I've noticed a few posts claiming sonar noise can be reduced by shielding the cable. What's the consensus on how to do this? I've seen some posts saying the ground the shielding at both ends, others just one. I'd appreciate any thoughts/feedback.
You need to be a member of diydrones to add comments!
Replies
The sonar is a pretty sensitive sensor it seems. Although not directly relevant to this discussion I wanted to share some experiences I've had with the traditional helicopter. I had bad sonar values even after putting cable shielding in place. I ran some tests where I moved the heli's tail servo around a lot (the tail servo is very close to the sonar). Then I ran an engine test where I powered up the main engine to full speed (the engine is far from the sonar). In both cases I got bad sonar values like you see below.
Then I unsoldered the SJ1 pin (so the APM power and the servo rails were no longer connected). I also provided power to the APM and sonar from a separate battery and had an immediate improvement meaning my problems were all caused by an unstable powersupply to the sonar. Interestingly though I did notice that while the blades were turning the distance returned by the sonar were slightly higher even though the heli was not moving.
I´ve tested both: isolating with shielding (as recomended in the post referenced here before) and with tree wires (servo cable one).
I didn´t see big differences (or nothing at all). The important thing is to get the sonnar far from ESC´s. This makes a dramatical improvement (as suggested in the wiki).
Using twisted three wire can certainly help a bit because the ground cable is acting like a partial shield. This is the technic used for Ultra DMA IDE connection inside PC computers. There is one ground cable between each signal wire to lower alien cross talk between signals.
Nevertheless, twisted cables can be very efficient when used with differential signals, like Ethernet cables. Unfortunately the RC world does not use differential connections, we have only sigle ended (asymetrical) one wire signal connections.
I think that a true shielded cable should be really better than a three wire twisted cable for long servo connections.
The ferrite core can help, for high frequency noise if this one is not very low impedance. But in some cases it will not work at all. Basically a core ferrite is an improvised LR filter. But true RC or RLC low pass filters will give a better high frequency noise rejection because they will be easily precisely adjusted to the needed bandwith and not more.
I've seen some cases, where only optical fiber was able to cleanly link two different systems, because of HF ground loops. When the ground loop do have high frequency content, even transformers do not work, because the noise can go through the transformer capacitive effect...
Globally, EMI problems are often difficult to solve, optical isolation is a radical solution but sometime difficult to implement.
Sometimes, EMI interferences are very complex, and not everybody has an RF engineer and a spectrum analyzer to solve them.
There is a lot of misconceptions in this domain. And every day i could read something opposite than what i learned at school.
Quite often with single ended signals, it's better to ground everywhere in a mesh topology. Then the noise currents are globally shared in the mesh and signals are less affected. This is the opposite that most peoples think.
For ESC, i would advice to use opto ESCs and a separate BEC.
Tests will be your best friend.
Theorically it's better to connect the shielding at the side where the ground plane is the best (obviously on the APM side).
But if you do this, you'll need another wire inside the cable to connect the negative ground pin of the sonar.
This method avoid ground loops in the shielding, producing voltage on the inside wires.
Depending on the EMI problems, it's sometimes better to connect the shield at the two extremities. this is because when the shield is open at one side, it becomes an long wire antenna, sensitive to a very large frequency spectrum and can introduce noise in the ground.
On the opposite, an open shield can transmitt radio frequency signals from the ground or from the wires inside, and trouble radio receivers. This problem is quite common on analog video cables for PC monitors.
Last, there are other complimentary methods to avoid noise :
1) use a differential connection (need a transmitter chip and a receiver chip) to reduce or cancel common mode noise. Using transformers is most of the time even better. In the audio world, the best noise rejection is reached with differential connections, helped by a transformer at each end.
2) use a lower I/O impedance. Using a high impendance to connect low voltage / high precision sensors is not a good idea. The line is very sensitive to noise. This is why in the professionnal audio world, and telephony world, microphones and even long lines are always connected using 600 ohms impedance.
3) add a low pass filter (basic RC or RLC filter) to avoid that high frequency content can enter in the input circuit.
4) use optical isolation : this is the best solution. Can be low cost for low bandwith, this is the solution used on consummer market Compact disc player : S/P DIF optical output using 1mm low cost plastic fiber optic. Digital transmission is simpler and lower cost on optical fiber. Analog can be done to but it's less common and not as simple.
optical isolation is fully resistent to noise. Today, it's even possible to power sensors through the optical fiber using the fiber as a power line, using a photovoltaic device at the sensor side to convert luminous power to electricity.
As it is not easy at all to diagnose from where does come from the noise without very expensive measure hardware, it is often more efficient to make tests with different solutions or rely on other people experience.
The first two solutions for the sonar should work, eventually lowering the APM input impedance and adding a low pass filter could be the ultimate solution if more noise resistence is needed. Do not forget that you will never get better noise performance than the sonar intrinsic noise performance even with the better transmission method.
I've grounded my sonar cable shield at the two extremities and i think it should work ok like this.
So related to this, where's a good place to buy shielded 3 or 4 core cable? I suppose I can repurpose old cat-5 cables or something, but I hate just destroying things.
I've been poking around online, but my choices seem limited unless I want to buy 100m or more.
Sadly my local Radio Shack is useless; I remember when they used to have stuff...
This is why I love this place. Thanks for your quick responses! I've got a shielded USB cable from a cheap old hub that I'd been hanging onto, I knew it'd come in handy one day. Time to start slicing it up.
Ground ONLY at the APM end. Never ever ground at both ends if the goal is to shield from RFI. Otherwise the shielding braid becomes just a part of the ground wire and the cable is as exposed to interference as before.
The first time I added a grounded sensor cable to my sonar, I used a 2 conductor cable with shield and grounded both sides. I've been flying that way for a while now. However, I did see the post from Pieter van Woerkom
http://diydrones.com/forum/topics/shielding-your-sonar-cable
which indicated I did it wrong by grounding the shield on both sides. I checked some EE sources on shielding which confirmed Pieter's post and I do believe it is necessary to get a 3 conductor cable with shield and ground only on the APM side. I've been lazy and haven't found a nice light 3 conductor cable with shield yet, but fully intend to replace my 2 conductor cable to the sonar and test the results in the next couple of weeks.
-
1
-
2
of 2 Next