The sensor shown in the above figure is a sourcing type, where the signal sources the 4-20 mA current which is then sent through
the shunt resistor and sunk into ground. Another type of 3-wire sensor is the sinking type, where the 4-20 mA current is sourced
from the positive supply, sent through the shunt resistor, and then sunk into the signal wire. If sensor ground is connected to U3
ground, the sinking type of sensor presents a problem, as at least one side of the resistor has a high common mode voltage
(equal to the positive sensor supply). If the sensor is isolated, a possible solution is to connect the sensor signal or positive sensor
supply to U3 ground (instead of sensor ground). This requires a good understanding of grounding and isolation in the system. The
LJTick-CurrentShunt is often a simple solution.
Both figures show a 0-100 Ω resistor in series with SGND, which is discussed in general in Section 2.6.3.4. In this case, if SGND
is used (rather than GND), a direct connection (0 Ω) should be good.
The best way to handle 4-20 mA signals is with the LJTick-CurrentShunt, which is a two channel active current to voltage converter
module that plugs into the U3 screw-terminals.
2.6.3.8 - Floating/Unconnected Inputs
The reading from a floating (no external connection) analog input channel can be tough to predict and is likely to vary with sample
timing and adjacent sampled channels. Keep in mind that a floating channel is not at 0 volts, but rather is at an undefined voltage.
In order to see 0 volts, a 0 volt signal (such as GND) should be connected to the input.
Some data acquisition devices use a resistor, from the input to ground, to bias an unconnected input to read 0. This is often just
for “cosmetic” reasons so that the input reads close to 0 with floating inputs, and a reason not to do that is that this resistor can
degrade the input impedance of the analog input.
In a situation where it is desired that a floating channel read a particular voltage, say to detect a broken wire, a resistor (pull-down
or pull-up) can be placed from the AINx screw terminal to the desired voltage (GND, VS, DACx, …). A 100 kΩ resistor should pull
the analog input readings to within 50 mV of any desired voltage, but obviously degrades the input impedance to 100 kΩ. For the
specific case of pulling a floating channel to 0 volts, a 1 MΩ resistor to GND can typically be used to provide analog input readings
of less than 50 mV. This information is for a low-voltage analog input channel on a U3.
Note that the four high-voltage channels on the U3-HV do sit at a predictable 1.4 volts. You can use a pull-down or pull-up resistor
with the high-voltage inputs, but because their input impedance is lower the resistor must be lower (~1k might be typical) and thus
the signal is going to have to drive substantial current.
2.6.3.9 - Signal Voltages Near Ground
The nominal input range of a low-voltage single-ended analog input is 0-2.44 volts. So the nominal minimum voltage is 0.0 volts,
but the variation in that minimum can be about +/-40 mV, and thus the actual minimum voltage could be 0.04 volts.
This is not an offset error, but just a minimum limit. Assume the minimum limit of your U3 happens to be 10 mV. If you apply a
voltage of 0.02 volts it will read 0.02 volts. If you apply a voltage of 0.01 volts it will read 0.01 volts. If you apply a voltage less than
0.01 volts, however, it will still read the minimum limit of 0.01 volts in this case.
One impact of this, is that a short to GND is usually not a good test for noise and accuracy. We often use a 1.5 volt battery for
simple tests.
If performance all the way to 0.0 is needed, use a differential reading (which is pseudobipolar). Connect some other channel to
GND with a small jumper, and then take a differential reading of your channel compared to that grounded channel.
The nominal input range of a high-voltage single-ended analog input is +/-10 volts, so readings around 0.0 are right in the middle
of the range and not an issue.
2.6.4 - Internal Temperature Sensor
The U3 has an internal temperature sensor. Although this sensor measures the temperature inside the U3, which is warmer than
ambient, it has been calibrated to read actual ambient temperature, although should only be expected to be accurate to within a
few degrees C. For best results the temperature of the entire U3 must stabilize relative to the ambient temperature, which can take
on the order of 1 hour. Best results will be obtained in still air in an environment with slowly changing ambient temperatures.
With the UD driver, the internal temperature sensor is read by acquiring single-ended analog input channel 30, and returns
degrees K. Use channel 30 anywhere you would use an analog input channel (e.g. with eAIN).
2.7 - DAC
The LabJack U3 has 2 analog outputs (DAC0 and DAC1) that are available on the screw terminals. Each analog output can be
set to a voltage between about 0.04 and 4.95 volts with 10 bits of resolution (8 bits on older hardware revision 1.20/1.21). The
maximum output voltage is limited by the supply voltage to the U3.
Starting with hardware revision 1.30, DAC1 is always enabled and does not affect the analog inputs, but with older hardware the
second analog output is only available in certain configurations. With hardware revisions <1.30, if the analog inputs are using the
internal 2.4 volt reference (the most accurate option), then DAC1 outputs a fixed voltage of 1.5*Vref. Also with hardware revisions
<1.30, if DAC1 is enabled the analog inputs use Vreg (3.3 volts) as the ADC reference, which is not as stable as the internal 2.4
volt reference.
The DAC outputs are derived as a percentage of Vreg, and then amplified by 1.5, so any changes in Vreg will have a
proportionate affect on the DAC outputs. Vreg is more stable than Vs (5 volt supply voltage), as it is the output from a 3.3 volt
regulator.
The DACs are derived from PWM signals that are affected by the timer clock frequency (Section 2.9). The default timer clock
frequency of the U3 is set to 48 MHz, and this results in the minimum DAC output noise. If the frequency is lowered, the DACs will
have more noise, where the frequency of the noise is the timer clock frequency divided by 65536. This effect is more exaggerated
with the 10-bit DACs on hardware revision 1.30+, compared to the 8-bit DACs on previous hardware revisions. The noise with a
timer clock of 48/12/4/1 MHz is roughly 5/20/100/600 mV. If lower noise performance is needed at lower timer clock frequencies,
use the power-up default setting in LJControlPanel to force the device to use 8-bit DAC mode (uses the low-level
CompatibilityOptions byte documented in Section 5.2.2). A large capacitor (at least 220 uF) from DACn to GND can also be used
to reduce noise.
The analog outputs have filters with a 3 dB cutoff around 16 Hz, limiting the frequency of output waveforms to less than that.
The analog output commands are sent as raw binary values (low level functions). For a desired output voltage, the binary value can
be approximated as:
Bits(uncalibrated) = (Volts/4.95)*256
For a proper calculation, though, use the calibration values (Slope and Offset) stored in the internal flash on the processor (Section
5.4):
Bits = (Slope * Volts) + Offset
The previous apply when using the original 8-bit DAC commands supported on all hardware versions. To take advantage of the
10-bit resolution on hardware revision 1.30, new commands have been added (Section 5.2.5) where the binary values are aligned
to 16-bits. The cal constants are still aligned to 8-bits, however, so the slope and offset should each be multiplied by 256 before
11
