Rancilio Silvia too hot with PID

[jfb_1973]

Hi, I just got my pid up and running and it seems like the recommended temp, 230 F is way too hot. Once it gets up to 200, if I press the brew switch I get a bunch of steam, as if I had the steam on.

I followed the instructions on Murph's site, with the exception that I have a different pid. I checked to make sure that the temp is in Fahrenheit. Is there anything I can check to make sure I have this set up correctly?

I'm using this pid

http://auberins.com/index.php?mai...h=1&products_id=14

Thanks

[another_jim]

If you're getting steam with the controller reading 200F, and presuming you're not living in the Andes, there's two big possibilities:

1. The sensor is off -- stick it in boiling water and see if it reads 212 -- also check that you haven't used a K setting in the controller for a J-type thermocouple or vice versa.

2. The sensor is not in good thermal contact with the boiler surface; check that your heatsink compound forms a complete seal between the sensor and the boiler surface, and that the sensor is fiirmly attached.

[jfb_1973]

I measured the temp of boiling water and it's reading 186 degrees. When I measured the temp I stuck the whole thing in the water. I wasn't sure if I was only supposed to put the washer part in there.

It looks like it's measuring room temp ok, around 80 degrees. I've got it set for the J thermocouple. Is there anything else that I could be doing wrong?

Thanks

[miKe mcKoffee]

I measured the temp of boiling water and it's reading 186 degrees. When I measured the temp I stuck the whole thing in the water. I wasn't sure if I was only supposed to put the washer part in there.

It looks like it's measuring room temp ok, around 80 degrees. I've got it set for the J thermocouple. Is there anything else that I could be doing wrong?

Thanks

Obviously it's not reading boiling water correctly unless you're many miles high elevation! (Not going to bother figuring out how high, but much higher than Mile High Denver:-) And hence the boiler will be about 25f higher than anticipated.

Are you sure you have a J type thermocouple?

If thermocouple type correctly set another possibility, have you played with the PID offset setting?

[jfb_1973]

It looks like I have a type t thermocouple. Duh. I changed it to type T and it looks more or less correct, around 205 degrees. Here in Austin at 500 ft it looks like the boiling point should be around 211. Is that normal for there to be such a discrepancy? I was thinking I could just set the offset to +5 degrees.

[erics]

Greetings Austin, Texas

"Calibrate" the sensor and the PID by holding the probe at the midpoint of a pot of vigorously boiling water or use a distilled ice/water slurpy mix and split the difference in corrections. Set your PSb parameter to about -6 based on what you have thus far said, i.e. PID reads 205 when it should have read 211. Keep in mind that these inexpensive PID's (while great values) do not display tenths so your 205 could be 205.01 or 205.99.

Set Sv equal to 228 as a STARTING point.

When Pv climbs to around 200-205 after you have turned on the machine, initiate the autotuning process.

Record all your parameters.

Change P, I, d, SF, & bb to 2.1, 28, 7, 6, & 10 respectively.

Did you buy the whole package from Auber Instruments - if so, you got a good deal as he is a reputable seller. Did you ever make up a pressure gage?

Eric S.

[jfb_1973]

Ok, so I got it all working. Once I autotuned it came up with these numbers:

P 14.5
I 346
D 86
SF 7
bb 124

Which are way off the numbers you suggested

P 2.1
I 28
D 7
SF 6
bb 10

Now, I have my sv set to 228. If I let it run for quite a while and it hovers around 218. It's also very slow to respond. When it's raising the temp it flashes every couple of seconds heating up the boiler very slowly. When I plugged in your numbers it raced up to around 255 before I shut it down to prevent a meltdown.

The whole thing is strange because I thought it would try and hit 228 and stay there. I assumed that the different value parameters would only affect the responsiveness of the system to temp change.

[another_jim]

Your I and D parameters are completely whacked, the P looks kind of high, something you'd only see in a super fast responding system.

Here's the manual tuning (Zeigler-Nichols) instructions:

1. put the setpoint to 180 or something reasonable
2. set the I & D to 0, the other parameters to their defaults,
3. start with the P at around 10 and let it stabilize.

You'll either see it settle to a fixed temperature, but somewhat off from the 180 setting, or you'll see it cycle around the setting.
-- If you see it settle and stabilize, lower the P setting till it starts cycling, make a note of this P value. Then time the cycle and make a note of the time period (the full sine wave, up from the setpoint, then down below it, then back to the setpoint).
-- If it is cycling, raise the P value till it stops cycling, then follow the instructions above.

-- Your P value is: 1.7 times the critical P value
-- Your I value is the cycle time divided by 2
-- Your D value is the cycle time divided by 8.
-- Check the controller documentation, the I and D values are usually in seconds, in that case use the cycle time in seconds for your figure. If the setting is in minutes or 10th seconds (yours looks like it may be in 1/10ths) use that unit for the cycle time.

I have no clue what the other parameters are; so I can't help you with setting them.

Many PID controllers have a moving average filter on the input to filter out noise. You should make sure this is set very low, to 0.2 seconds or less, since you are dealing with a fast process in a non-noisy environment.

[erics]

At last count, there were four inexpensive 1/32 DIN PID temperature controllers available on Ebay. In no particular order, they are the TET7100, SET712, SYL1512, and XMT7100. One look at all four of these controllers will tell you that they are all manufactured by the same company in China. They all have the same feature set which is dramatically reduced from that available with the Fuji PXR3 and other high quality controllers. Their available documentation is also dramatically less (in quantity and quality) than that available for the Fuji line. However, they do "do the job" and their pricing is, obviously, very attractive.

I have had much email exchange with great people at TTI Global and Auber Instruments in an attempt to FULLY understand the "P" and "bb" parameters. A better manual for these controllers can be downloaded from my "website":

http://users.rcn.com/erics/PID

Click on PID_Instruction_Manual and it opens a Microsoft Word Document.

With the Fuji line of controllers, including the PXR3, "P" is the proportional band entered as a percent of the input range. According to TTI Global, the default input range for the Fuji PXR3 is 32 to 1000 degrees F and I assume this is simply the "factory" setting for P-SL and P-SU. This can (and should be) changed to a much narrower band, say 50 to 350 degrees F when used on an espresso machine. The only reason I would take it up to 350 is to allow for a display reading of Pv when going into steaming mode (aren't those cappy's great?)

Understanding "P" with the Chinese controllers is a little more difficult because of a variety of reasons. In their case, "P" is defined as a proportional constant but is also entered as a percent of the input range from 0.1 to 99.9 percent. A good question (and the end result is that it really doesn't matter) becomes "what is the input range" because with these controllers, unlike the Fuji line and others, the input range is not user defined. The simple numerical range for Sv input with these controllers is minus 199 to plus 2999 and this numerical range is unchanged regardless of what temperature sensor has been selected.

With the Chinese controllers, the parameter "bb" is the proportional band, entered directly in degrees. As an example, if Sv is 200 and bb is 10, the controller will signal for power on full time at any temp below 190 and power off full time at any temp above 210. This is sorta equivalent to the hysteresis parameter with the Fuji controllers except that "bb" functions when you are in both PID and on/off mode whereas hysteresis in the Fuji only applies to on-off mode.

Still learning,

Eric S.

[another_jim]

With the Chinese controllers, the parameter "bb" is the proportional band, entered directly in degrees. As an example, if Sv is 200 and bb is 10, the controller will signal for power on full time at any temp below 190 and power off full time at any temp above 210. This is sorta equivalent to the hysteresis parameter with the Fuji controllers except that "bb" functions when you are in both PID and on/off mode whereas hysteresis only applies to on-off mode.

That's odd, since the P and bb are then two versions of the same P parameter. There has to be a third parameter enabling either one or the other. In any case, the same manual tuning method I gave before (Zeigler-Nichols ultimate cycle) applies to the bb parameter in that case, if it is the one being used.

[erics]

"There has to be a third parameter enabling either one or the other" - my thoughts exactly until I was given the description of the differences between the Chinese PID controllers and other controllers. I had thought that parameter "bb" was simply setting the deadband in on-off control mode, i.e. that it was exactly the same as the hysteresis parameter with the Fuji. But with the Chinese controllers (the four I mentioned), such is simply NOT the case. With the Chinese controllers, P is representative of the gain, hence their terminology "proportional constant" vice "proportional band".

I have one doing PID duty on Silvia and I gar-ran-tee you that the instant Pv falls 10 degrees under Sv, heat is on full time (with parameter bb set to 10, obviously). If I set bb at "x", heat comes on at an "x" degree differential - like clockwork. There is absolutely no overshoot and I return to within +0/-1 degree of Sv easily within 30 seconds of completing a typical double-shot.

Certainly the ZN method you gave is a well respected manual tuning methodology and was great of you to mention same but it is less applicable to these controllers than it is to one such as the Fuji or similar.

I am sensing boiler water temperature instead of boiler surface temperature as I am sure jfb_1973 is doing. It makes little difference as Greg S. has said many times - I'm just a hard head at heart.


Eric S.

[erics]

Now, I have my sv set to 228. If I let it run for quite a while and it hovers around 218. It's also very slow to respond. When it's raising the temp it flashes every couple of seconds heating up the boiler very slowly. When I plugged in your numbers it raced up to around 255 before I shut it down to prevent a meltdown.

I assume you are using a washer style thermocouple secured to the boiler top with one of the screws that formerly held the brew thermostat. It would be nice but not absolutely essential that you use some heat conductive grease between the washer surface and the boiler. Radio Shack sells this in a small tube.

Did you make the 5 or 6 degree correction to the display?

It has been a while but I seem to recall some big overshoots as you described when playing around with the various parameters, especially "P". No harm can be done as it is very easy to control Silvia manually either by flicking on the steam switch for more heat or hitting the brew switch to quickly cool things down.

My complete set of parameters for the controller are:

Inty: t; Outy: 2; Caty: 0; Psb: -3; Rd: 0; CorF: 1
P: 2.1; I: 28; D: 7; SF: 6; bb: 10; ot: 2; filt: 0

Set the initialization parameters first, then do autotuning when Pv is around 205. During autotuning, Pv should hit around 238, maybe 240 if Sv is 228. Autotuning SIlvia should take about 10 minutes if I recall.

Eric S.
erics@erols.com

[jfb_1973]

Yeah I adjusted the temp up 5 degrees. I'm still having a lot of problems. So last night I autotuned it again after playing around with manual tuning. After autotuning it stabilized at 238, which dosen't really make sense to me. I dropped the temp by running water through the group and it heated back up to 250. So I'm totally confused..

Is there anything I could have wrong with the wiring that would cause this problem?

[erics]

I can think of nothing wrong with your wiring that would cause the problems you are seeing.

What are the current values of the parameters? i.e. duplicate my list.

Perhaps post some pics of the installation? Power off the machine, unplug it from the wall and give a slight tug on all connections to the SSR, PID controller and any crimp connections inside the machine.

Power off the machine and unplug. Disconnect one of the connections to the heating element and set aside. Power up machine and PID. Try manually entering in all of my values except your own PSB which should be minus 5.0 or minus 6.0. After you have done this, double check that what you entered has taken by repeating the whole process. The Inty symbol for a type T thermocouple is not a full "t". Power off the machine and reconnect heating element.

Power on machine - temperature starts climbing - heating light should be full on until measured temp reaches Sv minus bb which should be 218 in your case. Light will go out at 218 and controller will initiate proportional control.

Eric S.

[jfb_1973]

I realized that the one difference between my settings and yours is that you have caty=0 and I had caty=2 (fuzzy logic). So when I put in your numbers the temp ranges from 219 to 230 and back down over a period of a little less than 10 minutes.

I just thought of one thing I've done that may be causing a problem. The line that runs from the ssr out of silvia is 12 gauge copper wire. Once it exits the back of silvia I have it joined with electrical caps temporarily to an extension cord I cut. The main reason being that the 12 gauge wire didn't fit into the terminals on the back of the pid. Perhaps this is causing problems?

[erics]

Well at least you're getting better but this is still not satisfactory. What exactly does Silvia do?

You said "So when I put in your numbers the temp ranges from 219 to 230 over a period of a little less than 10 minutes."

Does this mean the machine (from cold) zoomed to 219 in, say 3 minutes and then took 6.5 minutes to hit 230?

Or did Silvia reach 230 and then drop back to 219 in less than 10 minutes?

The control signal to the SSR from the PID can best be made with very flexible 18 gage extension cord wire and is polarity sensitive. At this point I would make straight connections to try and eliminate sources of problems.

At this point I would also visit Radio Shack and get a small tube of their thermal grease and apply it to the underside of the washer where it contacts the boiler surface.

The output of the SSR can be connected to the old brew thermostat female spades using 14 gage stranded wire and male spade connectors.

Lower P by a couple of tenths and increase bb to 12 and see what effect that has.

Eric S.

edit No. 1 - And if this fails, my only other suggestion would be to do a search on this site and on coffeegeek.com for PID parameters. It is not my intent that you use my parameters because we are sensing differently. However, they are a heck of a lot better than what you found by autotuning. I believe that the great majority of Silvia PID installations sense boiler surface temp with a washer style thermocouple.

edit No. 2 - try setting caty = 0 and then do your autotuning at Pv = 205, i.e. 10 percent below your chosen Sv. I am fuzzy on fuzzy logic but at least I sorta understand normal PID operation. "bb" is a very important parameter with the Chinese controllers - oftentimes overlooked because most people assume a controller is a controller whether it be Chinese or US or European. Such is not the case here.

[jggall01]

This spring I spent a fair amount of time evaluating the 1/16 DIN version (SYL-2352) of this same (similar?) controller for possible inclusion in PID kits. I experienced some of the problems described by the OP, in particular:
- runaway temperatures while autotuning
- inability to hold anything even close to SP after autotuning

The definitions of the P, I, and d parameters on the 2352 (and presumably the 1/32 DIN models) are very different than those on common controllers like Fuji, Watlow, Cal. So my careful attempts at manual tuning using the ZN algorithm described in Jim's post above were completely unsuccessful.

To make a long story short, I finally got the controller tuned to the point that it would hold set point on the Silvia within less than 1C. Overshoot after pulling a shot was still a problem (6C), but it recovered to stable control in around 3 1/2 minutes - not great, but usable.

The successful formula for me with the 2352 controller was this:
- manually reset P = 0 (instructions indicate that this also resets the offset??)
- keep the controller in its native Celsius units for autotuning (mine did not care for F)
- dial in around 4 units of digital filtering (my controller exhibited a great deal of noise in the temperature readings)
- run autotune from a dead cold Silvia

I had tried several variations of the above procedure with no success.

Successful AT results on my controller (very weird looking values, I agree) were:
- P = 108
- I = 260
- d = 66

I then manually changed to d = 33 and saw a very small decrease in the overshoot.

At this point I boxed the controller back up, satisfied that it was too finicky and sensitive to risk sending to a customer. (BTW, I still have the nearly new controller, still in its original box. Make me any reasonable offer and it is yours!)

I understand that my data, from a different Auber controller, may not be particularly useful to the OP. But Eric knew of my efforts and suggested that I post the info here, just in case.

Please don't interpret my comments as a slam on Auber Instruments. I had trouble getting mine to work as a temp controller for Silvia, but it could easily have been my too many birthdays at work again. Tech support from Auber was very friendly and responsive, but ultimately unsuccessful.

Jim

[another_jim]

I've gotten two Silvias, "Treatment Group"and "Control Group," pided by Dave B, for use as a two group espresso lab (hence the unromantic names). I spent yesterday playing around with the tuning. Dave uses T-type TCs screwed to the Tstat bracket, so the setup is thermally identical to yours.

The settings I use are P = 20C band, I= 120 seconds per repeat, D= 15 seconds, this is the ZN tuning with the I and D slightly softened (from 90 seconds and 22 seconds) to limit the overshoot. The TLC tuning algorithm used by the Cal is trickle slow.

The band on yours sounds like percent of range (unless it has a band width setting parameter like the Fujis), which would be 0 to 250 on a type T, so my setting would translate to 8. Given that the Cal autotuned this correctly (compared to the ultimate cycle band of about 12C to 15C), I'm not sure at all how the units on the Auber work. If they use proportional gain, rather than proportional band, everything on it is inverted (bigger P means more aggressive).

An alternative tuning is to emulate an ultratight deadband stat, basically letting it hunt inside a 3C band. Set the P to about 5C, the I huge to about 15 seconds, and the D very low, 3 to 4 seconds. Not sure anyone is interested in this approach. Setting the unit to be a stat doesn't hold it nearly as tight.

[jggall01]

Jim -

As you surmised, the P parameter on the Auber is a proportional gain. The I and d parameters are gains as well. I think the reason I struggled so much with this was inability to rely on "tried and true" values for these parameters. I had no choice but to rely completely on the AT. Oddly, I don't think I ever found a parameter that described the bandwidth outside of which proportional control was inactive.

OT - I have found that a slow responding control scheme works best for controlling Silvia. Intuitively, it seems that this might be because the slope of the curve on the heating side of the cycle is so much steeper than the slope on the cooling side of the cycle. Overshooting the setpoint carries a pretty severe time penalty as you wait for the boiler to cool (of course you can always pump a little water to speed that cooling up).

I have also found that the most difficult control situation occurs with a shot that is long, i.e. 30 seconds or more for a 2 oz double. It is very difficult to avoid overshoot here because the heating element is on for such a long time and the cooling effect of incoming water is small due to the low flow rate. I try to tune controllers to respond well to an incoming/outgoing flow rate of around 2.5 oz in 25 seconds.

FWIW, the parameters I find to work consistently well (on Watlow controllers) are:

P = 12 F degrees
rE = 0.35 repeats per minute (corresponds to around I = 3 minutes)
rA = 0.25 minutes = d

Using the ZN procedure suggests a higher value for P, around 18F, but I have not found this to be as effective.

Jim

[mike]

Wanted to wake up this thread, see if anybody has had luck manually tuning the Auberins PIDs. I'll keep a long story short and just say that I haven't had any luck with getting help from the seller on manually tuning, and have been living with some cantankerous PIDs for some time.

I'm running two of the PIDs on my Silvia, one on the boiler, one on a grouphead heater. Lino Verna is running one on his Silvia boiler. Both of us have seen extremely slow convergence, and never reach the SV (a few degrees short). Autotune definitely doesn't do it on these cheap pids - these aren't plug and play like the Fujis!

I have played with I and D (always in a 4:1 ratio), and got quicker response and reaching SV, but with a slow overshoot and correction. I've been leaving P at the AT value, and I think that is the value I need to tweak, so I'm thinking I might as well go back to scratch and determine them all properly.

I'm up for manually tuning using Z-N or the Scace post techniques, but can't figure out how to disable I and D to do the initial experimentation on P. The manual doesn't give any help here either (but thanks to EricS for making them available). Mine are the PIDs before 9-06.

I'll be happy to post my experiences with manually tuning the Auberins PIDs here.

I'm wondering if I can adopt the min or max of the range to make the I or D effect so minimal so as to be effectively disabled?

I range = 2-1999 (sec) - use max or min to "disable"?
D range = 0-399 (sec) - use max or min to "disable"?

(Sorry, I just can't remember my PID theory!)

As far as other settings, here is what I've found, but please correct me if I'm wrong:

CAty: I've not seen much difference between 0 and 2, so I will probably use 0

BB: as near as I can tell, we just want to shut off the full-blast heater soon enough to not cruise past the SV (ie. coast). The grouphead heater rope is moderately enough powered that I can make this a small number, the boiler I have to set a bit further out. 10-20F seemed to be a reasonable range, but this can be played with.

SF: I would have thought this would be the same as BB - as soon as the BB control is no longer in effect, you'd think you'd have the integration take over. Either way, I think if it is set less than BB and within the reach of the "coast", it should be OK.

FILt: this isn't a noisy environment, so I think 0 is fine.

ot: might as well set for 2s, the finest granularity

PS: there is some chance that the two heaters may interact and cause some interference, but nobody else has reported a problem in this config, and I definitely see poor boiler PID performance even when the grouphead is up to temp and steady, and vice versa. So, I'm pretty convinced that the problem is in the PID settings, not an interaction. Esp since Lino only has his on the boiler and his performance is poor too.