0
\$\begingroup\$

I'm having ringing-noise from a pressure transducer analog output. Here is the only information about the instrument: http://www.trescal.be/pdf/PPC500_GB_SI.pdf

Let me briefly explain the device and the issue:

This device has an internal analog to digital conversion at some stage and outputs that information on its display. The output from digital data displayed on the screen can be sent via RS-232 directly or as an analog output which is created using digital to analog conversion. The both RS-232 and analogue output sends the averaged voltages from the transducer at a rate of around three times per second.

For some reason I have to use the analog output. I record the readings from the analog output by a data-acquisition device which has 1GigaOhm input impedance. The DAQ device has many analog inputs and one of the analog inputs comes from this pressure transducer via a BNC cable which is between 10 up-to 15 meters long. All the analog inputs share the same ground and only at this transducer's channel I'm observing ringing-like noise. Same noise is observed if I hook the BNC up to an oscilloscope.

Below is an example of the transducer's analog signal in question. It shows the signal in time series for 60 seconds which is sampled at 8000 Hertz and plotted in LTspice: enter image description here

Here below if I zoom the signal, please see the three steps of discrete voltage sent by the transducer’s analog output per second: enter image description here

It seems like there is no reconstruction filter on the output of the transducer’s DAC. But actually it doesn’t matter in my case. What bothers is mostly those crazy jumps/ringings on the edges which sometimes exceeds one volt. And here below I'm directly zooming to the crazy ringing-edge and by using LTspice's cursors I find out that the ringing-noise is around 1kHz: enter image description here

So I decided to use a RC passive low-pass filter but only checked in LTspice and the simulation results are below (V1 is the output from the traducer; C1 and R2 represents the low-pass filter; R1 is the input impedance of the data acquisition channel; green plot shows after filtering): enter image description here

And for a better comparison before(blue) and after the filter(green) here: enter image description here

Here are my concerns and questions:

1-) I found the solution by filtering. It seems to me the loading is negligible due to the huge input impedance of the DAQ channel here. So do you agree in this case to use these R2 C1 values are optimum for a passive low-pass filter to eliminate 1kHz?

2-) Do you think this passive filter enough here? Is active filter necessary(I haven’t built any yet)?

3-) What could be the reason for this ringing? Could it be length of BNC or inductive effect? (It seems like we have an DAC outputting 3Hz discrete signal sent via a BNC coax cable)

\$\endgroup\$
  • \$\begingroup\$ Any square wave will have ringing if you zoom in enough. It is due to usually the inductance of the wires. I would say the filter is enough assuming the signal can still have a response time that meets your criteria. \$\endgroup\$ – MadHatter Oct 12 '16 at 19:28
  • \$\begingroup\$ so no need to go for an active filter i guess here. it is 1kHz. how about the choice of RC values? could there be a better combination? \$\endgroup\$ – user16307 Oct 12 '16 at 19:30
  • \$\begingroup\$ Seems reasonable to me... Active filters help if your ADC for example had a low impedance, also it is easier to build high order filters using only RC in active configurations. \$\endgroup\$ – MadHatter Oct 12 '16 at 19:32
  • \$\begingroup\$ Along MadHatter suggestion, just want to say that a quad opamp can house an 8th order Bessel, or Gaussian. \$\endgroup\$ – a concerned citizen Oct 12 '16 at 19:36
  • 1
    \$\begingroup\$ You may be loading the output stage of the transducer with too much stray capacitance (due to cable length) causing slight instability. Try adding series resistance (100 ohm to 1k) as close to the transducer as possible. Keep the capacitor at the ADC input for general noise immunuty. \$\endgroup\$ – Matthew Oct 12 '16 at 20:06
0
\$\begingroup\$

1). Confirm output impedance of Analog sensor output by load test for ~50% drop

  • complementary emitter followers used in Op Amps or discrete are notorious for resonance with capacitive loads from 100pF/m cables and current limiting causes low out followed by overshoot followed by ringing at loop gain BW of Op Amp
    • This may be what is happening
  • this is where current mode converters work better 4-20mA with separate return wire, shielded or twisted pair and CM ferrite sleeve.
  • add 100-300 R series at source to verify reduction in edge pulse
  • then load destination with 1k to reduce stray noise coupling and verify voltage drop if any. You may need a power op amp buffer with 120 Ohm source and low impedance load to reject stray coupling or whatever cable Z you are using.
  • check with proper high speed measurements for RF noise and LPF as required to rule out RF aliasing noise at 8kHz sampling rate

    • consider 10mH CM choke and RF cap across output for 50/60Hz and up noise

    • example of your CM 50Hz noise showing up as a differential mode (DM) signal

    • due to unbalanced source and load Z on 50 Ohm coax. enter image description here

    • Is a shame to degrade an instrument with > 60dB SNR down to 20 dB with improper analog cable maybe 30~40dB SNR with filter

\$\endgroup\$
  • \$\begingroup\$ 1k parallel load to reduce stray noise coupling would be less effective than having the OPs 7.2k and 47nF input filter at the ADC and a 100 ohm - 1k series resistor at the transducer output. \$\endgroup\$ – Matthew Oct 12 '16 at 20:41
  • \$\begingroup\$ perhaps same, I wrote that part before comments, but essentially matching source impedance and current limit with source R to cable C then "signal spectrum" matched filter at destination. Then verify with R divider of same equiv at source to verify noise only to Vdc, If noise is same then other noise sources exist. e.g. 50/60 Hz which is evident in filtered signal \$\endgroup\$ – Sunnyskyguy EE75 Oct 12 '16 at 20:47
  • \$\begingroup\$ @TonyStewart.EEsince'75 would it be too much demanding if you would draw an illustration/diagram/connections on what you mean or suggest? i'm kind of lost when i read you and cannot imagine the picture. the setup is: there is transducer there is the daq input and there is a BNC in between. otherwise thanks anyway. \$\endgroup\$ – user16307 Oct 12 '16 at 21:33
  • \$\begingroup\$ I dont have tools to show CM choke and twisted pair. Coax is not good for an unbalanced source and unbalanced high impedance load. Shielded twisted pair is better with balanced diff input. Multiple grounds from multiple locations also explains line noise. A Balun or CM choke can be a coupled coil pair. Unbalanced signals means CM coupling causes DM noise \$\endgroup\$ – Sunnyskyguy EE75 Oct 12 '16 at 22:10
  • \$\begingroup\$ @TonyStewart.EEsince'75 yes i knew about the 50Hz noise is there but since the average voltage matters in measurements, it cancels right? average of sinusoidal is zero. i would be very glad what you meant by "Is a shame to degrade an instrument with > 60dB SNR down to 20 dB with improper analog cable maybe 30~40dB SNR with filter" how did you come up with that(where does 30 40 come from what are they?)? and as last you wrote "I dont have tools to show CM choke". i meant maybe you can draw on LTspice or MS-paint the diagram you suggest not the actual choke. \$\endgroup\$ – user16307 Oct 12 '16 at 23:31
1
\$\begingroup\$

Assuming that the output of the instrument (used as a sensor) has some really noisy pressure input and the internal instrument workings do produce this kind of jumpy output, the answers are:

  1. Yes, a simple RC filter with 1kHz cut-off is an acceptable solution

  2. Yes, the data show that this is good enough;

  3. Most likely reason for the ringing at 1kHz is a flaw in instrument's design of output stage of internal DAC. The 1 kHz ringing cannot be explained by cable/load mismatches.

SIDE NOTE: It looks quite obvious that the selection of this sensor is inadequate to the nature of input pressure signal. The instrument states "settling time" of ~0.5s, which limits the spectrum of input signals to pretty much static pressure levels.

\$\endgroup\$
  • \$\begingroup\$ analog.com/library/analogdialogue/archives/38-06/… shows that capacitive loading of opamps can cause ringing and oscillation. It is a fact. A web search will show more evidence if you need it. 10-15 m of cable can have around 1nF of capacitance. This is more than enough capacitance to cause problems. Adding an "isolation" resistor in series with the output is one of the simple solutions, and particularly suitable for this application \$\endgroup\$ – Matthew Oct 12 '16 at 21:58
  • \$\begingroup\$ Still it points to the product design defect, if the DAC buffer failed to handle a reasonable load. However, I did miss the 15m BNC (RG-58U?) of analog cable, which cannot be named as reasonable connection either. Still I see the entire issue as misapplication of pressure sensor. \$\endgroup\$ – Ale..chenski Oct 12 '16 at 23:15
  • \$\begingroup\$ 0.5s may be large step response time but small signal and noise spectrum appears to be up to 50 Hz, so the -6dB spectrum may be somewhere in between. Generally optimal filtering matches the desired signal spectrum. It is apparent the OP did not do his homework on the instrument signals before design for transmitted measurements and EMI control. The digital interface would have been smarter. \$\endgroup\$ – Sunnyskyguy EE75 Oct 13 '16 at 13:21

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.