Community Bench Meter

[desolatordan]

As I said, I think our major concern should be designing a usable ADC!

The userinterface and interfacing with other stuff is not really complicated.

I don't have much time on my hands now but I will look into reverse-engineering and drawing schematics for something that is based on the Keithley 196. This should provide a good basis from which we can work on...

And I agree, only fancy graphical interfaces would require a bigger ammount of computation time, but people get along with just 8-seg displays

I agree, any of those microcontrollers would work fine. But because the analog front-end shouldn't be limited by the microcontroller, analog design can be developed parallel to microcontroller and user interface. I say whoever is most inclined to start programming with their favorite microcontroller be the deciding factor.

Get started with:
isolated external communication (ir, usb, wireless, ?)
physical user interface
graphical interface

[Mastro Gippo]

The main argument I see for using a 32-bit micro (ARM) is a graphical LCD (or OLED ). This might be nice for large numbers (your standard character LCD is a bit small for bench DMM IMO, and 7-segment is less flexible), dual display (like the Fluke 45) or graphs (both bar graph and histogram). I believe 8-bit micro's are a bit short on memory for this task (unless you go for the largest parts that costs more than a basic ARM).

*aehm*.. http://rossum.posterous.com/teeny-avr-media-thing-the-nanotouch-0
 

[GeekGirl]

If I want 32 Bit, I would go to the AVR32 line before ARM, but that is only due to the fact I have no experience with ARM although I am prepared to jump in an learn

[alm]

*aehm*.. http://rossum.posterous.com/teeny-avr-media-thing-the-nanotouch-0
 

Yeah, as long as you can avoid using a frame buffer it can work.

It would be possible to use two different micro's for measurement and control. The HP 3478A does this: one micro connected to the ADC, and one to the buttons/display/GPIB. I don't think it would be a good idea to mix two different micro families here, since only people that know both would be able to hack it, but it's possible.

I would isolate the signals as soon as possible, as opposed to isolating USB/RS232/Ethernet. Commercial meters are often rated for about 500V between input low and ground (and more between input high and ground). The less area you have to protect, the easier (i.e. connecting buttons to this potential would mean your buttons have to be rated for 500V). It's not practical to isolate the analog signals, so I'd put the isolation device right behind the ADC (SPI is easy to isolate, there are devices for I2C), or behind the front-end micro if we go that way. Relays are isolated, not sure what the isolation rating of small-signal relays is.

The analog part is definitely the hard part, and doesn't depend on the digital part, so it seems a good idea to start there.

Alson

[xani]

If I want 32 Bit, I would go to the AVR32 line before ARM, but that is only due to the fact I have no experience with ARM although I am prepared to jump in an learn

Well both are supported by GCC so probably biggest difference would be how to configure things like SPI/UART etc., probably some code can be even ported from avr ^^

[_Sync_]

After reading a bit, I think that we might get around using this chip for a start, it should provide satisfactory accuracy.
I'm still reading into integrating ADCs, but it seems that getting reliable results will require quite some effort. I will start experimenting as soon as I get the FPGA project for school done...

I'll keep you posted, if I get some progress!

[alm]

After reading a bit, I think that we might get around using this chip for a start, it should provide satisfactory accuracy.

From a quick glance at the datasheet, this device doesn't seem much better than the ICL7106 used by all cheap multimeters, just 2000 counts (decimal).

I'm still reading into integrating ADCs, but it seems that getting reliable results will require quite some effort. I will start experimenting as soon as I get the FPGA project for school done...

Building a high-accuracy ADC is probably quite hard. If we have to use an off-the-shelf ADC, there are plenty available with a higher resolution. A quick parametric search at Digikey shows several 24-bit ADC's below $10 (eg. AD7192), although getting 24 bits of usable data out of them won't be easy either. Not sure what the important specs in this application are, it may well be that what required a custom circuit in the eighties can be easily solved with modern parts.

Alson

[_Sync_]

Well the designs don't use regular ADCs because they are quite unstable and not very accurate in the low bit levels.

Yep, the chip isn't that accurate, I just had a quick looks in a coding break earlier.

I'll see what I can achieve with some crude design and then refine from there

[walterdelbono]

   ...woooooooooooooooooooooooooooooooooooooooooooooooooooooow...  

what an awesome topic god saint...

first of all i want to thank everybody for the information this topic is amazing...

i hope you finish this project some how because i think that it is one of the most wonderful things i found in internet...

and i want to build my own "eevblog community bench meter"

good luck...

[SockThief]

Hej everyone,

been a bit busy lately, so ive not been here much, and so i appologise if this has already been posted, but its well worth a look anyway

http://www.gadgetfactory.net/gf/project/butterflylogic/

cheers
SockThief

[GeekGirl]

Hi everyone,

I have not forgotten this project, just with the first week of Uni and a few urgent repairs, I have been waylaid. On Friday I hope to be able to do up a block diagram, so that we can split up the project in to manageable chunks

If anyone has any ideas, or has started anything on this. speak up

[free_electron]

Take a look at a 7129 for example. 4 1/2 digit dual slope converter.

Other solution may be to brew your own. These things are very sensitive and difficult to maintain. one good one is an ADuC624. these have 24 bit a/d convertor and integrated cpu that is running in time step with the convertor
( cpu sleeps while conversions is done).

The advantage is that you can do autoscaling algorithm and relay ( range select) on the cpu. You spit out your measurement through the built in uart , go over 2 optocouplers into the backend processor ( ground referenced electronics)

The backend processor handles user interface ( display / keyboard / usb interface / ethernet interface / gpib / whatever.

the measurement board has its own transformer winding powering it. communication goes between the measurement block and the main electronics through optocouplers. That was you have galvanic isolation. Your measurement electronics must be floating !. Otherwise there is no point starting this development at all. you will have groundloops all over the place.

if you want to learn how to make good input circuitry ( scalers / protection / current sense / true rms whatever ) go to agilents website and download the servicemanual for the 34401. they have fll schematics published. You won't be able to copy since there are a couple of asic's in it, but you can learn a lot from it.

i see a lot of posts in this topic about cpu selection. that is all irrelevant. the hardest part is getting the analog done right. You need to come up with a list of requirements there

voltage range for DC and AC ( flashover protection )

current range for AC and DC ( overcurrent protection. current sense resistor and gainstage ...

resistance range .. ( 4 wire ? ) protection against users that want to measure the internal resistance of their wall socket while it is live ...

Diode check ( use current source with voltage scaling so you can test zener diodes )

continuity check

Frequency count ?

capacitor / coil test ? ESr metering ?


I myself don't care about being able to measure 1 kilovolt and 10 ampere. give me something that is accurate in the 0 to 5 volt range with 3 useful digits after the comma.
current wise i'd like to see microamps to 1 amp range..

and then there is another biiiiiiig question. How do you calibrate this thing ?

The measurement logic should NOT use floating point anywhere. everything should be scaled integer math, preferrably even bcd encoded.
you can solve some of the problems by selecting the correct reference for tha a/d . analog devices has the REF19x family. there are reference diodes that produce exactly 2.048 and 4.096 volt. ideal to hook up to a/d convertors since they prodice directly  binary weighted scales.

feed that with a stable suppy from a current source (lm319 for example. that thing has an oven on board ) and you are now very accurate. You can self calibrate the inl dnl of the on board a/d and compensate with some smart software.

for current sensig you will need a kelvin contact setup with a gainstage. susumu makes great kelvin resistor in the 20 milliohm to 50 milliohm range with 0.01% accuracy. for half a dollar you have one.

sense the drop over it using something like an AD8220 instrumentation amp and you can get somewhere..

These are the first things to consider. the instrument stands or falls around the aquisition section. get that one right and the rest is smooth sailing.

[dds]

Hi GeekGirl, there is already a block diagram at the TI site:

http://focus.ti.com/graphics/blockdiagram/blockdiagram_images/6279.gif (for handheld DMM)

http://focus.ti.com/docs/solution/folders/print/591.html

http://focus.ti.com/docs/solution/folders/print/593.html (bench DMM)


I don't know if you have already seen those docs, if yes - sorry

[_Sync_]

Now that my FPGA project for school is over (well unfinished but that's another story) I am able to more on this project.

From what it looks like a dual-slope ADC is the standard used in common multimeters. I can understand that, because it cancels possible errors through integration. The issue with those ADCs is that you need very good parts in order to get high resolution and accuracy.

I found that todays common ADCs e.g. delta-sigma ones (e.g. LTC2410) should be able to produce a acceptable performance over a period of time.

The LTC2410 in particular is a 24 bit delta-sigma ADC (with even five more bits after the LSB!) with SPI interfacing and good accuracy. Everything is contained in a SSOP-16 package, which is well, quite small but it should be possible to use.

After I got my hands on this guy I'll see what I can do on the input side.

TrueRMS should also be possible with the help of the AD637 that seems to be widespread in this application.

AD637 Datasheet
LTC2410 Datasheet

[logictom]

Read this thread couple of days ago, any news on the specs being written up? Interested to see where this goes

[gbowne1]

I love the bench meter design, but you could optionally add a remote panel in handheld format or make the remote panel moutable.

Make sure you have a good input protection as you don't want to have a bench meter explode on ya either.  Dave, your video blogs on that were 'spot on'. 

[Bloch]

When it comes to the enclosure I am going to suggest that we use Farnell / Newark, this is so that we can all have everything FIT, Front Panel overlays the right size etc.

When doing a project like this I like to start with the form factor and user interface first and how that's all going to work, and then work backwards to the detailed schematic. Because a schematic does not a project make.

http://www.hammondmfg.com/
http://www.serpac.com/
http://www.polycase.com/
are some of my favorites.

Also, I'd consider a budget too, too high a price and it's going to be a show-stopper, so cost should always be factored in.

I like the hammond 1455 series. But I dont think they can "stack" or have a option for a foot.

So that about bopla.de the Ultramas series have a folding foot set =21899]http://www.bopla.de/index.php?id=228&L=0&tx_commerce_pi1[showUid]=21899

And Farnell have them.

[Psi]

Its probably been said before, but i dont have time to read 8 pages to cheak

It would be cool to have duel,tri or even quad input. Being able to plug in 2,3, or 4 sets of probes (isolated from each other) and measure multiple things at once.
Not only can you then do 4 wire resistance measurement like in the recent blog but you can program the mcu to do lots of other interesting things too like...

* measure volts and current and the meter displays watts, as well as all the other combinations of V= IR  and P = VI
* measure temperature and current and the meter displays the relationship
* measure two voltages and two currents and the meter displays the input power/output power and efficiency of a power supply

etc..

[mindspan]

First off, let me say I ran across this forum and site while looking to see what others were doing in the way of building a bench DMM, and I am bookmarking it as a new interesting source.

I end up doing a lot of traveling for work and have been wanting to put together a combined set of test instruments in a single bench package for when I travel. One of the key pieces will be a DMM since I hate having to leave my HP 3457A behind when I travel (it has actually come with me around the world at least once). Anyway, i figured I'd share where some of my thought processes on this type of project have gone.

1. Case- I agree with many on here that have discussed picking a form factor and building around it. I find this is key to not later having to try and stuff things in badly or ending up with huge cases. The simple act of defining a board or panel outline with constraints placed by the enclosure is to me very helpful. As to case manufacturers I have used most that have been listed before and I would add one more. For handhelds (although bench is definitely my preference I build a lot of hand held prototypes) I recommend Box Enclosures. My preference for them is they are one of the few that have rubber boots for most of their handhelds.

2. Design style -  normally I try to make things as modular as possible, especially on prototype or low volume things. I actually was just looking at a spectrum analyser design that takes this to an interesting level http://www.scottyspectrumanalyzer.com/slim.html is a page on their modular approach; essentially small modular boards with mounting solder pads on the edges that can be tacked to a large pcb with cutouts for a grid of the smaller pieces.

3. Included pieces - Personally my goals is to include a lot of different pieces of test equipment into on travel package, so not really applicable to the goals you have here, however I would suggest maybe adding one thing that for me is a big part of the design. I plan on adding a digitally controlled DC power supply since a supply is the second most used piece of test equipment on my bench after the DMM (maybe first). I was looking at doing something similar to http://www.tuxgraphics.org/electronics/200707/bench-power-supply-unit.shtml but using whatever micro I settle on for the DMM and other functions. For my work the biggest upside to this is being able to see input voltage, current, and power all at once, and I will add logging since I build a lot of wireless battery powered items and need to see average power usage logged over various power states (sleep, transmitting, etc.)

4. MCU - For the DMM section I have narrowed my choices down to 3 possibilities. The first two are basic - ADUC845 or similar from analog (similar to the ADUC7060 mentioned before but 8052 based), one of the new PSoC's based on 8052 from cypress. Both of these offer precision ADC on chip with PGA's and MUX's as well as plenty of DIOs for controlling input, range and measurement type switching. The larger PSoC's also offer various filters and opamps on board, USB, and lots of flexibility for pinout and pin function  (nice if you want to just break out a block of pins to cover various functions dependent on code). Partially driving these choices is that I have dev kits for both, and in the case of the aduc I have considered just integrating the dev kit board into the meter and saving a lot of design on the analog layouts. PSoC also has a new high precision analog dev kit that would make a good starting point for the meter.

My third possibility is something interesting if you are looking for very high precision. TI has a 32bit adc chip that they designed for very high precision measurements (ADS1282) while I personally would not even attempt to integrate this chip due to the LVDS signalling and need for a DSP to read it, they have a reasonably priced dev kit (http://focus.ti.com/lit/ug/sbau144b/sbau144b.pdf) that includes a USB interface and sample program to do voltage measurements ($170 from mouser, not CHEAP, but considering the cost of 6+ digit bench meters a bargain compared to commercial meters). This board would be an interesting basis for a 6+ digit meter that has all the analog backend and computer interfacing done and just needs the front end range switching and whatnot developed. I am still strongly considering this as I do a lot of very low level measurements from sensors.

Anyway, I will be following this thread with interest as I think having more open source test equipment designs for hobby/small business electronics is a great idea. Computing power and precision measurement IC's are to a point now where there is not much test gear that can't be built on the cheap.

-Aaron

[Zad]

No onboard ADC is 'precision' or 'low noise' no matter how many bits or nV of noise they claim. It just isn't possible. You might like to compare the data sheets for the TI 32-bit ADC to their 24-bit parts. The 32-bit parts are several bits worse than the best 24-bit ADCs.

[Bloch]

An onboard ADC will also lock the project to a single cpu.

The more modular the better.


Front end
adc
cpu
IO (to PC)
Display

[Neb]

I don't see a need for a 31-bit adc since they are so expensive.  I think a 24-bit adc would be more than enough and the saved money can be spent on better input circuitry.  The real culprit will be getting the noise floor down and keeping it calibrated.  Ideally it would not need calibration, the references should be accurate and not drift.  If i try to calibrate it with a store bought DMM with 1% accuracy, i can't expect the device to produce better than 1% accuracy and the extra counts/adc bits are wasted.

[firewalker]

Is this project dead?

[Bored@Work]

Is this project dead?

It was dead right from the beginning. Because

a) It is difficult. Despite the wet dreams of some, a good meter can't be done with just cobbing together a cheap ADC with a cheap MCU.

b) Because of a) it requires a good amount of work.

c) Most people discussing here were after getting it without any own effort, in the hope all they had to do was to add to the wishlist. This behavior is not compatible with b).

d) Most people discussing were just after getting something cheap. This won't happen. One can't beat the price of cheap multimeters, and one can't bet the price of expensive multimeters. The cheap ones are, well, just too cheap. The expensive ones require expensive parts and expensive equipment, e.g. to adjust them.

e) There was and is no forceful project leader. Someone skilled enough and with enough time on his or her hands  to push the project, to make decisions, to attract contributors, to guide contributors towards the common goal (heck, a common goal wasn't even defined), and keep contributors working for the project, even if their opinions are overruled.

f) No one is or was willing to fund the effort. See d) and e).

Realistically, it never had a chance of taking off.

If you want a community project, fix a) ... f), and probably some more issues I forgot to mention. If you want an good chance of having a DIY meter, forget about the community project, do it completely on your own, and once done brag about it. The leechers at HackADay will welcome you with open arms and make you their hero (for a day).

[NiHaoMike]

It might not be worth the effort to build a high accuracy meter or an extra cheap meter, but there are some niche applications where it makes sense. Such as a meter that automatically calculates average and peak power. Or a datalogger that records voltages to a SD card at a sample rate of a few Hz or kHz. In any case, we'll make the assumption that the accuracy needed is comparable to or less than that of the under $20 meters and that the builder already owns a meter (and calibration reference) much more accurate than the one being built.

One extremely high accuracy measurement that is relatively easy to make is time/frequency. Put a GPS in your homemade frequency counter and you'll very likely have more accuracy than you'll ever need.