Teardown & Cleaning a Laser Dust Sensor PMS7003 After 2 Years

[@moeburn]

THANK YOU THANK YOU THANK YOU!!! My PMS7003 died and I had no idea why, I didn't have the courage to take it apart without a guide. Yours is the only teardown guide of the 7003 model ON THE ENTIRE INTERNET. Upon tearing it down I was able to discover there was nothing wrong with the board, just that stupid fiddly connector, and I instead removed the connector and soldered wires directly to the PCB, and now it works again! HOORAY! thankyou!

[@davidefano]

Thank you for showing how to orient the sensor ( 'connector down' )

[@alexgoldstein7997]

Excellent video! I really appreciate the info!

[@ryanbrooks1671]

Thanks for doing this. I'm interested in the power efficiency of the 7003 and this was helpful.

[@WizardTim]

Glad to hear the video was useful! During the 2019-2020 bushfires here in addition to the one shown in the video that’s setup inside I also put one on a breadboard on my back deck along with an ESP8266 and an 18650 cell (an old 1,800 mAh one) so I’ve had to deal with the power consumption of it as well. Some things that might be useful to you: • It needs a good 5V supply, feeding it 4.4 V after a diode drop will result in the fan not starting reliably or not spinning fast enough. • The manufacturer states you should let the sensor stabilize for 30 seconds after waking it from sleep mode before trusting readings, this is certainly true however you can probably get away with 15 seconds. For my application sampling every minute I didn’t bother to get the sensor to sleep but I did put the ESP8266 to sleep between transmissions as the WiFi consumed a lot of power. • With the ESP8266 mostly in sleep mode only transmitting every 60 seconds and the PMS7003 constantly running I got a little over 12 hours of battery life so I changed the battery at midday and midnight. The whole setup worked out to an average of about 100 mA @ 5V (500 mW) which is pretty much what the datasheet claims.

[@alhajee]

This video deserves more likes 👍

[@alhajee]

I’m working on an air quality module that comprises of the PMS7003, and I’m looking for good sensors for O3, NO2, CO, & CO2. Please, which sensors would you recommend?

[@WizardTim]

@@alhajee I have several gas sensors manufactured by Winsen, I would recommend looking at them. I like the MH-Z19 modules that use the NDIR method to measure CO2, accurate and well priced modules. I also have a ZE07-CO which works well however it is an electo-chemical sensor. I haven't bought any O3 or NO2 sensors as they're somewhat expensive to get good resolution and low interference. Note that cheaper gas sensors often have similar interference gases so you might find it better to use a hotplate air quality sensor like the CCS811 or SGP41 and a CO2 sensor if you just need an AQI.

[@jaro6985]

Didn't realize they were modulating the fan, I guess they do a one time flow calibration at manufacturing. Good explanation.

[@WizardTim]

When you turn it on you can actually hear the fan go to 100% and then ramp down, my GPU does a similar thing and the marketing says it’s "to blow dust out of the heatsink" and it actually works somewhat well, makes me wonder if this does it for the same reason or if it’s just to make sure the fan starts reliably? I also wonder how the factory calibration is done; I suspect they probably use the same calibration constants for all units and test them on a pass/fail basis as they're not super accurate and tend only to agree with each other to about 10%.

[@jaro6985]

@@WizardTim oh good idea with ramp, yeah they are not super accurate but l suspect you can use serial commands to change the constants. Of course you'd have to have some $10k+ reference meter for that to be relevant.

[@sofronius2]

Thank you for thorough vid & explanation, nicely done

[@ansible9340]

Thanks for the video mate, there was a lot of white noise in the background but everything else was perfect and thanks for subtitles too.

[@mortezasafai7983]

Thanks Tim

[@charlottepepin4860]

Thank you so much for the video! This may sound like a silly question, but how do you know the fan pulls in 0.1L/min? I couldn't find the information on the data sheet of the sensor. All I could find is that the volume of air is 0.1L.

[@WizardTim]

That's a good point. Unfortunately I don't remember where I got that number from. I can only assume I misread the datasheet's 0.1 L volume as a flow rate of 0.1 L/min as that seems suspiciously low considering a typical 20 mm blower fan like that can do 50 L/min, although they're not running it at full speed in this sensor the fan couldn't operate at 0.2%, it would stall before then. Taking a complete guess I would say somewhere around 10 L/min is a more reasonable flow rate to expect.

[@scelikr]

Hello! Thanks a lot for the detailed video! I am using PMS7003 and its readings are jumping very much. The readings are usually about 10-15 ug/m3, at the same time, about 1 time per minute they jump ~ 200 ug/m3. Could you guess why this is happening? What do I need to do?

[@WizardTim]

Hmmm, it could be a lot of things so unfortunately I can only give you a couple pointers of what to investigate, could be: - 5 V supply is low, it needs a solid 5.0 V, it won't work properly at for example 4.6 V - 5 V supply could be noisy and causing interference, likely if you have other power electronics sharing the same 5 V or maybe a loose wire - A programming issue or noise on the TTL serial TX/RX lines possibly from a loose wire (I think there's a checksum in the packets it sends, not sure if most libraries check it's valid) - Dust on the internal photodiode (unlikely if you've just bought it)

[@zen123w]

Great video! If you had this running continuous in a home/office setting and cleaned it once a year, what do you think would be the life expectancy of one of these devices?

[@WizardTim]

The manufacturer says the Mean Time To Failure (MTTF) is > 3 years without cleaning in a typical indoor application. The sensor in the video I still have and it's working fine now 5 years old, along with 2 others that aren't quite as old. The fan inside is the thing that will probably break first, in a normal indoor application with ~5 μg/m³ of dust I think 10 years would be a reasonable expectation for the limit of it's useful life.

[@WizardTim]

If you want a dust sensor with a very long life expectancy there are draft type sensor available, they heat the air inside to make the air rise though the measurement chamber without any moving parts, but those devices consume more power and are typically less accurate.

[@rolandberendonck3900]

Great video Tim! Thank you for being so informative about this sensor. However I am still not completely convinced whether this is the best sensor to use. Currently I have used the SDS011 sofar, but that one has also got his issues. For instance it produces much more noise which is annoying if you use it inside. Otherwise is it somewhat bigger and is it more fit to be used in a more turbulent environment and even a hose can be applied. The teardown and cleaning up will also be much more easier since it is bigger and has a simpler construction. What I do not like about these sensors is that the output is all but steady especially in the PM10 range. So anticipation by a MCU is quite difficult because the measurements are goint up and down just in one minute. If you would have a DMM like that you would just throw it away. Do you have any experience with the Winsen dust laser sensors. It seems to me that they are exactly the same as Plantflower ones but they are much cheaper.

[@WizardTim]

I haven’t used the SDS011 from Shandong Nova Fitness, but I can say the 4 I have from Plantower are pretty quiet (at least a lot quieter than the 2U server in the same room). The values I get from them have a clear trend and are smooth with very little random noise unless I introduce dust then it changes rapidly, although I haven’t done much sampling faster than 1 minute intervals but I have had bad connections causing random data before. Zhengzhou Winsen makes some good sensors, I have a few of their CO, CO₂ and CH₄ gas sensors but I haven’t used their dust sensors, after a quick look I see they have the ZPH02 which isn’t a laser dust sensor but an LED dust sensor which only gives very rough data and can’t determine particle size, they also use a draft induced by a heater instead of a fan so it would be silent, they are also a lot cheaper but I imagine not very accurate or precise.

[@rolandberendonck3900]

@@WizardTim The silence of the Plantpower sensors is indeed amazingly silent!!! Perfect for indoor use. I blew one up; a pms3003. Got the leads connected the wrong way :( But in your video you mentioned it could probably be fixed by soldering a new voltage controller. I have another 5003 and a 7003 is on the way so I can compare them and see if they suit me. Winsen has also got laser dust sensors type ZH03B and they are a lot cheaper than the Plantpower ones and according to me they are exactly the same as the PMS3003. The lifetime of all these sersors is around 10000 hours, but I think that number does not take into acccount when the sensor is cleaned in the way you showed in your video. Maybe the laser itself can have an even longer liftime than 10000 hours. At least I hope so because cleaning these sensors is well worth it if it works again like new. And it is probably be possible to find new fans for these things since that should just be a standard kind of fan.

[@WizardTim]

@@rolandberendonck3900 I’d say you have a good chance of repairing it from a reverse polarity however the internals of the PMS3003 are different, I believe it uses a SOT-223 AMS1117 regulator rather than a SOT-23-5 one. I expect there are 3 main contributors to the life expectancy: - Dust accumulating affecting the flow rate giving falsely low readings - Laser aging resulting in dimmer reflections thus fewer and mis-classified particle sizes - Failure of fan bearings resulting in complete failure Plantower says it should last >3 years in normal conditions and I certainly believe you could get 5 - 10 years out of it indoors. And I see what you mean the ZH03B looks identical to the PMS3003 but personally, the PMSA003 is my favourite, but it’s of course the most expensive.

[@rolandberendonck3900]

@@WizardTim The PMSA003 is about 13 Euro shipping included. Is it the best Plantower sensor? I thought the PMS7003 was the latest type and the best? An SDS011 costs about 17 Euro. Still not very expensive for such a sensor. And does this PMSA003 has the stable and smooth readings or do they all have that? I am using my own moving average code to smooth the readings out, but the readings are so jumpy it almost makes no sense. Another thing that surprised me about the PM5003 sensor was the resolution which seems less than the SDS011. I get readings in 0.1 ugram/m3 resolution whereas the Plantpower just jumps from 1 to 2 which is quite a big step. That is a factor 10 difference. I don't get that. Just to make things clear Tim, I am looking and searching for the best sensor with the most stable readings. And unfortunately there is very little content and info about these sensors I think.

[@rolandberendonck3900]

@@WizardTim Typical lifetime of laser diode modules are 25,000 to 50,000 hours. www.worldstartech.com/what-determines-the-lifetime-of-a-laser-module/ Since laser diode lifetimes can be into the 100,000 hour range it is generally not practical to test the laser diodes at normal operating ranges due to the long test time. www.newport.com/medias/sys_master/images/images/hbc/h43/8797050241054/AN33-Estimating-Laser-Diode-Lifetimes-and-Activation-Energy.pdf

[@alhajee]

I’m curious how you got the thermal image. That’s genius of you!

[@WizardTim]

The image was taken with a FLIR E8 thermal camera with a custom macro lens. Thermal camera is often step 1 whenever I troubleshoot something now, super useful.

[@Muonium1]

I don't think you have a counterfeit, I think they're just iterating their designs very quickly. I have one from a few years back and remember them coming out with a new model number immediately after I got it even though the one I had was really new. I really think there's no wayyyyyy these cheap things are actually "counting" anything. The number of particles in the beam at any given time, the total lack of focusing optics before the photodiode, etc. all point to the certainty that what they're actually doing is just periodically grabbing a light intensity value from the photodiode and shoving it into an algorithm that lets them guesstimate the PM2.5, 10, etc. values based on overall light scatter at any given time. The question is, did you see any statistical difference in detection efficiency after cleaning?

[@WizardTim]

Yes I agree it probably isn't a counterfeit, the plastic injection molded case and folded metal parts were identical, it was just strange to see the older datecode one was the one with a custom MCU and overall better mechanical design, my guess is I got two units from when they were switching production lines. After cleaning it there was pretty much no change in readings, maybe a slight increase of 1 μg/m³, but nothing major.

[@ddlyhm]

that vid is very helpful! thx a lot

[@BorisSiber]

thx

[@rubensantanarodriguez2001]

Your explanation is excellent, it has helped me a lot. However, the fan does not turn and it gets very hot. Do you know why it doesn't spin? Thanks a lot. 🥰

[@WizardTim]

Thanks! Hmm, I assume this is one you're using yourself with an Arduino or something? Typically when something doesn't work and just gets hot it's usually because it's electrically damaged which is what happened to mine at 3:09, I plugged it in backwards and instantly blew up that regulator IC. I'd double check it's the motor and not one of the ICs that's getting hot, also check the fan rotates by hand without power to check it's not mechanically jammed. Another thing I have noticed with them is they really need a good 5V supply, the 5 V pin on an Arduino or similar will often cause problems because they have a diode between the USB port VBUS and the 5 V header, check with a multimeter it's like >4.90 V and not something like 4.4 V (5 V minus a diode drop). Cheap thin breadboard jumper wires can also make poor connection or break internally so check them over as well.

[@serenityslife4910]

Are these dangerous? I have a mila air purifier. It says it has a laser sensor. I am always next to it. Is it bad for my health? Aren’t lasers radiation?

[@WizardTim]

Those aren't dangerous and your Mila air purifier is perfectly safe. "Laser radiation" is just a technical term for a special type of focused light that's "non-ionizing", the laser in those sensors are what's called "Class 1", which is perfectly safe. You're probably thinking of "ionizing radiation" (like X-rays) that's dangerous but that's very different, you very likely don't own anything that emits ionizing radiation.

[@alanmangroo3656]

Interesting video. What did you use to log your data over the years?

[@WizardTim]

The sensor is connected to an ESP8266 via UART and I have a cron job running on a server (same one as in my “ZFS Disk Failure and Recovery” video), every minute it sends a HTTP GET over WiFi to the ESP8266 which returns a CSV string which is then saved to a .CSV file, the plot is generated in MATLAB. I’ve been considering switching over to JSON, MQTT and an SQL database to make it more reliable, let the sensors sleep between transmissions and make querying data easier and faster but I haven’t had the time.

[@alanmangroo3656]

@@WizardTimThat is a nice and simple solution. I've been working on an SDS011 based sensor writing JSON data to AWS DynamoDB mainly as a learning project. I've tried Thingspeak which works very well and lets you present some simple charts.

[@kamiko-i5g]

@@WizardTim I just set the same sensor at my place. I used the esp8266 on which I installed tasmota firmware. The tasmota sends data via mqtt to a domoticz installed on a router with openwrt. This does not require a large configuration and you can immediately see graphs and export data e.g. to csv

[@WizardTim]

Nice! I love the Tasmota firmware as well, one of the best open source community driven firmware projects I’ve used. I have a SONOFF Pow R2 which runs Tasmota, it’s configured to send MQTT messages to the same server which runs a Mosquitto MQTT broker and Node-RED which timestamps it and saves it to a .CSV for later visualization in MATLAB, all 401 MiB of it! (yes it's slow to parse hence the desire to go to an SQL database)

[@morpheus3671]

эх скетч бы для этого датчика , для ардуино uno .

[@WizardTim]

Sorry I don't understand Russian but if you're looking to use this sensor in an Arduino sketch with an Arduino Uno try this library: github.com/avaldebe/PMserial

[@morpheus3671]

@@WizardTim thank you very much, but I don't understand this at all, all I can do is upload a sketch to arduino =), no one wants to help, even for a fee, everyone either does not answer, ignore, or don’t want to help. very bad people and society, lazy and selfish ... a big problem, neighbors are burning fires, the stench goes into the supply ventilation, need to control the plasma filter through a relay

[@WizardTim]

Sadly it’s very difficult and time consuming to teach someone how to do a project like that especially over the internet in a different language. To make it easier to follow you should make a post on the EEVBlog forum and explain what you need to do, what you have, attach the sketch you’re using, what you’re stuck on and attach a picture of the hardware. www.eevblog.com/forum/beginners/ I might be able to help so leave a link to the post but the regular members usually know more than me.

[@morpheus3671]

@@WizardTim thank you

[@rolandberendonck3900]

@@morpheus3671 Hi Morpheus, to make this sensor work with Arduino you should download the Adafruit software ("Project Bundle") and tweak it a little which is very well explained by Adafruit itself. Here is the link: learn.adafruit.com/pm25-air-quality-sensor/arduino-code