# Power supply noise problem in a measurement system

This might be a bit long to explain but actually a power supply noise is affecting a system's measurements. I will briefly try to explain the system and the noise issue with several pics. This is directly related to my previous question: How to verify observe or check the source of AC line noise

I'm using this [system][1] together with this [module][2](SECTION 4 shows ELECTRONIC COMPONENTS). This is a pressure scanner system. Basically the ZOC modules amplifies and multiplexes 64 pressure channels(500Hz sampling freq. for each channel) and sends the particular channel's analog voltage output(chosen by a decoder code sent from the system) to the RAD_unit which then converts this voltage to a digital by its ADCs and sends to a PC.

Here is the basic illustration of the system setup:

[![enter image description here][3]][3]

All system is fed from one AC mains outlet using a power strip.

Depending on the day, time of the day or powering the system through different AC mains outlets I was getting different magnitude of spiky noises in my measurements. Below shows this difference. Both voltage is on vertical axis represents sensor data and scaled with factor 100.(For example 10 in vertical pressure axis is 0.1 Volt). Here is a comparison of two measurements where the bottom is when there is spiky noise:

[![enter image description here][4]][4]

Later on I checked the +-15VDC power supply outputs(RPM100 in above illustartion) which powers the module and sensors, and I found out there is noise similar to the one in measurements.

Here is how the power supply connects to the module:

[![enter image description here][5]][5]

And here are some screenshots of the power supply's +15V to GND measurements from scope in AC coupling:

A very noisy situation:

In FFT view:

A less spiky/noisy situation:

[![enter image description here][8]][8]

And here when there is almost no significant noise(the gap is scope's artifact):

Question:

It seems to me there is a big correlation between the measurement noise and the power supply noise here. Power supply is very old at least a decade old but a special one and powers all the units with its special connector.

1-) What kind of power supply noise can it be? It is obviously effected by the AC lines)

2-) For a quick solution, Im planning to cut the cable and place 100nF caps between +15V to GND and -15V to GND. Do you think it can be a solution? And if so would 100nF be okay?

• You really should have edited your original question to add the additional material, rather than starting a new one. However, I have closed the other question as a duplicate of this one in order to avoid confusion and duplicated effort. Commented Dec 12, 2016 at 22:33
• yes sorry, i just found out the new data thats why. Commented Dec 12, 2016 at 22:40
• How the cable #155900-1 (10ft) is designed? Do I understand that you are seeing 100mV noise spikes. How much is this relative to the entire measurement scale? Commented Dec 12, 2016 at 23:06
• Also, do I read this correctly that the RADBASE ADC+MUX is connected to USB hub via cable #155899-1 that is 50ft long? Commented Dec 12, 2016 at 23:14
• @AliChen I dont know how it is designed but actually in my case bth sides are round connectors the rest is the same. Yes 100mV when the noise gets wild. Im almost sure this is related to the power supply noise I added as screenshots. Its old system. Power supply looks old. Would it be safe to add 100nF caps as I asked in in my question just to try? It is tricky to add caps to this cable though.. Yes there is a USB extender that long "I guess". Im not in that location right now. Commented Dec 12, 2016 at 23:16

Your oscilloscope scans are helpful. I see groups of noise spikes repeating at 10 millisecond intervals. This suggests a brushed motor running at 6000 rpm. Your oscilloscope time base is accurate enough that the 10 ms. rep rate is certainly not related to 60 Hz. line frequency.

But from the FFT plot, those noise spikes have consistent frequency content at about 68 kHz. There are many spikes on that FFT plot because the noise is discontinuous. As suggested by Ali Chen, noise sources at this frequency might emanate from switching supplies, or possibly a faulty light ballast.
Perhaps a combination of the two?...A motor driven by a Pulse-width-modulated power source (PWM @ 68 KHz.)?

• Thanks for the suggestions. When I switch off the VFD and AC motor the noise is still there. If I even monitor a disconnected power supply in the 2nd floor I see this noise. I think the noise is in the earth bouncing and if the power supply is using the earth it is effected from the earth. And if disconnected it is picking up noise through the air. But surprisingly less noise if I use 2 prong connected to mains but still there. So should I unplug all the equipment from the mains to find it? Should I use a ferrite bead or something like this: uk.rs-online.com/web/p/iec-filters/0238817 ? Commented Dec 13, 2016 at 20:18
• I dont also get why it is repeating each 10ms. Commented Dec 13, 2016 at 20:38
• @doncarlos, The groundloop possibilities of your complex instrumentation sends a chill up my spine. As Dave pointed out in his video, on a very simple noise demo, this isn't easy. But note that he sniffed out the noise source. Adding line filters did nothing in his case. Just hope the noise source isn't in the power supply(s) to your various instrument modules. Commented Dec 13, 2016 at 21:11
• I've suggested that 10mS could be a 6000 rpm motor, but this 10mS rate could also result from the digitizing rate of an Analog-to-digital sampler beating against a nearby frequency. Your 'scope shots are signals going into A/D from ZOC33? Commented Dec 13, 2016 at 21:23

To re-formulate the question, the system has the following structure:

1. a bunch of pressure sensors are combined into analog block "ZOC33", with addressable multiplexor into a single analog output;

3. The analog link between ZOC33 analog muxer and the ADC is 10ft (3m) long. While the analog link does have a separate analog ground (-OUT) that is separate from digital and power return ground, the +OUT analog line does not have any shield.

The system seems to be distributes over an area of 20-30ft, and the overall grounding wiring is missing.

The max scale output is +-2.5V, and the observed episodic "noise" is 100mV (as compared to presumed "normal" 20mV), or about 4%, which is too much for (I guess) a 12-bit ADC.

My guess is that the problem is related to unprofessional ground wiring across the whole data acquisition system, which has ground loops, and thus is susceptible to external interference from some heavy electrical equipment (motor/pumps, welders, etc.) nearby, or radio emissions. The correlation between noise on power rails and noise in ADC channels is likely because the entire system receives EMI, and all grounds are bouncing everywhere.

Solution to the problem is to hire a seasoned professional engineer to examine the entire setup, check/monitor quality of AC supply network, identify sources of high-power EMI, and suggest correct grounding schema (or power filters on AC) for the entire data acquisition system.

P.S. From the power spectrum of noise (FFT picture), it looks like the building has some poorly designed switching high-power supply unit running at 80kHz, with extremely sharp edges. And sitting on Floor2 AC mains.

• Even there is no earth there is the same issue(by using 2 prong connector). There is no ground loop. Btw is how about the 100nF decoupling caps? Wouldnt that eliminate the DC power supply noise? Commented Dec 13, 2016 at 0:33
• do you mean such filter: en.wikipedia.org/wiki/Line_filter ? Commented Dec 13, 2016 at 0:41
• No, I meant something like this, astrodyne.com/emi-filters/pdf/F1900SP-W.pdf Commented Dec 13, 2016 at 0:46
• I don't think any hack of the form of 100nF here and there will solve your problem. You need to find the source of your heavy interference. Look around, make some sniffer, sniff around. Commented Dec 13, 2016 at 0:54
• When you deal with noise do you "always" try to find the source? I dont get why decoupling caps would no effect. Anyway probably you are right. Commented Dec 13, 2016 at 1:00