The Multi-Ref is not a test device as such, but the tool I use to test my test equipment. I got the idea for this from the user [matseng] as the Dangerousprototypes.com forum, and found it very useful.
MrRef, as it was called, is just two coin cells powering a precision voltage reference. The intention is to allow checking your own multimeters for accuracy and drift. Not everybody has a 6.5 or better DMM at home, with a accuracy of 0.01% or even better. Back then my best DMM was a TekPower TP4000ZC, 4000 counts and 0.5% accuracy at the best, anything else was just cheap 3.5 digit DMMs, some of them more than 15 years old (but I got a UT61E shortly afterwards which claims 0.1% accuracy). So it was interesting to see how accurate the meters still are. This is something important - you want to know that you can still trust the stuff you measure.
So I built a small MrRef by myself, using REF5020 from TI. This is a low noise precision reference with 2.048V output. Its 0.05% accuracy result in at most 1mV error - one digit basically. But this value creates a problem: its too high to properly check a 2000 count DMM since these go only up to 2.0V.
So when I got my UT61E (two actually) I decided that I needed more than one reference voltage. The best would be to cover the range from 1 up to 10V so you can test at least two ranges and also see how good the linearity is. In the end I selected mostly references from Intersil since they came with the highest initial accuracy. I have no way to go to a calibration lab to get the actual reference values so I need to rely on the manufacturer. here is what I selected:
|Reference||Voltage||Initial Accuracy||Drift (ppm/°C)|
|10V||0.025% / 2.5mV||2.5|
|ISL21009BFB850||5V||0.01% / 0.5mV||3|
|ISL21009BFB825||2.5V||0.02% / 0.5mV||2|
|ADR4520B||2.048V||0.02% / 0.4mV||2|
|ISL21090BFB812Z||1.25V||0.03% / 0.375mV||7|
Together these did cost about $30, depending on where you get them. The best one is 10 times as accurate as the UT61E and the worst one still three times. I deemed this good enough to check how accurate my meters are and how much they drift over time.
I put all of the references on a small PCB, together with some 78xx regulators so the references don't need to dissipate so much power and heat up. Everything was powered from a 9V AC supply (whose voltage was doubled).
The PCB itself was designed to be generic so I can use other references. Most of them follow a common SO8 footprint, and just differ in whether they need input and output capacitors, and sometimes an optional capacitor.
(The resistors in the middle look a little bit funny - I forgot that the 78xx regulators need some minimal load so I needed to add them. You can see the free footprints for the capacitors, and how they are populated differently for the different references. Whats not to be seen here is that I also have a DPDT switch so I can easily switch polarities to the multimeter so I can test both positive and negative ranges.
Later on I also bought some precision resistors (with 0.01%, and I also some older ones with 0.1%) to test the resistance ranges. I also used them to manually check the current ranges. The most difficult range to test are capacitances - the best you can get are 1% foil capacitors. They easily cost 1 up to 4 dollars each, depending on the capacity. I ended up with a 1nF and a 4.7nF 2.5% capacitor.
Since I have everything to test my meters, why a second version? Well, there are two reasons.
The first one is that I killed the REF102C by accidentally connecting the 9V AC power supply to its output. It didn't like that. I made the big mistake to have all connectors look the same, and AC in and 10V out are just on opposite corners of the board. Its actually surprising that it took me several year to make this mistake. So I replaced the chip, and looked for a way to prevent that in the future.
The second reason is that everything is fiddly. You have the flying DPDT switch, need to place the pin header on a separate output for each voltage (in the correct position). The resistors and capacitors are not mounted to anything but I just kept them in a bag. And the power supply was also in a bag together with the PCB and the caps and resistors.
So my second version was not about a redesign, but about making a real project out of it: place everything in a proper case, with a proper range selection switch and just a pair of binding posts. And since there was some room to spare, I added a precision current reference to.
For the latter one one I used the design from the Scullcom video blog - its easy to build and quite accurate. I wanted to build it since I saw it for the first time but never came around.
For the voltages, resistors and capacitors I found an old 2x24 rotary switch - unfortunately the current reference needs a three-pole switch so I put it on an separate pair of binding posts. Time to start wiring everything:
The REF200 current source is just on that small SOIC8 breakout board, attached with magnet wire to the range switch. On the large switch I already added the wires for changing the polarity of the voltages.
Then I mounted the existing PCB and the 9V transformer 8without its case) to a small acrylic plate which is then mounted in the enclosure:
The current reference is powered from the existing 9V regulator so I don't need an additional battery. The resistors and capacitors go to a separate board, wrapped in heat shrink tube. One of the resistors (the 100k one) is on the bottom side since its a SMD one.
If you count very closely you can see that there are too many wires. This is due to a small trick: I added one pair of wires for the capacitors, which I left open. And another wire for the resistors which gets shorted to the common wire. These additional wires end up in a '0nF' and '0 ohm' selection, which can be used to zero out any capacitance and resistance in the measurement wires. Since they are of the same length and placement as the regular wiring this relative measurement is hopefully quite accurate.
Everything wired up:
The front plate was designed in Draftsight (a free CAD program for Linux, Mac and Windows). I printed it on paper, glued it to the plastic front plate and them covered it with self-adhesive foil:
I found a nice trick to get the lines for the scales all in the right place:
- draw two circles, centered on the hole for the rotary switch, which mark the inner and outer end of where the lines should be
- then create points to split the circles in segments (Draw / Points / Segment - select a circle and it gets marked)
- configure Draftsight so points get a large cross (to see them on the drawing)
- Use the "Element Snap" function to snap start and end points of each line to the correct segment points
- afterwards, delete the circles and the points, and you have just the scale lines left
I think it ended up quite well. No need to fiddle with wire grabbers, re-positioning connectors and looking for parts in a small box: