ELK M1G & RCS TR40 thermostat error in Celcius

sandman

Member
Hi ALL

I was hoping to get a bit of advice here. I am using the ELK M1G and RCS TR40 thermostat in celcius mode. I am seeing some weirdness with the heat setpoint value as in the heat setpoint is 20C on the thermostat but when I call the settings via the ethernet port, its 1 degrees less ie: 19C not 20c as shown on the thermostat.

Also if I use Homeseer to set it to 20C, I then poll the unit, it give me 19c back.....

Should I raise this issue with ELK or bring it to RCS's attention ? I dont have anything I can connect directly to the thermostat to see what it is sending...

I dont know anyone else with this configuration in celcius that I can compare notes to.... bugger

Cheers..Pete
 
...
Also if I use Homeseer to set it to 20C, I then poll the unit, it give me 19c back.....
....

Surely this error occurs when the scale is converted. Any error while rounding.
One or more device along the infomation chaine (thermostat - ELK - Homesser) is computing internal in scale 1 and had to reply in scale 2.

And while converting "centigrade" (as americans often calling °Celius)
to "funnygrade" (as I am calling °Fahrenheit),

any rounding error happens. Such an error will also happen if the internal scale is the other!

Unless the converting-rules are working very precise. So it is not enough to convert by the usual formula

°C = (°F - 32) * 5 / 9

but to avoid (most, - not all!) roundig errors

°C = int ( ... same conversion ... + 0.5)

where "int" means the integerpart of the value which is computed with some decimals

and vice versa.

Sorry, I'm afraid about it, but I think, you are forced to live with this until America is complete metric and there is only one scale.
 
...
Also if I use Homeseer to set it to 20C, I then poll the unit, it give me 19c back.....
....

Surely this error occurs when the scale is converted. Any error while rounding.
One or more device along the infomation chaine (thermostat - ELK - Homesser) is computing internal in scale 1 and had to reply in scale 2.

And while converting "centigrade" (as americans often calling °Celius)
to "funnygrade" (as I am calling °Fahrenheit),

any rounding error happens. Such an error will also happen if the internal scale is the other!

Unless the converting-rules are working very precise. So it is not enough to convert by the usual formula

°C = (°F - 32) * 5 / 9

but to avoid (most, - not all!) roundig errors

°C = int ( ... same conversion ... + 0.5)

where "int" means the integerpart of the value which is computed with some decimals

and vice versa.

Sorry, I'm afraid about it, but I think, you are forced to live with this until America is complete metric and there is only one scale.

When the TR-40 is set to operate in Celsius mode, it also reports the temperature values in the same mode over the serial link. Thus any conversion would have been done either in the Elk or somewhere else. But I would tend to believe that the Elk simply reports the temperature value that it gets from the thermostat to any other device.

One thing that I find annoying with the TR-40 is that it always maintains a 3 degree spread between heat and cool setpoints, even in Celsius mode. Celsius degrees are "wider" and a 3 degree spread corresponds to 5.4 degrees F. This is too wide for comfort and makes it impractical to use Auto mode. I have reported this to RCS and suggested that in Celsius mode, the allowed spread should be 2 degrees.
 
One thing that I find annoying with the TR-40 is that it always maintains a 3 degree spread between heat and cool setpoints, even in Celsius mode. Celsius degrees are "wider" and a 3 degree spread corresponds to 5.4 degrees F. This is too wide for comfort and makes it impractical to use Auto mode. I have reported this to RCS and suggested that in Celsius mode, the allowed spread should be 2 degrees.

Can't you change the spread in the TR40 installer settings screen? I think it is called Delta-something?
 
Guy,
It sounds like round off error in the Celcius to Fahrenheit conversion.
Yes Spanky,
but let me add some thoughts to my first reply to this point which was rather superficial:

1. of course any device must use internal the finer scale(=°F), or another, even more finer as 0.1 °K or any artifical scale. Don't laugh, but this would solve the problem and this more of accuraty can reached very easy.

2. also possible is a complete other math-approach of conversion: using a table inside the most common temp-zone and using formulas only outside this range. If there is a range outside offered.
 
HI Guys

Thanks for all of the helpfull posts, I tend to agree its a rounding issue which seems a pain in the a$$ around the 19c mark :) I think I'll fire up the M1 SDK and do some testing again to see if I can plot whats going on. It seems that if I change it 2 degrees ie: 18-20c on the heat setpoint thats fine. If I change it from 20-19, the Thermostat will show 19 but on the serial port, I get 18....

Everything is point towards me having to pack up the house and move to the US :)
 
I still don't know where you're seeing a deg F to deg C conversion. First of all, the DS1721 temperature sensor chip used by the thermostat works in deg C internally, so if anything it should be more accurate in Celsius mode since it needs to do a conversion to report in degrees F.

Then there's the fact that the temperature it will report over the serial connection will be the same one displayed on its screen, at least that's what I observed as I was developing an application that talks to it over the serial port.
 
Back
Top