With the electronics done, they needed to be tested with the system, and
I was now able to see timing affects of various delays, and tank
To simplify this description, we need to remember that there are two outputs of the tank. One vents straight up with nothing in the way, and no ignitor. This will vent the propane, and not ignite it. I will refer to this as vent. The second output has hot surface ignitor's at the opening which will ignite the propane that leaves. I will refer to this as ignited.
The sequence I started with, opens the vent, then after a 750ms delay, opens the ignited. Both are left open for another 750ms, then both shut off. The purpose of this was to get a volume of propane in the air above the candle before igniting. Being able to do this was a design feature from the beginning, which I came up with from researching. All of the examples I found on the web produced higher velocity flames, some even approaching "flame thrower" type looks. I've never found any examples of this pre-ignition venting technique, so as far as I know I came up with it just from thinking of various techniques for achieving the appearance I was after. This is part of the reason I wanted to log this project, to share the idea to others.
This 750-750 sequence worked well right off the bat. The accumulator tank pressure after running was <10psi(gauge is not very detailed below that), meaning nearly the entire volume was released, a positive sign. I went through a bunch of pressures, and when it came down to it, higher pressure was better for flame height and size as expected. I had a concern that high pressures would lead more readily to the flame thrower effect, as well as make the result quite loud. Given the pre-ignition vent, this wasn't a problem. In the end, I decided on running at 60psi.
Shortening the vent time reduced the flame size and height, and trended towards more of the flame thrower effect. Lengthening it didn't seem to help as I could tell that it ended up reducing the accumulator tank pressure a lot prior to the ignition release. I suspect this also would be impacted by too much of the vent propane drifting away and not being used.
The ignition time at 750 also seemed pretty good. Going up to 1250ms made for an obvious flame thrower effect at the end, which was undesirable in this situation. Shortening made for smaller flames. In the end, I ended up at 850ms as maximized without getting the flame thrower effect. That also points to the actual time that there is flame in the air.
The resulting effect is quite impressive. The flames reach more than 30ft in the air, and do look like the desired fireball. At about 30ft away, you easily feel the heat. The couple were present for some of this testing, and squeed with joy when they first saw it. I'll try to get a video up soon.
It is also noted that the dual stage release means that the accumulator tank is at a lower pressure before ignited release. This probably also has a large help avoiding the flame thrower effect.
At the higher pressures, it is noted that there is an extra delay prior to ignition, perhaps in the order of 250ms. That is, the ignition propane is released, and 250ms later, it is actually ignited. I believe this is due to needing to end up with the correct propane-oxygen mixture at the HSI. Possibly due to some of the high velocity propane falling back down or swirling down towards it. I think it may be a good idea to mount the HSI higher up than I did. I will look at modifying the mounting of the second HSI to be higher.
Though not yet tested in stronger breezes, I think it is best to substantially reduce the vent time as most of that propane will just be blown out of the area before ignition, and not serve any purpose.
The no wind/slight breeze sequence I intend to go with is: open vent, 750ms delay, open ignited, 850ms delay, close both.
The stronger breeze or more sequence I intend to go with is: open vent, 100ms delay, open ignited, 850ms delay, close both.
In testing, probably 20 nearly full accumulator tank releases over about 20 minutes, and the source LP tank was starting to slush up and the evaporation rate observed to be reduced. (The tank was about half full during this testing)