Homeroasters - Home Roasting Coffee Community » DATALOGGERS - CONTROLLERS - RATE OF RISE METERS » Dataloggers/Controllers/Rate of Rise Meters
Who is here? 1 guest(s)
Page 2 of 5: 12345
 Print Thread
TC4 - Bourbon application software
JimG
Top  #26 Print Post
Posted on 04/08/2011 8:39 PM

Quote

Bhante wrote:I'm wondering what my best solution for the filtering is in this case. I had BT_Filter and ET_Filter both on 10, and Rise_Filter on 85

First, get the latest aBourbon here:
https://tc4-shiel...hes/RB-200

Then open the user.h file and edit the values for RISE_FILTER and ROR_FILTER.

RISE_FILTER acts on the values of BT, smoothing them out before the RoR calculation is performed.

The ROR_FILTER is new, and is the result of working with Bill on his Gesha program. This new filter acts on the computed RoR values themselves.

We found that a combination of these filters gives a good result with not quite the lag penalty.

I'd suggest maybe trying 85 for both filters, and then moving up from there.

Jim
 
Web
JimG
Top  #27 Print Post
Posted on 04/08/2011 8:57 PM

Quote

bvwelch wrote:
With just slightly more effort, you could define an equivalent message for each of the 4 tc button operations. I'm not sure how useful that is, but t wouldn't be too hard.

Useful, I think, and already done.

Button 2 marks first crack.

Button 3 marks second crack.

Button 4 marks eject, and now also causes the screen to be captured for posterity smile

Jim
 
Web
bvwelch
Top  #28 Print Post
Posted on 04/08/2011 11:14 PM

Quote

JimG wrote:
Useful, I think, and already done.

Button 2 marks first crack.

Button 3 marks second crack.

Button 4 marks eject, and now also causes the screen to be captured for posterity smile

Jim


Let me try again -- I was suggesting, that in a manner similar to pBourbon sending RESET, pBourbon could send additional messages if you type F,S, E, or whatever on the PC keyboard -- send them back to aBourbon, which might light an LED or something???
 
Bhante
Top  #29 Print Post
Posted on 04/09/2011 3:56 PM

Quote

JimG wrote:
Then open the user.h file and edit the values for RISE_FILTER and ROR_FILTER.

Many thanks.

Is there a formula I could plug into excel for experimenting, or is it more complicated than that?
 
Bhante
Top  #30 Print Post
Posted on 04/09/2011 3:59 PM

Quote

JimG wrote:
Button 4 marks eject, and now also causes the screen to be captured for posterity smile

You know, it would be nice to have the screen to be captured automatically when closing the window, when it saves the data.
 
JimG
Top  #31 Print Post
Posted on 04/09/2011 8:34 PM

Quote

Bhante wrote:
Is there a formula I could plug into excel for experimenting, or is it more complicated than that?


Yi = ( 1 - beta ) * Xi + beta * Yi-1

Y is the series of filtered values.

X is the series of raw data.

Beta is the amount of filtering. If beta is zero, then no filtering is done. As beta gets close to 1, then the data become very heavily filtered.

In the aBourbon code, beta is actually specified as a percent, but you get the idea.

Jim
 
Web
JimG
Top  #32 Print Post
Posted on 04/09/2011 8:35 PM

Quote

Bhante wrote:
You know, it would be nice to have the screen to be captured automatically when closing the window, when it saves the data.

Piece of cake ;)
 
Web
greencardigan
Top  #33 Print Post
Posted on 04/10/2011 1:52 AM
I've just finished a roast using the latest Release Branch of pBourbon (RB-200). I was using pBourbon together with the TC4/PICAXE combination rather than the usual TC4/Arduino combination.

The latest version allows the option of loading up an old logfile as a background as seen in the snapshot below. The logfile loaded here is Jim's test roast as posted earlier in this thread. smile

www.greencardigan.com/misc/air_roaster_build/data_logger_controller/roast201104101548-6891.jpg

This test highlighted a few issues.

1. The background plot of ROR is not showing the exaggerated values, just the raw values out of the log file. I can fix this easily.

2. The ROR for this roast is also not showing exagerated values. This is due to a coding error in my PICAXE code only. An Arduino with running aBourbon will not have this issue.

3. Negative ROR values are getting plotted as positives. Again, this is a coding error with my PICAXE code only. But I think I like this error! Maybe the line could just change color for negative values? Any thoughts on this?


EDIT: I forgot to mention I was not recording ET as my ET probe is playing up.

Brad
Edited by greencardigan on 04/10/2011 1:53 AM
 
Bhante
Top  #34 Print Post
Posted on 04/10/2011 7:00 AM

Quote

JimG wrote:

Quote

Bhante wrote:
You know, it would be nice to have the screen to be captured automatically when closing the window, when it saves the data.

Piece of cake ;)

Your cake always tastes good Jim!
 
Bhante
Top  #35 Print Post
Posted on 04/10/2011 8:11 AM

Quote

greencardigan wrote:
But I think I like this error! Maybe the line could just change color for negative values? Any thoughts on this?

Interesting, but I suspect it would be rather confusing.

Talking of colours I am somewhat colour-blind and I don't get on too well with the coloured lines on a black background. Especially these dark red lines for me are so faint they are barely to be seen at all. Maybe for people with normal colour vision they seem quite bright and clear!? Dark red, dark blue and dark green (for example) lines on a white background would be much easier to see (at least for my eyes), and they would also print much better. The templates would be good as a relatively faint gray, and the grid could be a very very faint gray, so faint that it almost fades into the background but still aids reading off temperature without interfering. The text font for the markers could afford to be about half the point size.

Bhante
 
Bhante
Top  #36 Print Post
Posted on 04/11/2011 3:55 PM

Quote

JimG wrote:

Quote

Bhante wrote:
Is there a formula I could plug into excel for experimenting, or is it more complicated than that?

Yi = ( 1 - beta ) * Xi + beta * Yi-1

Y is the series of filtered values.
X is the series of raw data.

Thanks Jim. I've tried it on the data from the last roast. Plotting it turned out a bit complicate - I'll have to install the earlier version of Excel that I was using before, the 2003 version is far too difficult to format sensibly. Values for beta of .60 seem to give reasonable results, higher and the distortions become much greater, the closer you get to 1. That is with data that has already been filtered once though, I'm not sure how that affects the results. I'll have to try and generate some unfiltered data and then try again.
 
JimG
Top  #37 Print Post
Posted on 04/11/2011 5:40 PM

Quote

Bhante wrote:
That is with data that has already been filtered once though, I'm not sure how that affects the results. I'll have to try and generate some unfiltered data and then try again.

In user.h, change BT_FILTER and ET_FILTER to 0. This will give you unfiltered values in the log file for BT and ET.

The log file is in CSV format, so is easily used by Excel.

Jim
 
Web
drip_kid
Top  #38 Print Post
Posted on 04/11/2011 11:30 PM
Thanks all for contributing to this thread. I've been waiting on thermocouples to come in and in such a short amount of time this has gotten super organized and lots of input added. By the time I actually get this running it's going to be great. Wish I could buy you a beer!
-Mat
 
JimG
Top  #39 Print Post
Posted on 04/15/2011 11:57 AM
Bourbon Roast Monitor Release 2.00 is now ready for download.

Hats off to Brad (greencardigan) for significant improvements on the PC side of things. He added several new features, and together he and I were able to fix the timing issues that had bothered me since the first version of Bourbon.

Thanks to Bhante for his helpful suggestions, and to Bill for keeping us on the right track.

Please give the newest release of Bourbon a try on your TC4/Arduino, either as a standalone roast monitor, or with pBourbon running on your laptop for realtime plotting and logging.

The roast markers are a very nice addition.

Probably my favorite addition is the ability to plot the log of a previous roast (nicely faded to the background) as a guide for the current session. (The original feature of plotting a guide profile for BT is maintained and made a little easier to activate).

http://tc4-shield...EL-200.zip

There is a long README in the zip file with more details.

Jim
 
Web
Bhante
Top  #40 Print Post
Posted on 04/15/2011 12:21 PM

Quote

JimG wrote:
Bourbon Roast Monitor Release 2.00 is now ready for download.
Please give the newest release of Bourbon a try on your TC4/Arduino
Jim

Nicely timed! I'll try it out now.

Bhante
 
JimG
Top  #41 Print Post
Posted on 04/16/2011 5:12 PM
I am proceeding to work on a few features intended for the next (3.x) release of Bourbon.

Initial thoughts:

1. We should also release standalone executables that do not require Processing to be installed on the user's computer.
2. If a saved log file, "logfile.csv", is present, it will automatically be read and plotted. Same deal with a guide profile, "profile.csv". If they are both present, they both get read.
3. If the config file indicates Celsius operations, then the program looks for "logfile_c.csv" and "profile_c.csv" instead.

Other ideas?? -- keeping in mind that Bourbon is intended to be the entry level application for the TC4 ;)

Jim
Edited by JimG on 04/16/2011 5:13 PM
 
Web
greencardigan
Top  #42 Print Post
Posted on 04/17/2011 7:40 AM
A standalone executable would be good.

As far as features go I think pBourbon is just about done. The only thing I don't like is the traces starting behind the axis text. But in can live with that.

Brad
 
JimG
Top  #43 Print Post
Posted on 04/18/2011 4:21 PM
Bourbon Roast Monitor 2.10 is ready for download from the tc4-shield.googlecode.com project site.

There were timing issues at startup with the PICAXE and Uno that sometimes caused the timer reset to happen late.

New code (discussed in the tech thread) was added to make the communication with the remote more robust.

Link for the 2.10 zip file:
http://tc4-shield...EL-210.zip

Jim
Edited by JimG on 04/18/2011 4:32 PM
 
Web
bvwelch
Top  #44 Print Post
Posted on 04/18/2011 6:10 PM
Well, it may not be of interest, since version 2.00 is already obsolete, but here's a roast I did this afternoon. Seemed to work fine, except it looks like I'll need averaging on the Processing side to smooth out the RoR noise...

bvwelch.com/tc4/newclassic.jpg
Edited by bvwelch on 04/18/2011 6:11 PM
 
greencardigan
Top  #45 Print Post
Posted on 04/18/2011 7:30 PM
Maybe just try increasing the filter % in aBourbon? I did a roast last night with the rise filter at 90% and the RoR filter at 94%. The lag in the ror trace wasn't as bad as I expected.
 
Bhante
Top  #46 Print Post
Posted on 04/19/2011 12:25 PM
I have done a roast on the hottop with Bourbon 2.0 with no environment sensor, and I have found a rather troubling roughly linear relationship between T0 and T1. Is there any reason to expect such a relationship? Of course, when there is no T1 sensor plugged into the TC4, its value is to be expected to be somewhat undefined, but such a specific relation to T0 is unexpected to me. If the channels are not fully independent when one sensor is disconnected, why should they be fully independent with the relevant sensors connected? Possible explanations that spring to mind include: some kind of memory effect in the ADC between conversions; capacitance effect; or some strange effect mediated through the calculation of temperature correction. The ambient temperature however is virtually constant, and does not appear to be the cause.

up.picr.de/6875729bnt.jpg

up.picr.de/6876683cmw.jpg
Edited by Bhante on 04/19/2011 1:12 PM
 
Bhante
Top  #47 Print Post
Posted on 04/19/2011 1:52 PM
I've done two roasts with Bourbon 2.0, both on the HottoP: first with all filtering disabled (to collect data so that I can experiment with the filtering in excel), and then the second with all filtering on the default settings.

Unsurprisingly the first plot with no filtering does not allow RoR to give much useful information, but the purpose was only for data collection (I haven't had time to experiment with the data yet).

up.picr.de/6876900guk.jpg

The plot from the second roast is with the default values for filtering:

up.picr.de/6876931cog.jpg

I have changed the graph colours, which my eyes get on rather better with; the colours were done somewhat quickly by trial and error and there is naturally room for improvement.

It would be nice to attach the time and temperature to the event markers. Earlier in this thread I attached a picture from GreenBean's roastlogger program, which marks the events in a very nice way. The template events are shown in gray with labels below. The current roast is shown in colour, with the events in red above, so that they do not conflict. In each case there is an arrow from the label to the event, either from above or from below as required.

up.picr.de/6773506pex.png
 
JimG
Top  #48 Print Post
Posted on 04/19/2011 4:03 PM

Quote

Bhante wrote:when there is no T1 sensor plugged into the TC4

Not sure we can draw any conclusions from readings made with an open circuit. Try looping a wire between the open terminals and see if the behavior is repeated.

Jim
 
Web
JimG
Top  #49 Print Post
Posted on 04/19/2011 10:44 PM
If there are any Ubuntu users out there, I am finding that my machine (an older Dell laptop) definitely does not like framerates faster than 1 in pBourbon. Besides that, pBourbon is running very well on Ubuntu 10.10 with Processing 1.2.1.

Jim
 
Web
JimG
Top  #50 Print Post
Posted on 04/23/2011 10:18 AM
First roast of the day (Ethiopian + Yemen mix). Monitored with Bourbon running on Ubuntu laptop. Roasted using manually controlled heater/fan on Hottop:
JimG attached the following image:
roast01.jpg
 
Web
Page 2 of 5: 12345
Jump to Forum: