9
\$\begingroup\$

I'm using a Shinyei PPD-60PV particulate sensor in a product and I've noticed something very odd in testing, and I don't know how to explain it. It's connected to a WildFire board by way of cabling to an interface adapter board. The WildFire is powered by 5V through its USB port. The PPD-60PV has two 5V/GND connections made through the interface adapter board, and an analog output that is wired to the A7 ADC input of the WildFire by way of the interface adapter board.

My product supports two fundamental modes of operation: (1) Wi-Fi connected, and (2) Offline. What I've discovered is that in Wi-Fi mode, the analog output of the PPD-60PV sensor appears to rise by about a volt. What I discovered (and painstakingly narrowed the symptom down to) was that this voltage rise happens somewhat gradually (over several seconds), only after the ESP8266 connects to a Wi-Fi network. It also recovers to a normal baseline value gradually (over a similar period of time) upon resetting the ESP8266 (and hence disconnecting it from the Wi-Fi network).

Further diagnostic experiments reveal that this voltage rise on the analog output of the sensor occurs even if I disconnect the analog output from the WildFire altogether leaving the 5V / GND connections in place and probe it with an oscilloscope.

Also if I have two assemblies plugged into the same power source, with one of them in Wi-Fi mode, and one of them in offline mode, the offline mode unit exhibits the voltage rise phenomenon. The rise is certainly there, and also noteworthy that it's to a lesser degree than when the unit itself is in Wi-Fi mode, e.g. 600mV - 700mV.

An offline unit connected to an isolated power source (e.g. a battery pack) does not experience the voltage rise, e.g. despite physical proximity to a Wi-Fi connected unit.

I wondered if maybe it was a ground path resistance issue, but everything is pretty short lengths here, and I measured the resistance from both sensor ground connections back to the WildFire ground at 0.2 ohms each, and I measured the total current of the system at about 300mA (displayed on a the LCD of a conventional benchtop 5V power supply). That certainly doesn't account for a 1V rise as far as my reasoning goes.

My understanding was that the PPD-60PV analog output is a low-impedance buffered output, but that's not entirely clear from the datasheet. I'm kind of stuck / perplexed at the moment, and I'm not sure what to do next.

So, on to my forlorn question. What could be the root cause of what I'm observing here? What advice do you have as to what I might do next to drive this issue to ground, as it were?

\$\endgroup\$
  • 1
    \$\begingroup\$ Maybe the sensor is EMI sensitive. \$\endgroup\$ – Spehro Pefhany May 17 '16 at 19:08
  • \$\begingroup\$ EMI or fields are out of suspect, because of what the OP mentions about the isolated supply there. @SpehroPefhany Vicatcu, did you observe the 5V supply with oscilloscope? I suspect that it is all about voltage referencing circuit on the sensor, directly related to supply voltage, which has voltage drop spikes. Recall that USB power has a narrow linear range. The slow react can be explained if referencing circuit has some passive filtering. You don't mention anything about the battery powered device's wifi mode, if it experiences the same disorder. \$\endgroup\$ – Ayhan May 24 '16 at 11:11
  • \$\begingroup\$ Please provide more information about your setup, like a schematic or PCB drawing picture. If your using off the shelf components, post a block diagram. Also, what is the time constant of your sensor (How fast can it respond to changes) Thanks \$\endgroup\$ – Voltage Spike May 25 '16 at 16:14
  • \$\begingroup\$ @vicatcu - Hi - Nice problem description. "What advice do you have as to what I might do next [...]" - I see some "missing" (or at least, not mentioned) tests, whose results would progress finding root cause. However, the lack of any response (positive or negative) to the 2 previous comments, suggests you might not need further help or perhaps have even solved the problem? So to avoid me wasting time on suggestions that are no longer needed, could you give an update? Thanks. (Also more info about num of available power sources and available 'scopes & num of channels on each, would help me.) \$\endgroup\$ – SamGibson May 27 '16 at 21:32
  • \$\begingroup\$ Just bad timing with memorial day and other stuff going on, I'll get back to it \$\endgroup\$ – vicatcu May 27 '16 at 22:23
3
\$\begingroup\$

If your system uses a photodiode for detection, it is attached to a relatively high gain amplifier/integrator, and strong electromagnetic fields (wifi) can result in induced AC voltages that are rectified by the diode junction and appear on the output. If this is your problem, you can solve it by increasing the distance from your wifi transmitter or additional shielding around the photodiode. I'm betting your sensor has some shielding around the photodiode already.

\$\endgroup\$
3
+100
\$\begingroup\$

For some reason, the particle sensor is prone to picking up high frequency noise from the 2.4Ghz band. Since you don't have any control over the PCB layout or circuit of the particulate sensor, the options you have for EMI control will be restricted. There are a few things you can do.

1) Let the manufacturer know. There is a remote chance that they might help you with the problem

2) Shield the unit
First put the unit in a metal enclosure with only holes for the analog and power signals. The best metal enclosure would be made from copper, use copper tape to close up any unnecessary holes. Aluminum can work but is not as good shielding material. There are two ways the 2.4Ghz signals could be effecting your sensor. One is conducted emissions through the power and analog wires that connect to the board, the other way is through the air.

If you put a metal enclosure (no holes except for power and the analog signal), and you still get the signal rise. This would suggest that the signal is getting through the wires. If its getting through the wires, then increase the inductance by adding ferrites and filter capacitors. Ferrites increase the inductance of the wire, and can be added to the outside of the wire. High frequency signals always take the path of lowest inductance, increasing the inductance will "alter the current pathway" of the signal similar to the way increasing resistance decreases current in the situation of a parallel resistive load.

If you don't have a problem with conducted emissions, great. The particulate sensor won't be able to operate without access to air. So then you will need more experimentation with putting holes in the box to allow for sufficient airflow while blocking high frequency signals. Grounding the box may help, experiment with grounding it at different points some will be better than others. Since I can't see your setup, I can't comment a good position for the ground.

EMI problems take testing and patience, good luck.

\$\endgroup\$
  • 1
    \$\begingroup\$ If there was an easy and reliably way to solder Aluminium, it'd make a great EMI shield... \$\endgroup\$ – Sam May 28 '16 at 6:27
  • \$\begingroup\$ I think I'm going to award the bounty to this answer as time is running out on it, but I'll keep the question open at any rate until I get to the bottom of it. Passing the cable to the sensor through a ferrite core is an interesting idea I hadn't thought of trying. Not sure where to get such a thing off-the-shelf, but I'll have a look. \$\endgroup\$ – vicatcu May 31 '16 at 19:25
2
\$\begingroup\$

It would appear that your problem is conducted EMI (not radiated) from the WiFi module. Try blocking any RF currents in the power and signal leads with ferrite beads. Better still, construct a pi-network filter for each lead by also adding capacitors to ground on either side of the bead.

schematic

simulate this circuit – Schematic created using CircuitLab

Keep all leads, especially on the ESP8266 side, as short and direct as possible.

\$\endgroup\$
  • \$\begingroup\$ The ESP8266 is not connected (directly) to the sensor in question. The Sensor and a separate microcontroller (ATmega1284p) are powered by 5V, the ATmega1824p is connected to the sensor and to the ESP8266, and the ESP8266 is powered by 3.3V derived through an LDO from the 5V... \$\endgroup\$ – vicatcu May 31 '16 at 19:21
  • \$\begingroup\$ Yes, I understand that. The box labeled "Your Circuit" encompasses all of that. The point is, even a ground wire or power through a regulator can be carrying stray RF interference. The pi network filters keep it confined to just the ESP8266. \$\endgroup\$ – Dave Tweed May 31 '16 at 19:24
  • \$\begingroup\$ Ah I see what you mean... unfortunately that would require me to re-spin WildFire which I can't manage at the moment. Certainly it's worthy of consideration for v4.1. I've never seen Ferrite beads used in series with ground either, that's interesting. \$\endgroup\$ – vicatcu May 31 '16 at 19:28
  • \$\begingroup\$ It's equivalent to what happens when you put a ferrite ring around an entire cable. \$\endgroup\$ – Dave Tweed May 31 '16 at 20:14
0
\$\begingroup\$

The sensor may be affected by RF radiation. I have seen this effect at work on factory mass-production product.

One way to test is

a) connect power to sensor

b) monitor output by battery operated multi-tester

c) use a separate USB LiPo battery pack to power the ESP8266 and put it in Wifi connected mode. Since there is no physical wire connection between the ESP8266 and the sensor / sensor powersupply / multi-tester, any effect can only be through RF radiation

d) vary distance between the ESP8266 and sensor, says, from 3 meters to a few centimeter

e) observe if voltage rise occurs when the distance is small

EMC Susceptibility is a known issue. It is common for mass-produced electronic products to go through EMC Susceptibility testing as part of the certification process. see wikipedia "Radiated field susceptibility testing typically involves a high-powered source of RF or EM pulse energy and a radiating antenna to direct the energy at the potential victim or device under test (DUT)."

The test transmitter creates field strength at xxx V/meter and sweep over a wide frequency range. For example, EN61000-6-3 is 30 MHz— 230 MHz, 30 dBuV/m and 230 MHz— 1 GHz, 37 dBuV/m.

\$\endgroup\$

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.