Gas Aga
Control (Mk 1)

(an electronic solution)
  Markland Lodge

  This article describes an electronic solution to the control of a Mark 1 Gas Aga (3-door, 2-oven), allowing the temperature of the hot oven to be monitored, adjusted and stabilised as required. It may well be useful for other models of Gas Aga also.
   
  Background
When I moved into my house (Markland Lodge, Huyton), it had a 1970s gas Aga (Mk1). I confess that it has always been difficult to regulate the temperature. Even Aga engineers who have looked at it seemed to have left it worse off than they found it. I realised last year (2013) that it was the Teddington gas valve that was faulty: the thermocouple sensor had become detached from the 'cable'. I decided to buy a new valve and thermocouple (see picture) which cost me £150 including VAT (2013). It was duly fitted - with the original bypass screw, as recommended.
  In order to test how well it was working, I not only threaded the new thermostat lead of the Teddington valve from the hot (top) oven to the burner assembly, but also took a digital thermometer lead - so that the temperature of the top of the hot oven could be accurately monitored. From early measurements with the thermometer, I noted that the top of the hot oven is about 25-30C hotter than the bottom of it - something that is probably already well known amongst Aga aficionados.

Even after much adjustment of the Teddington valve, however, I was disappointed that control of the Aga could not really be maintained by it - at least not as well as I would have expected. The setting-up process involves adjusting a very small 'screw' on the Teddington valve (inaccessible without taking the burner out) so that the valve opening/closing comes into play in the range of the outer dial (1 to 5). Even getting this right - after multiple extractions of a 'hot' burner assembly - it still didn't really give the predictable setting of temperature that one might have expected. I decided that there had to be a better way - perhaps with the use of electronics.

What follows is a method of doing just that - in a system where the Teddington valve is more or less inoperative. In other words, the Teddington valve is fully open and not trying to restrict the gas flow to the burner in any way.
   
  Controlling the Gas Flow
A solenoid gas valve would, I thought, allow me to switch on and off the gas to the Aga, but I couldn't really see a way of siting it near the burner assembly in the supply to the burner. Also, if I put the solenoid valve in the main gas supply, it would have the effect of switching off the pilot light. However, I thought, what if I simply set up a system where, some distance from the Aga, the incoming gas supply is, under normal circumstances, restricted, and then I use a solenoid valve to by-pass the restriction? Provided I had a 'robust' pilot light, which remained lit whether the gas supply was restricted or not, then perhaps I could organise a degree of programming around this.

To test the feasibility, I decided to ignore temperature control (for the moment) and simply 'lower' the Aga temperature overnight using a solenoid valve connected to a timer. This would have the added benefit of saving money overnight. The situation for the gas input is shown here:
 
  The gas enters though the 'main gas stopcock' at the top right. From there, it has three ways it can go to the Aga. If the 'restriction by-pass valve' is open, then this bypasses the other two routes, and the Aga works normally. If, however, the by-pass valve is closed (which I will call 'normal operation'), the gas either goes 'down' through the restriction or 'down' through the solenoid valve. If the solenoid valve is 'off' (closed), then the gas is restricted and the Aga flame is low (pilot still lit). If the solenoid is 'on' (open), then the gas can flow freely to the Aga, and the flame will be full on. For this test, the timer was set so that the solenoid valve closed at 8 pm at night, and opened at 4 am in the morning.

It turned out that the pilot light was indeed robust enough to cope with the reduced gas flow through the restriction that I had made, and the Aga temperature dropped accordingly overnight. You can see from the wiring on the above photograph that the solenoid (at this stage) was controlled only by the timer. It is therefore possible to set the timer so that it restricts the gas flow for a given period overnight, and turns it up again for daytime use. Of course there is a great deal of hysteresis in the system because of the high thermal capacity and insulation of the Aga but if, as mentioned above, the gas is restricted between the hours of 8 pm and 4 am, the graph of temperature (top of the top oven) against time looks as shown in the graph immediately below. Here, temperatures have been taken over a few days, so there may be differences between adjacent points (as the oven is used differently on each day). Also, this graph applies to a particular setting of the 'restriction'.
   
 
  At 4 am (start of graph) the solenoid opens and the temperature rises. By around 8 am, the temperature is near to 260C - and still rising. It remains high all day until the solenoid is switched off by the timer at 8 pm, at which point it falls overnight. Notice that the maximum daytime temperature (around 280C) is rather high. Indeed, this was the reason for investigating this system; the Teddington valve could not be relied upon to work effectively in controlling this temperature.
   
  Controlling the Temperature
What I really wanted to do, therefore, was to 'control' the temperature of the Aga. I decided to do this by starting with the solenoid/timer system above, and adding to it a 'PID' controller and a K-type thermocouple. Now, strictly speaking in what is described below, the 'PID' nature of the controller (Proportional, Integral and Differential) will not be used - I shall use only the simple relay output of the 'PID' (to switch on and off the solenoid).

I threaded a K-type thermocouple with a long (5m) 'sheathed' lead through from the top of the hot oven (where it was 'fixed') down to the burner assembly. The thermocouple lead was then taken to the outside world, and on to the controller, via the bottom left hand corner of the burner door, and around the side (photograph below shows this). This is a little unsightly. I had considered drilling through the side of the Aga, but thought better of this for now.
  The thermocouple lead was taken to the 'PID' controller, which was sited some distance from the Aga (still in view in the kitchen). The positioning of the controller was such that the temperatures could be read clearly (once again, this can be seen in the photograph below). I say 'temperatures' (plural) because the controller shows the temperature you desire to set the Aga at (SV - in green) and the temperature it is actually at (PV - in red) at any particular time. The temperatures can be shown in Centigrade or Fahrenheit.
  By putting the (normally closed) relay output of the 'PID' controller in series with the output of the timer to the solenoid (not shown in the photograph above), it means that, during the day (when the timer is 'on'), if the temperature of the Aga (PV-red) is below the set temperature (SV-green), it will hold the solenoid valve 'on' (open) and the Aga will warm up. Once the set temperature is reached, it will turn the gas off (or rather, allow only the restricted flow), and the temperature will start to fall. The 'PID' controller also has an indicator light showing when the relay is open/closed.

'PID' controllers of this type have a number of settings that can be adjusted. You set the SV (desired temperature when the solenoid will switch off), and you also set a number of other parameters, one of which is the drop in temperature from SV at which it should switch the gas (solenoid) on again.

The following graph shows temperature against time for a number of particular settings. These are:
1)    A particular restriction/valve setting
2)    SV = 250 (degrees C)
3)    Drop in temperature = 10 (degrees C, before switching back on)
   
 
  The temperature scale is different to the previous graph (140-260 degC), but the time scale is the same (4 am to 4 am).

You should be able to see that the controller does exactly what I wanted; the temperature during the day is maintained at around 244C on average - but I could change this by adjusting the controller SV.

Note that readings for the above graph were taken over a number of days, and this 'masks' the fact that the rather 'random' points along the 'flat top' of the graph are actually oscillatory - as the Aga heats up (solenoid open) and cools down (solenoid closed). The period of oscillation is almost 3 hours. On this test, the maximum value of temperature was 251C and the minimum is 235C.
   
 
  Photograph showing the 'PID' controller (top) and the exit point of the thermocouple lead (bottom)
 
Two other things should be noted. The solenoid definitely did close at 250C, but there was obviously 'heat flow' in the system which took the maximum up to 251C. Similarly, the solenoid definitely did open at 240C (250-10), but the Aga continued to cool to 235C (even though the gas was on full) before it began warming up. It is well known that there is a good deal of hysteresis in the system. A smaller value than 10C would perhaps help to reduce the amplitude of variation (and time period), but the hysteresis of the system will inevitably mean it cannot be eliminated.

Also, it is clear that if the Aga were being used a lot for cooking, it is likely that the minimum temperature of the above graph would go even lower than 235C at different times of the day. The very 'nature' of an Aga is such that you cannot really 'control' the temperature so that it is entirely stable. However, when the temperature does go very low for some reason, the gas would be on full (solenoid open).

I shall continue to experiment with various settings (restriction setting, SV temperature and temperature 'drop' etc) until I find the results I want. However, I am now confident that I can control the Aga (inasmuch as it is 'controllable'), and the whole thing has cost me less than the £150 that I paid for the Teddington valve. I also have a clear indication in my kitchen of what the Aga is doing and what the temperature of the top of the hot oven is.

Eddie Boyes, January 2014


PS: After further experimentation, I have settled upon,


1)    A particular restriction/valve setting
2)    SV = 240 (degrees C)
3)    Drop in temperature = 2 (degrees C, before switching back on)

This gives (assuming no cooking) an oscillation period of about 1 hr 15 min, and a variation in temperature of about 6C, with a mean of about 238C. The following shows a simulation of the daily changes in temperature (assuming no cooking) when the timer is set at 8pm (off) and 4am (on):