Looking for a Mechanical engineer to join the Ardupilot/PX4 Hardware team

[Philip Rowse]

Hi all
    As we ramp up the Pixhawk 2 effort, there is a need for some mechanical work.  Pixhawk 2 is more than an EE effort, and to maximise its abilities, we really need someone, or a group who have the FEA skills to model the effects of vibration on the system, and to assist in making some choices in the mechanical aspect of the design.

    Basically, I need assistance in verifying the Isolation system on the stand alone version of this board.

    The current setup is optimized for Solo, this may or may not be suitable for the bulk of other use cases in the market.

We also need someone who is good at injection moulded plastics etc.  

Please contact me via this list if you have any of these skills, or you know someone that does.

Thankyou

Philip

[Nikolay Arsov]

Hi Philip,
There is such a guru, but he's living here in Smolyan, Bulgaria. He's a real guru in mold and stamp design and is using SolidWorks and Proengineer.
He's also a guru in injection molding. He's a mechanical engineer with > 35 years experiance. Currently he's designing molds and tools for Phoenix Contact.
His name is Georgy Kasabadjakov. You can freely contact him at georgi.ka...@gmail.com
Best regards
Nick

[Philip Rowse]

Thanks Nick!

is he interested in Open hardware?

[Philip]

I need to emphasise... This is not a paid position, this is developing mechanical parts that are for the community.   The primary need is somebody who has the ability to do FEA of proposed solutions for dealing with vibrations.

Sent from my iPhone

On 7 Nov 2015, at 12:17, Philip Rowse <phi...@3drobotics.com> wrote:

Thanks Nick!

is he interested in Open hardware?

--
You received this message because you are subscribed to a topic in the Google Groups "drones-discuss" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/drones-discuss/tO8_hhSH_Jo/unsubscribe.
To unsubscribe from this group and all its topics, send an email to drones-discus...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[David Kraus]

Hi, Im mechanical engeneer but now,but I dont have access to FEA SW. How complicated is dumping system and, maybe could be possible to compute it with simpler methods by hand.

David 

Dne sobota 7. listopadu 2015 7:43:32 UTC+1 Philip Rowse napsal(a):

[Inverted CNC]

I've been learning vibration simulations using simscale. Willing to take a look at it. Also pretty accomplished when it comes to CAD/CAM. 

Regards

Er...@aspire-suav.com

You received this message because you are subscribed to the Google Groups "drones-discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email to drones-discus...@googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

--
Sent from Gmail Mobile

[Jiro Hattori]

Hi, I am also mechanical engineer, some sort;-)
Do we really need FEA to mount IMU and main flight controller?

[Jonathan Challinger]

I think the difficulty is that the mount needs to consistently achieve the desired frequency response in 6 degrees of freedom, while staying compact and cheap.

On Sat, Nov 7, 2015 at 2:53 PM, Jiro Hattori <jiroh.yk....@gmail.com> wrote:

Hi, I am also mechanical engineer, some sort;-)
Do we really need FEA to mount IMU and main flight controller?

[Jiro Hattori]

Yes, we need some mechanical band pass filter that should be better than good old 3M sponge tape.
I do not know the actual chip on Pixhawk 2, but should have some noise level.

[Jonathan Challinger]

It's so easy in code, you just type "butter2(20,1000)" and you're done! We just need perfectly linear sensors with ridiculous full scale range...

On Sat, Nov 7, 2015 at 3:55 PM, Jiro Hattori <jiroh.yk....@gmail.com> wrote:

Yes, we need some mechanical band pass filter that should be better than good old 3M sponge tape.
I do not know the actual chip on Pixhawk 2, but should have some noise level.

[aitor...@gmail.com]

Hi there,

I'm mechanical designer, skilled in designing parts for plastic injection manufacturing processes, and FEA analysis. Have also demonstrable experience in designing multirotors and many other professional and also DIY gadgets, etc.

If I understand correctly, the goal is to design the box containing the PCB and its damping system. If not, forget the following :)

In my opinion, it doesn't exist the universal solution you are looking for.

Your main goal would be to determine the resonance frequency of the firsts main vibration modes to which the pcb is subjected. The quad's main vibration source are the motors. Leaving aside the starting phase of the motors, the main work regime (r.p.m ->  frequency source) of the motors will be KV and also voltage dependent. In addition, you have to consider the arms size that hold the motors and the material that these are made of (FEA program will need it to calculate the structure's vibration modes frequencies)

Furthermore, in the most conservative case (assuming you have a particular type of engine, working at a certain voltage, hovering in a given amount of throttle and a mechanical structure with a well known stiffness...), for an accurate FEA analysis, and supposing that you want to use some hyperelastic material for pcb damping, you should be able to calculate the Mooney-Rivlin Constants of this material (needed for most FEA programs when it comes to nonlinear analysis of hyperelastic materials) and this will be possible only in the assumption you have a complete material characterization and usually you won't have this.

Summarizing, I think there is not a universal solution for this, and the attempts to find a good solution involving FEA analysis won't throw an accurate solution. There are too many degrees of freedom out of your your control in this problem: KV, Voltage, structure stiffness, damping material...

I suggest you to use a well known damping material, filtering the most sensitive frequencies for the different IC in the pcb and... that is. Vibration control rests in frame designer's side more than in yours.

Best regards,
Andrés

[Jonathan Challinger]

I'd have to agree there to a point. You might design a solution for a 450-size multicopter and have it resonate at the lower frequencies that a larger tradheli vibrates at.

However, we can probably do something that handles the 95% case which means attenuating frequencies higher than something like 50hz.

--

[Philip]

Yes, what Jonathan said...

We are looking for the 95% solution... And assuming that, there are a bunch of ways to achieve this. We already do! 

What we are looking for, is short cutting the process of guessing, and ensuring that designs meet at least the logic of the goal, before moving to the expensive prototyping stage.

The other thing is existing designs.  There are a bunch of isolation solutions for Pixhawk that are out in the community that are really just making things worse... And there are specific issues, that would benefit from having a vibration and resonance specialist on the team. 

Phil 

You received this message because you are subscribed to a topic in the Google Groups "drones-discuss" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/drones-discuss/tO8_hhSH_Jo/unsubscribe.
To unsubscribe from this group and all its topics, send an email to drones-discus...@googlegroups.com.

[john...@gmail.com]

Looking at the sensor characteristics and Nyquist ect, could you not turn the question around and design the sensor cage dampening to prevent frequencies that will cause aliasing? The goal is not to prevent any and all vibrations from the copter, but to make sure the sensors data is usable.

 

[Jonathan Challinger]

The problem is not aliasing but clipping. We need to design the mount to attenuate vibrations such that they never peak higher than 16g.

On Nov 8, 2015 3:42 PM, <john...@gmail.com> wrote:

Looking at the sensor characteristics and Nyquist ect, could you not turn the question around and design the sensor cage dampening to prevent frequencies that will cause aliasing? The goal is not to prevent any and all vibrations from the copter, but to make sure the sensors data is usable.

 

[Philip Rowse]

What Jonathan said....

Regardless as to why... being able to understand the hardware before prototyping is just a basic function of modern day engineering.  FEA tools are so simple to use these days.... but getting access to them is not.

◤PHILIP ROWSE

    LEAD SYSTEMS ENGINEER

    3D Robotics Australia

    +61 (0)427326366

    website | facebook | instagram

[Andy Little]

On the vibration issue

The simplest way to provide vibration damping is to add mass, so clamp the autopilot rigidly to something heavy like the flight battery.

regards

Andy

[aitor...@gmail.com]

Bigger copters use low KV motors and high voltages. Lets say as a first approach from 150 to 350 KV, and 6s to 12s batteries. This results in ~1700 (lets hover at 50% throttle) to 15540 rpm: 28 -> 259 Hz . On the other hand, smaller quadcopters use higher KV motors and lower voltages. Lets say as a first approach from 1000 to 1200 KV, and 3s to 4s batteries. This results in frequencies from ~100 to 296 Hz.
Assuming that frame stiffness is high enough (which usually meets when it comes to quadcopter's assemblage, there is not loose parts, etc), its resonance frequency would be much more higher than this. In this case, frame's resonance frequency won't see motor's vibration: We can ignore the frame and we can focus in isolating the pcb from frequencies from ~20 to 300 Hz (very nice bass frequencies :))

Important: If you place baro in pcb's bottom side, you should isolate this from damping material: placing baro in direct contact with damping material would be the same that placing this direct over a subwoofer! :)

Regards,
Andrés

[john...@gmail.com]

One thing to remember is that since the motors are constantly changing rpm, you get lots of different vibration frequencies and amplitudes as the arm vibrations meet at the center of the frame and mix together. So It's not a simple as just looking at the motor kv.

[aitor...@gmail.com]

Jonathan: Consider a Spherical Cow... ;)

[Robert Lefebvre]

I had to look that up.  Thanks!  That's great.

On 9 November 2015 at 07:04, <aitor...@gmail.com> wrote:

Jonathan: Consider a Spherical Cow... ;)

--

[Robert Lefebvre]

It's a really bad assumption that the frames are stiff enough that they don't need to be considered.  

[aitor...@gmail.com]

El lunes, 9 de noviembre de 2015, 14:29:47 (UTC+1), robert.lefebvre escribió:

It's a really bad assumption that the frames are stiff enough that they don't need to be considered.  

It is true Robert. Fortunately I didn't say that. I said, IF frame's stiffness is high enough... I am not assuming it's true.
As I said in my first post: frame's construction (in all its infinite forms and possibilities) is out of your control, so I'm just trying to set an approaching to the problem in the most ideal conditions. If helps.

Regards

[Tom Pittenger]

With a multi-rotor, don't you get higher frequencies due to inter-motor interaction? At the very least you would have highest-RPM times motor count.

--

[Ryan Linderman]

The FEA tools are relatively simple to use but they are quite expensive licenses (~$40k) they require a hefty desktop computer to run with much detail and if you don't know the material properties like the frequency dependent soft squishy plastics and foam tapes, fiber composite frames, etc. then garbage in = garbage out.

If you want to take the mechanical engineering of this system to the next level then there would also need to be a fair bit of testing to vet out the new solution to the old so we could improve the models for the next design round. We would also need to define a 95% standard air frame configuration and vibration profile to use with future models. This implies we know the vibration generated by a "standard" motor and frame configuration quite well. If every time we start a new project we change the input parameters it will be very hard to know if the results are improving.

Probably the easiest way to start would be to impose a vibration profile to the pixhawk board enclosure and ignore the rest of the system -- this is where a standard vibration profile that is somewhat representative is important. We could then try different suspension systems between board and enclosure to gauge how effectively we are damping vibration. It's up to the end user and their combination of airframe and motor to control what vibration gets to the pixhawk enclosure. Future designs modeled with the same standard vibration profile on the enclosure would allow an indication of how much things are improving.

I can imagine there is a simple spring feature that is molded into the enclosure that allows the board to float on the spring mechanically -- which wouldn't allow some frequencies to propagate to the board. When cables are plugged in, the springs deflect to a hard stop to allow enough force to insert the cable. One would have to be careful that cable tension is not so great that is pulls the board to the hard stop after assembly.

I have access to an under utilized Ansys Mechancial license as well as Solidworks if you want to try some simple runs. The 3D CAD would need to be cleaned up to remove lots of fine details like fillets and every single board component and tiny feature (again we need a kind of standardized simple board with basic mass and stiffness).

-Ryan

[Robert Lefebvre]

Yes, exactly.  You have vibration at every motor frequency, then each of those times two for two blade propeller passage frequency and flapping frequency.  Those RPM's are constantly changing.  Then you have the harmonics of all of that.  Natural resonant frequency of the airframe.  Then the beat frequency of all of those coming together at the central hub.

It's really a wonder these things fly at all.

But that is why I disagree with the concept that all would be fine if we have perfect sensors with high sampling frequency, and just digitally filter everything out.

For all intents and purposes, the boards are subject to "white noise".  And when the signal/noise ratio gets too low, all you're doing is trying to filter mud.  

[Inverted CNC]

Out of curiousity has anyone looked inside gyro/fbl units from nitro helis? They are internally dampened and never seem to have issues in a much more harsh vib environment. I have a couple old spartan and vbar sensors packed away. Ill dig them out and post some internal pictures.

And FYI... There are FEA alternatives on the cloud now that solve the problem of needing heavy cpu horsepower onsite and they are subscription based.

--
Sent from Gmail Mobile

[Robert Lefebvre]

Yes, I have a Fubar which is a Vbar clone, and a Skookum 720 which I have opened.  There is nothing inside them.  No isolation anyway.  The board in the Fubar is simply wedged between stanchions in the plastic case.  The 720 is similar, except the case is aluminum, and the processor on the board is semi-bonded to the case with heat sink compound.

However, these systems don't have accelerometers, and don't attempt to self-level, and that is key to understand.  When mounted on a firm surface, there is not much issue with rotational vibration.  For the most part, these systems use "open loop" angular heading control, meaning, they can tolerate some amount of angular precision loss (whether due to heavy filtering, or noise), because the pilot will correct it.

It is the accelerometers and attempts to keep track of "which way is up" which cause problems for us.  There is quite a lot of linear vibration, which makes the signal to noise ratio lower.  Also, we rely on the accelerometers more, since the whole point of the autopilot is that the pilot shouldn't be required.

There are of course some heli FBL systems that have self-leveling, but it's well known that these are problematic to use with nitro or gas engines.  Most seem to have some provision (bright LED) to indicate to the pilot that the vibrations are too high and it has lost reference heading. They also use powerful and easy to use FFT to provide vibration histograms to help the user solve vibration problems.

I've got a Pixhawk working well on a gasser (with an unbalanced engine no less).  I've actually not had many problems with it due to the physical isolation system I set up.

[Inverted CNC]

Looked but couldn't find the vbar without digging real deep. But I did come across a Naza that had a bit of a wet landing ;-).

It uses very soft foam similar to 3M tape to isolate the sensor. It is also slightly compressed when the lid goes on and it looks like they added some mass in the form of

[Inverted CNC]

[Robert Lefebvre]

Yes, I wanted to mention the DJI units.  Their other systems such as the A2 and Wookong, to even further, using a cube of metal, with the circuit boards arranged on each of the 6 faces.  Then that is mounted between 8 points supports, one at each corner.  So again, they use mass to get the natural frequency into a workable range.  I also think that the mass on the IMU is helpful to reduce the effect of vibration energy transmitted along the required ribbon cable. (ie: more vibes transmitted through a given ribbon cable onto a 1g board, compared to a 10g board). 

[john...@gmail.com]

I did a teardown of the ZeroUAV X4 in later 2012. Same principle. Sensors placed in a steel cage (mass and rigidity) and then isolated from the rest.

http://diydrones.com/profiles/blogs/zerouav-ys-x4-teardown

To unsubscribe from this group and stop receiving emails from it, send an email to drones-discuss+unsubscribe@googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to the Google Groups "drones-discuss" group.

To unsubscribe from this group and stop receiving emails from it, send an email to drones-discuss+unsubscribe@googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to the Google Groups "drones-discuss" group.

To unsubscribe from this group and stop receiving emails from it, send an email to drones-discuss+unsubscribe@googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

--
Sent from Gmail Mobile

--
You received this message because you are subscribed to the Google Groups "drones-discuss" group.

To unsubscribe from this group and stop receiving emails from it, send an email to drones-discuss+unsubscribe@googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to the Google Groups "drones-discuss" group.

To unsubscribe from this group and stop receiving emails from it, send an email to drones-discuss+unsubscribe@googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

--
Sent from Gmail Mobile

[Jiro Hattori]

I think that is logical for vibration dumpning.
Firstly, we need lager mass in CG.
Secondary, we need springs and dumpers for any directional vibrations.
Most of the failure on DIY flight controller mounting system may be thinking of one directional vibration input.

[Inverted CNC]

I just sat in on a FEA webinar by Autodesk. Their setup for FEA using Fusion360 is pretty slick. 

So if there are CAD files availble as a starting point so I dont have to start from scratch I will donate my time to setup sims.  I just dont have time to do the CAD from scratch right now.

Eric

--

You received this message because you are subscribed to the Google Groups "drones-discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email to drones-discus...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Paul Riseborough]

Jon,

Clipping is bigger problem than aliasing for most Copter setups, but for systems with high speed motors or gears, aliasing still remains an issue and i have seen logs which demonstrate this effect clearly.

The job of the isolation system is to reduce the transmission of frequencies to the sensor that alias back  to produce an offset. This occurs when there is vibration energy around any multiple of the IMU sampling frequency. For the majority multi-rotors and planes with direct drive motors, most of the energy is at the motor and blade passage frequency, so there is relatively little vibration energy at the 800 and 1000 Hz sampling frequencies for our sensors. However a geared motor is capable of producing a lot of energy at these frequencies as is a high speed propeller or ducted fan.

That means that the isolation mount design has to address both clipping and aliasing which means:

- Its natural frequencies cannot coincide with a motor or prop blade passage frequencies

- Its transmission of vibration at the sampling frequency needs to be low enough so that aliasing doesn't occur for geared setups. This is what drives the design to lower frequencies.

- there needs to be enough damping in the mount to prevent clipping due to mount resonance, but not so much that the isolation of vibration at higher frequencies is compromised. 

- It natural frequencies cannot be so low so that they couple with the autopilot.

Personally I don't think all of the above are achievable until we increase the sampling rate to 4kHz.

-Paul

[Paul Riseborough]

Andy,

I know this is semantics, but the primary job of the mounting system is not damping vibration, it is isolating the sensors from the vibration. I agree with you on the benefits of mass loading the autopilot to improve isolation, however if the rotational frequencies of the autopilot and battery get below 30Hz, they will cause problems with autopilot tuning.

-Paul

[Robert Lefebvre]

Don't forget that an important reason to employ mass loading, is to reduce the effect of vibration energy bypassing the isolation mount by way of the wiring connected to the autopilot.  This is why I have harped on the wiring being supplied with these autopilots.  The wires need to be as low gauge as possible, multi-strand, with soft insulation such as silicone.  And we need to reduce the amount of wiring as much as possible.  For example, on helicopters I always use a servo breakout board, which means the Pixhawk only has 4-6 PWM signal wires coming off, and does not have 8-12 heavy power and ground wires in addition.  My next large heli will use Futaba SBUS servos to help even more. 

Directionally, this is why it makes some sense to have the IMU unit on it's own isolation system, with only signal wires to the processor. I realize this is what the PH2 is doing.  I'm just pointing it out for completeness of the discussion.

Another point about the isolation system is that the IMU CG needs to be in the same location as the centroid of the isolation mounts.  This is another place many many people are getting it wrong.  This is needed to avoid linear-rotational vibration cross-coupling.  I see too many autopilots perched high on top of vibration isolators which are also narrow-set.  This is awful. 

My gas heli is successful because I took all of these things to extremes. The Pixhawk is mounted inside a weighty aluminum structure with wide-set dampers, using as few wires bridging the gap as possible.  All wires have a strain-relief loop, and I even bird-caged the ribbon cables to reduce their stiffness. 

--

[Tom Pittenger]

I hear Mark Colwell has some vibration testing experience.

[a a]

I studied numerical methods 25 years ago.
For the last 10 years I study vibrations generated by wind turbines, wind farms, infrasounds, brain waves, FFT,
vizualize vibrations in 2D/3D, study seismography and infrasounds generated by vibrating wind towers, transmitted through the ground kms away.
I go to Vienna Conference on Seismography with my papers.
Develop 100 seismographic sensors array, build sonars, georadars.
Built small wind tunnels to test drones indoor.

Since vibrations can be visualized I am open for any cooperation.

I need budget to buy hardware to build stand for tests, install sensors, board and more.

Just let me know what hardware should be tested and what is a source of vibrations in the tests, full frequency spectrum (rpm) and number of vibrators.

darius

mant...@gmail.com

[Ryan Linderman]

I am a mechanical engineer with lots of experience developing hardware/electronics for outdoor applications.
However I'm pretty new to drone engineering and am interested in studying the vibration modes in my Iris+ airframe with Pixhawk controller.
What is the best way to access the IMU sensor data? I'm assuming we will have to run it through an FFT filter to identify intensity of different modes.

Thanks,

Ryan

On Thursday, November 5, 2015 at 7:36:49 PM UTC-8, Philip Rowse wrote:

Hi all
    As we ramp up the Pixhawk 2 effort, there is a need for some mechanical work.  Pixhawk 2 is more than an EE effort, and to maximise its abilities, we really need someone, or a group who have the FEA skills to model the effects of vibration on the system, and to assist in making some choices in the mechanical aspect of the design.

    Basically, I need assistance in verifying the Isolation system on the stand alone version of this board.

    The current setup is optimized for Solo, this may or may not be suitable for the bulk of other use cases in the market.

We also need someone who is good at injection moulded plastics etc.  

Please contact me via this list if you have any of these skills, or you know someone that does.

Thankyou

Philip

[Randy Mackay]

Ryan,

 

     In the log bitmask it’s possible to set the IMU logging to RAW (i.e. check the IMU_RAW checkbox).

            http://copter.ardupilot.com/wiki/common-downloading-and-analyzing-data-logs-in-mission-planner/#setting_what_data_you_want_recorded

 

     This should output the raw data from the IMUs at a ridiculously high rate (1kHz for mpu6k, 1.6kHz for lsm303d) and the mission planner has a built in FFT viewer (I forget where it’s hidden).  Make sure you have a high quality SD card!

 

-Randy

[philip]

Tridge can weigh in here… but the quality of the card is quite specific, some of the expensive cards actually don’t work as well for high speed logging I believe….