Steam Captain wrote: ↑Mon Jun 14, 2021 2:24 pm
Correct, a syphon won't work. The elasticity brought into the equation by adding a syphon would completely ruin any recognizability of a pV-graph.
I've tried to get my hands on a pressure transducer, but the answer I got was always "We only sell to commercial buyers" - Those pressure sensors would have been perfect. They worked up to 400°C and covered all pressure ranges below the Gpa scale. And they were small (M5 thread). BUT they didn't want to sell.
I secretly settled for spark plug pressure sensors right now unless there'll be a source of a good pressure sensor to switch them out. The challenge is, that almost all pressure sensors on the buyable market are not suitable because they simply measure too slowly and not just a little, but several orders of magnitude too slowly.
A siphon filled w/ water should have very little effect on the response of the pressure transducer.
I've had very little difficulty buying components of all sorts, esp. over the web.
The above claims a response time of less than 1 mS; this should be sufficient to characterize performance at 300 rpm (5 rev/sec ) or so; it'll be a bit squirrely around valve opening events.
''A siphon filled w/ water should have very little effect on the response of the pressure transducer''
I thought that too first reading that a while back and there's likely a sleight of hand involved, but how could it be kept completely full?
An idea I had but never posted, was a bleeder at the top of the loop, mount the siphon upside down and plug it's entry with a hard neoprene ball. As long as there as there is very little air in the siphon the ball would barely move. That would do well to eliminate any compression, hold the water and isolate the temperature.
Plugging it with a rubber ball, a siphon would likely not even be needed, a straight pipe would suffice.
The above claims a response time of less than 1 mS; this should be sufficient to characterize performance at 300 rpm (5 rev/sec ) or so; it'll be a bit squirrely around valve opening events.
Pretty decent for so cheap. I got the 0-300psi version.
I am using an arduino with a data logger shield(SD card slot and a button cell backed up Real time clock for timestamps). The arduino only has 10bit Analog to Digital convertors, but it already seems sufficient to test further under steam.
This is the arduino data capture with the pressure sensor hooked up to an air compressor.
Very nice results. I'm not sure how fast your Arduino can sample, but I'd make sure to low-pass filter your pressure signal, w/ a cut-off frequency of half your sampling rate (https://en.wikipedia.org/wiki/Nyquist_frequency). Even a simple RC circuit will help reduce aliasing. W/ an inexpensive encoder driving sampling on each count, your indicator should work pretty well.
barts wrote: ↑Wed Jul 14, 2021 4:28 pm
Very nice results. I'm not sure how fast your Arduino can sample, but I'd make sure to low-pass filter your pressure signal, w/ a cut-off frequency of half your sampling rate (https://en.wikipedia.org/wiki/Nyquist_frequency). Even a simple RC circuit will help reduce aliasing. W/ an inexpensive encoder driving sampling on each count, your indicator should work pretty well.
- Bart
The points in that graph are averages of 10 acquisitions.
I'm not really sure yet the exact aquisition rate, and I'm sure it will change. Each analogRead() function takes only 0.0001 s to complete. Right now I'm printing some very long well formated string csv records, that's pretty cpu intensive. If I just printed the timestamp and raw values I'm sure I could get a faster rate.
It may be a full indicator eventually, but for this summer I'm going to get another pressure sensor on the other cylinder, so 2 total. I'm going to try and get a baseline of hopefully like 5 hrs of steaming.
If you can acquire the piston position via an encoder pulse and trigger sampling with that, you may find it easier to analyze the data w/o needing to infer position via valve events - particularly if you change cut-off during the sampling period. Some incremental encoders have a Z or index pulse; this can be used to compare results for a complete revolution easily. In fact, this would let you compute the area under the curve and provide a real time display of indicated HP (assuming appropriate constants were entered). If need be, you could sample every other revolution and do computes in the other.
cyberbadger wrote: ↑Tue Jul 20, 2021 1:57 pm
Interesting, the sensor shows after the engine is shut off a vacuum of 10psi develops and then dissipates.
Given that you're watching the cylinder cool down and the steam inside condense, this makes perfect sense.
