Mexican Stir-Fry?

I had a can of beans that was going to expire soon, so I figured I would make some dinner revolving around it. When I went to the store I was thinking Mexican. At the store I picked up a shallot and some steak figuring I would try to make one of those dishes where there is the steak smothered in a spicy red sauce. Which naturally become I’ll cut up the steak and cook it like a stir-fry so it’ll be easier to make. And here is my creation:

Lessons learned

1. Read the can before you plan around it because Baked Beans are not equal to Beans.
2. Any “sauce” that I make always ends up brown. (Ok, maybe this isn’t a lesson, but a quandary left unsolved)
3. I really need to learn how to stir-fry. My meat always ends up dry and over-done and my veggies are always underdone.
4. I should figure out which spices are actually in Mexican food. The only part I hit was spicy.

A new project emerges

So when it rains it pours, but is a good way. I have felt like the last couple years I have been rather dry of projects. I wanted to do something, but could never come up with an idea. Well now I feel like I have too many projects and no where near enough time. Some of this might also be having more than enough motivation to actually work on them all.

So the new project… I always forget if I have closed the garage door, so I wanted to hook up a small computer and camera to let me see if the door is open and then close it. Hopefully I should be able to figure out how to controller the lights on the Garage Doors easily. And lastly if at all possible, I want to power the whole thing off of the Garage door itself to make it small and easier to hook up.

Understanding the Garage Door

Step 1 was clearly to understand the garage door opener and understand how to control it. I poked around online a little and saw very little information. The one piece of information I did find was that it was a 24v system and that you needed to send certain width pulses at specific frequencies. And that the light button and door button it sent slightly different patterns. Because there wasn’t enough information out there I figured I needed to at least test for myself.

After measuring everything it looks like when the default buttons are attached, the voltage is around 5v. But when they aren’t, the open voltage is closer to 20v. So that is different. I calculated back and found the source impedance was around 1.2kΩ. After a little more digging it turns out that my system just uses a simple voltage signal. Anything greater than 3v is “do nothing”, 1.2-3v is “turn the light on/off”, and everything below 0.5v is “open/close the garage door”. (Note: those are slightly rounded as the accuracy I had with setting the resistance was about ±0.2v.)

Also to get to the 5v that the normal buttons were at, it was drawing about 10mA. So since I have 2 garage doors, I might be able to draw up to 20mA total. So if I had some super capacitors, I could set it up to handle more than 20mA burst as long as the majority of the time I am below 10mA or so.

WiFi Boards

I don’t really want to run cables all through my house to do this, and I really want it accessible remotely. This leads me to logically thinking that I want something that can either connect via WiFi, Bluetooth, or RF to my server. Looking at what boards are our there I found a bunch of different Bluetooth and RF boards, but I could get a whole WiFi board for just as cheap. So I settled on trying out the Photon and the ESP8266 based ESP-12F board.

Photon

The photon board showed up first. It took me a second to get everything powered on and connected. They did an amazing job getting that ecosystem very easy to use. The base firmware lets you connect from your phone, register the device to your account, and then you can do everything from their website to write and program your device.

But that then got me thinking. Everything I was doing was going through their servers. From compiling, to programming, to even telling it to turn on the LED was going up to the cloud and back.This seems really inefficient. I played with it a little more and was able to program directly from my PC, but I still needed to compile in the cloud.

So I then measured the current draw… It was hovering around 100mA! That was way higher than I expected. Apparently when they said deep sleep, they meant off. I was used to from the MSP430 days that deep sleep meant that you could wake up on interrupts and get back to things. I was hoping with a really simple software like what I was running we could get down to 12mA, or at least something around 50mA that I could then work on getting the power settings right.

Poking at trying to get the device to sleep revealed some bugs or at least non-obvious design patterns around low-power modes. There is a sleep command that just messes up the WiFi, It doesn’t appear to actually sleep the processor itself. So I have yet to find a way to say sleep the processor and wait until I get a packet.

ESP8266

I am not sure what to call the board. It looks like the ESP-12F board is part of a family of boards all based around the same general layout with the same processor, but all slightly different. This also took much longer to get my environment setup and running. Which I expected to some point because it doesn’t have a built it USB programmer/serial port.

I struggled to find good documentation on how to get started with Arduino and the ESP8266 boards. And by that I mean I found a ton of blog posts and instructables, but nothing seemed too concrete.  After I got it all up and running, well it felt exactly like the Photon except for how you program it. It seems very obvious that Photon was heavily based on Arduino (or the other way around, I don’t know which one came first). After playing with it for a while, it seems like https://github.com/esp8266/Arduino is the best reference. Again the current is way higher than I would like, but maybe I can remove the Arduino layer and write directly for the processor and then get the power down.

Updating the Website

In a recent post I mentioned that I had done the Cruise Against Hodgman decks. Part of this required me to get a website. So I put something together and have slowly updated it as I go. However, I am now thinking about expanding into more board game design. So my existing “hardcoded” website isn’t going to cut it.

I have started looking at using Umbraco, a CMS solution. I picked it because it was based on C# and MVC, which I already have experience with. It looks nice from the standpoint of being able to create Document Types with different data pieces and then a template that explains how to display it. I have a nice editor that allows me to easily tweak pages without republishing everything. Some of the other features look promising too, like setting publish/unpublish dates.

However, I am not sure how to work around some of the patterns I am used to. Being able to independently work on multiple different computers. The configuration on how the Document Types are setup are paired with files on disk that are created at runtime. This just makes it very hard to publish a site to a new location without wiping everything there.

Do you have a CMS system in place? How do you handle development, code changes, and publishing? How do you handle backup and restores?

Cruise Against Hodgman

Leading up to JCCC4 I kept hearing people talk about Cards Against Humanity. It is a popular game on the boat to the point that there was talk of a celebrity match. I wanted to be involved, but not being a celebrity I wasn’t sure how. Finally, I decided that I wanted to print up a deck based on jokes, quotes, and themes from JCCC and the surrounding nerd sphere. But I wanted the deck to be nice. I mean really nice. Like matching-the-Cards-Against-Humanity-quality nice because if you are going to go through that process, why not?

Along with James and Beth, we started looking into printing options, ideas, coordinating, etc. After a couple meetings brainstorming and refining we had a deck of cards, it ended up being not as hard as I expected. Last thing was to coordinate with a bunch of people on the cruise to chip in and I was able to printed enough to make it reasonable to print at that quality.

I got the decks and carefully packed them into a duffel bag and headed to the airport. Of course I was going to carry the decks onto the plane because why would I want to pay to check them? Apparently, TSA doesn’t like large bags of decks of cards. They pulled me to the side, and opened up the bag and the proceeded to pull out each deck and wipe it down. Apparently decks of cards look a lot like C4 and so they needed to test each one to make sure that they weren’t explosives. That took a while.

I ended up trying to get all of the people I based cards on to sign a card.  It was an awesome way to get up to meet them, although I never really knew what to say to them which sadly ended up being more about getting the signature than getting to really meet them.

John Hodgman was the “villain” of the cruise on JCCC3. So when I was trying to figure out what to name the deck, I obviously needed “Cruise” as the first word, but what was a good H? Then it hit me, Hodgman. I met him early on on the cruise and I asked him to sign one of the cards about him. He looked at it and made a comment about misspelling his name. I went on to explain how I was so worried I would do that and constantly was checking again. And then he showed me the card. The ONE spot I misspelled his name was on his card. I was so embarrassed. Later in the cruise Hodgman came out onto stage before someone else’s performance and went on little rant about how I misspelled his name and how he was going to sue me out of existence. It was oddly scary and awesome.

Another person I got to meet on the cruise was Max Tempkin! Cards Against Humanity was a sponsor for the cruise, so it made things a little awkward. Max was a very nice guy and we had a good chat. I’m really sorry if I ended up preventing him from getting to leave because a group of people formed around us as we were starting to leave obviously looking to talk with him.

For JoCo Cruise 5, work kept me busy till the last second so there was a mad rush at the end to get everything in and done. But I did it right this time, I pre-shipped the cards down to Florida and I brought a hand truck to move the cards around. This made the whole trip and experience much easier. Hodgman was unable to make the cruise, so I am not sure who really won there. Again I made a few small mistakes on the cards, because of course.

For JoCo Cruise 2016, I wanted to expand my crew a little to ensure that we had a bunch of good cards so I asked my friend Nick to join us. We came up with a ton a new cards and a lot of great feedback. However, that was a learning experience as I believe everyone had their own idea of what we were making and how. So at the end when we needed to lock down it become obvious that we had misconceptions. We resolved our differences and came to a common ground and were able to print the deck on time. Everyone seemed to really like the deck.

It was a lot fun to get to create something like this. I hope everyone enjoyed their decks. You can see the cards that we printed up here, http://hyperentertainment.org/.

Hungry Hungry Hippos

A long time ago, John Scalzi posted something about being bad at remembering names.

So naturally I tried to response with something I thought would be funny with a slight nod to The Simpsons.

@scalzi challenge accepted, I will get you to recognize meeting me sometime. Ok, this is hard, I’m bored. Let’s play hungry hungry hippos!

— Derek Mantey (@doolbneerg) April 4, 2014

I was surprised when he responded.

@doolbneerg If you bring Hungry HungryHippos when we meet, I will definitely remember you.

— John Scalzi (@scalzi) April 4, 2014

And as my earlier post with the Menorah has shown, I like doing silly things. So the next time we were in the same area I brought along my (newly purchased) Hungry Hungry Hippos to challenge him to a game.

@scalzi Thank you for the good game! I had a lot of fun, even if you beat me 4 to 1… #JCC5 pic.twitter.com/fRd5JnVIhG

— Derek Mantey (@doolbneerg) February 10, 2015

I am fairly certain he did forget about the whole thing, but then again why would he remember.  It was fun. I enjoyed the interaction and hope he didn’t think me too weird.

Reaction Time

I feel a little ashamed to admit this… I am really curious about how people react. Like probably unhealthfully so. I very rarely end up satisfying this curiosity because I have no way of knowing how people are actually reacting, or what the final outcome is. It is like in Patch Adams where he surprises the lady and then a while later she turns around and smiles. Would that have been interesting to do if there was no recognizable reaction? What if it had the same effect but they weren’t around to see it?

A long time ago in college, back when everyone was still using AIM a lot, I was in a computer lab. I noticed a girl near me on AIM and quickly noted her screen name. When I got home I IM’d her, said that I saw her in the lab and then asked if I could have her screen name so we could talk more. She responded with No. I never talked with her again, but I always have wondered what it was like from her side. Was I just another gross guy? Was it weird but in a charming way? Did it make her smile at all? Or the second the window closed did she forget about it and it had no more effect on her life?

GoBe Data

After having my HealBe GoBe for about a month I attempted to gather some good data on how it performs. This however has proved to be a challenge. First, it seems that the sync between the GoBe and the website occurs asynchronously on the phone. Early on I kept loosing data during the day. At the time my best guess is that this is because I would open the app for a second to make sure the UI came up and then turn the phone off. The phone would show the data that one time, but then the next time I would open the app some of the data might go missing. I had one day where probably 90% of the data was lost because of this. Recently I saw a note about fixing a bug around losing data, so who knows what the root cause really was.

Secondly, the app asks you questions about when you ate to make sure that it is doing the right thing. It took me a while, but it looks like there are two different phrases it will use: “Please confirm that you were eating food” and “Are you sure that you were eating food”. It seems like those imply if it auto-detected the eating versus you pressing the button. It would have been nice for them to actually just call that out. This was compounded by the fact that during my tracking the time ranges presented were about an hour off from reality.  If I ate breakfast at 7:45, it would ask if I ate between 8:30 and 8:50.  Again, this made me lose data early on.

So here is some of the data that I gathered that seems valid enough… On August 1st we ate out at a restaurant that didn’t have nutrition information, so it is possible the calories I counted there are off. I have a FitBit One that I use to track steps, and I used their food database to track my calorie intake. I tried to measure out all of the food I ate using a digital scale when possible to get the more accurate data.

Part of the way through the experiment I realized it might be interesting to get ultra-data-collecty… I tried to capture what GoBe thought through-out the day, right before I ate. Unfortunately if I ate two meals or snacks very close together it wouldn’t be done measuring the previous one so that would cause some offset, but that only happened a couple times.  Hopefully that’ll give some insight into how much it thinks I am eating a meal.

8/4/2015	FitBit	GoBe	Error	Time	Diff FitBit	Diff GoBe	Diff Error
Before Lunch	415	745	79.5%		415	745	79.5%
After Lunch	1526	1734	13.6%		1111	989	-11.0%
Before Dinner	1796	2555	42.3%		270	821	204.1%
Total	2830	4238	49.8%		1034	1683	62.8%

8/5/2015	FitBit	GoBe	Error	Time	Diff FitBit	Diff GoBe	Diff Error
Before Lunch	529	756	42.9%	11:25 AM	529	756	42.9%
After Lunch	1573	1966	25.0%	3:45 PM	1044	1210	15.9%
Before Dinner	1843	2945	59.8%	6:56 PM	270	979	262.6%
Total	2717	3691	35.8%		874	746	-14.6%

8/6/2015	FitBit	GoBe	Error	Time	Diff FitBit	Diff GoBe	Diff Error
Before Breakfast	210	448	113.3%	9:04 AM	210	448	113.3%
Before Morning Snack	584	1241	112.5%	11:20 PM	374	793	112.0%
Before Lunch	687	1565	127.8%	12:17 PM	103	324	214.6%
After Lunch	1722	2409	39.9%	3:40 PM	1035	844	-18.5%
Before Recovery Drink	1992	3184	59.8%	6:40 PM	270	775	187.0%
Before Dinner	2112	3704	75.4%	8:20 PM	120	520	333.3%
Total	3127	4434	41.8%		1015	730	-28.1%

8/7/2015	FitBit	GoBe	Error	Time	Diff FitBit	Diff GoBe	Diff Error
After Breakfast	421	368	-12.6%	8:54 AM	421	368	-12.6%
Before Second Snack	501	1083	116.2%	10:52 AM	80	715	793.8%
Before Lunch	592	1434	142.2%	11:42 AM	91	351	285.7%
Before Afternoon Snack	1650	1952	18.3%	3:13 PM	1058	518	-51.0%
Before Dinner	2020	3314	64.1%	7:23 PM	370	1362	268.1%
Total	2999	4080	36.0%		979	766	-21.8%

So far it actually seems more like it is just giving me estimates on how much I ate based on my activity and current “deficit” of calories than actually measuring calorie intake. You can see that the average meal size it “reads” is actually fairly constant even though my meal sizes were not, especially when I started to try to add a lot of small snacks.

They have continued to pushed out updates to the firmware, but doing this type of data collection is very time consuming, so I haven’t gotten any more. Many of their release notes talk about trying to handle various sizes of meals.  Once they actually called out that they are calibrated for 400-800 calorie meals, which fits with the data I have been seeing. Until I see something better I am guessing that yes they aren’t actually able to measure calorie intake, they just found a sensor that helps them be good at guessing normal meals.

What are your thoughts?