I’m visiting this particular system at the moment, and there’s no visible sign of sensors on the internal pipework. I assume they’re buried somewhere in the external unit?
Also, wondering if the Vaillant attempts to maintain a particular dT? Would an under-reporting return sensor (so over-inflated dT) cause the unit to behave differently? I’m seeing persistent cycling here, and rarely any steady state running.
The sensors are in the outdoor unit:
The Vaillant does not try for a certain dT but rather aims for a specific flow temperature. If the current flow temperature is above the target value, an energy integral is counted up and the compressor turns off when it reaches zero. As soon as flow temperature drops below target, the integral decreases and when a target integral value (default -60°min) is met the compressor turns on again. So a flow temperature error would lead to potentially erratic regulation.
But an inaccurate observed dT would still result in an inaccurate reported COP, wouldn’t it?
yes, which is what various people see. Some over and some under. My dT is under-reported and so is my COP.
Thanks for confirming.
Such a shame that Vaillant is saving money at the cost of their customer satisfaction and brand reputation to save literally a euro worth of better sensors out of a multi-thousand euro sale price.
They could easily have saved more elsewhere. (For example the removable all-metal front of the uniTOWER.)
Unfortunately, good sensors don’t sell devices, as opposed to shiny metal fronts. I’m sure some beancounter sat down to calculate that replacing the sensors for the few people who care is cheaper than having good sensors in the first place. My pump significantly overestimates COP so if I trusted those COP figures I would tell everyone how efficiently my Vaillant heat pump works! Finally the aroTherm plus are marketed to achieve high flow temperatures for old houses; whicht comes also with high dT. At 10K dT noone even cares about a 1K error, much less a 0.5K error. It’s us nerds with the super optimized flow temperatures that see and care about those inaccuracies 
.
It is possible that the sensors are “good” and it’s just the installation of them is poor and inconsistant.
Sure, even the best sensor needs proper installation and calibration. For me, the sensors don’t even equilibrate after 8 hours of the compressor being off. That really points towards a bad sensor quality or bad calibration. At that point they should show the same temperature even dangling in free air.
Where from? Do you have a link? Is that a photo of them?
I ordered them from here (site in German only it appears) where you can customize the sensor to your liking.
Rohranlegefühler mit Edelstahlhülse (sensorshop24.de)
I have two PT1000, 4-wire, waterproof (for calibration in a water bath) version with 1/10 DIN accuracy which costs around 25€ apiece (depending on cable length and insulation type). They come with proper fastening clamps and I put thermal grease (Arctic Mx5) under them. They are properly thermally isolated by the Rockwool insulation around the pipes.
I read them with two Adafruit MAX31865 boards I bought off Amazon):
The sensor boards are connected via SPI to a NodeMCU ESP32 where I am doing some custom processing and averaging and then push the data via MQTT to emoncms (happy to share code/schematics if desired).
The sensors were extremely close in measurements out of the box (<0.05K difference over the whole range from 0 to 70°C). I still put them in an ice bath, put my sous vide in there and gradually raised the temperature in 5°C steps from 20°C to 70°C. Based on the datapoints from the temperature plateaus I calculated a 2nd-order correction function to match up both sensors over the whole range, i.e. only a relative correction since I don’t have an accurate calibrated temperature probe as ground truth.
@glyn.hudson and I have been involved in a remedial project locally, a poorly performing LG high temp heatpump has been replaced with a Vaillant arotherm, the good news is that the COP has improved from 2.2 with the LG to 4.8 so far with the Vaillant 
It’s been interesting comparing the Vaillant app figures to the class 1 electric and class 2 heat meter that we provided to the installers. Here’s a screenshot from the app:
and comparison with electric and heat meter values:
| Vaillant app | Class 1 electric & Class 2 heat meter | Error | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Date | Electric | Space Heat | DHW | Combined | COP | Electric | Combined | COP | Electric | Heat | ||
| 29/03/24 | 30.4 | 175.1 | 7.9 | 183 | 6.0 | 30.1 | 144.8 | 4.8 | 1.12% | 26% | ||
| 30/03/24 | 21.9 | 100.1 | 16.9 | 117 | 5.3 | 21.9 | 94.4 | 4.3 | -0.15% | 24% | ||
| 31/03/24 | 14.4 | 91.1 | 1.8 | 92.9 | 6.5 | 14.0 | 73.3 | 5.3 | 3.20% | 27% | ||
| 01/04/24 | 19 | 107.7 | 5.7 | 113.4 | 6.0 | 19.0 | 90.9 | 4.8 | 0.18% | 25% | ||
| 02/04/24 | 13.3 | 77.4 | 6.3 | 83.7 | 6.3 | 13.4 | 67.0 | 5.0 | -0.46% | 25% |
The electric values are pretty good but looks like a temperature offset error is pushing up those heat values again. This has to be a relatively easy one for Vaillant to fix.
@glyn.hudson got some good pictures of the temperature sensors I think on his parents system, they look like they are in pockets and so should be good but perhaps they are just not calibrated pairs??
Here’s what I think are the temp sensors, they look like they are in pockets with two red cables attached:
It looks like one of these Vaillant sensors, the same as they use for boilers (part No. 2415622). Can anyone confirm that this is the exact part? Since it’s an OEM part I can’t find any spec data:
Update 5/08/24: the sensors are actually clip-on, see recent post
Here’s the flow sensor:
It’s hard to see how pocket sensors could produce errors as big as I’m seeing…
30 day COP of 6.55 doesn’t seem believable, and the trend doesn’t match the outdoor temperature like others do. MyVaillant is overestimating heat by about a third, but not consistently.
Could there be air or something messing with the flow sensor? Is there a setting within Vaillant to tell it about glycol?
I guess it’s possible that sensors in pockets could be very poorly calibrated.
Can you compare the flow sensor reading with the flow reading from your heat meter? The Vaillant flow meter looks like a Vortex for meter which are usually pretty good in my experience.
Not that I’m aware of. Have you got glycol?
No heat meter on this system (my father’s).
Yes. If the system expects pure water, then it would assume a higher specific heat capacity, and therefore over-estimate the heat delivered?
Looks just like my system.The inconsistent variation with temperature comes from two points in my opinion:
This doesn’t appear to be systematic enough to be malicious in any shape or form - just marginally viable components without enough of a disclaimer on the instructions.
A boggo uncalibrated NTC is “good enough” for setting a flow temperature (any error you just move the heat curve for that specific unit) and “good enough” for deciding when to start and stop the compressor.
You can try to estimate COP with these and the result is “good enough” for checking that there’s not something catastrophically wrong but not enough to hit the last 20%.
20% price point 80% function?
L0 monitoring = Vaillant estimate
L1 monitoring = eBus with zero offset in the calc (run heat pump circ for long enough with compressor off to establish zero offset then use this in yield calcs)
L2 monitoring = eBus elec/flowrate and direct-immersion onewire sensors
L3 monitoring = MID elec/heat?
Good to see that that the electricity data is relatively accurate, given that’s the data vaillant are reporting back to OVO for this new heat pump tariff. At 15p/kWh for heat pump power usage it seems like a good option, it reduces COP requirement for gas combi cost parity to ~2.3 rather than ~3.6 on their standard tariff. I know many on here will be on more adventurous ToU or agile tariffs but it seems a good option for average users without the dangers of peak pricing.
If such schemes are extended to other HP brands and energy suppliers it might reduce some of the nervousness, particularly in social housing, around heat pump affordability.
I was wrong about these 
pocket sensors, I’ve heard from an internal source at Vaillant that the sensors they use are clip on sensors with tiny copped tipped NTCs, they clip onto the outside of the pipe. This is the exact part 
:
The fact these are clip-on sensors explains probably explains why some systems have significant temperature sensing errors, possibly the sensor is not fully clipped on?
The good news is that it should be a relatively easy fix to remove the side panel of the HP and check the sensors are clipped tightly to the pipe. Can anyone experiencing significant errors confirm if the issue is indeed a poorly clipped sensor? @Timbones @Andre_K @dfeist. Removing the side panel is quite straightforward, from memory you’ll need to remove the top cover first, which will then expose the screws on the side cover. You’ll need to remove the side cover closest to the flow/return connections. Please take some photos if you find any issues with how the sensors are attached,
It does seem that if you get lucky and have a system where the sensors are clipped on properly, the onboard monitoring can be pretty good. Here’s another data point: Comparison between the onboard monitoring and MID meter data from a 5kW Arotherm+ that I installed for my parents, the data from this system is on heatpumpmonitor, the onboard heat & elec data matches the MID meter within about 3-4%. However, I’m still very glad that I installed 3rd party monitoring, the data granularity has provided a lot of additional insight.
Interestingly, the Vaillant ‘Energy efficiency’ figure shown in the app had average 7.3% error compared to MID and was different to the calculated COP = (environmental yield + electrical energy) / electrical energy, I decided to use the calculated COP figure since it was more accurate but most users would probably just look at the ‘Energy efficiency’ number. Can anyone explain this discrepancy?
All these figures were taken from the MyVaillant app, I’ve not yet compared these figures with the working figure displayed on the Vaillant controller.
How would I identify a poorly clipped sensor? Is it wiggly? And more importantly what can I do about it? I just want to make sure I do everything I could do once I open up the device.
