Slow Maverick probes – impact on control?


 
W

WayneM

New member
I'm currently using a PID from Omega with a bare thermocouple to control my temperature. It does a decent enough job but I'm intrigued by the HeaterMeter and features like the open lid detection, web interface, alarms, ramp down, cook logging, etc and I'm itching to build one.

That being said, I already have an ET-73 and I find that the probes react very slowly to changes in temperature, especially compared to the bare thermocouple I'm using with my PID that has a time constant of 0.01 seconds. I understand that this is due to the thermal mass of the Maverick probe and that it is essentially acting like a moving average filter of the temperature. What I don't know is how that would affect temperature control.

It seems like you would need to log temperature with a separate device that has a very short time constant in order to see what the effects are. Has anybody tried this? I realize that this is probably not an issue but I'm curious what you guys think about it.
 
"React very slowly" is a relative term. Sure, if you stick a probe into a pot of boiling water it takes 20+ seconds for it to come to temperature but that's not at all the conditions inside a smoker. The probes can adjust to the current temperature as fast as the temperatures change inside of a 10lb piece of meat or a smoker. They're read 128 times per second per probe if I remember correctly and the results or oversampled/averaged.

I'm curious what you would be able to do with temperature information 100 times per second. The temperature is up 0.01 degrees increase fan speed by 0.01%! The temperature is down 0.01 degrees decrease fan speed by 0.01%!
 
I think you misunderstood the time constant. The 0.01 seconds is the rating of the time it takes the thermocouple to achieve 63.2% of it's final value. It has nothing to do with sampling (which is 10 Hz on my PID, fwiw).

It takes 5 time constants to reach the final value which means the thermocouple I'm using takes 0.5 seconds to come to temperature. For comparison, the Thermapen has a time constant of 0.6 seconds which means it takes 3 seconds for full temperature. The time constant of a Maverick ET-73 probe is unknown but certainly much higher due to the thermal mass of the probe.

My point in mentioning the time constant was to illustrate that my temperature input for my PID has essentially no lag. I have observed that my ET-73 in the same smoker and right next to my thermocouple will take several minutes to arrive that the same temperature. Surely that input lag would have an effect on control. I was just wondering if anybody had characterized that with an external, fast-response temperature measuring device.
 
Oh yeah I see what you're saying. There's definitely some lag due to the nature of the probes but the control loop has a lag of a similar magnitude. I've never measured it (I do have a thermocouple and PID though) but I would say at least 99% of any control error is in the PID constants or there being too much natural draft through the smoker.
 
I'm with Bryan, I don't see the benefit of such a high sampling interval. temperature of the probe isn't really important, it's temperature of the system including all the mass of the smoker, meat, water pan, and whatever else is in there. I think that the amount of maximum heat possible to be produced would take much longer to effect a significant change of the system than even a 1 or 2 second sampling interval would make a difference for. In other works, the probe itself would not be exposed to a more rapid delta in temperature than the sampling interval could accurately measure.
 
RJ, I think you're falling into the same trap Bryan did. The sampling interval isn't the point and not anything I originally mentioned. Both systems have a sampling interval that is far higher than necessary.
 
No, I think I read it properly, your supposition that "Surely that input lag would have an effect on control." is what I was addressing and my opinion is that I don't think there would be a significant impact on temperature control due to the mass of the system relative to the amount of heat being generated would be in time units orders of magnitude larger than the sampling interval.

My .02.
 
The Maverick probes are indeed a bit slow to react to major temperature changes. If, for instance, you have a hot smoker already and then install a Maverick probe it will take a little while for the probe to read the actual temperature in the smoker, meanwhile the fan will probably run because the HM thinks the temp is low.
That said, if you get your probe into your smoker right away it will rise to temp right along with the smoker, the lag is less of an issue if the probe doesn't have to come from behind with major temp changes. Once the probe is warmed to pit temp the reaction to smaller changes in temperature is plenty fast to achieve solid temperature control, and the probes track plenty fast to measure the changes in temp of a piece of meat....
That said, I have at times wished the probes could achieve an accurate temp in a matter of seconds, were able to handle higher temperatures and were a bit more resistant to water....
 
I came across a post about control design while researching this.

For pressure, flow, and inline composition and temperature control, the measurement time constant is probably already the largest time constant in the loop. An increase in this time constant due to coatings or filter times not only makes the trend chart smoother but allows the user to increase the controller gain which furthers the deception. You and the controller are seeing an attenuated version of the real world.
Click to expand...

I am by no means an control theory expert but it stands to reason that a slow input could negatively impact the ability to maintain a stable temperature. Worst case you would end up with an run-away process or a significant oscillation. Best case is that it takes much longer for the process to settle at the set point.

A PID simulator makes this easy to instrument and visualize. For example, look at the following simulation where the only change is to the thermistor lag. Remember that the blue trace is the actual temperature, not the smoothed-due-to-lag sensor temperature.

With a lag of 16 seconds, the process arrives at the set point in less than 10 minutes and is very stable.
vif.png


By increasing the lag by 2 seconds to 18, the process takes closer 20 minutes to settle down.
4pz.png


When the lag increases another 2 seconds to 20, the process permanently oscillates.
5t3.png


Please understand that I am in no way trying to impugn the HeaterMeter. I think it's awesome and I'm looking forward to building one. It's just that I find this a fascinating topic that doesn't appear to have been addressed by the community so far and is somewhat hidden since all the cook logs are using the same temperature source as the control loop.
 
In addition, since HM is an open system, you can use whatever probes you may want. All you would need to do is use whatever appropriate connectors on the HM(there is even a header to connect a set of jacks via a ribbon cable), use the appropriate value resistor, and change the probe values in the GUI.
 
Ralph, I already believe that HM can do a great job of controlling a pit temperature. I also realize that this is almost certainly not going to significantly affect the end product. As some of the commercial vendors have reminded us, home ovens have far poorer control and we seem to do OK with them.

I'm just intellectually curious about the issues as summarized in the blog post I linked to. Your graph, while awesome, does not really address the question of smoothing due to the thermal mass of the probe. The only way to gather that data would be with a bare temperature measurement device with the smallest thermal mass possible.
 
RJ Riememsnider said:
In addition, since HM is an open system, you can use whatever probes you may want. All you would need to do is use whatever appropriate connectors on the HM(there is even a header to connect a set of jacks via a ribbon cable), use the appropriate value resistor, and change the probe values in the GUI.
Click to expand...

That's well beyond what I'd plan on doing. I already have the ET-73 probes and would plan on using them as-is. But you bring up a great point and a strength of HM.
 
Hey Ralph, quick question about your pit. Obviously it's very efficient as evidenced by the low fan speeds and steady temps. What kind of pit is it?
 
Those graphs are very interesting and it makes me wonder how we get such good results. It must have something to do with the fact that the lag between any adjustment of the output and a corresponding actual temperature change taking so long (at input lag = 0).

As is pointed out, the thermistor-based probes are the most commonly available without resorting to specialty places like Auber. Most BBQers also already have some sort of remote read thermometer so they've probably already got existing probes, and this can cut the cost of the project by probably 1/3rd.
 
It could be that the smoothing effect is masking temperature fluctuations. For example, here is an oscillation and the resulting damped signal using a 120 second moving average filter to approximate the effect of the probe's thermal mass.
zp52.png


That's why I'm curious if anybody has logged temperature with a more responsive thermometer.

Also, it's definitely possible to tune a control loop with lag. I know the autotune function of my Omega PID has a damping setting for low, medium and high depending on the input power and load/sensor coupling. Here is the diagram they use to describe the three levels.

qpm.png
 
The smoker that I made that graph with is NOT a quality pit, it's a cheapo Char-Broil H2O smoker I got at a rummage sale for $10 bucks. I think the fan speed runs so low because of the air delivery system I designed for it....

BurnPlate.jpg


This pit is great for low and slow, the HM can get rock solid temps even down below 200 degrees. My Kamado is harder to control because it holds the heat so well, but the Kamado works great for cooking pizza's!

sPizza.jpg
 
Ralph, that air distribution system is pretty cool. Is that under an ash pan or will those holes get covered during a cook?

Tasty looking pizza. Do you use parchment to get the pie off the peel? I used to do that but didn't really like how the bottom of the crust turned out so I use rice flour and corn meal to dust my peel and build the pies quickly. Haven't had a problem since I switched to this method.
 
One thing we have not mentioned that that temperature measurement speed is dependent on the coefficient of convective heat transfer for substance being measured. Liquids are much faster than gases, as most of us have undoubtedly observed already.

I've been digging some more and found an interesting article in the journal The Physics Teacher that address this.

In the article an example is given of a thermometer going from 20 ºC in air to 50 ºC in water. It takes 8.5 seconds to reach the final reading which means the time constant (τ) is 1.7 s in still water.

That same thermometer going from 50 ºC in water to 20 ºC in air takes 625 seconds to reach the final temperature which means τ is 125 s in still air. That's almost two orders of magnitude slower.

What really caught my eye is this quote:
If the experimental design is such that the temperature to be measured will not remain at this temperature for a time greater than about 5τ, then the temperature measured by the thermometer will be in error.
Click to expand...
Thus, we remain with the unanswered question: what is the time constant of a Maverick probe in air? I don't think I can fully answer that question but I have an idea involving time lapse photography and my oven. If it works I'll let you know what I find.

I'd recommend reading the article. It's short and there are some other interesting points like how a solid such as a thermometer probe has a thermal resistance equal to 1/(coefficient of convective heat transfer*surface area) and a thermal capacitance equal to (mass*specific heat), thus making temperature measurement analogous to an RC circuit.
 
WayneM said:
Ralph, that air distribution system is pretty cool. Is that under an ash pan or will those holes get covered during a cook?

Tasty looking pizza. Do you use parchment to get the pie off the peel? I used to do that but didn't really like how the bottom of the crust turned out so I use rice flour and corn meal to dust my peel and build the pies quickly. Haven't had a problem since I switched to this method.
Click to expand...

The air delivery tubes are inside the burn pan, the coals go right on it and it works great.

I'm using parchment paper on the pizza for now because I don't have a pizza peel, I would prefer to use corn meal flour on the pizza stone with no parchment paper but I have no other way to get the pizza on and off the stone currently...
 
You must log in or register to reply here.

 

Back
Top