight off) even though the battery is small (IIRC, it's 450mAh). The meter will use something bigger, perhaps 2200mAh since that's the capacity of a laptop-type battery cell. Or some AA NiMH batteries, which are commonly available as high as 2700mAh.

Potentially, E-ink could be used, but what's available is too slow for a general purpose meter.

[KTP]

I thought briefly about making a multimeter type module with a bluetooth interface.  The module would not need buttons or a display, these would be handled by a laptop or smartphone that is paired to the device.   A smartphone would have a display with enough resolution to show nice waveforms and enough processor power and memory to do about whatever you want to the data it would receive over the bluetooth connection.  I might play around with this during the summer since I am not taking an EE course then.

[tecman]

Nice idea, but a lot tougher than you might think.  Your specifications/concepts are very tough:

- 16 or 24 bit converters at 200 KHz are not cheap if they are truly accurate to the desired level.  There are faster A/Ds on the market, and they may be monotonic, but actual accuracy is lower that you might want

- Input divider will need capacitive dividers as well to preserve accuracy at the frequencies mentioned

- Hall effect sensors are a nice idea, but temperature effects and drift limit the accuracy to a few percent, well short of the goal

- Electronic range switching is available, but JFET or CMOS switches have 10-300 ohms on resistance, so the divider design is a challenge to try to maintain the level of accuracy you indicated.

Actually the display, communications, etc that you mention is the easy part.  High accuracy, wide bandwidth (relative speaking) measurement circuits are a real challenge.  Your 16 bits is 15 uvolts on a 1 volt range. Nearly all devices, even resistors get horribly expensive to handle these accuracies, and have good temperature characteristics.  This is why lab meters of this quality are several thousand dollars.  You will see a surprising amount of discrete components and circuitry to meet these requirements, as integrated circuits often run into trouble.

paul

[NiHaoMike]

Nice idea, but a lot tougher than you might think.  Your specifications/concepts are very tough:

- 16 or 24 bit converters at 200 KHz are not cheap if they are truly accurate to the desired level.  There are faster A/Ds on the market, and they may be monotonic, but actual accuracy is lower that you might want

Virtually all meters are most accurate at DC, which obviously only needs a few Hz sampling rate.

- Input divider will need capacitive dividers as well to preserve accuracy at the frequencies mentioned

Not if you derate the accuracy for AC. 20kHz is only audio frequency, so capacitance shouldn't be very significant.

- Hall effect sensors are a nice idea, but temperature effects and drift limit the accuracy to a few percent, well short of the goal

How do the clamp meters do that? Do they just deal with the reduced accuracy?

- Electronic range switching is available, but JFET or CMOS switches have 10-300 ohms on resistance, so the divider design is a challenge to try to maintain the level of accuracy you indicated.

Discrete MOSFETs easily get less than 1 ohm of resistance. Although they might be too capacitive for our uses.
Or perhaps instead of a voltage divider, have the input resistor connect to an active I-V converter. A multiplexer can select the scaling factor of the converter in such a way that the resistance of the multiplexer is automatically compensated for.

can it survive in a month for a single charge? for normal usage? or backlighting off? how long?

I think it's something like 14 hours of normal usage, which is with the backlight off. 14 hours should be plenty for lab use (when was the last time you were constantly working for 14 hours in a lab?), although it could be an issue out in the field. Perhaps other power sources like solar cells could be added?

[tecman]

Nice idea, but a lot tougher than you might think.  Your specifications/concepts are very tough:

- 16 or 24 bit converters at 200 KHz are not cheap if they are truly accurate to the desired level.  There are faster A/Ds on the market, and they may be monotonic, but actual accuracy is lower that you might want

Virtually all meters are most accurate at DC, which obviously only needs a few Hz sampling rate.

- Input divider will need capacitive dividers as well to preserve accuracy at the frequencies mentioned

Not if you derate the accuracy for AC. 20kHz is only audio frequency, so capacitance shouldn't be very significant.

- Hall effect sensors are a nice idea, but temperature effects and drift limit the accuracy to a few percent, well short of the goal

How do the clamp meters do that? Do they just deal with the reduced accuracy?

- Electronic range switching is available, but JFET or CMOS switches have 10-300 ohms on resistance, so the divider design is a challenge to try to maintain the level of accuracy you indicated.

Discrete MOSFETs easily get less than 1 ohm of resistance. Although they might be too capacitive for our uses.
Or perhaps instead of a voltage divider, have the input resistor connect to an active I-V converter. A multiplexer can select the scaling factor of the converter in such a way that the resistance of the multiplexer is automatically compensated for.

can it survive in a month for a single charge? for normal usage? or backlighting off? how long?

I think it's something like 14 hours of normal usage, which is with the backlight off. 14 hours should be plenty for lab use (when was the last time you were constantly working for 14 hours in a lab?), although it could be an issue out in the field. Perhaps other power sources like solar cells could be added?

Even at audio frequencies, the capacitance of a resistive divider can result in errors beyond your desired specs  Even the few PF at audio frequencies will cause errors with a 1 meg resistor.

As far as the DC accuracy, you can average readings, but that only works on DC or LF.  Again you want 20 KHz with no mention of lower accuracy.

Low resistance MOSFETS are plentiful, but generally not suited for signal level operation.  Bipolar signals and non-linearaties at very low signals present an issue.  The idea of compensating for the resistance is also an issue as the on resistance varies with several factors.

Lastly clamp meters are mostly AC and use transformer coupling.  If you look most are only rated at 2% accuracy.  Hall effect is usable on DC, but again the accuracies are again around 2%

I am not trying to beat you up here, but just to try to have you see that your dream project may be a dream.  If these things were as easy as it seems, they would be readily available commercially.  In my work, we see a lot of people overspec their requirement to beyond what is practical or cost effective.  One of the biggest jobs is to bring reality into the picture.  Most of your goals are achievable, but not without considerable complexity and cost.

paul

[alm]

I agree 100% with tecman, it's pointless to talk about trivial details like the communication interface before you got the basics (analog part) down.

Virtually all meters are most accurate at DC, which obviously only needs a few Hz sampling rate.

Some bench DMM's have an adjustable rate setting, to trade off resolution versus speed.

- Input divider will need capacitive dividers as well to preserve accuracy at the frequencies mentioned

Not if you derate the accuracy for AC. 20kHz is only audio frequency, so capacitance shouldn't be very significant.
[/quote]
Unless it's in parallel with resistors in the megaohms. For a one megaohm resistor, 8pf of parallel capacitance would half the impedance at 20kHz.

- Hall effect sensors are a nice idea, but temperature effects and drift limit the accuracy to a few percent, well short of the goal

How do the clamp meters do that? Do they just deal with the reduced accuracy?

The first AC/DC clamp meter I found (Fluke 336) specifies 2% accuracy, so yes . And it probably contains some tricks like temperature compensation to improve the accuracy of the Hall sensor. I'm not sure if I see the point in using a Hall sensor. At low currents, you can use something like a feedback ammeter. At high currents, you can just use a low-value shunt and a high-gain amplifier (like Dave's ucurrent).

- Electronic range switching is available, but JFET or CMOS switches have 10-300 ohms on resistance, so the divider design is a challenge to try to maintain the level of accuracy you indicated.

Discrete MOSFETs easily get less than 1 ohm of resistance. Although they might be too capacitive for our uses.
Or perhaps instead of a voltage divider, have the input resistor connect to an active I-V converter. A multiplexer can select the scaling factor of the converter in such a way that the resistance of the multiplexer is automatically compensated for.

What's wrong with relays? That's what all the big boys use, and they're actually concerned about switching speed (automated testing in system applications). If you want a wide input range (eg. 100mV full scale to 600V full scale), you need a wide range of gains (with constant bandwidth, or so much bandwidth that the roll-off is way beyond your frequencies of interest).

can it survive in a month for a single charge? for normal usage? or backlighting off? how long?

I think it's something like 14 hours of normal usage, which is with the backlight off. 14 hours should be plenty for lab use (when was the last time you were constantly working for 14 hours in a lab?), although it could be an issue out in the field. Perhaps other power sources like solar cells could be added?

I think a precision DMM and a scope are two very different instrument, the latter could probably do with 8 bits of vertical resolution, unless you want to do FFT.

[NiHaoMike]

What's wrong with relays? That's what all the big boys use, and they're actually concerned about switching speed (automated testing in system applications). If you want a wide input range (eg. 100mV full scale to 600V full scale), you need a wide range of gains (with constant bandwidth, or so much bandwidth that the roll-off is way beyond your frequencies of interest).

The main problem is power usage. A 5v signal relay from an old ADSL modem draws about 40mA when on. In contrast, even a 4.3" video game LCD only uses 10mA at 5v and 1.2mA at 2.5v. (http://www.sparkfun.com/commerce/product_info.php?products_id=8335) There might be ultra low power relays out there, which could be considered.

I think a precision DMM and a scope are two very different instrument, the latter could probably do with 8 bits of vertical resolution, unless you want to do FFT.

FFT would probably be one of the features, given that the target bandwidth is designed for audio use.

Perhaps the voltage input resistor could be paralleled with about a 10pF or so capacitor (roughly the same as a scope probe), then each attenuation factor resistor paralleled with a capacitor of the appropriate value. Then perform frequency transformation in software to flatten the frequency response.