Best option for sensor input simulation for noise filter design.

[v8dave]

I am working on a number of designs with analog sensors, mainly 4-20mA types and I want to design the hardware to handle noise through either hardware filters and or software filtering. The ADC itself has some noise but I've managed to get this nice and quiet now with good PCB design and power line filtering. Now I need to look at the input signals with some noise on them. 4-20mA is generally quite good for this but some sensors I've tested have a noisy output and I'd like to clean this up so need some form of simulated input to do so.

Is there anything on the market, reasonably priced that would do this. Budget would be up to around US$2K, possibly stretching to 2.5

[T3sl4co1l]

What's the noise environment? Quiet, inside a shielded enclosure? Long wires? Industrial switching noise?

Generally, you can't go wrong with LC filtering (on both terminals of the device), but this is limited by the bandwidth of the device, capacitive loading, etc.

Tim

[nctnico]

For simulation you could use any function generator with noise capabilities. Basically you will want to filter signals in hardware so they are well below the Nyquist frequency (half the sample frequency) and from there you can use digital signal processing techniques to do further filtering.

[v8dave]

Generally, you can't go wrong with LC filtering (on both terminals of the device), but this is limited by the bandwidth of the device, capacitive loading, etc.

I have a low pass RC filter on the input just now and this is working quite well but I am seeing some noise from some sensors and not all so I need to get a test setup to try and improve the filtering. LC filtering was next on the list. The ADC works really well and I would like to have something better in terms of clean readings. Something that gives a small change on the 4th decimal place for the 4-20mA input would be ideal. This is because the sensor range is up to 10,000 PSI (pressure) so small changes on the ADC input cause large readings on the display.

[nctnico]

If you can add a low noise opamp you can drive an LC based filter (these need to be driven & loaded by a specific impedance) or use the opamp to implement a sallen-key filter. If your ADC is 10 bit you have 60dB (6dB per bit) of dynamic range so I'd design the filter to have more than 60dB attenuation at half the samplerate.

[v8dave]

If you can add a low noise opamp you can drive an LC based filter (these need to be driven & loaded by a specific impedance) or use the opamp to implement a sallen-key filter. If your ADC is 10 bit you have 60dB (6dB per bit) of dynamic range so I'd design the filter to have more than 60dB attenuation at half the samplerate.

I have 16 analog inputs across 2 ADC devices and limited board space to add an op-amp so I have to consider other options. With the RC or LC filter on the hardware side with digital filtering on the software side I am hoping I can get good enough filtering and resolution. The ADC's are 16 bit.

[ADT123]

To simulate the output of noisy sensors how about an AWG - most will have more than enough bandwidth.  You dont need a 16 bit AWG as you can just reduce the amplitude of the noise signal you are injecting.

There are plenty of benchtop AWGs in the budget you are talking about, alternatively there are high resolution PC oscilloscopes with built in AWGs so you can output the noise and measure back after the filter. 

[macboy]

Generally, you can't go wrong with LC filtering (on both terminals of the device), but this is limited by the bandwidth of the device, capacitive loading, etc.

I have a low pass RC filter on the input just now and this is working quite well but I am seeing some noise from some sensors and not all so I need to get a test setup to try and improve the filtering. LC filtering was next on the list. The ADC works really well and I would like to have something better in terms of clean readings. Something that gives a small change on the 4th decimal place for the 4-20mA input would be ideal. This is because the sensor range is up to 10,000 PSI (pressure) so small changes on the ADC input cause large readings on the display.

How frequently do you take ADC readings? How quickly must the display track changes? I would recommend digital filtering, but for it to be effective you will need to take at least, say, 100 readings per second from the(each) sensor. Basically you will average multiple successive readings to get a more slowly changing display with finer apparent resolution. In order to quickly track to large changes, you can have a delta threshold, where if the change from sample-to-sample is big enough, then you display the raw values until you have a "stable" reading again, then display the averaged/filtered values.