During a hot vibration discussion, one respondent posted about overlapping props. If I remember right, for one particular ship, he felt that the optimal overlap was about 30% for an octa H.
Previously I had only tested up to a 5% overlap, assuming that more would be bad. So much for assumptions. The data is shown below. The data is the result of taking:
- over 400 data points
- using a T test stand where four motors/props were mounted on top spaced vertically at 0, +16mm, 0, and +16mm
- horizontal overlap amount was variable for each series of tests from 0% to 55% relative to prop radius.
- props tested were 11x4.7 fiber reinforced plastic slow fly.
- prop rate was 0% (mid throttle) generating about 960 grams of lift.
- motors were 750 KV
- the vibration sensor was located at the top of the neck of the T oriented to measure x/y
- the dB meter was located 1 m distance from motors
Test equipment was:
- APM 2.5 controller
- gram scale
- T test stand
- Mastech HY1530EX DC Power Supply
- Radio Shack Sound Level Meter
- Extech SDL800 Vibration Meter
Test Setup:
showing the two left motors and scales in % overlap
showing the four motors on top of the T test stand
The results were:
Note the following for this particular setup:
- Interestingly, thrust (the blue line) might increase for low overlaps. Need more samples to confirm or reject this possibility.
- Frame vibration (the red line) might not increase much at all for low overlaps, but does increase dramatically after a point (about 12% ish overlap)
- dB (noise level) increases dramatically right away.
- Not shown on the data was that I had planned to test at even greater overlaps. But, I started to hear what might have been occasional sonic pops by turbulence breaking the sound barrier (or some other phenomenon). This caused me to also test for vibration and dB.
Limitations of Test: I'm going to try alternative custom frames to the V using low KV motors. When the new test props arrive, I'll test the lot and select the optimal one for the most efficient low KV motor. Then I'll retest overlap using that setup with multiple vertical options also. Until we see at least one more setup, caution is advised on accepting the results here as applying to more than this particular setup. Only shown are relative changes, not the actual numbers. The actual numbers (grams thrust, dB, and frame g forces) are impacted by this setup (1" bar, mass of the T-frame, etc.). In layman terms, the slope directions can be generalized within this context, but not their magnitudes.
Of Interest to Vibration: With overlapping props, optimal is based on the criteria of the pilot. So in that mix, consider that sound pressure on an exposed APM will go up 30% ish at the frequency of rotation, about 200 Hz, when props are greatly overlapped.
Sound pressure is more likely to impact APMs that are in free space (mounted on moon gel or double sided foam). On the other hand, the forces from frame vibrations also go up with overlap. If frame forces are too high, then some decoupling of the APM from the motors/frame might be required if free access to more mass isn't available. This might explain the alt-hold fly-away experienced by the pilot that posted a reference to that event. This can also explain why with stiff frames, the least vibration occurs when the APM is hard mounted to the frame with the addition of moon gel or double sided tape worsening the condition.
For an APM adequatley shielded by stiffness and mass, to unnecessarily expose that APM to 60 to 90 dB sound waves by putting the APM on pliable material could be as harmful as not supporting the APM with pliable material when the frame has high vibrations that are not mitigated by stiffness and mass. It is up to the pilot (or commercial enterprise that made the copter) to make that trade for the safety of the ship, pilot, bystanders, and property.
Conclusion: Some gains in reduced frame size and efficiency might be achievable by a stiff frame with small prop overlap when the APM is hard mounted to frame/mass (so as not to be affected by the increased forces from air pressure). But more sampling and testing will be required before this hypothesis can be accepted with reasonable certainty.
More testing will follow to answer some of those questions and others.
Replies
Just saw this comment from way back, and had to respond.
Yes, overlapping props create noise and vibration. The question is, does it really matter? The answer is no, not with my frame design. It effectively isolates the vibration, and the sound pressure was never shown to be of any consequence outside of a number showing on a Radio Shack SPLM in a laboratory. It's loud, but it doesn't really matter in any way. The machine takes excellent video.
good insight Rob. the only reason to test noise is for those interested in photographing birds and other animals (my favorite subjects). but like you've experienced, landscapes, homes, farm flora, and power lines don't care.
Added the spacing side. Added more data points to areas in need. Interesting results.
Conclusion:
As I stated before, optimal spacing of props has many parameters that vary from pilot to pilot. But if vibrations transmitted to the APM via the frame or via the air is a concern then:
o For air transmitted vibration, the farther apart the props the better.
o For frame transmitted vibration, there may be a sweat spot that significantly decreases frame vibration.
o The noise decrease can only be only partially accounted for by the change in distance to the APM, so something in addition to the increase in distance is happening.
How To Turn Your Copter Into A Test Bed
There are probably better and easier ways to do this, but in order to get good throttle control out of the APM, it can't be done by simply placing the copter level. One needs to actually turn off all pitch, roll, and yaw. The way I did this (there are many ways to do this like using the passthrough function but this allowed me to arm/disarm and turn on and off and immediately power up), is to:
}else if( _frame_orientation == AP_MOTORS_V_FRAME ) {
// V frame set-up *****original numbers*****
// add_motor_raw(AP_MOTORS_MOT_1, 1.0, 0.34, AP_MOTORS_MATRIX_YAW_FACTOR_CW,7);
// add_motor_raw(AP_MOTORS_MOT_2,-1.0,-0.32, AP_MOTORS_MATRIX_YAW_FACTOR_CW,3);
// add_motor_raw(AP_MOTORS_MOT_3, 1.0,-0.32, AP_MOTORS_MATRIX_YAW_FACTOR_CCW,6);
// add_motor_raw(AP_MOTORS_MOT_4,-0.5,-1.0, AP_MOTORS_MATRIX_YAW_FACTOR_CCW,4);
// add_motor_raw(AP_MOTORS_MOT_5, 1.0, 1.0, AP_MOTORS_MATRIX_YAW_FACTOR_CCW,8);
// add_motor_raw(AP_MOTORS_MOT_6,-1.0, 0.34, AP_MOTORS_MATRIX_YAW_FACTOR_CCW,2);
// add_motor_raw(AP_MOTORS_MOT_7,-1.0, 1.0, AP_MOTORS_MATRIX_YAW_FACTOR_CW,1);
// add_motor_raw(AP_MOTORS_MOT_8, 0.5,-1.0, AP_MOTORS_MATRIX_YAW_FACTOR_CW,5);
// V frame set-up *****Frantz V 7/15/2013*****
// add_motor_raw(AP_MOTORS_MOT_5, 0.945, 1.000, 0.801, 5);
// add_motor_raw(AP_MOTORS_MOT_7, -0.945, 1.000, -0.801, 7);
// add_motor_raw(AP_MOTORS_MOT_1, 0.757, 0.333, -0.801, 1);
// add_motor_raw(AP_MOTORS_MOT_6, -0.757, 0.333, 0.801, 6);
// add_motor_raw(AP_MOTORS_MOT_3, 0.570,-0.333, -0.629, 3);
// add_motor_raw(AP_MOTORS_MOT_2, -0.570,-0.333, 0.629, 2);
// add_motor_raw(AP_MOTORS_MOT_8, 0.383,-1.000, 0.629, 8);
// add_motor_raw(AP_MOTORS_MOT_4, -0.383,-1.000, -0.629, 4);
// Test Bench *****Frantz 8/6/2013*****
add_motor_raw(AP_MOTORS_MOT_1, 0, 0, 0, 1);
add_motor_raw(AP_MOTORS_MOT_2, 0, 0, 0, 2);
add_motor_raw(AP_MOTORS_MOT_3, 0, 0, 0, 3);
add_motor_raw(AP_MOTORS_MOT_4, 0, 0, 0, 4);
add_motor_raw(AP_MOTORS_MOT_5, 0, 0, 0, 5);
add_motor_raw(AP_MOTORS_MOT_6, 0, 0, 0, 6);
add_motor_raw(AP_MOTORS_MOT_7, 0, 0, 0, 7);
add_motor_raw(AP_MOTORS_MOT_8, 0, 0, 0, 8);
... the last eight lines inserted just before the }else {
So no matter how hard the floor vibrates, the throttle is in control. I also use the Monitor menu on my radio to nail the same throttle level on each test.
Prop Overlap Test 2: Data for props CW-CW props going in opposite directions as the pass each other have behavior that is different than props CW-CCW going in the same direction when they pass each other.
Unlike props moving the same direction at the overlap, props that move in opposite direction at the overlap (CW-CW or CCW-CCW) the noise and vibration levels are proportional to the decrease in lift (the exception being right at the start where lift isn't impacted but noise and possibly vibration are).
Next test will explore 'negative overlap' or 'positive spacing' and the impact that has on lift, vibration, and noise. I'll conduct this in a manner that, while not usable for my ships, will be useful to wider spaced quads.
Franz, your bias is showing. Speaking of shattering beliefs...
I have never had a fly-away on one of my overlapping prop machines. That occurred on an F-330 frame which is a conventional non-overlapping quad. Coincidentally, on that machine, the PX4 was rigidly mounted to the same plate as the battery. I had had the same theorem that you did that simply battery mass on the same plate as the APM would be enough to help damp vibrations. I was wrong.
The data you present is very interesting. First, because it confirms again, that there is virtually no loss of thrust efficiency with the overlapping props. It also seems to confirm that this setup is noisier, and causes more vibration than non-overlapping. Something I suspected (based purely on analogue-biological test methods...)
But that's about where your hypothesis falls apart.
The ultra-low vibration logs and video I have posted are from this overlapping H-8 frame. That was accomplished with a decoupled APM. Apparently the vibrations dampers do a good job of isolating the APM (and camera) from the frame vibrations, and sound pressure is not causing a problem.
As far as I remember the old reports on coaxial systems, you should play with the vertical distance as well (increase it).
The other referenced pilot is Rob Lefebvre and you can read all about his design using overlapping props here. He also experienced a fly-away by hard mounting the APM. He's just not a pilot he has a name, experienced heli pilot and developer of the Trad Heli code. He brings not only RC experience but mechanical understanding of materials and their properties. He's an awesome person to have on the team. (oh, and passionate!)
Can you not be so marred to scientific process. Please undertake the full experiment first and show results that will benefit diydroners. Complete your hypotheses and present your findings . Not half baked thought processes laced with underhand comments.
400 data points is objective only in that it is a number, but subjective if they are not relevant. And i can't see the increase in dB right away either.
In all my other responses to you that ask technical questions about Nyqvist sample rates. Measurements using indicative data sampling as a tool for debugging problem frames, and your own use of a high quality sensor, you don't reply.
And the reason why many experienced people (and they really are experts in their field) are getting frustrated, is that you are ignoring engineering knowledge that has been learnt from education and said experience.
There is a dedicated group of people who are obsessed with making this the best platform possible. Who are followed by an enormous talent of customers and innovators who can see it's application in the real world. The developers want this to be as safe and as perfect as possible. You have valuable input, but you seem to be going too far. You are undertaking interesting experiments, but you are not listening or even responding to others valid concerns.
You are smart guy, but your being as stubborn like bull, as many engineers are. This breaks down in the common collective of Open Source, as there really is not a hierarchy, just consensus.
So tone it down, present your findings, and if they are really are all they are presented to be, it will be accepted, adopted and even probably refined.
And the final point I have to say, this forum has been one of the best forums I have been involved in. There are so many people with great experience that you can learn so much. People here are so patient and just answer questions about things I know nothing, and I learn so much. I try and put back as much as I get. It's a good and interesting place to be.
At best your data suggests (assuming it is statistically valid) that we can obtain a 1% thrust increase by paying for it with a 1% vibration increase. More vibration is bad. Of course, given the number of sample points and no measure of variance/deviation about each sample point, it is impossible to conclude if these are even statistically significant results. More testing is clearly needed.
If you do further testing, also please test the hypothesis of separating the blades. I.e., having a negative overlap as defined above. I'd like to see if we can gain more power and lower vibration in that case, or if there is a symmetry (I don't expect there to be).