Honeywell Valencia problems

[paulockenden]

My main bedroom is our first zone to come on. I hear the TRV open, and then a short time after I hear the pipework start to 'click' as it expands, so I'm pretty sure that hot water is flowing as soon as it's needed.

Did you give your HR92 time to learn the new valve? I'm wondering whether, during the two week (or so we're told) learning process, the start and end points of the pin movement actually change? So what you're seeing as 75% might be 30% in a couple of weeks time?

Not saying it will, just suggesting that this could explain what you are seeing.

Of course, if you HAVE given it time to learn then just ignore me!

P.

[DBMandrake]

My main bedroom is our first zone to come on. I hear the TRV open, and then a short time after I hear the pipework start to 'click' as it expands, so I'm pretty sure that hot water is flowing as soon as it's needed.

I would assume in that scenario there is a sudden jump in set point from a night time setback to a daytime temperature ? If the set point jumps past the current temperature by at least 1.5C the HR92 will just go from 0% pin position straight to 100% and call for a 100% heat demand as well.

So if the night time set point is 15C and morning wake up time is 20C, if the actual temperature is anything below 18.5C when the set point changes it will just go straight to 100% and you won't see any delay in the initial boiler firing.

The scenario I'm painting is more where you have a fixed set point and the room temperature gradually falls below it - say 21C set point but due to a warm day the room is 22C and the boiler is off. As the temperature slowly drops in the evening by the time it gets to about 20.5C the HR92 will generate enough heat demand (10%) at a pin position of approximately 35% to cause the boiler to fire, but the valve wasn't flowing until 75% pin position. This would require the temperature to fall to say 20C.

So the whole time the room is gradually falling in temperature from 20.5C to 20C, which might be half an hour or even an hour in the afternoon/evening, the boiler is cheerfully heating and pumping hot water around the ABV loop and the radiator remains cold. This is not just theoretical it's what actually happened to me.

The following morning when it was set to come on to 21C by 7am, two hours after that the room was still 20C, yet the radiator was cold and once again the boiler was running with the flow temperature pegged at it's 65C flow temperature. I checked the pin position which was 74%, and the reported heat demand on the controller from the zone was 42% despite the cold radiator. That's the point where I really decided "no, that's not right" and started the whole investigation that lead to this thread...

Did you give your HR92 time to learn the new valve? I'm wondering whether, during the two week (or so we're told) learning process, the start and end points of the pin movement actually change? So what you're seeing as 75% might be 30% in a couple of weeks time?

Not saying it will, just suggesting that this could explain what you are seeing.

Fair comment.

No, I didn't give it two weeks to adapt, but it did get 48 hours, after which time it showed no signs of improving it's valve position calibration. I wasn't keen on leaving it for days/weeks where it was running the boiler without heating the radiator. Our gas bill is already high enough.

The long term adaption you're referring to I believe is more about the tuning of the PID controller to learn the response time of the whole radiator/room system. So learning not to overshoot/undershoot etc. Yes, this does take a few days and in fact even after changing stroke mode it has been undershooting slightly in the living room the last couple of days but it is now coming right. This adaption process can't be sped up.

However calibration of the pin position happens a lot quicker - it just takes a few full open/close cycles and can be hastened by repeatedly setting a low set point and high set point and giving the valve time to respond - the pin calibration does "settle in" a little bit in the first few cycles, especially in stroke 1 mode where it is trying to operate over the full available travel. But once you've cycled the valve through it's full range of movement a few times the pin calibration doesn't seem to change any further - I checked that again to be sure on the spare valve/HR92 that I have.

If you weren't to manually cycle the set point up and down a few times then yes, in a typical schedule it would take a few days for the pin calibration to settle in just because it does take 3 or 4 full pin cycles to do so.

I can't be certain of course, but I'm reasonably sure that the pin calibration problem I was seeing would not have gone away by itself if I'd left it a couple of weeks. The set point would have eventually been reached and maintained but only by running the valve near the upper end of its pin position with a high heat demand from the boiler and plenty of occasions of calling for heat when not needed.

[rotor]

Surely there has to be a fault? The HR92 has no idea what the valve is doing "under the bonnet", it just knows that anti-clockwise = more heat, and clockwise = less heat. If it's demanding heat while having the pin in a position where water is unable to flow, then it has got things wrong, i.e. it hasn't learned the behaviour of the valve. As for the difference between the two stroke types, that is not something that has ever seemed a problem to me (all my radiators have Valencias), as I suspect the range of movement is usually absolutely tiny. I only hear the HR92 motors going continuously (more than a second or so) at the beginning or end of a session, and when wide open (i.e. at the beginning of a session) any given radiator is piping hot, so there's no flow restriction worth worrying about in the "wide-open" position. If the HR92 is able to sit (fully screwed on) where it is just beginning to depress the pin, then it should have more than enough room in both directions in the normal stroke configuration.

I've had a couple of problems with my Valencias getting stuck closed, so could it be that?

[DBMandrake]

Surely there has to be a fault?

You mean I have 7 faulty HR92's and/or 7 faulty brand new Valencia's ? Seems unlikely.

The HR92 has no idea what the valve is doing "under the bonnet", it just knows that anti-clockwise = more heat, and clockwise = less heat. If it's demanding heat while having the pin in a position where water is unable to flow, then it has got things wrong, i.e. it hasn't learned the behaviour of the valve.

But how could it learn that ? How would it actually know at what pin position the valve starts flowing when the only feedback it has at all is the flimsiest of indirect evidence of a change in temperature in the room ? The answer is, it simply can't know.

I posted this graph in another thread a while back where I empirically derived the relationship between valve pin position and heat demand sent to the controller. This relationship is fixed and does not change with stroke mode, nor does it learn or adapt over time or differ from one HR92 to another. It's always the same on all the HR92's I tested.



It's pretty clear from the graph that there is an implicit assumption built into the device that a valve will start flowing at about 30% of the calibrated pin position, and hence pin positions below this don't call for any heat.

Sure, if the valve starts flowing at a significantly different pin position it will still eventually adapt and regulate the temperature somewhat, however that doesn't mean that you won't have problems like calling for heat when the valve isn't flowing or overshooting when other zones come online. Which has been my main point in this thread.

As for the difference between the two stroke types, that is not something that has ever seemed a problem to me (all my radiators have Valencias), as I suspect the range of movement is usually absolutely tiny.

It certainly is in stroke 0 mode - it only lifts the valve about 1mm off the limit stop for "100% open" despite the valve mechanically having 4mm of travel available. Stroke 1 increases that to about 2mm.

I only hear the HR92 motors going continuously (more than a second or so) at the beginning or end of a session, and when wide open (i.e. at the beginning of a session) any given radiator is piping hot, so there's no flow restriction worth worrying about in the "wide-open" position.

For your radiators maybe! Not mine. Not being able to get full heat from my living room radiator in stroke 0 mode was an ongoing issue with my old valve bodies, and one of the major reasons why I changed to the Valencia in the first place, thinking all would be perfect with the valves that the HR92 were designed for. But nope! In stroke 0 mode the same problem was there, but even worse.

If the HR92 is able to sit (fully screwed on) where it is just beginning to depress the pin, then it should have more than enough room in both directions in the normal stroke configuration.

But it's already pushing the pin down by 2mm just by screwing the HR92 base on... refer to the photo I posted. Unless you screw the internal balancer in, but either way you're still halving the available pin travel, and then by using stroke 0, halving it again.

I've had a couple of problems with my Valencias getting stuck closed, so could it be that?

What made you think they were stuck, and what did you do about it ? If you mean the room was say a degree below the set point but the radiator remained stubbornly cold for an hour then this is exactly the issue I was having that drew my attention to this whole situation in the first place. The pin position was 74% and heat demand about 42% (notice these still correspond exactly on my graph that was derived on my old valves) but a stone cold radiator and after an hour of being a degree below the set point with the boiler running uselessly I decided to investigate.

I didn't go looking for problems, they came looking for me!

[blowlamp]

If you say the introduction of a spacer has no effect because the HR92 just re-adapts back, then I wonder if the soft rubber seal is the main problem with this setup.

If you need to open a valve by 30% of its total travel just to be sure that it's open at all, then something isn't quite right.


My I-Temp/Tower combo adjusts in 5% increments and I'm sure there is flow at a pin position of 10%.

The I-Temp measures pin position by index impulses off a drive gear and knows the approximate count for a full stroke.
Critical dimensions of my valve bodies must have been significantly reducing the stroke length/count causing I-Temp to error out, but introducing the spacer brought the stroke length/count back within tolerance.

I-Temp uses the fairly crude (but effective) means of finding a 'valve closed' position by detecting when index pulses slow down or stop.



Martin.

[rotor]

But how could it learn that ? How would it actually know at what pin position the valve starts flowing when the only feedback it has at all is the flimsiest of indirect evidence of a change in temperature in the room ? The answer is, it simply can't know.

I have no knowledge of how HR92s work, but from a purely common sense point of view, how could you design a system that "learns" without any feedback? "Assuming" how a particular radiator valve works --of which there are bound to be hundreds of different designs-- would be pointless. Here's my quick back of an envelope attempt at designing a self-learning HR92:

- Set the pin at 50% and request heat
- Wait 5 minutes
- Did the temperature increase? By how much?
- Change the pin position (close it "a bit" if the temperature increased too fast, open it "a bit" if it either decreased, didn't increase, or increased too slowly)
- Make note of the rate of temperature increase relative to the pin position. Use this information in future decisions
- Continue adjusting the pin position until I understand what pin positions create specific temperature changes; e.g. the two main ones I'm interested in are: a) maintaining room temperature, b) increasing room temperature by 1 degree over a 15 minute period. Notice that these settings are specific to each room (is it a big room or a small one, are there two radiators or just one, are the radiators super efficient or not, do the radiators have a massive lag or not, etc.), and the type and behaviour of the radiator valve is just one more factor that the HR92 has to learn.
- The end

I know the above doesn't help troubleshoot what is wrong with your setup, but given how flaky Evohome is, I would be tempted to just factory reset everything and start again. The above description should be able to easily achieve the result of getting your radiators to warm up.

[DBMandrake]

I have no knowledge of how HR92s work, but from a purely common sense point of view, how could you design a system that "learns" without any feedback? "Assuming" how a particular radiator valve works --of which there are bound to be hundreds of different designs-- would be pointless. Here's my quick back of an envelope attempt at designing a self-learning HR92:

- Set the pin at 50% and request heat
- Wait 5 minutes
- Did the temperature increase? By how much?
- Change the pin position (close it "a bit" if the temperature increased too fast, open it "a bit" if it either decreased, didn't increase, or increased too slowly)
- Make note of the rate of temperature increase relative to the pin position. Use this information in future decisions

I appreciate the sentiment of what you're saying, and it makes perfect sense to us humans, but trying to implement such a complex and indirect system of inference in a simple device like this using only a single feedback parameter (measured room temperature) just wouldn't work.

Some of the hurdles I can think of:

1) Whenever the HR92 thinks the room should be hotter, it changes both the valve position and heat demand to the boiler at the same time. How does it know which one of the changes was responsible for an increase in room temperature ? From about 30 to 70% pin position (on a properly calibrated system) both are responsible, however only if this zone is the zone demanding the most heat. If there is another zone that is demanding a lot more heat, only the valve position is affecting the temperature and the change in heat demand is more or less ignored, since the system heat demand to the boiler is simply Max(Zone1,Zone2,Zone3....) etc... From about 70% up it is mainly changes in heat demand that are responsible as most valves are already giving maximum flow by that point, (notice the increase in steepness in the heat demand curve above 70%) however if there is another zone calling for 100% heat at the same time, the 70-100% range is effectively only deadband.

2) During the sampling period between making a change and waiting to see what the result is, there can be a lot of outside influences (noise) that can confound making any sense of the result, including changes in heat demand from other zones which will affect the boiler duty cycle, doors being opened and closed, people entering and leaving the room, cookers being turned on, showers being turned on, windows being opened or closed etc. In a busy house there are a lot of confounding variables that make it difficult enough to keep the temperature steady, let alone probe to see where the opening point of the radiator valve might be.

3) Radiators heat up quickly and cool down very slowly. If you open it to 50% then wait a few minutes to see what happens (remember it only samples the room temperature roughly every 4 minutes) and you've opened it too much, closing the valve down again will only result in a very slow decline in temperature or rate of temperature rise even. So you'd need to do it the other way around - open it a very small amount, wait, open it slightly more, wait, until finally there is a rise in temperature. This could take a very long time, during which time the house owner is cold and annoyed that they turned their heating on 45 minutes ago and the radiator is still cold!

- Continue adjusting the pin position until I understand what pin positions create specific temperature changes; e.g. the two main ones I'm interested in are: a) maintaining room temperature, b) increasing room temperature by 1 degree over a 15 minute period. Notice that these settings are specific to each room (is it a big room or a small one, are there two radiators or just one, are the radiators super efficient or not, do the radiators have a massive lag or not, etc.), and the type and behaviour of the radiator valve is just one more factor that the HR92 has to learn.
- The end

A lot of the learning process that you describe is actually implemented, so I think you may be misunderstanding my complaint.

1) The warm up rate of the room to hit a target at a certain time is learnt over time - this is optimal start, and this learning process is handled by the controller. It uses the learnt thermal response time of the radiator/room and the difference between current temperature and set point to decide how early to change the set point.

2) Thermal lag of the radiators/room in terms of overshoot tendency is learnt and adapted to - this is done by the individual HR92's by heuristically self tuning the differential part of the PID controller tuning. So if you make a step change in the set point from say 15C to 20C and the room temperature is 20C, and it goes flying right past 20C because it's rising quickly and still putting out a lot of heat at 20C it will recognise this overshoot, adjust the differential parameter and next time it's in the same situation it will start closing the valve position down well before it reaches the set point in rooms that heat quickly, "coasting" to reach the target. It does a pretty good job of this in most situations.

This can happen in reverse too where it undershoots the target, leveling out 1-2C below the set point if it has previously compensated for overshoot. I notice this particularly with a radiator that often has a towel on it which increases the thermal lag to the room considerably - initially that will cause overshoot because it slows the heat output into the room so that its still putting out heat long after the radiator closes, however after removing the towel it then undershoots until it re-learns.

3) Offset error is learnt and adapted to automatically as well. This is also done by the individual HR92's by using the integral parameter. For example say it's a warm day so it only takes a valve pin position of 45% to maintain a steady 20C in the room. It learns and adapts to that and will keep the room precisely at 20C. The following day is much colder, now it would require a valve pin position of 65% to put out more heat to maintain the same 20C. When it initially warms up the room it may undershoot slightly however over time (30 minutes to a few hours) it will gradually adjust and hit the target spot on.

4) The HR92 can adapt to differences in the valve pin length - to a certain degree. There is about 3.3mm of usable pin travel from the HR92 base, and the calibration process basically winds the pin down until a set force is reached, that is considered to be closed. It then winds it out a set number of turns to reach 100%. In Stoke 0 mode this is typically about 1 turn less than the maximum movement available, in Stroke 1 mode it always winds right out until it hits the outer limit stop in the base.

The pin range is then calibrated 0-100% using these two limits. If at any time during operation it reaches one of these limits before it thinks it has, it recalibrates and rescales the range on the fly. So say it is trying to open to 100% and it gets to what it thinks is 90% and then hits the limit stop. It immediately recalibrates this point as 100% and scales the rest of the range proportionally. You can see this happening if you're in the pin position mode when it hits the limit stop. I've found it takes 2-3 full excursions from closed to open to fully "settle" the calibration and stop making small changes to it.

Because of this calibration process if the pin is slightly shorter (emulated by adding a shim as discussed earlier) it adapts to this because during the calibration process it pushes the pin down until a certain force is reached or the limit stop is hit. Normally the bottom limit stop is never hit, the force limit is reached first. With the shim it would have to wind down further which means that 0% is slightly further down the range of travel the base allows.

That means 100% is further away from 0% hence the valve can open slightly more with the shim. But the amount of pin travel from the closed position to where the valve starts to flow - which is dependent on the compressibility of the washer, doesn't change.

Also the relationship between calibrated pin position and heat demand seems to be fixed. It starts calling for heat at about 30% - end of.

I'm not saying it couldn't be done better than it is now - one improvement might be for it to have a sensor to actually measure the flow temperature of the pipe that it's sitting on top of to have a better idea of when it starts to flow - using a similar algorithm to the one you describe. However I don't think this would work by trying to sense air temperature in the room, especially with a remote sensor. With a sensor for the pipe temperature, yes.

The problem with that idea is that it then makes a huge difference whether the HR92 is on the flow or return side of the radiator. On the flow side it will very quickly detect the pipe is getting hot and get an accurate idea of when it started to flow - on the return side there will be a huge delay due to the cold radiator continuing to pump out cold water for a long time! Another possibility would be an IR sensor aimed at the radiator. But since you can orientate the HR92 in different orientations, and some radiators have side covers that insulate the heat quite well that could also be problematic, and putting it on the return side of the radiator (which is best for air temperature sensing) would introduce the same response delay as directly measuring the pipe temperature.

I don't think this is an easy problem to solve, in short, other than having a more versatile Stroke mode setting that allows finer calibration than just two different modes.

[DBMandrake]

A lot of the learning process that you describe is actually implemented, so I think you may be misunderstanding my complaint.

1) The warm up rate of the room to hit a target at a certain time is learnt over time - this is optimal start, and this learning process is handled by the controller. It uses the learnt thermal response time of the radiator/room and the difference between current temperature and set point to decide how early to change the set point.

2) Thermal lag of the radiators/room in terms of overshoot tendency is learnt and adapted to - this is done by the individual HR92's by heuristically self tuning the differential part of the PID controller tuning. So if you make a step change in the set point from say 15C to 20C and the room temperature is 20C, and it goes flying right past 20C because it's rising quickly and still putting out a lot of heat at 20C it will recognise this overshoot, adjust the differential parameter and next time it's in the same situation it will start closing the valve position down well before it reaches the set point in rooms that heat quickly, "coasting" to reach the target. It does a pretty good job of this in most situations.

This can happen in reverse too where it undershoots the target, leveling out 1-2C below the set point if it has previously compensated for overshoot. I notice this particularly with a radiator that often has a towel on it which increases the thermal lag to the room considerably - initially that will cause overshoot because it slows the heat output into the room so that its still putting out heat long after the radiator closes, however after removing the towel it then undershoots until it re-learns.

3) Offset error is learnt and adapted to automatically as well. This is also done by the individual HR92's by using the integral parameter. For example say it's a warm day so it only takes a valve pin position of 45% to maintain a steady 20C in the room. It learns and adapts to that and will keep the room precisely at 20C. The following day is much colder, now it would require a valve pin position of 65% to put out more heat to maintain the same 20C. When it initially warms up the room it may undershoot slightly however over time (30 minutes to a few hours) it will gradually adjust and hit the target spot on.

Saw a perfect example of this learning process when I checked my graphs tonight.

Earlier in the week evening temperatures were quite low around 5-8C and the system had adapted perfectly to this with the evening living room warmup hitting it's target temperature bang on without any undershoot or overshoot, then yesterday it was suddenly unusually warm in the evening - 16 degrees. This caused a minor overshot as highlighted on the left:



The overshoot was only half a degree, and considering the outside conditions had changed from 5-8C to 16C that's relatively minor - with my old sticking valve bodies it would have overshot by 1-1.5C in similar changeable conditions.

Last night it would have quickly detected the overshoot and learnt from that. Had the temperature tonight been similarly warm it would have hit the target perfectly without overshooting.

However tonight was colder again - about 12C, and you can see on the right hand side that due to learning from yesterdays warm conditions it has actually undershot the target and flattened out about half a degree below the set point. Interestingly at the time I took the screenshot the boiler was completely off and the 15 minute pump overrun had stopped as well. So despite being slightly under the set point it actually wasn't calling for any heat.

However over time it will adapt again to the colder temperatures and being slightly under the target - probably by the time we go to bed tonight, if not by tomorrow. In fact while I've been writing this I've just noticed the radiator starting to warm up a bit...

[dty]

What they need to do is design a combined HR92 with valve body. Then they could include a flow sensor, and even make the whole system self-balancing.

[DanD]

I know this is an old thread, but I thought I'd add my recent positive experience with my Valencia valves after re-reading this thread. I replaced about 5 of valves about a year ago to help fix some problems I was having with valve banging and humming sounds from open valves when the majority of the valves were closed. This fixed most of the problems except for one radiator, in a bathroom. This radiator is often the only one calling for heat as it has a higher setpoint and it continued to produce an occasional humming noise for a few minutes which I believe coincided with the valve attempting to close. I decided to try adjusting the valencia 'balancing' valve insert which was set to near the maximum open setting. By closing the adjuster by only about half a turn it has cured the humming noise and my assumption is that this has helped the valve to close more firmly. I didn't need to mess with the stroke setting on the HR92 which remains set to the default, but this was next on my list, if the adjuster didn't fix the problem.

PS when I installed the valencia valves, I decided to use the inserts to balance each radiator and I opened up each of the lockshield valves fully.