Evohome - Minimum on time explanation

[DBMandrake]

Hi All,

Can one of the Honeywell guys (or anyone who knows for certain) explain the algorithm behind the "minimum on time" system setting ?

Yes, it's fairly obvious that minimum on time is the minimum amount of time that the boiler will be fired in each TPI cycle, to protect the boiler against damage or inefficient operation. What is not stated however is what happens when the demand is above zero but below the set threshold.

For example say the system is set to the default of 6 cycles per hour giving a 10 minute cycle, and minimum on time of 1 minute. This means the minimum duty cycle that the boiler will be fired at is 10%.

The total boiler demand seems to be the combination of all the different zones. Eg, each zone stat sends a demand figure from 0 to 100% to the main controller (based on how much heat it thinks it needs to reach/maintain the set point) which then aggregates them (sums and truncates to 100% perhaps?) and then sends the duty cycle to the boiler relay.

So what happens if the total system demand is below that which the minimum cycle time would allow for - for example if the demand is 5% due to only one zone being active and needing a small amount of heat to maintain its set point.

Will the boiler not fire at all until a demand of 10% is required, or will any demand from 1% to 10% cause the boiler to fire at a 10% duty cycle ? (with 0% not firing at all)

The reason I ask is I have noticed an interesting/annoying phenomenon on our system. In the early evening multiple zones are active and during this time the boiler is typically operating at roughly 20-30% duty cycle (estimated by timing the BRD91 cycle time) and during this time the living room maintains a perfectly stable 21 degrees.

At 8pm all the other zones in the house go off but the living room zone remains at 21 degrees until 11pm. Roughly half an hour after the other zones go off the temperature in the living room starts to drop significantly due to the boiler not firing sufficiently (or at all, hard to tell without watching it constantly) and the flow temperature dropping way down...

The temperature will typically drop as low as 20 degrees before the system makes any attempt to rectify the drop in temperature - checking the valve position (option 10 on the HR92) shows that the valve is fully open, eg >80% which if the boiler was running would warm the room back up again easily, however the boiler is either not running at all or only at the minimum 10% duty cycle.

Once the boiler does eventually come on enough to heat the flow temperature back up the radiator gets hot and the room then overshoots the set point, usually by half a degree, sometimes by a full degree. This is after the room had been previously stable within half a degree for hours on end.

The overshoot I think is caused by "integral windup" - the HR92 knows the temperature is too low and progressively opens more and more but because the flow temperature is cold it has no real effect. The longer the room is below temperature with the valve open the more the algorithm increases the integral which means when the flow finally heats up it overshoots while the integral settles down again.

To see if it helps, I have increased minimum on time to 2 minutes. Whether this helps or makes things worse, depends on whether boiler demands below the minimum on time are completely ignored, or trigger the minimum on time, hence this post posing this question.

If demands below the minimum on time are just ignored, increasing the minimum on time will actually make this problem worse because now the temperature in the living room zone will have to be even lower before sufficient heat demand is generated by the zone stat to trigger the boiler to be fired for the minimum on time.

However the other possibility is that the boiler was firing at a 1 minute / 10% duty cycle, but that is simply not enough warmup time for my boiler to generate any usable increase in flow temperature. Thus increasing it to 2 minutes (or maybe 3) would mean that even if the actual demand is very low (say 5% to maintain the temperature in one zone) the boiler would fire on a 20% duty cycle and thus heat the flow quick enough to maintain the temperature in the room when the demand from the other zones goes away...

Anyone else notice this phenomenon when all but one zone shuts off ?

Edit: Here is a graph from today, showing that the temperature came quickly up to the 8am set point in the morning without any overshoot and apart from one small blip remained dead on 21 all day until 8pm - a remarkable feat when the door was open some of the time, closed some of the time, people were coming and going and so on... Then after 8pm when the other zones went off the temperature drops a full degree while the radiator goes cold due to insufficient firing of the boiler until it eventually wakes up and starts firing the boiler enough to heat the radiator up again.

[top brake]

http://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=2&cad=rja&uact=8&ved=0ahUKEwiB0pn_-JLKAhUBxmMKHcYsADwQFggmMAE&url=http%3A%2F%2Fwww.honeywelluk.com%2Fdocuments%2FAll%2Fpdf%2FFAQ%2520TPI%2520Control.pdf&usg=AFQjCNHxpv5xPBaYhbp5xS1ZI337vRSjlA

[DBMandrake]

Thanks, I'm quite familiar with how TPI works in general, but that leaflet doesn't actually answer my question of whether the boiler fires at all if the demand is less than the configured minimum on time, and doesn't even mention minimum on time in fact.

There are only really two possibilities - either it doesn't fire at all unless the demand is greater than the minimum on time, or it does fire for at least the minimum on time if there is any demand above zero.

I can figure out which strategy it uses myself with a bit of empirical observation by setting an unusually high minimum on time setting but I was hoping someone knew the answer off hand.

[DBMandrake]

Nobody has any suggestions ?

I am definitely seeing an issue where the boiler is being "under fired" when only one zone requires heat.

When at least two zones are calling for heat and the boiler is running at at least 30% duty cycle or so, our living room is able to maintain a perfectly steady 21 degrees with the HR92 making very small adjustments to keep the balance.

However at about 8pm when other zones go off the living room is the only remaining zone active for a couple of hours until the bedroom comes on. During this time the boiler is hardly fired at all, causing the temperature in the living room to drop significantly.

From my anecdotal observation it doesn't seem that the boiler fires at all until the room has dropped to 20 degrees, and even then only starts to fire at a very low duty cycle of maybe 10%. (1 minute on out of 10) Meanwhile the HR92 reports that the valve is 80-90% open but the radiator is barely warm to the touch as the flow temperature is so low. Eventually (it can take more than half an hour) the boiler duty cycle increases and the flow temperature comes up enough to start warming the room again, it then overshoots by up to 1 degree.

With only the one room active I get a cyclical under and overshoot of about +/- 1 to 1.5 degrees due to this issue. However if I turn on any other zone in the house which causes the boiler duty cycle to increase, the temperature in the living room will once again remain stable.

I tried increasing the minimum on time to see what would happen, but it definitely makes it worse not better - with it set to 4 minutes for example, the living room had to drop to 19 degrees before the boiler fired at all.

From what I can see, minimum on time does not behave the way it should. The way it should work in my opinion, is that when there is any demand for heat, it will come on for this minimum amount of time each TPI cycle. So if your boiler takes a couple of minutes running to produce any usable heat you could set it to 2 minutes to make sure that when the demand is low it still runs sufficiently to avoid the problem I'm seeing.

The way it seems to behave now, is if the demand is lower than the minimum on-time it simply doesn't fire at all. This leads to the room temperature continuing to drop despite the HR92 being wide open until finally a high enough "demand" figure is produced to start firing the boiler, but by then it is too late and the room has gone well below it's set point.

Edit: here is another graph demonstrating the issue:



The temperature came quickly up to the correct temperature without overshoot at around 7am and remained steady until about 2pm when all the other rooms in the house were manually turned off leaving only the living room set to 21 degrees, at which point it kept dropping by up to 1.5 degrees before recovering, sometimes with overshoot, sometimes without. All due to the boiler not firing at all until the zone was well below the set temperature.

[The EVOHOME Shop]

Nobody has any suggestions ?

It doesn't look that way. You seem to know what you are talking about and have the time and effort to monitor your evohome system. If I had the time, I would monitor my own system to compare a system that has been in for a significant amount of time vs one that has been in a few weeks. I am obviously not sure at this stage what it would conclude.

I find the fact that we can see new insight into how are rooms are actually 'heated' extremely interesting. You have knowledge that you could never have dreamed of having before. I appreciate as an engineer myself, not knowing why something is doing something is frustrating, but TBH not many people would 'feel' the effects of what the above graph is showing.

I would be very surprised if anyone was put in a room and knew exactly what the temperature was - all they would know is if the room was too warm or too cold and 1 degree would not be noticed in comfort terms IMHO.

[DBMandrake]

but TBH not many people would 'feel' the effects of what the above graph is showing.

I would be very surprised if anyone was put in a room and knew exactly what the temperature was - all they would know is if the room was too warm or too cold and 1 degree would not be noticed in comfort terms IMHO.

In this case I did actually notice the effects of the temperature drop before seeing it on a graph though, it wasn't until I went back and looked at the graph that I saw there was a pattern to it.

I agree that most people probably wouldn't notice a difference of 1 degree, but only if the temperature was steady. Any temperature from about 19 to 21 in the Living room feels comfortable to me if the temperature is constant, however if the temperature keeps oscillating above and below the average by +/- 1.5 degrees then the room does feel too cold as the temperature is dropping and then too warm as it's rising again, as people are more sensitive to changes in temperature than the absolute temperature due to our ability to acclimatise to different temperatures. (Change in blood flow to the skin etc)

Conventional wax pellet TRV's are very poor at maintaining a steady temperature (not to mention conventional non TPI wall stats) so you do feel this constant see-sawing of temperature, hence the tendency of many people to be constantly adjusting their TRV's/thermostats when they feel too hot or cold.

With the HR92 it's so good at maintaining a very steady temperature under most circumstances once at equilibrium that you can go for hours on end without feeling the need to make any adjustment, and I've got used to that accuracy and stability now... so if the temperature does start see-sawing it makes it that much more obvious.

I think I have an idea what could be causing the problem though, and that is that I only have 3 zones controlled with HR92's at the moment, Hall, Living Room and Bedroom. The other rooms are still using manual TRV's. Of those the Bathroom radiator is currently out for renovating so doesn't count, and two of the bedrooms are semi-permanently set to frost protect so don't count either.

The remaining two rooms are the kitchen with one radiator and dining room with two radiators. Both rooms tend to be set to about "2" when the heating is on during the day thus are parasitic zones that are "piggybacking" when the boiler is running, without the knowledge of the evotouch controller.

When the evotouch believes that only one room (Living room) is calling for heat, and only a small amount to maintain a steady temperature in the proportional band it runs the boiler at a very low duty cycle, like 10-20%. If only one radiator plus the automatic bypass valve was flowing this 10% duty cycle would be enough to give a useful flow temperature, however if you hang an additional 3 radiators off that 10% duty cycle they'll drag the flow temperature right down to the point where it's inadequate to maintain the living room temperature.

I presume if those three extra radiators had HR92's then the evotouch would realise that there was actually more than one radiator worth of heat demand (it adds the heat demands of each zone when calculating boiler duty cycle?) and thus fire the boiler to a much higher duty cycle more appropriate for four radiators instead of one.

Or, if I fully close all the manual TRV's so that the Living room is actually the only one flowing, in theory that would also allow the flow temperature to increase enough to maintain the temperature ? (Something that's easy for me to test)

Does it seem feasible that the underestimation of the total radiator heat demand due to some zones not having HR92's could explain the problem I'm seeing ?

[The EVOHOME Shop]

Does it seem feasible that the underestimation of the total radiator heat demand due to some zones not having HR92's could explain the problem I'm seeing ?

It could, but I would only be guessing as I don't know exactly how Honeywell manage the evohome Controllers calculations or what assumptions they make when it comes to evohome system design (i.e. do they expect a minimum or average number of HR92's on a system).

[DBMandrake]

That's really good news to hear that you will be selling the HGI80's

I plan to get one eventually, largely because I just like to see how things work, so being able to monitor the communications between the devices is quite appealing, as well as being able to graph temperatures without relying on the Web API. I already have a spare Raspberry Pi doing the current Evohome graphing via the Web API so the HGI80 could plug it directly into that and set up the proper drivers/software for it...

I too would be interested in knowing what algorithm they use to combine the heat demand of each HR92. From what I understand the BDR91 is sent a heat demand as a percentage which translate to an on/off duty cycle, and each HR92 also sends a heat demand as a percentage to the Evotouch. So the Evotouch must combine those in some way to produce a total system heat demand that is sent to the BDR91, but the algorithm is not clear.

What I have observed is that any single zone being more than about 1.5 degrees below its set point (outside the proportional band ?) is enough to generate a maximum heat demand (100%) at the boiler relay.

A single zone in the proportional band near the set point seems to generate a very low demand of maybe 10-20% hence the issue I'm seeing. Multiple zones being within the proportional band does cumulatively add up in some way - for example with 3 zones (I only have 3 zones) all regulating in their proportional band boiler duty cycle is more like 30-40%, and definitely increased over one zone.

As you say - how do the calculations allow for the number of zones in the system, and does it use number of zones to figure out how "large" the system is ? For example if I have only 3 zones in the system all in the proportional band, does that generate more heat demand at the boiler relay than if I had 9 zones and only 3 of them were active ? I don't know...

Last night I tried manually closing the radiators that I suspected were causing the issue I'm seeing - it did help somewhat. It reduced the undershoot with only one evohome zone being active (and several manual TRV zones being active) from about 1.5 degrees to 0.5 degrees. The radiator was starting to heat up again when about half a degree under the set point instead of 1.5 degrees. This suggests that the boiler was firing at a very low duty cycle, maybe the minimum 10%, but that the additional load of the unaccounted for manual TRV radiators was preventing the flow temperature from increasing to a useful temperature. With only the one radiator actually flowing the radiator did start to heat up in a much more timely fashion.

I suspect using TPI with a BDR91 is always going to be a bit problematic when only one zone is active though, as it's difficult to regulate the flow temperature that way with such a low duty cycle when there are too many unknown variables, such as the response time of the boiler. I suspect that an OpenTherm interface may solve this issue as the Evotouch could ask for a specific flow temperature and know that it is going to get it rather than trying to guess what duty cycle might achieve that...

[paulockenden]

An HGI80 linked to Domoticz is really useful.

P.