easy-forum
English Support - Hardware => easyControl => Thema gestartet von: Elka am November 09, 2014, 17:29:48 Nachmittag
-
Hello,
I am a coffee roaster from Poland who bought a machine for roasting cofee
driven by Moeller PLC and MFD-TAP13-PT-A.
My temperature readings were heavily smoothed by PT1 filter. We switched
the filter off but would like to know the exact formula of the filter which was damaging our data in the past.
Any help will be welcome.
-
Hi Elka,
For help, we need the PLC type (800, EC4P, ...).
PT1 filter FT in easy800 PLC:
Y(t) = [Ta/Tg] x [Kpx (x(t) - Y(t-1))]
regards Thomas
-
Thank you so much! Nobody has been able to answear this question so far, even then I had asked Moeller in Germany and in Poland and send my PLC documentation to Moeller offices.
My PLC is Easy M800 and MFD Titan.
We had 2 temperatures readings, corrisponding to BT (bean temperature) and ET (environmental temperature).
When BT was not in oscillations mode due to burner way of working (1 minute ON, 1 minute OFF) a delay in BT reading was as big as 60 s. In oscillation mode strong smoothment was observed (the amplitude of BT oscillations was cut by 2/3) . When the burner was OFF BT/ET temperatures seemed to rise and when it was OFF they seemed to decrease.
Near local extrema the delay seemed to rise and decrease..What is K in your formula? What is Ta and Tg?
Thank you for your explanations once more..
-
We have found such an item:
ftp://ftp.moeller.net/DOCUMENTATION/AWB_MANUALS/h1365g.pdf
and read on the page 33 about PT1 smoothing filter but we still don't understand what is Kpx?
we know that: (t) is probably the real temperature registered by PT100. y(t-1) is the display reading just one sample time before. Kpx is any amplification...
Ta-actual time?
Tg delay time
-
we have found also an information (on Polish Elektroda.pl) tat K=0,38 for an "ideal" low pass filter..
-
In attached file you may find a comparison beetween real measurment from Pt100 via Phidget 1046 and smoothed "mesurment" by our ex-PT1 filter.
If we could know the exact formula of our filter it would be GREAT!
-
Here is the correct file. The previous one seemed to be empty
-
Everything is described also here:
http://www.home-barista.com/home-roasting/2014-world-coffee-roasting-competition-profiles-t31029-20.html
and you may see plots posted by my husband under nickname "Antonio"
-
Here is a screen of our PT1 filter:
http://datasheet.octopart.com/MFD-TAP13-PT-A-Moeller-datasheet-5399030.pdf
-
Hi,
here is the User Manual for easy800:
ftp://ftp.moeller.net/DOCUMENTATION/AWB_MANUALS/MN04902001Z_EN.pdfPlease read page 192-195 to understand the FT function block.
-
thank you very much :-)
I have got it finally
Goodnight :)
-
I would like to ask if from my pdf file with data (I would like to send also another one representing a full roasting session) is possible to simulate/calculate (?) the exact formula of MY filter? its Tg and Kp. How?
I am a biologist, non an electronic engeener so I cannot do it by myselve..
It would be possible to find my Kp/Tg somewhere in the easy project also?
I have a copy of it.
All
-
I would like to ask also if this fiter may be described as lower pass filter or inertion filter?
-
Hi Elka,
FT is a low-pass filter. Your data table shows it.
It is possible to simulate/calculate FT funktion with EXCEL. If you use Easysoft, you can log and show function values online in software oscilloscope. Here you can see the step response of signals in your program, to optimize the process.
Professional function simu software are SIMULINK, Mathlab ...
I think, Kp (proportional gain factor) shold be 100 (=factor 1)
Here is a screen of our PT1 filter:
http://datasheet.octopart.com/MFD-TAP13-PT-A-Moeller-datasheet-5399030.pdf
MFD-TAP13-PT-A is not a filter, it`s an I/O modul with PT100 sensor inputs
Thomas
-
Hi Thomas,
my data were already simulated by a Polish engeener and he obtained K=0,38
He thinks a low pass filter does not describe this problem. He made my a spectrum analysisis of signals, but I would like to know the EXACT formula..
When the coffee is roasted and the roasters does not try to keep the temperature stable, data are delayed by 60s..
We switched OFF the filter now..
-
Hello Thomas
Do you think a symbol of PT1 filer in the Easy Project may be: PTO1/1, PT02/1
-
You may find a file in the attachment. A Polish automatic engeener told me that he cannot find any software filter in this project..
-
Here is another one..
-
..the third one:
-
and the last one..
-
I have obtained such an e-mail from Eato:
"Dear Mrs. Elzbieta Wierzchowska,
please excuse my late reply. What exactly do you want? Read temperature
and record? This filter (low pass and PT1 behaviour) smoothes the analog
input signals. This suppresses variations in the temperature sensor
values that are caused by high-frequency interference. Note: More
effective filtering, however, is provided by a hardware filter, which is
implemented by the A/D converter. You can set this hardware filter via
the Filter area.
The staggered use of hardware and software filters drastically increases
the recovery time. This should not, however, be greater than necessary,
otherwise the signals will be delayed more than is required by the
smoothing. (This is an excerpt from the Easysoft help!)"
They have not sent me any useful information so far.
Unfortunately my data seem not to switch a formula of the filter..
-
Hi Elka,
I could see the code this morning. Without comments, it was difficult, but I understand it.
As your engineer said, there are no FT blocks in the program.
Unfortunately, I do not know the exact formula for the modul filter. If it does not tell EATON, one could only determine approximately with traces.
Some parameters of the filter I've marked you in the picture.
For a higher resolution of the sensor values, I would switch to 0.1 ° C. There are but then the program several parameters to change.
The filter can be set to 50ms sampling rate.
regards Thomas
-
Hello Thomas,
thank you for your answear. We were trying to change scaling while switching off the filter, but our Giesen coffee roaster: GW15 have not wanted to work after this change..
Today I was contacted by a Polish Eaton representative, mister Jacek Zarzycki who told me that Eaton Central Office rapresented by mister Roberto Ariosto considers parameters of lower pass-filters are "prioprietary" so they cannot be announced to anybody..
From the same Roberto Ariosto I recevied a month ago the following e - mail:
"Dear Mrs. Elzbieta Wierzchowska,
please excuse my late reply. What exactly do you want? Read temperature and record?
This filter (low pass and PT1 behaviour) smoothes the analog input signals. This
suppresses variations in the temperature sensor values that are caused by
high-frequency interference.
Note: More effective filtering, however, is provided by a hardware filter, which is
implemented by the A/D converter. You can set this hardware filter via the Filter
area.
The staggered use of hardware and software filters drastically increases the
recovery time. This should not, however, be greater than necessary, otherwise the
signals will be delayed more than is required by the smoothing.
(This is an excerpt from the Easysoft help!)
Should you have any further questions, please do not hesitate
to contact us.
With best regards
Avec de sincères salutations / Con cordiales saludos
Roberto Ariosto
Applikations-& Produktsupport Automation
Produktsupport Safety Technology
************************************************************
Eaton Industries GmbH
Electrical Sector / Industrial Automation Business
IAB-MTS / Applikations- & Produktsupport
Hein-Moeller-Str. 7-11
D-53115 Bonn"
Today a Polish represnative of the same company told me that filtering "improve precision of measurments"..
-
Hi Elka,
I wrote, if you change the resolution, you must also change the program. The machine could not work like that.
I send you a new program file with the changes, so it works.
The changes are indicated in the images.
My idea was to determine the temperature in high resolution read by MD30 / 31 for the SCADA and the values in 1 ° C resolution output to MD10 / 11 on the old program.
Please change the SCADA system to the read access of MD24 (it`s ° C / ° F) -> MD30 (only 0.1 ° C) and MD25 -> MD31 to.
I hope your measurement process is better!
Thomas
PS: I could not install the sensor offset for a PT100 calibration in the program because of insufficient memory.
-
Dear Thomas,
You are so nice and helpful! Thank you so much!
Elisabeth
-
After thinking and calculation I was able to give to a student who is writing a software for us the following instruction to estimate my roasting profiles registered with PT1 filter ON for T-bean..since time 30s after charge:
Ror=Rate of rise of T-bean (calculated by 30s)
T-bean_new=estimated "right", non smoothed T-bean
T-bean="old smoothed temperature)
T_bean_new (30s)=T-bean-2/3 Ror
T_bean_new (60s)=T_bean-Ror
T_bean_new (100s/1min.40s)=T_bean+Ror
t100s=Phidget "real" minimum for our profiles
T-bean_new (140s /2 min 20s)=T_bean+1/2 Ror
T_bean_new (150s /2 min 30s)=T_bean (!)
T_bean_new (165s /2 min 45 s)=T_bean+6C
6C is a half of the BT amplitude which was "allowed" by PT1 filter during auto-mode oscillations.
T_bean+new (180s /3 min)=T_bean + 6C
T_bean_new (205s /3 min 15 s)=T_bean + 3*Ror
T_bean_new (210s /3 min 30s)=T_bean+2*Ror
after..to the the end of the roasting curve:
T-bean_new(n)=T_bean+2*Ror..
We are still looking for a right formula to estimate our past charging temperatures.
It is possible to say more about formula of the PT1 filter from my calculations?
-
Hi Thomas,
Somebody from the Polish Forum Elektroda wrote that for a graphic rapresentation of PT1 filter this page may be very useful:
http://en.wikipedia.org/wiki/Bode_plot
What do you think about it?
All the best,
Elisabeth
-
Hi Elka,
I tried to replicate the heating curve of the roaster mathematically.
In the program, a two-position controller works.
The diagram is not entirely correct, but it should show about the heating curve.
(1-e^(-t/T))repesents a PT1 delay.
If you want to have an exact curve of the filter, you need the sensor to the hardware simulate and record a step response.
Thomas
-
Dear Thomas,
you are incredible! Thank you so much!!!
Elisabeth
-
Hi Elka,
can you send me a set of measurement data of the step response?
Unfortunately I have no MFD hardware to make it yourself.
please proceed as follows:
To Build a circuit with changeover switch by an electrician.
0 --- --- R100,00 \ =0°C
| switch --- 0
| _R175,84_ / =200°C
The resistors have to be as exact as possible.
(possibly from potentiometers build)
terminal of this circuit instead of the PT100 T1.
Set PT filter ON.
Start easysoft in online mode with oscilloscope
Create the record of the MD30 and turn the switch to 175 Ohm.
stop the measurement when the value is stable. It should not take longer than a minute.
Now you have a step response of the filter only. With dataset can derive a formula from.
Note: Do not start the roaster, only the PLC must be in RUN mode!
-
Dear Thomas,
Thank you very much for the instruction. It is possible we will make what you suggest, even then for me personally it seems difficult (for my husband not)
I have publsihed your suggestions on my fb profile:
https://www.facebook.com/elzbieta.wierzchowska.96?fref=ufi
Giesen roaster was an ufficial roasting tool during last World Roasting Championship and competitors were penalised for differnces between their "roasting plans" and their real drop values: 1 C of differnce, 1 point of penality. I wrote:
"By now with PT1 filter switched off "Giesen BT/ET readings" are still ALWAYS lower than Artisan/Phidget readings and the differnce is not so small: (3-6 C)..so we do not stick to "absolute ET/BT values" and try to concentrate more on NAVIGATION itself: a shape of BT/ET curves, an inclination of the Ror curve and on everything COFFEE ITSELF tell us.
We hope only somebody will double check a roasting tool for the next World Roasting Competition and will think twice before setting up the rules about "roasting plans"..
We have changed the original VERY SLOW PT100 with a new quicker one, so we have one more problem with calibration between these two thermoprobes..I will came to this subject later.
Have a nice weekend and thank you for EVERYTHING:-),
Elisabeth