Quick Mill Carola PID Test & Analysis:

In this post, I will describe a test while I monitored the Carola's heater continuously for an hour. This test reveals:

1. How the Carola's PID behaves.
2. How quickly the machine's boiler warms up.
3. How quickly the machine recovers after pulling shots.


Intro:

There's been some interesting discussion about the Carola's PID behavior and performance. For example, is the Carola's heater controller a true PID or something less (like an electronic thermostat)? In the end, I'm hoping for some of you to help shed some light on this.


Test Method:

The easiest way to track the PID's behavior would have been to hook up a data logger to the heating circuit. What I had at hand was an iPhone.

So, instead I took a 72 minute video of the heater light and analyzed the brightness of the pixels in each frame. The video below video depicts this test method as Carola transitions from the "heating up" to "maintaining set temperature."

Raw Data:

The raw data reveals how the heater pulses. For example, this graph shows in more detail the transition mentioned above.



The PID consistently works on a 2 second cycle, during which the heater will pulse for a certain amount of time. The length of this heater pulse ranges from 1.5 seconds (while heating up) down to 0.1 seconds.

Definition: Duty cyle - the percentage of time that the heater is on.

Thus, the numbers indicate that the duty cycle ranges from 75% (1.5/2) down to 5% (0.1/2). Below we see the PID switch from a 40% duty cycle to a 5% duty cycle.




Results:

Below I have graphed the duty cycle for the entire 72 minute test.



This graph has a lot of interesting information. Some notable observations as we move from left to right across the graph:

• The PID reaches a maximum duty cycle of 75% (which is 1.5 sec ON / 0.5 sec OFF, repeated).

• The boiler reaches temperature after about 10 minutes. The heater's ramp down only takes about 30 seconds, but there are a few intermediate steps as it ramps down (i.e. it doesn't just immediately switch between two modes.)

• To maintain the set temperature, the PID alternates between a 5% and 40% duty cycle roughly every minute. This is the main part that concerns me regarding whether it's truly behaving like a PID and/or keeping temperature very constant.

• As the overall machine continues to heat up (from 10-28 minutes,) the 40% duty cycle time durations gradually become shorter and more consistent. The overall machine is reaching steady-state and requires less energy to stay at temperature.

• After pulling one shot at 29 minutes, the machine does an extra long 40% duty cycle and appears to recover after about 2 minutes. The effect of this shot can be seen for about 10 minutes, as the 40% duty cycle time durations taper off.

• After pulling 2 successive shots at 53 minutes, recovery takes about 5 minutes, and the machine even kicked into a 70% duty cycle for a bit. The overall machine seems to get back to steady-state after 10 minutes.


Summary:

In summary, the PID does seem to have some level of intelligence going on, and it's always actively doing something to try to hit a target. But it seems a bit crude, so I'm not sure if it's really doing what most PIDs do.

1. Behavior
It only has about 3 types of behavior - a 75%, 40%, and 5% duty cycle, all of which operate on a repeating 2 second time period. The PID alternates between these three duty cycles for various amounts of time.

2. Warm Up
As described in my review above and in other reviews, the boiler water gets up to temperature after about 10 minutes. But the overall machine has not yet reached steady-state until about 20-30 minutes. So while you can pull a shot after 10 minutes, it may be advisable to do a flush to help the brew head get up to temperature.

3. Shot Recovery
The boiler temperature takes 2-3 minutes per shot to recover, and after you are done pulling shots the overall machine seems to take about 10 minutes to get back to steady-state.


Please Share Your Expertise or Experience!

So that's more than anybody ever wanted to know about the Carola's PID behavior! But obviously I find this kind of thing fun to tinker with.

What I'm really curious about is whether this type of behavior is typical of all espresso machine PIDs, or if it's just how the Carola's cheapish "PID" works (or if it's not really a true PID or more like a proportional controller with a few extra tricks up its sleeve.)

Please feel free to offer your opinions, expertise, or experience with the Carola or other machines! Thanks.

Thanks for the analysis of the warm up and temp stability of the Carola. Oddly and reassuringly the quick short pulses of the indicator light are what I've been using to know the machine is up to temp, and the total warm up times are what I've experienced -- so I am glad the science agrees with my seat of the pants experience!!

You are seriously a saint. Nice to get some real data on the behavior.

One suggestion for an additional project if you are up to it, to get a decent measure of what the actual temperature in the boiler and in the grouphead are doing, put a skin temperature sensor under the insulation on the boiler and on the grouphead and a thermocouple wire on the shower screen above the puck (k-type thermocouples connected to a readout device). Then you could video this data on your phone as well and see how the entire system is acting. This would help yourself and everyone understand what the actual boiler water temperature, grouphead temperature, and temperature at the puck are doing.

However, you have already put so much time into this, and probably already given us enough info to help us make buying decisions. So, I don't blame you if you doing wish to take this on. It would also cost about $50 to buy everything on Amazon. But, if you were up for it, it would be a cool project.

That's some fine investigative work on the temperature control for the machine. The question is how tight does the temperature have to be controlled on a small home model espresso machine? I suspect that very few users would taste the difference with this machine and one with a full blown PID.
We are not landing an aircraft on autopilot, just making a cup of espresso. It does the job.
To appreciate a true PID, and a rather ruthless one at that, watch this video... http://www.youtube.com/watch?v=j4OmVLc_oDw
Enjoy your coffees,

Dave

Don't overlook the importance of the group temp here. Most e61s need at least 45 minutes to heat up.

The temp of the group e61 has a signifiant impact on the shot. I would wager it also has a large influence over the actual thermal stability during a brewing session. More so than a boiler recovery which the PId is focused on.

Thanks for the nice comments so far.

Agreed, it would be fun to take it to the next level, in which case I'd use a proper data acquisition unit. My day job is as a test engineer, so it would just be a matter of whether I get curious enough someday down the road. But it would more be for curiosity/tinkering purposes than for practical purposes.

Honestly, if I were to spend more money on the machine, it would probably be for one of these as a practical everyday tool. It would also give the "final answer" as to how consistent the temperature is at the puck, regardless of how the machine accomplishes it. Maybe my next birthday

In the end, I'm still quite happy with the Carola, even though I suspect the "PID"' isn't the smartest.

Greetings,

After 15 years my Zaffiro became too costly to repair and I purchased the Carola. The Carola was an excellent replacement.

Cliff

Also. Have you tried using the boiling point to determine the proper setting? Jim S. Posted more about that somewhere while he was dialing his Bianca pid offset

mlunsford27 wrote:Thanks for the recommendations on PID controllers. The website does say that the Carola has a PID, so I think that it does but just not a digital display. Therefore, it will still control the temperature without overshooting and then dropping down in cycles causing deadband. In this case it just appears to be a pain to change the temperature.

I was also confused about whether my machine (A QM Carola) came with a PID or not. The Carola EVO seems to have the same device.

As you point out, the documentation says it is a PID. But I am an electrical engineer, and after looking at the behavior of the device, I spoke with the folks at Chris' Coffee, and they confirmed that it is indeed an digital thermostat. Your graphs above give us some more clues about what it is doing.

Here is a guess:
Mode 1: Stone cold
Run at a 75% duty cycle

Mode 2: Within 'n' degrees of set point
Switch to a 40% duty cycle

Mode 3: Within '(smaller than 'n')' degrees of set point (we are basically there)
Switch to a 5% duty cycle

Simple, but quite effective.

I have owned my Carola for several years now, and while I am sure there are some here who will disagree, for my application, I have never had any problems I could blame on poor temperature regulation. But - and this is a big but, I make one to two espressos per day. If I was using the machine in a commercial application, I absolutely could see where a PID controller would be the way to go. For me, the thermostat gets the E61 head up to temperature in about 20 minutes, but it is not stabilized yet. (But in an emergency you can get a decent cup of espresso out of it - YMMV). And because of the huge thermal mass of that head, after about 35 minutes, it seems very stable, although I agree that 45 minutes or more is ideal.

I did experiment with the dip switches a little bit in the beginning to get it dialed in, but I have not touched it since then. For my money, the number one thing that affects the quality of your cup is the quality and freshness of your beans. The number two thing is getting the grind right (including spending enough on a grinder to get good results) so that the number three thing, the timing of your pull, is right. Just about everything else is in the noise.

Oh yeah - and one other thing. You will enjoy the beans you like (doh).

You can (and I have) spend hours or days trying to get things to work out so you get a good cup, only to discover in the end that you just don't like the beans. Even if other people swear that these beans make the Best Espresso Ever! If you have fresh beans, have messed around with all kinds of grind settings and still end up throwing the cup in the sink, try some radically different beans. Always buy Colombian or South American beans? Try making espresso from some Ethiopian beans. You get the idea.

There are plenty of good articles out there on the difference between a digital thermostat and a PID. Probably the biggest difference is that the PID "learns" the thermal characteristics of the system, and then uses what it knows to apply just the right amount of heat for just the right amount of time to keep the boiler at the setpoint.

Think of it this way - it learns the thermal ballistics of the system. From what it knows, it calculates a trajectory and then gives the heater a "nudge". Then it observes how the system behaves, recalculates the ballistics and then gives another nudge. Or doesn't. When you get to the set point, you may overshoot. Then the PID begins calculating when and how fast the temperature will begin to drop, waits a bit, and then gives the boiler a quick zap to try to hit the set point exactly. Of course then it observes that it has missed slightly, recalculates, etc., etc., etc. It is a very busy thing.

Think calculus. Don't think about the PID measuring the temperature, think about it measuring the RATE at which the temperature is changing - and in which direction with relation to the set point. Then think about the PID looking back at what it has learned about how the rate of temperature change has responded to different TIMES of application of boiler voltage in the past in order to estimate how long it needs to apply voltage this time in order to hit the set point.

You need a solid state relay with a PID controller because the controller may interrupt power to the boiler hundreds of times per minute (it depends on how the PID is configured as to how "fidgety" it is). A mechanical relay simply could not keep up with the demands of a PID and would fail prematurely.

Hope this helps.
Brad

narwhal wrote:Honestly, if I were to spend more money on the machine, it would probably be for one of these as a practical everyday tool.

Yep - did that. I don't have the results anymore, but on my QM Carola, the temperature stayed within a few tenths of a degree. The thing is that the Carola has that big honking piece of metal screwed to the front, with water from the boiler constantly circulating through it due to thermo-siphoning. If you think about it, to budge the temperature of the E61 head, you would need to a) make a substantial change to the temperature of the water in the boiler, and then b) circulate that through the thermo-siphon long enough to c) move the temperature of more than a pound of metal. It is gonna take quite a while for a temperature change in the digital thermostat to make it out to the group head.

In other words, once the temperature stabilizes, you got to push really hard to move all that thermal mass. And the digital thermostat just is not going to let that happen.

If you pull eight shots in a row, that is different. Now you are into boiler temperature recovery time, water capacity of the boiler, etc. There is a reason that commercial machines are commercial. But if you are making one, maybe two cups at a time, that head temperature is very consistent.

Brad