SimFeedback-AC DIY Motion Simulator 10Khz Research

[scornflake]

Hey all,

I'm really keen to build one of the SFX-100 rigs, but I've heard (been in the presence of) my brothers and the 10khz servo/driver noise really does my head in, so much so that I would not be able to put up with it. 5-10m of it, and I'm done. This is a darn shame

So, I'm going to see if I can get rid of it. Either electronically (preferred) or with accoustic treatment(s).

In this thread I'm simply going to document my journey. At this point I am unsure if it'll come to anything. Maybe I won't succeed. Maybe I'll get half way. Who knows.

The story begins with Neil buying himself a single driver and servo, a 3d printer, and saying to himself "hmm. what next"

My full notes are here, but I can't promise you'll understand them: https://docs.google.com/document/d/1J4AUnKNYtFRujTR9BFBB_lJudZRP_02BOt6lfAIN45I/edit?usp=sharing

Look below, I'm going to try to post a more understandable version of my journey!

 

[scornflake]

My initial ponderings:

1. If I can remove the noise from source, this would be far preferred. If I have to spend money to do so, it's still far far preffered. My thinking is that a dbox costs upwards of $30-40k NZD. If I have to spend $1-2k to kill the noise; wouldn't that be a *good* outcome?
2. Look at all options. While removing the noise from the source is ideal, it might not be possible. What other options exist? What can be done with accoustic treatments?

First steps are to measure the noise and see what we're dealing with. More on that below.

 

[scornflake]

Test environment:

1. Single servo motor, connected to a driver, as far away as possible. For me this is just on the other side of a door that I can *almost* close. I want to see if the driver or motor or both are making the noise.
2. Using the motor and the driver alone. No USB connection. Tests are done with the motor in Jog_01 mode, value about 30. Custom parameters for SFX-100 are set on the driver as well (e.g: rotation 300, etc https://opensfx.com/testing-and-configuring-servo-drives/).
3. AKG-200 condensor mic, connected to a Lexicon Omega, via XLR cable. Using iSpectrum software to measure the noise at about 150m away from the motor.
4. Keeping motor within "chequebook" distance had to use somehting to ensure the motor was at a consistent distance to the mic, a real old chequebook was the ticket. There you go, you can go lookup what one of these ancient paper devices actually is if you don't know

The room. The video below gives you an ambient noise profile of the room.

Single servo, about 150mm away from the mic, mounted to a motor mount, sitting on a carpet floor.

From this we can see that the rough difference is about 30db between ambient and average volume of the 10khz tone that some people hear. For those that *can* hear it, it is darn right painful.

There are various posts saying that 10db is "somewhat equivalent" to a doubling of audible volume (it's WAAAAY more power; but we don't perceive power linearly as volume it seems).

So, at first glace this seems to be 3x louder than the background noise (At 10khz) or about 5x more power). Either way. It makes my ears bleed!!!

(ok, not really, but I hear the same ringing in my ears for 30-50m after listening to one motor for 20-30m while doing testing). It's baaaaad.

 

[scornflake]

In my reading I've seen that polyester fibre is excellent at blocking high frequency sound. I will be aiming for that. In the meantime I've done some tests with a sleeping bag mat (med density 10mm foam) and some flexible floor tiles (med density 600x600m tiles).

Results (including the layering if microfibre cloth inbetween layers) show that the sleeping bag foam wins. Here's what I have so far (using above videos as reference).

- Single layer foam gives ~ 5-8db reduction depending on "the day of the week". I did a number of tests and got slightly different results. Nothing above a 10db difference though. Some tests included using microfibe cloth either in front of or behind the foam. The MF cloth made little difference. It made some, but not significant.

- Foam, mf cloth, foam, mf and then a 0.8mm PLA shell

I layered some materials together, as above. It was better, but I would not consider it a "solution", given that we still expect to have 4 (four) units in the room, and that the effect of bolting them to the P1 chassis is still unknown (I'm expecting more noise; not less). Even so, it's encouraging:

 

[scornflake]

A final test from today was with polyester fibre.

Worth mentioning that in ALL tests so far, the wrap (foam, mf cloth, nothing, shell alone) has been around the motor itself. The shaft of the motor is pointed towards carpet, and the head/top of the motor has been exposed.

In the next test, I got our "MotoGP Blanky" (100% Polyester Fibre sir!) and carefully folded it into a "motor height sized thing". I then wrapped that around the motor TWICE.

Results, I think, speak for themselves:

It appears the poly-fibre will be a great solution for building a motor enclosure. I'm going to try to source some suitable material, and will also model what I do.

Oh; btw: my existing models are all publicly available anyway; via OnShape:

Onshape

Sign in to Onshape, the #1 fastest growing CAD system in the world with over 4 million users.

cad.onshape.com

I warn you now: I made asymetric shells thinking about reflections. I'm not yet convinced that a PLA shell will cause significantly noticable reflactions to matter, so I'll very likely remake these to be symetric, look way better; and be faster to print. I'm printing using PLA with a 0.8mm nozzle and 0.64 layer height.

 

[scornflake]

I'd be very happy to receive informed advice relating to countering this 10khz noise (whether that be accoustic or electrical). Of course I'm doing what I can, but I'm not an expert in accoustics nor generation of noise from electrical circuits.

 

[Mascot]

As a fellow sufferer who hardly uses his SFX due to this issue, I wish you Godspeed and all the luck in the world with finding a solution.
In the main thread (I think - from memory..!) I'm sure it was stated that a 110v supply (with the driver settings adjusted accordingly) vastly reduced the coil whine. This seemed to be the conclusion from our American cousins, anyway.

There was detailed chat about this issue and various potential solutions in the SFX discord, but the discussion was unfortunately shut down, ostensibly for 'safety reasons'. My own attempts to contain the noise in an insulated cooler box were a dismal failure. There's still debate about whether the 10Khz whine is coming from the drivers or from the motors themselves. It might be one, the other or both.

 

[saxxon66]

Like Mascot said, we, the SFX community, did some detailed investigations regarding the 10k noise.
We do know why and where it comes from.
It is the current loop freq. of the driver. It is a very common freq. for industrial servo drivers, because it is the sweet spot between performance and parts costs.
If you search the web you will find many papers regarding this.
http://www.ti.com/lit/wp/swpy031/swpy031.pdf

The sound can travel (wires, housing, ...) so we do not know exactly where it is emitted.

We stopped the research at a point where it was no longer safe because you need to disassemble the driver and do some desoldering and reattaching of parts. On a board with big 400v capacitors this is not a good idea for a common DIY person and as a general DIY solution. I do not know what "ostensibly" means but it sound like he think it was not for a good reason to stop it.
Anyway, we did not find a solution. The "real" solution would be to find a driver that has a higher control loop freq., but that means also higher costs (like 4-5x).

I did some test with a cardboard box going over the driver, that worked quite well, but I do not think that a cardboard can be a real solution.

We have updated our website (https://opensfx.com/about-the-sfx-platform/) to include a note about AC Servo Drives "The AC Servo Drives emit a high frequency whine, the majority of users are not troubled by it and is not noticeable under motion. There is a small %age of users (approx 1-2% without a formal survey) who find the whine very uncomfortable under motion.

 

[Mascot]

We stopped the research at a point where it was no longer safe because you need to disassemble the driver and do some desoldering and reattaching of parts. On a board with big 400v capacitors this is not a good idea for a common DIY person and as a general DIY solution. I do not know what "ostensibly" means but it sound like he think it was not for a good reason to stop it.

I wasn't the only one who thought that shutting down ALL discussion about the problem was heavy handed when it could have been limited to discouraging anything that involved opening the drivers, especially as some good ideas were being proposed. It seemed at the time that discussing anything that could be viewed as a 'negative' (the coil whine, in this case) was taboo. Anyway, ancient history, the rule has been set on your discord and will no doubt be adhered to and enforced.

Also, I really don't know why you quote this 'approx 1-2% of users without a formal survey' when there is no basis whatsoever to quote mythical numbers in the absence of hard data. 'A small percentage' suffices and keeps things appropriately vague. I suspect the number is actually much higher based on PMs I've received on the subject, but that's just my opinion based on my own experience. It's not a criticism of the SFX system, it's just an unavoidable side effect of the cost-effective components to keep the project affordable for many.

I would strongly suggest that at the very least people listen to a 10khz signal on YouTube before committing to the SFX, just to get an idea of what to expect during operation. Some people cannot hear it at all, and other people can. Of those who can hear it, some are able to live with it, but others sadly are not.

I'm still hoping for a solution that'll enable me to enjoy the SFX as intended, so thanks again to @scornflake for starting this thread and for investing his time towards that end.

 

[scornflake]

There's still debate about whether the 10Khz whine is coming from the drivers or from the motors themselves. It might be one, the other or both.

I can verify that noise most definitely comes from the motor. 100% guaranteed. It probably *also* comes from the driver & the frame when the motor is fixed to it. Again, yet to get there.

I'm starting simply with one piece at a time, to hopefully get a more accurate picture of where the noise is emitted. I 100% agree witih saxxon66 that it would be better to kill it at the source, and also agree that carries more risk (at least with existing gear that we have).

Here's some pics showing how I'm testing at the moment; sort of forgot to put those up before!

 

[scornflake]

Anyway, we did not find a solution. The "real" solution would be to find a driver that has a higher control loop freq., but that means also higher costs (like 4-5x).

It's worth mentioning that for some of us this added price may be something we're willing to pay. If you know of a driver that you thought might work, I'd even considering buying one for you if that would be useful for testing. While it adds to the cost, my perspective is that I could not build or use the system at all in its current state given my sensitivity.

Sure, with better (more expensive) drivers it'll cost more. But if it works, then I'm still not paying *anything* like dbox prices, so for me personally I'd probably just buy the drivers that didn't make that noise. Then maybe I wouldn't have to worry about sound absoption solutions at all.

 

[DotComRich]

Very interested in this as well as I am one that isn't bothered by the sound to the point of not being able to use it, but when I take my hmd off and waiting between races browsing or whatever, the whine always starts bugging me. Obviously once im racing its a non factor but still I want to find a solution or easy way of reducing it. If a simple cover drastically reduces it then sign me up.

 

[Tronicgr_6DOF]

On a quick acoustic measurement with audio spectrometer (spectroid), the motor with replaced power cable with Shielded cable, has 5-10 dB less noise than the unmodified cable... Also in commercial building electrical installation I tried the servos, I don't hear any whining at all as the mains power is very clean and has real ground.

 

[scornflake]

On a quick acoustic measurement with audio spectrometer (spectroid), the motor with replaced power cable with Shielded cable, has 5-10 dB less noise than the unmodified cable... Also in commercial building electrical installation I tried the servos, I don't hear any whining at all as the mains power is very clean and has real ground.

Thanks Good idea!

So you have the shield on the motor drive cable connected to the same grounding point on the servo drive, right?

 

[Tronicgr_6DOF]

Thanks Good idea!

So you have the shield on the motor drive cable connected to the same grounding point on the servo drive, right?

Yes, for the shield to work and block EMI (electrical noise) as well the switching 10khz noise, it needs to be connected only on one side of the shielded cable. To keep the ground from looping, its recommended to wire the shield ground on the servo drive side, that is closer to the mains power ground.

 

[scornflake]

Yes, for the shield to work and block EMI (electrical noise) as well the switching 10khz noise, it needs to be connected only on one side of the shielded cable. To keep the ground from looping, its recommended to wire the shield ground on the servo drive side, that is closer to the mains power ground.

Got it. I'll get some cable, and repeat the tests.

 

[Insert Coin]

My 2c: I have moved the servo drivers to the other side of the wall. The drivers are in an old server cabinet (closed, but with air gaps and fans). I drilled a hole in the wall for the cables. Servo cables were already shielded with anti-EMI tape. The 10 KHz whine went down by approximately 50% according to my son, who is very sensitive to this. I did this drastic solution for him; I can't hear the whine at all - I guess I'm too old . My son could hear the 10 KHz coming from the drivers especially, but since the volume only dropped 50% I guess it either still whines through the wall or the servos themselves are whining too (couldn't measure that with Spectroid though: the 10 KHz peak is the same everywhere in the room).

 

[Mascot]

My 2c: I have moved the servo drivers to the other side of the wall. The drivers are in an old server cabinet (closed, but with air gaps and fans). I drilled a hole in the wall for the cables. Servo cables were already shielded with anti-EMI tape. The 10 KHz whine went down by approximately 50% according to my son, who is very sensitive to this. I did this drastic solution for him; I can't hear the whine at all - I guess I'm too old . My son could hear the 10 KHz coming from the drivers especially, but since the volume only dropped 50% I guess it either still whines through the wall or the servos themselves are whining too (couldn't measure that with Spectroid though: the 10 KHz peak is the same everywhere in the room).

I was considering something similar a while ago (after the failed cooler box experiment), as my home office is adjacent to my games room. I think what stopped me was the doubt about exactly where the whine was coming from. I didn't want to go to the mess and expense of cutting and dressing a sizeable hole in the wall only to find that the noise wasn't coming from the drivers after all.

Were there any definitive conclusions about the benefits of running at 110v? I remember some discussion about it and a couple of case studies. If this really works then a £100 step-down converter could be a neat and cheap-ish solution for 220v users.

 

[Furnace Inferno]

I can hear the noise but easy enough for me to ignore and once I’ve got the Oculus on it drowns out the sound anyway.

I know you like to use your speakers whilst you drive Mascot but why not just get done noise cancelling headphones since you have a headset on anyway and then problem solved?

The only time that wouldn’t work is if it’s annoying someone else in the room!

 

[Mascot]

I can hear the noise but easy enough for me to ignore and once I’ve got the Oculus on it drowns out the sound anyway.

I know you like to use your speakers whilst you drive Mascot but why not just get done noise cancelling headphones since you have a headset on anyway and then problem solved?

The only time that wouldn’t work is if it’s annoying someone else in the room!

I have thought about that and even bought a noise-cancelling headset (with mixed results), but a lot of the time I really need to hear the phone and the doorbell. I also really like having beefy surround audio coming from the 5.1 amp with Crew Chief coming through the Rift headset just like pit comms with an engineer being piped into my helmet, if you'll pardon the expression.