This video has been supported by Tektronix.
Hey guys, are you up to date? In the last video we've repaired and rejuvenated a fluke
5700 a multifunction calibrator or ruined what was left of it. Today we are going to
determine which and how much with the help of a majestic Keithley 2002. Tektronix was
kind enough to send that over just for this purpose. But since the opportunity to work
with Keith Lee's highest resolution multimeter doesn't present itself all that often,
I'd like to give you a little tour first. First and foremost, this is of course, an
eight and a half digital multimeter, the king or queen of the 2000 series. It has a nano
volt resolution in the lower DC voltage ranges, which on its own doesn't mean all that much.
At least four other meters in this shot can display nano volt. What counts other more
subtle properties of Keith Lee's custom ADC noise stability, linearity, and useful sample
rate. The letter also known as DC v measurement speed is a really important parameter,
it can be configured to hold a number of power line cycles
or presets like fast or high accuracy. high accuracy means 10 power line cycles are about
200 milliseconds per eight and a half digit reading where I live. That's what they recommend
to achieve the uncertainties specified in their data sheet plus another digital filter
that displays the average out of every 10 readings. Bottom line, we are looking at about
two seconds per high accuracy reading, whereas solar Tron 7081 needs almost a minute. Both
of these meters are synchronized to the main frequency so that the positive and negative
half waves of power line interference cancel each other out, or at least every reading
gets the same amount of power line interference. Other than that, no schematics or anything
have been released for the 2000 to two. So I can only guess based on internet rumors
what makes it good. This is a simplified version of what I think their analog to digital converter
looks like. It's based on a programmable logic device and a stable clock signal. For a specified
time they let the voltage to be measured charge a capacitor via a precision resistor, only
to discharge it again via one of multiple precise current sources. In a perfect world,
the time it takes to discharge that cap would only depend on the initial input voltage.
If you ever find that perfect world, let me know.
But while we are stuck here, it's an absolute marvel that they managed
to make this circuit work so well, in spite of jejuri clocks,
drifty resistors, leaky caps and analog switches that behaves differently
from one another and from time to time. Look, even my simulation consisting of purely ideal
components somehow manages to have noise. The three different rundown current sources
are what gives this arrangement it's named multi slope ADC. They are derived from an
LTC 1000 day this multimeter. That way it inherits the famous stability of that part.
And I can confirm that now with a measurement of my calibrator set to 10 volt for four days
straight, including some harsh climate conditions in my apartment. These are all pretty good
results within roughly one ppm the blue trace solo Tron 7081 is going a bit crazy in the
first half because I've disabled its internal drift correction in favor of low noise. But
the green trace representing q3 2002 is the most stoic in spite of being exposed to the
same ambient temperature swings as the others pretty good.
Speaking as a connector purist, there's only one thing I don't like about the 2002 It's
a recessed routed banana jacks, I got some special Pomona low thermal EMF test leads
for these but not sure if those are even necessary. It's probably all fine for measurements above
nanobot. But man would I be thrilled to see a success or with copper binding posts. Otherwise,
no surprise I really like the 2000 to two. It's the smallest and lightest multimeter
of its caliber. It has a beautiful two line alphanumeric display into the user interface,
as well as the GP IB language are exactly the same as for the 2001. So I didn't have
to learn anything or rewrite my software. It's quiet to not completely silent there
is a little cooling fan. But I could definitely work in its vicinity
all day long. Unlike my fluke 5700 hairdryer. Listen when I wake it up.
No kidding. Nothing can really work in its direct vicinity. It's hard to exhaust what
scare away all the precious ppms in neighboring test equipment. That's why it has to live
on the ground for the time being the 2200 The relative internal temperature sensor with
Richard can tell the difference between its momentary internal temperature and the one
it had during calibration. 2001 doesn't have that I just checked it. Off the MLS can read
the most common types of thermocouples at resistance thermometers with a fantastic milli
Kelvin resolution. To determine the absolute value of a thermal coupler reading,
it would have to add the reference junction temperature,
which would be the front input connector temperature in this case, they don't seem to have sensors
they are so absolute thermal couple of measurements are only possible with the appropriate thermocouple
scanner module which can be plugged into the rear. Not a big deal. I think in precision
applications, platinum resistance thermometers are the way to go anyway. And for those of
the absolute readings are included Of course,
whatever it is, you want to measure with a 2002 to be at temperatures, voltages DC, AC
or resistances. There's always data storage and a sophisticated set of math operations
available to determine often needed values like min max, average standard deviation,
and more without involvement of a computer. Alright, let's get on with what we are here
for. Of course, computer involvement is where we
are heading in a moment to plot data over a longer
time. And to verify that the 5700 is working more than temporarily. I'm also setting up
my own eight and a half digit meters to come along for the ride. But those haven't seen
a calibration in years and are nowhere near as trustworthy as a latest generation 2000
tools straight from the manufacturer. I was a bit sad when that turned out to be just
out of Cal two when it arrived here July the 16th.
Otherwise, I could have adjusted to the calibrate arrangers manually to match Keith
Lee's and from their inherited a quasi traceable accuracy to all my other meters. Oh well,
as long as we can verify stability and linearity, we can always import the absolute accuracy
later. While everything is still warming up, I'm making some momentary measurements over
10 polar line cycles with a repeating averaging filter that starts over every 10 measurements
and manual triggering so that I can make adjustments on the calibrator at my own pace.
While the two champions don't agree perfectly, and that's no surprise after such extensive
repairs and renovations, but I'm getting similar results, which is the most important part
here. I can now use the rotary encoder to fine adjust the output and show how far apart
of the two devices are approximately point oh, four 4% in this case, that's 44 ppm, very
unseemly under normal circumstances. All the current ranges seem to be perfectly stable
at a glance all the way from micro Empire
to the highest to Empire current range, which still has that cracked resistor network. Now
for the voltage ranges, I do care about every last little nano vault I can get. But somehow
I've acquired yet another meter with recessed banana jacks and I have only one pair of these
Pomona low thermal test leads. Generally these hollow ones are also regarded as pretty good
in the low thermal EMF department. But these happen to be attached to a solar Tron input
cable. Luckily, I have another very reliable connector available.
ubiquitous Ethernet cable is usually shielded quite well.
And to the super thin pure copper data lines in combination with some metrology
grade q tips make excellent low thermal EMF connections. When there is
no thermal mass, any temperature difference won't last long. Whenever connecting two or
more mains powered devices like this common mode voltages are a topic to keep in mind.
The calibrator has a special blue output terminal called guard. Well, that's not the same kind
of guard we are used to seeing and Keith Lee's high resistance ultra low current designs.
In this case, it's a shield that surrounds the calibrator power transformer as well as
it's appropriately named in guard section. Since Keith Lee's inputs are not mains Earth
referenced Anyway, I'm not using the
external guard function in this case. Otherwise, I would have connected to the cable
shield to guard on the calibrator side and to input low on the device side to
give common mode voltages somewhere to go where they don't influence my measurements.
Okay, I think every participant has warmed up now with the GBI B cables to but they don't
care. They are so stiff as always. But let's try anyway to get some automation going. Somebody
gave me a hint about this awesome device which is unfortunately known Longer produced as
far as I can tell, it's a GPI B to F net gateway. And it looked even less remarkable in its
original enclosure. I had to destroy that to satisfy my curiosity, and find out how
hard it could be to replicate such a thing. Oh,
well, my conclusion is that Keysight makes something that does the same thing for $1,500.
And of course, you can always just use a standard GP IB two USB adapter.
Not sure who is stiffer the cables are mean. Eventually, I want all my instruments connected
and automated like this. I've started to make a minimalist multi instrument data logging
software for this. So far, I've only implemented my favorite multimeters. Notice how Keithley
2002 just inherits everything from 2001. These in quotation marks modern or not ancient devices
are relatively easy to talk to Luke commands understood by brown shoe box Keith leads from
the 1980s. However, oh boy, oh, those are absolutely cryptic and take a long time to
implement. Here's my 237 SMU for example. And here's the fluke 5700 pretty easy.
I'm using a new revision of an old acquaintance as a
Raspberry Pi container, the Ross pet three from San founder. It's a bit of an overkill
for just writing down the occasional eight and a half digit reading into a database.
But it's quite a convenient mobile package for every project that would otherwise be
a raw Raspberry Pi board and an unmanageable nest of cables to peripherals and monitor.
This thing is the monitor and we are going to set it up to keep an eye on our precious
ppms not to use an ordinary USB GP IB
adapter, you'll have to download and compile the Linux GP IB project, which needs kernel
headers and more and isn't even compatible with all of them. My lucky Ethernet gateway
needs only the Python v x 11 package, and we are ready to talk to GP IB devices. We
could immediately write data into CSV files. But I want something a little bit more fancy
than that. So I'm going to install influx DB and grafana. Those are pretty popular for
data logging on the Raspberry Pi, so you can just follow tutorials on the internet.
I also added periodic reading of tour TMP 117. i squared c temperature sensors to my
software with their eight milli Kelvin resolution. Those are just the right kind of overkill
for ambient temperature measurements during long term experiments. Remember, this is how
I used to document these long processes, just a GoPro in time lapse mode, primitive and
hard to evaluate. Now I've got a Python script that sweeps the calibrator output voltage
from minus 10 to 10 volt in 50 millivolt steps while triggering and reading all three volt
meters once per step. The results as well as temperature information is saved in a time
series database from where it can be accessed by grafana. For example, that's a really cool
browser based visualization software. dark mode by default, Need I say more? My simple
linearity tests featuring three volt meters and three thermometers are really just scratching
the surface of what's possible with grafana. But hey, it's a huge improvement and a lot
more insightful than the time lapse isn't it? I can use math operations in the database
queries to turn the absolute Voltage Readings that are stored they are into ppm deviations
for example. This reveals a number of things. First, my calibrator is doing well, and has
always Keithley 2002. The yellow trace. It has the best linearity out of the three meters
I've tested here. This inverted bell curve reassembles typical results for that meter.
Solar Tron 7081 isn't too bad either or be a bit noisy. Advantest rs 6581 t as a bit
of an ugly zero transition. I haven't had a close look at that one yet. Maybe we can
fine tune it a bit. But that's it for today. Mission calibrator rescue has been a huge
success. Thanks to Tektronix for helping me verify that. And a huge thanks to my PDB friends
for making all this possible in the first place.
