Are Hall sensors really needed for joystick’s?

[Cloakedwand72]

Are hall sensor’s really needed for flight joysticks? That’s one of my deciding factors on a HOTAS.

[Nuke]

no.

i got some really high quality potentiometers from the '60s that still display good linearity and low noise. they were made in the usa, so good luck finding that now. ive actually tested this by connecting a stepper motor to the pot and stepping through at regular intervals. the data looked about as clean as what i was pulling off of the hall sensor. 

the quality of potentiometers in joysticks ranges from excellent to awful, and its not always in the ones you would expect. i found really crappy pots in a fairly high end stick (an old saitek hotas, which also had awful wiring), and my ch controllers had really solid pots as well. however durability is also an issue. with older ceramic or metal cased pots, they are damn near immortal and they come from a time when analog was king. but the newer pots which are made with conductive ink and pcb (usually even crap like bakelite pcbs) are very subject to wear over time. they have exposed copper contacts which can also corrode. 

hall sensors are not all rainbows and unicorns either. the sensors i have evaluated are only liner through about 80% of their range. others have dead bands on the top and bottom of the analog range, some as large as a volt off the top and bottom at 5v, which is 2/5th of your range. you can compensate for bad linearity (for any sensor) by generating a look up table of measurements at known angular displacements (eg 1 degree per index), store the raw measured value along side a corrective calibration value. for any value, find the 2 measured values that bracket it and lerp the associated corrective values, add the result to your data.  measure each sensor because its a good probability that 2 different sensors (even of the same type) will need different corrections, and number your sensors so you can associate the correct lut to the correct sensor. the best way to make the measurement is with a stepper motor, well isolated from the hall sensor to prevent magnetic interference. 

there is also a matter of range. most of the hall sensors i have are good for 180 degrees before any dead or nonlinear zones are delt with. while most pots do 270 degrees (though other ranges exist, also make sure they are linear and not logarithmic). pots do not usually have dead zones and you can use the full range. either range is more than what a joystick uses (30-60 degrees). gearing or mechanical linkages can be used to reduce the range, but add extra play. you can also use a higher resolution adc and direct drive to reduce slop. hall is better in this regard as you are throwing away less useful range, especially when you can only use 2/5th of it (thats 72 degrees). 

the most innovative joystick sensor ive seen are the optical sensors used on various sidewinder sticks (lare '90s early '00s). these were the same sensors used on early optical mice (usually a low res camera + pre-programmed dsp). id love to see a modern refresh of this using better sensors, as you can get mice with stupid resolution now. 

that said it all comes down to the engineering and any corner cutting they made in manufacturing (which are ever present in todays tech markets). get something you can mod. reviews and teardown videos are critical in making a decision. 

Edited October 6, 2022 by Nuke

[Dafni]

2 hours ago, Nuke said:

**bookmark-worthy information**

Thank you for sharing this detailed insight. I appreciate your time!

[Nuke]

i remember the first edition of the x52 (saitek branded) it was technically a hall joystick. but it was implemented in such a shoddy way as to be worse than a pot joystick. it was actuated through a plastic rack and pinion. there was a sliding peice with a single magnet. that moved over the sensor. this no doubt lead to a lot of nonlinearity that was questionable. i dont know how their b-fields were oriented, turns out that's really important for linearity. best results ive had were placing 2 magnets in north-south north-south order on either side of a rotating armature, which would rotate about the sensor (an ss49e in a to90 package with the center pin roughly in line with the rotational axis). such that you get straight flux lines through the center axis where the sensor is placed. the stick also had a single spring system which i didnt really like. on the ch fighterstick usb, i can feel how much x and how much y i have applied, you just cant do that with the x-52

the saitek x-36? i think was the stick i mentioned in my previous post. just a terrible pot stick. its only redeeming feature was the throttle paddle it had behind the throttle grip.  the ch fighterstick is legendary, those will last a decade easy, though i wish they would have used a material other than abs plastic. it makes it look cheap (its not), and has lead to the hardware trim being a bit sloppy. and i wish they would have used a better adc (its only 8-bit).

[boriz]

Hall sensors, being non-contact and immune to wear 'should' always be better. Any non-linearity can easily be linearised in firmware. The problem is the rest of the hardware and how it is effected by wear and tear. A good Hall joystick can become a turd after six months if the hardware is weak, and a good Pot joystick can remain good for years if the hardware is strong.

No substitute for a proper review when choosing peripherals. And by proper, I mean non-sponsored and after lots of use. Early adopters beware.

[Nuke]

in theory but like i said i have well used vintage potentiometers with better linearity than hall and have stood the test of time.

can you get an equally high quality hall sensor. well, yes, yes you can. you can get high resolution angle sensors with a digital interface so you dont have to deal with signal noise in the wiring and corrected for linearity at the factory (they have a dsp build in to apply signal corrections). but consumers dont buy sensors, they buy joysticks. those sticks come from manufacturing companies with shareholders, and those shareholders want a bigger roi. so its in their best interest to use the cheapest sensor possible, even shaving a few cents off per unit will have an impact on their bottom line, and also an impact to the performance for the end user. you can always spend a premium for what amounts to using a part that costs less than a dollar more. 

the thing im trying to hammer home is that it varies from stick to stick. my ch controllers have pots and crappy adcs, yet have lasted over a decade. my record survival rate for any saitek stick (one of which was a hall stick) was 2 years. bad engineering and cost cutting will kill any theoretical benefit to be gained by using one sensor type over the other. a good sensor wont fix a bad gimbal design. the best sensor tech ive ever used were microsoft's sidewinder line, with their optical sensor (basically a mouse sensor with a dsp on chip). all modern smd boards, state of the art sensor, force feedback on one model. the non ff models had a single spring mechanism which i do not really like, i like separate spring tension on each axis. so dispite good electronics and sensor tech, a mechanical design choice can ruin things (saitek in a nutshell). it is a bloody shame nobody uses that technology anymore, except with a modern high resolution mouse sensor with a decent gimbal set.

if i was in the market for a new stick (and ive been looking for something better), id want solid gimbals (preferably metal construction), high modability with a community providing upgrades and modules as well as parts for maintenance, and no dependendance on proprietary software. the sensor tech might be the 4th thing i want, but if its a high moddable stick you can usually put whatever you want in there. i was looking at a warthog, thrust master has a good modding community (proprietary software though), or if you want really high end go virpil and pay over $200 for the gimbal alone. it is a sexy gimbal though, and its their low end model.  

Edited November 7, 2022 by Nuke

[Nuke]

after a series of videos im starting to get seriously interested in building a resolver or synchro based joystick. these are ac analog devices capable of voodoo the likes of which us dc sensor enthusiasts can only dream of. resolvers emit two signals proportional to the sine and cosine of the input shaft. i figure sticking a 16 bit double ended dac on each axis needing 2 channels to read each axis and a little trig, i can get really high resolution input out of the thing. synchros on the other hand have a cool property where by they can not only be read, but driven, enabling force feedback, or amplified and sent to a bigger synchro.

these things are pretty common in avionics and were used heavily during wwii. you can still find them in surplus shops and modern examples exist if you want to pay a lot of money for them. anyway you need an ac supply to drive them. not the usual 60 hz you get from mains, lots of them run at 400 hz and at voltages ranging from benign to insane. not really hard to drive with an h-bridge and a micro controller, just pwm out the sine waves while rotating through polarity, though there is probibly a cleaner way to do it. anyway im going to research it while i try to remember the ac unit i never finished in my electronics classes. before i can do anything i need to locate the parts.