« Close

Datasheets and User Guides

App Notes

Software & Driver

 

A-3-1 Noise And Resolution

A Note About ADC Noise and Resolution

Analog voltages measured by the T7 are converted to digital representation via the T7's analog to digital converter (ADC).  The ADC reports an analog voltage in terms of ADC counts, where a single ADC count is the smallest change in voltage that will affect the reported ADC value.  A single ADC count is also known as the converter's LSB voltage.  The ADC's resolution defines the number of discrete voltages represented over a given input range.  For example, a 16-bit ADC with a +/-10 input range can report 65536 discrete voltages (216) and has an LSB voltage of 0.305 mV (20V ÷ 216).

The stated resolution for an ADC is a best case value, assuming no channel noise.  In reality, every ADC works in conjunction with external circuitry (amplifiers, filters, ect.) which all posses some level of inherent noise.  The noise of supporting hardware, in addition to noise of the ADC itself, all contribute to the channel resolution.  In general, the resolution for an ADC and supporting hardware will be less than what is stated for the ADC.  The combined resolution for an in-system ADC is termed effective resolution.  Simply put, the effective resolution is the equivalent resolution where analog voltages less than LSB voltage are no longer differentiable from the inherent hardware noise. In addition to defining the smallest measurable analog voltage,  the effective resolution also defines the RMS peak-to-peak noise on a given analog channel.

Closely related to the effective resolution is the error free code resolution (EFCR) or flicker-free code resolution.  The EFCR represents the resolution on a channel immune to "bounce" or "flicker" from the inherent system noise.  The EFCR is not reported in this appendix. However, it may be closely approximated by the following equation:

EFCR = effective resolution - 2.7 bits [1.]

The T7 offers user-selectable resolution through the resolution index parameter on any one AIN channel.  Internally, the ADC hardware uses modified sampling methods to increase measurement resolution beyond the ADC's base resolution.  Valid resolution index values are 0-8 for the T7 and 0-12 for the T7-Pro [2.][3.].  Increasing the resolution index value will improve the channel resolution, but doing so will usually extend channel sampling times.  See section 14.0 AIN for more information on the resolution index parameter and its use.

 

Noise and Resolution Data

The data shown below summarizes typical effective resolutions and expected channel sampling times over all resolution index values.  Data for the T7 and T7-Pro data are combined and presented together for convenience, where resolution index values 9-12 only apply to the T7-Pro.

The AIN sampling time is the typical amount of time required for the ADC hardware to make a single analog to digital conversion on any channel and is reported in milliseconds per sample.  The AIN sampling time does not include command/response and overhead time associated with the host computer/application.

 

Noise and Resolution Test procedure

Noise and resolution data was generated by collecting 512 successive voltage readings, using a short jumper between the test channel and ground.  The resulting data set represents typical noise measured on any one analog input channel in ADC counts.  The effective resolution is calculated by subtracting the RMS channel noise (represented in bits) from 16-bits.

Effective Resolution = 16 bits - log2 (RMS Noise [in ADC counts]) 

 

Table A.3.1.1.  Effective resolution and sampling times for various gains and resolution index settings. Resolution index settings 9-12 apply to the T7-Pro only.
Resolution Effective Effective AIN Sample
Index Resolution Resolution Time
  [bits] [µV] [ms/sample]
Gain/Range: 1/±10V
1 16.0 316 0.04
2 16.5 223 0.04
3 17.0 158 0.06
4 17.5 112 0.09
5 17.9 85 0.16
6 18.3 64 0.29
7 18.8 45 0.56
8 19.1 37 1.09
9 19.6 26 3.50
10 20.5 14 13.4
11 21.4 7.5 66.2
12 21.8 5.7 159
Gain/Range:10/±1V
1 15.4 48 0.23
2 16.0 32 0.23
3 16.5 22 0.55
4 16.9 17 0.58
5 17.4 12 1.15
6 17.9 8.5 2.28
7 18.3 6.4 2.55
8 18.7 4.9 3.08
9 19.5 2.8 3.50
10 20.5 1.4 13.4
11 21.4 0.7 66.2
12 21.7 0.6 159
Gain/Range: 100/±0.1V
1 13.3 21 1.03
2 14.2 11 2.03
3 14.7 7.8 5.05
4 15.2 5.5 5.08
5 15.7 3.9 5.15
6 16.3 2.6 10.28
7 16.7 1.9 10.55
8 17.2 1.4 11.08
9 18.3 0.6 3.50
10 19.1 0.4 13.4
11 19.6 0.3 66.2
12 19.7 0.2 159
Gain/Range: 1000/±0.01V
1 10.9 11 5.03
2 12.3 4.1 10.0
3 12.7 3.1 10.1
4 13.3 2.1 10.1
5 13.8 1.5 10.2
6 14.4 1.0 10.3
7 14.7 0.8 10.6
8 15.0 0.6 11.1
9 15.4 0.5 3.50
10 16.1 0.3 13.4
11 16.4 0.2 66.2
12 16.4 0.2 159

Figure A.3.1.2.  Analog input effective resolution over various gains and resolution index settings.

Figure A.3.1.3.  Analog input LSB voltage over various gains and resolution index settings.
Figure A.3.1.4.  AIN sample times for analog inputs over various gains resolution index settings.

Notes:

 

1.  The equation used to approximate the EFCR is determined using +/-3.3 standard deviations from the RMS noise measured on an AIN channel.
 
2.  The default value for ResolutionIndex is 0, which equates to 8 for T7 command-response reads, 9 for T7-Pro command-response reads, and 1 for T7 & T7-Pro stream reads.
 
3.  The T7-Pro is equipped with a 24-bit delta-sigma ADC, in addition to the standard 16-bit ADC.  Analog conversions occur on the 16-bit ADC when resolution index values 0-8 are used.  Analog conversions occur on the 24-bit ADC when resolution index values 9-12 are used (command response mode only). 
 
4.  The hi-resolution 24-bit ADC is not supported in stream mode.

 

Support - Table Styling Fix

4 comments

So the latency is around 0.16 seconds for the highest possible uV resolution.  What real sampling interval does this equate to for a typical application, given communication time and processing, etc.?  Is there a minimum overhead -- are we talking 0.3 sec or 1 sec or what?

What's the transient voltage swing of the analog output on T7/T7 Pro?

mfalchetta's picture

Dear all, we are recently beginning to use a Labjack T7 board for the purpose of connecting it as I/O board to Simulink RT application to control an experimental solar Dish driving a microgas turbine, that we are now commissioning.  The board has been delivered "as it is" with very little instructions on how to operate it.  We are now at a good point in commissioning, but I observe that in many cases (basically when we operate three-phase electric motors in the vicinity of the board) the AI input signals are disturbed significantly.  Indeed we have 5 Voltage signals and 9 TCs connected at the AI ports of a DB37 additional board.   I believe the problem could come From e.m. noise, in fact due to the hurry in completing the plant in due time, the connections to the primary instruments have been performed using normal, not shielded cables, either for the voltages and for the TCs.  The question is on how to perform a proper connection, in practice:

1 should we connect any GND or the SGND to a proper "physical ground" item, e.g. the metallic structure of the Dish  ?

2 should we use shielded signal cables ? where to connect the shield sides ? one side only or both ?

3 Which checks could be advisable ?

Regards 

Massimo Falchetta

LabJack Support's picture

1.  There are so many things that can affect this it is hard to provide a simple answer, but usually the answer is no.  Are you using USB or Ethernet?  What connections do you already have to GND.

2.  They can sometimes help, and certainly don't hurt.  You would typically connect the shield to GND on the T7 side only.

3.  This is the big one.  I would do more tests to understand the problem before you think too much about #1 and #2 above.  Remove all connections except comm/power, and then just jumper a channel to GND.  Look at the noise levels on that channel and compare to the expect noise levels from Table A.3.1.1 above.  If it looks good, even with your machine on, then try a similar test with one of your voltage signals.  It that voltage signal looks good remove it and try with one of your TCs.  If that looks good start to add more signals.  Let us know what you see.