Evaluation Type: Independent Products
Did you receive all parts the manufacturer stated would be included in the package?: True
What other parts do you consider comparable to this product?: gp2y1010au Sharp dust sensor
What were the biggest problems encountered?: OLED screen failed following a few drops, assembly of sensors to shield is not very positive, humidity probe wire is very stiff
I have used this kit in a real world situation for several months and it has proven to be not only a useful tool but incredibly educational. The range of the sensors within the kit make it an excellent choice as a starting point for understanding air quality measurements. The kit offers extremely good value compared to buying the parts individually and is supported by clear, well documented and considered code for the Arduino. There is some room for improvement but my use of this kit is far from over!
This has been a great experience and I would like to thank neilsroberts for providing the kits
Let see how I got on. . .
Background and Application
I have been working in the automotive sector for ten years, and my company have been asked by a customer to investigate opportunities for improving air quality in the cabin. This is especially important in China where air quality monitoring is as common as checking for rain in the UK. My electronics (hobby) background gives me a good basis with Arduino, metrology and sensor kit design and packaging.
I am well experienced with cabin temperature and humidity and have a good understanding of what creates ‘comfort’ but my experience of air quality is a little limited. I expect to improve my understanding of typical cabin air quality and how it changes during a drive cycle. I will compare this against the air quality of the building I work in. I will explore the effect of some basic changes to improve cabin air quality with before and after evaluation of the changes. If time allows, I will compare the readings of the sensors against metrology systems commonly in use within the automotive.
You can read more about my back ground here Begin at the beginning. An Introduction.
The EPA Air Quality Index is a great place for acceptable limits of substances in the air.
Unboxing and First Use
Temperature and Humidity
The T9602 temperature and humidity sensor is an excellent package size for use in the cabin. Typically I would use a much large probe which is intended to be used outside the vehicle and becomes quite clumsy inside the cabin, so the T902 is a really good solution for me in every day testing. Operating range and accuracy for humidity is right where I need it to be for my application, although a temperature operating range of -20c and +70c would probably cause me issues. I regularly see cabin temperatures exceeding these limits (real world) which suggests the sensor is geared at indoor measurements only (despite being IP67 rated).
One issue I did find, repeatedly, with this sensor was the cable. It was far too stiff for the way I wanted to use it and I wanted to measure temperature and humidity in the same air stream as the dust sensor, this became quite a problem. Ultimately I resorted to a spot of carpentry to wrestle the cable under control.
The SM-PWM-01C dust sensor is a far superior looking sensor to many of the cheaper rivals, perhaps simply because of the black PCB or neatly arranged components with clear and contracting silk screen. It has very good performance to match. While more than double the price I have paid for an older designed Sharp sensor, GP2Y1014AU0F, the SM-PWM-01C reading was more stable and consistent perhaps simply because it is digital.
Like the temperature and humidity sensor, the dust sensor is also specified to be used indoors in a more controlled environment than I would normally test in. An operating temperature between -10 and +60c just won’t cut it in an automotive application. I have now tested at -20c in a controlled environment without issue, although I have no reference sensor to test against. It is also not clear how the sensor would react to sudden temperature changes. I suspect the readings might be out if condensation forms on the reflector mirror or IR sensor.
Without an aspirator fan, the dust sensor must be located in a vertical position and directly in the flow of air to be measured. For an automotive application this simply won’t be possible to package, so it is likely that that an aspirated version from a competitor will be required.
Despite the limitations for application, the learning derived from this sensor has been significant and invaluable as we will see below. I suspect a laser version of the sensor would potentially overcome many of the limitations.
The T6713 carbon dioxide sensor is truly remarkable and the part of the kit that I have most enjoyed working with. To think that so much information can be gleamed from such a small package is astounding. Again operating and storage temperatures would cause concern for automotive application, but the potential for this sensor in the cabin makes me think this limitation could easily be overcome. I’ll come back to this sensor later.
Okay, the 128 * 64 OLED is not a sensor, but it is a significant part of the package. I’m used to using Arduino with screens and have plenty lying around, but to include it really does make the kit a full and complete package. I like the colour combination and it is very clear in all ambient lighting conditions.
Unfortunately in my case, a few unexpected trips around the foot well of the car didn’t do the screen any good and it failed. It actually took some time before I found this out as I had only been running with the logger and not the display output.
~$20 from many sources
Value for Money
I've put the DigiKey current prices in to show how good the value of the kit is compared to buying individual sensors is.
Total price for sensors alone is $181, the Digi-Key price for the whole kit is just $143.55, and the includes the shield and Arduino compatible with great source code to go with it!
For an initial test, I set the kit up on my desk and let it record for a day. I expected to see nice stable conditions as the desk is far from any doors (or windows) and well-conditioned. I wasn’t disappointed, the logger worked well, the traces are relatively flat and the spike in CO2 around 3000s can be attributed to a desk based meeting of minds (or lunch).
The two spikes are caused by movement rather than a bad sensor.
The values meet previously defined expectations giving good initial confidence in the readings.
At the time of writing I have logged over 50 hours of data in car with the sensor mounted between the head rests of the car (yes that is a stick and cable tie assembly).
I don't intend to bore the reader with a graph of every single trip, but I thought it would be interesting to see some of the real world testing
The graph below shows a typical drive cycle, short town traffic followed by highway (60 to 70 mph) for 40 minutes. At initial look it would appear that dust level is proportional to speed, but further data analysis has shown that the other vehicles on the road at the time have a bigger influence. For example if I follow a couple of lorries then the dust level is far higher than if I am on an empty highway. I put this down more to up-draft of dust from the road surface than vehicle emissions.
It can also be seen that the climate control system is doing a good job of managing the temperature and humidity of the cabin.
For similar driving conditions, with AC off (compressor disengaged) we see a notable increase in humidity as the air is not being dried.
Again for similar driving conditions, with the climate system locked in re-circulation mode, we see a steadily growing CO2 level which reaches levels consistent with poor concentration, loss of attention etc. This is not a surprise as the cabin air is not being replenished, but does offer an opportunity to improve cabin air quality by limiting the duration the system can stay in recirculated air mode for example.
It is clear in all of these traces that the CO2 level starts quite high and drops down. The technical note for the sensor does specifically note that the sensor takes 5 minutes to warm up and give a stable reading.
I don't have access to a calibrated CO2 or dust monitor, comparing the AAS dust sensor to the Sharp sensor while comparable levels the AAS sensor gave a more robust value (not cutting out) with far less noise on the reading making it more reliable for making decisions from.
I was able to use a calibrated Testo meter for temperature and humidity, the results where very pleasing, the AAS sensor having slightly more noise but both well within measurement error.
I would like to compare against calibrated or reference air sample in the future.
What is Next?
Well I have by no means completed the testing I want to do on this kit, it will continue to run in the vehicle for a good while yet. It is also likely that my company will look to buy another set of kit to use in our overseas development centre. Personally I would like to get a case designed and made and have this series of sensors as a standard part of our instrumentation list.
Pleased you like the kit, as you say it is principally designed to evaluate sensors in industrial applications, and is not designed as a product per se, interestingly we are working on something that is…
Very good Road Test Report.
You did a great job of reviewing the device and showing use data over time.
Well written and you clearly made your points about the device and software.
Hi Neil, will you be adding instructions on how to add the VZ-89 VOC sensor to the kit? I have a tiny bit of Arduino experience (use it for building keyboards/keypads), and I'm about to purchase your kit for monitor things in my home and office. I'm a bit unclear about how to step down to 3V, and how to add the VZ-89 code to the original ino file. I understand this kit isn't meant for total newbies like me, but it'd be great if there are some concrete instructions in the future for adding your VOC sensor.
Couple of answers to comments earlier.
Your Github load worked, this is exactly our intention for the code, its a starting point for others to develop into what they need.
The temperature and humidity probe is normally found in industrial ventilation assemblies, which is why it is a little less flexible that you needed, but we have assemblies for other uses.
The dust sensor in the kit is the lowest cost version we do, it uses an infra red LED as light source and aspirates by convection. In order to make sensors more accurate a number of things need be done, the first of which is a brighter light source to get more light scatter, so for this we use a laser. Next on the list is controlled aspiration, dependent on application we use either an axial or centrifugal fan, these then control the air volume more tightly. On industrial level our SM-UART-01C part has these features, but of course is more expensive; but more accurate.
Automotive application has many other challenges, the operating temperature is much wider than the industrial sensor, we need to consider humidity to compensate the signal, and of course testing and longevity are prerequisites. Some Auto makers are also considering measuring outside values as well, which adds another set of environmental challenges for the operation of the sensor. We have sensors that meet these needs, but obviously we needed start at low end for users to get experience of the indices, expectations and values.
How did you pull the speed data from the OBDII port with the Arduino board. I like how that helped you correlate some data and looks like something I would Ike to do for other projects.
Without an aspirator fan, the dust sensor must be located in a vertical position and directly in the flow of air to be measured
Why the need for a fan, the internal heating element is used to get a slow and steady air flow through the sensor. I would thing a fan may cause some undesired affects and affect the readings as well. Just my thoughts.
Quick observation, why the use of Digikey prices and not those of Newark.com ? Otherwise interesting review, still working through it though.
Fantastic review. We appreciate the level of detail provided!
For completeness I have now uploaded the Arduino code (without libraries) that I used for the test into GitHub;
I've never uploaded anything to Github before, let me know if there are any problems.
Thank you for the feedback neilsroberts
Using a rolling average for the dust sensor, in my experience, is needed for automotive use. Ideally the vehicle occupants don't want to know changes are happening so they have to be made slowly. I'll have to have a play with the rolling average duration.
We know we can use alternative sensors for total dust consumed (pressure drop across a filter) but typically the cost is too great for our customer to add to their product. A dust sensor is easily sold to the end user as it is common place in the worst case market, they are used to having them around and expect to see them in the vehicle. I would be interested in more information about the automotive version of the dust sensor. We are currently looking at adding one to a customer package, and have identified three which we will be bench marking before the end of the year.
Thanks for the notes on the VOC sensor, I had realised that it was 3V, but didn't realise it would alter the calibration, I expected a release of smoke. I'll get some assignments out of the way and pick back up with the VOC measurements.
Pleased you like the kit, as you say it is principally designed to evaluate sensors in industrial applications, and is not designed as a product per se, interestingly we are working on something that is.
As some parts are sourced from different vendors (we don't make screens) there is some differences in nomenclature, hence pictures..........
And we deliberately chose low cost PCB headers to keep the cost right down, unfortunately this does have the disadvantage that things can be connected wrong way around. We are just starting another shield design for other sensors, so your comments are noted.
The dust sensor has a rolling one minute average built into the code, else dust values do jump about - unlike other measurands dust does not go through a sensor in a nice even stream, government statistics are based on 24 hour averages but this is obviously impractical, I've done some work on this and longer is definitely better, in industrial products I'd recommend a 2 minute rolling average. Incidentally the value in brackets on the display is the calculated value used in µg/m3, however the sensor used is more qualitative (is it dusty?) rather than quantitative (how dusty is it?), but works sufficiently well to give a colour code. Our soon to be released SM-UART-01L is more accurate than the 01C used here, but is unsurprisingly more expensive. We are developing an automotive dust sensor that has better characteristics, and is qualified for that application, as we also have auto qualified temp, hum, and CO2 sensors.
I note you have a MiCS-VT89-TE VOC sensor to add, there is some code for that on GitHub, be aware it is a 3V device, and putting 5V on it will alter calibration.