Electric radiator with internal thermostat vs TPI (noise & power behavior)

[Dj0udj0u]

Hi,

I’m using Versatile Thermostat (thermostat over switch) in Home Assistant to control a 1500W electric radiator (Sauter, oil-filled) via a pilot wire module.
When the radiator is in external control / pilot wire mode and managed by Versatile Thermostat (TPI), I observe:

• frequent ON/OFF cycles
• audible metallic clicking noises (thermal expansion)
• power oscillating roughly between 300W and 1200W (never full power)

I tried increasing cycle_min to 60 minutes, using only Comfort/Eco modes (no hard OFF), which reduced the noise but did not fully solve it.

When I switch the radiator back to its internal thermostat / schedule mode (no external control):

• noise is much lower
• heating phases always reach ~1500W continuously
• behavior is clearly smoother and closer to the manufacturer’s design

From what I understand, with a “thermostat over switch”, Versatile Thermostat cannot force full-power heating, only a time-based duty cycle.

My question is:
Is there any recommended approach (current or planned) to better handle electric radiators with internal thermostats, where full-power continuous heating is preferable to frequent TPI cycling?

Thanks for your work on Versatile Thermostat, and for any insight you may have.

[pathumzcode]

Hi,

I’m using Versatile Thermostat (thermostat over switch) in Home Assistant to control a 1500W electric radiator (Sauter, oil-filled) via a pilot wire module. When the radiator is in external control / pilot wire mode and managed by Versatile Thermostat (TPI), I observe:

• frequent ON/OFF cycles
• audible metallic clicking noises (thermal expansion)
• power oscillating roughly between 300W and 1200W (never full power)

I tried increasing cycle_min to 60 minutes, using only Comfort/Eco modes (no hard OFF), which reduced the noise but did not fully solve it.

When I switch the radiator back to its internal thermostat / schedule mode (no external control):

• noise is much lower
• heating phases always reach ~1500W continuously
• behavior is clearly smoother and closer to the manufacturer’s design

From what I understand, with a “thermostat over switch”, Versatile Thermostat cannot force full-power heating, only a time-based duty cycle.

My question is: Is there any recommended approach (current or planned) to better handle electric radiators with internal thermostats, where full-power continuous heating is preferable to frequent TPI cycling?

Thanks for your work on Versatile Thermostat, and for any insight you may have.

You're experiencing the common challenge of using TPI (Time Proportional Integral) control with an oil-filled electric radiator that has its own internal thermostat. The symptoms you describe are typical:

Frequent ON/OFF cycles creating thermal stress
Metallic clicking from expansion/contraction
Power oscillating between 300W-1200W instead of full 1500W
Much smoother operation when using the radiator's native controls

Why This Happens
The core issue is that TPI control is fundamentally incompatible with radiators designed for continuous full-power heating:

TPI works by duty cycling: It rapidly switches the heater ON/OFF to achieve an average power output
Your radiator expects continuous power: When its internal thermostat is active, it wants sustained 1500W until the setpoint is reached
Pilot wire mode bypasses internal logic: The radiator loses its designed thermal management behavior

When you use the internal thermostat, the radiator:

Heats continuously at full power until satisfied
Uses its own temperature sensor and hysteresis logic
Minimizes mechanical stress through longer heating cycles

Recommended Approaches

1. Use Native Radiator Control (Best for Your Hardware)
   For radiators with good internal thermostats, this is often the optimal solution:

Let the radiator manage its own heating cycles
Use Home Assistant only for scheduling (via pilot wire comfort/eco commands)
Accept that you lose precise temperature control but gain reliability and efficiency
Use the pilot wire for presence detection (comfort when home, eco when away)

2. Increase Cycle Times Further
   If you must use external control:
   yamlcycle_min: 120 # Try 2 hours or even longer

Longer cycles = fewer mechanical stress events
Trade-off: slower temperature response
Still won't achieve true continuous heating

3. Consider Alternative Thermostat Integrations
   Versatile Thermostat has multiple control modes:

Thermostat over valve: Designed for valve position control (not applicable here)
Thermostat over climate: If your radiator exposes a climate entity with its own logic
External temperature sensor: Use a more accurate room sensor, but TPI still applies

4. Hybrid Approach
   Use pilot wire commands without TPI:

Map Home Assistant automations to send Comfort/Eco/Frost commands only
Don't use Versatile Thermostat's temperature regulation
Create simple schedules that switch modes rather than cycling power
This respects the radiator's internal design

5. Hardware Alternative: Smart Relay with Temperature Feedback
   For those wanting smart control without TPI issues:

Use a smart relay that can read the radiator's actual behavior
Implement long-cycle bang-bang control (e.g., 30+ min minimum cycles)
Or use predictive algorithms that learn the radiator's thermal mass

Current Versatile Thermostat Limitations
As of now, Versatile Thermostat doesn't have a "continuous heating mode" specifically for resistive heaters with internal thermostats. The TPI algorithm assumes the heating element responds linearly to duty cycle percentage.
What would help (feature request territory):

A mode that respects minimum ON times (e.g., "never cycle faster than X minutes")
Integration with radiators that expose their internal state
A "bang-bang with long hysteresis" option for thermal mass systems

My Recommendation for Your Setup
Given your observations, I would suggest:

Accept the internal thermostat's superior performance for your specific radiator model
Use Home Assistant for scheduling only:

Send Comfort mode when you're home
Send Eco mode overnight or when away
Use Frost protection mode for extended absences

Monitor with power sensors but don't try to micromanage the heating cycle
Save the wear on your radiator - those thermal stress cycles will eventually cause mechanical failures

Your radiator is designed as a thermal mass heater (oil-filled) with internal temperature regulation. Fighting that design with external cycling will always be suboptimal.
Additional Resources

Versatile Thermostat documentation on TPI tuning
Consider posting in the Versatile Thermostat GitHub discussions for feature requests
Look into the "self-regulation" capabilities of modern pilot wire modules

Bottom line: Sometimes the best smart home solution is knowing when not to override manufacturer-designed behaviour. Your radiator works better on its own because it was engineered for exactly that use case.

[jmcollin78]

Hello, if you have a pilot wire (a black cable oftently) you can avoid to have multiple on/off but control the preset directly.
See Nodon in the documentation. Of course your radiator should have the pilot wire, else you have no choice with VTherm.

The other choice is to remove VTherm...

[maxenceleduc92]

Hi,
Did you try increasing minimum activation and deactivation delays in the TPI configuration ?
I think increasing cycle_min might actually increase thermal stress, as it has plenty of time to fully heat up and fully heat down. By reducing the cycle_min, you will take advantage of the heater thermal inertia and will smooth out the heating cycles.
If you reduce cycle_min, minimal_activation_delay and minimal_deactivation_delay too much, it might mess with the heater's internal electronic though, as some heaters wait for a relatively long time before applying the pilote wire order

[Dj0udj0u]

Hi everyone,

First of all, thank you all for taking the time to reply — I really appreciate the quality of the explanations and the different points of view.

I’d like to add a few clarifications about my setup and share some additional observations, to make sure we are all aligned and to get your thoughts on a possible next step.

Context reminder:

• Electric oil-filled radiator (Sauter), 1500 W
• Pilot wire (black wire) connected to a Nodon pilot wire module
• Home Assistant with Versatile Thermostat

I already tested using only Comfort / Eco pilot wire modes (no hard ON/OFF relay control)

Even when using Comfort/Eco via the pilot wire (instead of pure start/stop), the radiator still produces loud metallic clicking noises when externally controlled, especially during transitions.

Key observation:
When I fully hand control back to the radiator’s internal thermostat (and only monitor power consumption via the Nodon module), the clicking noise becomes much weaker and almost disappears.

I analyzed the power consumption curves in that mode and noticed something very consistent:

Comfort mode:

• Heating ON ≈ 30 seconds
• Heating OFF ≈ 10 seconds

Eco mode:

• Heating ON ≈ 10 seconds
• Heating OFF ≈ 30 seconds

In both cases, instantaneous power is close to full power (≈1300–1500 W).

This suggests that the internal thermostat applies a simple, stable time-based modulation with short and regular cycles — and this behavior is clearly much quieter mechanically.

Regarding Versatile Thermostat configuration:
My current TPI settings are:

• cycle_min ≈ 40 minutes
• minimal_activation_delay = 10 seconds
• minimal_deactivation_delay = 0 second

Based on the internal behavior observed above, I’m wondering whether it would make sense to experiment with longer activation/deactivation delays and shorter effective cycles, to better leverage the radiator’s thermal inertia and possibly reduce the clicking noises.

My question / hypothesis:
Do you think approximating this internal timing externally (within VTherm limits) could realistically help, or do you believe — as some of you already suggested — that even with similar timings, pilot-wire-based external control will always remain mechanically suboptimal compared to the radiator’s internal regulation (fluid temperature awareness, internal hysteresis, hardware-level control, etc.)?

I’m fully open to the conclusion that internal regulation may simply be better suited for this type of radiator, even if it limits how much Home Assistant can act on scheduling and comfort overrides. Before closing the topic, I wanted to share these measurements and get your feedback.

Thanks again for your time and insights — happy to test and report back if useful.

[maxenceleduc92]

It looks like your heater has an internal cycle of 40 seconds, but you did set a cycle_min of 40 minutes. The difference is very big, that's why you don't get similar performance.

I have a few questions:

• When using VTherm, what temperature do you put for Comfort and Eco on your Sauter heater ? (I guess Eco is Comfort -3.5°C, but some heaters allow to set a specific temperature)
• With your power meter, can you check how much time the heater takes to start heating when you switch the pilote wire command from eco to confort ? And from confort to eco ?

[caiusseverus]

This suggests that the internal thermostat applies a simple, stable time-based modulation with short and regular cycles — and this behavior is clearly much quieter mechanically.

If you have 40 minutes cycles, then the output is going to be on and off for some period during that time. At 50% power you will have the radiator on for 20 minutes, and off for 20 minutes. That's long enough for it to heat up and cool down significantly which will cause it to expand and contract, and make the clicking noises.

The other thing with oil filled radiators is that they can almost never emit as much heat as electrical power that they can draw at safe temperatures. That means that if the element were on at 1500W continuously, the oil would get dangerously hot without some safety cut off. Most of these heaters will use their full power to get up to the set operating temperature then switch on and off to keep at a sensible radiator temperature.

What is probably happening is that the internal control logic, by using a 75/25% or 25/75% on/off duty cycle, is keeping the radiator at a fairly constant temperature - you get the clicks when it heats up initially, but once it's hot it stays that way until you turn it off. If you use the pilot wire, it applies a 100% duty cycle until it reaches the temperature cut off which may lead to a higher temperature, but then if it is using a simple mechanical temperature switch, has to cool of significantly before the power cuts in again leading to a large hysteresis of the radiator temperature and hence increased clicking from expansion and contraction.

I would suggest using a much shorter cycle time, such that the radiator can get up to and maintain a relatively constant temperature below its safety level. If you choose something like 10 minutes or maybe even shorter like 5 minutes or less, with a minimum on/off time of something like 10 seconds, VTherm will have quite good control over the average power level. The internal controller has a cycle time of 40 seconds for comparison.

That means with a 5 minute cycle time (300 seconds), a 10% power level would be on for 30 seconds and off for the remainder). The minimum effective power available with a 10 second minimum activation is about 3%.

You will have to experiment to see what gives the best results, but a long cycle time isn't likely to help you.

[Dj0udj0u]

Sorry for the delay in my response. It took some time to perform controlled tests and gather reliable observations.

I was able to run several measurements using a power meter and by observing the internal PWM behavior of the oil-filled radiator.

Observed behavior

• The internal control cycle of the radiator is ~40 seconds.
• Changes via the pilot wire (Eco ↔ Comfort) are applied on the very next internal cycle, with no additional latency.
• Eco → Comfort: duty cycle changes from ~30s OFF / 10s ON to ~20s OFF / 20s ON.
• Comfort → Eco: reverse behavior, also applied immediately.

This confirms that the radiator reacts deterministically and without hidden delays.

Regarding cycle duration
Using long VTherm cycles (e.g. 30–40 minutes) indeed caused large thermal swings of the radiator body, which resulted in audible expansion/contraction noises. Shorter VTherm cycles (5 minutes) combined with minimum on/off delays slightly above one internal cycle (~45s) provide much smoother thermal behavior.

Current configuration

• VTherm cycle: 5 minutes
• Min activation / deactivation delay: ~45 seconds
• Radiator internal PWM remains stable and consistent
• Mechanical noises are almost gone

About internal temperature settings
One key observation is that setting the radiator’s internal Comfort temperature too close to the target room temperature reintroduces a secondary regulation loop. This leads to temperature overshoot and instability.

Raising the radiator’s internal Comfort temperature above the VTherm target effectively turns it into a thermal limiter rather than an active regulator, allowing VTherm to fully control the average power without interference.

Based on the latest adjustments, I’m now going to observe the system over the next few days to confirm two things:

• that the room temperature remains reasonably stable around the target values, and
• that the mechanical clicking noises remain almost completely gone, especially during night periods.

So far, the results are encouraging, but I want to validate this over a longer period before drawing definitive conclusions.

If you have any additional remarks, suggestions, or see potential improvements in this setup, I’d be very happy to hear them. Thanks again for taking the time to share your expertise.

[maxenceleduc92]

45 seconds min activation delays with a 5 minutes cycle results in a 15% minimum power, which could give you some instability/overshoot I think. 15% is probably a lot more than what you need to maintain your room's temperature.
Try with a VTherm cycle of 10 minutes at least.

Based on your observations, if you set the heater's internal comfort temperature way above Vtherm target, then trigger pilot wire to Comfort for 20 seconds and then trigger pilot wire to anti-frost, it will still wait the end of the 40 seconds internal cycle to stop heating?