Andrew Que Sites list Photos
Projects Contact
   Brynden celebrated his birthday and graduation this afternoon.  In the fall he is off to collage.  I remember attending an other graduation for Brynden—preschool—and it doesn't seem that long ago.  And if I don't get back on track, he'll end up finishing collage before me.

July 18, 2014

Estimating the Curve



   Did some experimentation with the early data from ππ.  The light curve looked like it followed a the Gaussian function.  I was wondering if I couldn't curve fit this function to fit my data to get a theoretical light estimate.  Long story short, the data I am logging seems asymmetrical and the Gaussian function is not a great representation.  I might continue to experiment just to see if I can't use the Gauss-Newton algorithm to do this fit for me,


   The graphing of lux over time done by ππ has been moved to the Blue-Dragon.  The data is still obtained from the Pi using sshfs but the rending of this data is done on a much faster machine.  The graph is now usable.  It had been designed so you could click on the graph and drag it to the right or left.  However each update took over one second to render and return, which made this feature pretty useless.  Now that the graph is rendered on a machine several times more powerful it updates about 4 times a second (at least locally). 
   This presents a major question: can the Pi be a good web server?  I had considered replacing the server at (known as the Micro-Dragon) with the solar powered Pi.  However, I think it may end up being too slow.  Most of the sites would be fine including the photoblog.  But the math demonstrations might not be functional without (or fast enough to be useable anyway).  More experimentation is needed.  On the plus side, there are other single board computers that seem to be following in the Pi's footsteps.  Some use a quite powerful mutli-core ARM CPU, and the specs are better than the server we have now.  The down side is they also draw more power.  So we shall see.


   Seems ππ is back to full operation and doing fine.  All the data from the couple of days around the wireless crash have produced mostly useless data.  The Raspberry Pi has no battery backed clock so when the power is cycled it losses it's time.  What's nice is that the system knows this and synchronizes the time over the Internet when it boots.  However when I cycled power trying to get the wireless back, the system started with no idea what time is was because it could not connect to the Internet.  So all the logging has invalid timestamps for those time periods.  The data is pretty much useless.
Elmwood Park

Elmwood Park

   So the wireless communications to ππ were bad, but this weekend they changed to non-existent.  After some investigation I found the USB wireless adapter were completely dead.  It wasn't recognized by the Pi, or my main machine, and acted as if it had not been plugged in.  The wireless aspect of this project has turned into more of a pain than it was worth, so my solution was to run an Ethernet cable out a window and to the roof.  After I did this, the Pi is back to logging data.

July 13, 2014

Inside the Clut House

When we left off the group had confronted a cloaked figure associated with necromancers. Ellenoria had unconvincingly tried to gain his confidence using Andel's unconscious body stating she was trying to use a raise undead spell. After being pushed out of the way she called for Marv to be a little more persuasive. Annilis stepped into the figure's path and had been thrown back over 100 feet.

Today the encounter continued. Just as this took place, the wizard suddenly appeared between the cloaked figure and Marv. The cloaked figure seemed to be interested in this sudden appearance and asked if the wizard had come from The Void. Since Joking is crazy he didn't answer directly but was kind, sorta, but not really entirely sure he did. The cloaked figure referred to the wizard as brother and invited us back to the black house. The others in the group were asked about what we wanted. After some chatting Ellenoria explained we had tried to free some Void Creatures in a dungeon we came across. We were asked the name of what we tried to free, but we only knew the numbers. There was mention of She Who Burrows and the cloaked figure said they were looking for her.

We were left to reconvene at the black house in a few hours. The Solarian, Gaius Dermious, was contacted via the Noitacinummoc stone to and given an update about a group interested in Void Creatures. We asked about She Who Burrows but Gaius did had never known their names. We suspect She Who Burrows might be the missing creature IX.

A cover story was needed. Clearly they thought the wizard was one of them. So it was decided to have everyone pretend Joking was our leader, and we were anxiety groupies who hated the world and wanted chaos, death, and destruction. After getting to Kobold's up to speed we hung outside the black house, Ellenoria played and sang some punk rock while Marv (who is 7' tall) and Sparky (2' tall) danced in a circle pit.

When the cloaked figures returned we were invited inside. The key to get in was like a sheet of paper but sometimes appeared in multiple dimensions—it was clearly magical. Unlike the outside, the inside was not painted black. There was the smell of sweet rotting. It was run down like any abandon building and had nothing inside. The group was shown to a parlor and served cookies and tea. Ellenoria was able to watch closely enough to tell the otherwise identical figures apart from they movements. A lot of magic was detected in the house. There was a strong protection spell guarding the place from outside entry, and a lot of magic we could not identify. However, there was defiantly something building up in power.

After refreshments which only some of the group ate we were lead to a basement, and then a sub-basement. Other cloaked figures were making bricks and the sub-basement looked to be something they constructed themselves. It seemed to be centered directly under the house and contained a circle of seating around a center alter. As we entered so did several other cloaked figures, 10 in all. All looked the same and only be watching their movements could one be distinguished from the other.

The group was asked if we wanted to join with the cloaked character's group. Anytime the group was asked about anything, Ellenoria started ranting about how the world was just full of conformist pigs that all worshiped stupid gods, and that chaos and death was the only thing everyone deserved. This seemed convincing enough for them and Andel agreed to be the first to be initiated.

The ritual began and Andel began to levitate partly over an open pit. The cloaked figured removed their hoods to revel they only had half a head. On the top just above the mouth their head was gone and there was a black flame. It seemed obvious that Andel was about to loss half her head in this initiation process and it was time to take action. Protesting began and the cloaked figures said a sacrifice of part of yourself was required to become part of their group. Ellenoria said that was stupid conformist garbage. Battle commenced.

Greatly outnumbered by magic users far more powerful than ourselves was not exactly what we had in mind. However we decided to concentrate our efforts on their leader—the one running the ritual. Their attacks were powerful and magic in nature. Annalis made a beautiful critical role and bloodied the leader within the first couple rounds of the battle. The enemies, other the other hand, rolled terribly. In the first round one fumbled and fell over for a two rounds. And the next round one critical fumbled and knocked themselves out for the remainder of the battle. Marv ended up killing the leader but taking a huge amount of damage in the process. Ellenoria sang songs of battle and healing, but one was knocked unconscious before the battle was complete. After the leader and two others had been killed the remaining figures fled. We had won, but something was charging up in the pit by the alter.

Repair complete

Repair complete

   After a comment by my roommate and a little reading I found that the Raspberry Pi has trouble with USB wireless adapters because it can not supply enough power.  This sounds like it is my issue and I will have to come up with a method to solve that.  There are several options, and I just have to figure out which I want to use.
   I did some experiments after the sun went down last night to see about putting in some kind of light reducer on the lux sensor.  I found a small piece of white paper has a roughly 1/4 light reduction ratio.  My tests couldn't really tell if it was linear or not so I am gambling on it being the case.  In addition, the 1/4 reduction I am not entirely sure about either but it should be close.
   Today it was cloudy all day, so not the best test for my light reduction, but the numbers seem similar to my other cloudy day.  Cloud cover really reduces the amount of light.  Yesterday we were over 20,000 lux, and today under 6,000.
   Pictured is the finished dresser.  All the gluing is complete and the back tacked back on.  Seems quite solid and is ready to be returned to service.

July 11, 2014

Understanding ππ's Stability Problems

Clamps and Setting

Clamps and Setting

   We have found that ππ likes crash periodically making Operation Lux difficult.  Turns out the Raspberry Pi was not crashing, just the network connection.  For whatever reason the Pi losses the wireless network connection and never gets it back.  Fortunately now that the lux logging happens as a start up service the device keeps logging data even without a network connection.  The reason it stopped logging before was that the logging was invoked via SSH, and once the SSH session terminated so did the logging.  Why the wireless networking card is causing a problem will have to be investigated, but it is good to know what is causing the problem.
   Today was a somewhat cloudy day averaging just under 14,000 lux for the 24 hour period.  (While the day isn't technically over, now that the sensor is reading 0 lux I can analyze the data.)  With the 30 watt solar panel I normally use for calculations I estimate that only about 80 Wh of energy would have been produced, falling short of the 84 Wh needed.  Tomorrow is suppose to have thunderstorms and my guess is considerable cloud cover.  Should be a good day for data collection.
   Pictured is the dresser I am repairing.  The side has been reglued.  For clamps I am using tie straps.  Scrap wood keeps the metal parts of the straps from damaging the dresser.  They make is easy to apply constant pressure over a large area.  There are four glue joints setting under the strap clamps.

July 10, 2014

ππ Location change

Elmwood Park with Operation Lux on the roof

Elmwood Park with Operation Lux on the roof

   The other day I moved ππ to a location on the main roof.  Initially I thought this part of the roof would be harder to access, but it turns out it is fairly easy to get to from the garage roof.  With plenty of extension cord it was easy to relocate the experiment.  Being higher up ππ should receive more light and will get no shadow from the rest of the house.  So how is the data logging going?
   Well, we have 3 uninterrupted days.  Two of these days were sunny, and one rather cloudy from the old location.  The cloudy day averaged just under 7,000 lux for a 24 hour period, but the two sunny days were 23,900 and 21,800 for average lux over 24 hours.  What does this mean for available solar energy?
   The math involves summing the total light throughout the day.  This value is then the lux day value.  This can be converted into watt days per square meter (Wd/m2) by multiplying the conversion factor of 0.0079.  A watt day isn't a normal unit, but dividing by the 24 hours in a day will convert a watt day into a watt hour.  Doing so will convert watt days per square meter (Wd/m2) to watt hours per square meter (Wh/m2).  Now a solar panel needs to be selected where the area of efficiency of the panel is known.  Multiplying the watt hour per square meter by the light collecting area will produce the amount of energy available to the solar panel in units of watt hours (Wh).  This needs to be multiplied by the efficiency of the solar panel to get the watt hours actually produced.  To run the Raspberry Pi for 24 hours requires 3.5 W * 24 h = 84 Wh of power.  As long as the watt hours produced exceeds the watt hours needed, the solar panel will produce enough to run the device.  However, this assumes the energy is constantly available and battery efficiency needs to be factored in.  That is for an other article.
   I have been doing my calculations with a 30 watt solar panel that has a 0.1875 m2 area and 16% efficiency.  For the 23,900 lux day I have calculated this panel will produce 136 Wh, and 123 Wh for the 21,800 lux day.  Both of these cover the needs for the Raspberry Pi (without battery and cable efficiency factored in).  However, the 7,000 lux day is only calculated to produce 38 Wh which is far short of the power needed.
   Presently there is too little data available to make any conclusions.  However, we are getting a feel for the the number we have to work with.  There are two problems that need to be addressed in this setup.  First, why ππ keeps crashing.  And second (as of yet unmentioned) is that the lux meter saturates at 70k lux and this happens much of the afternoon.  We need to somehow add a light divided so we can measure to over 100k lux.  Otherwise we are not getting good data for the afternoon.