Hello and welcome to our community! Is this your first visit?
Results 1 to 7 of 7
  1. #1
    AeroQuad Lead Software Architect Kenny9999's Avatar
    Join Date
    Jun 2010
    Saint-Jean-sur-Richelieu, Quebec, Canada
    Reputation Points (Add)

    Analog sensors vs Digital sensors

    For sometime now I hear Honk saying that analog sensors are better, they can be sampled faster and some others argument,

    There is a couple of new member here that are Hardware wise guy, Martyn and mindThomas, and there was an interesting conversation going on the IRC. So, I find it interesting and would like to start a thread about it to see people argue to have a better idea by the end.

    [18:08] <mindThomas> Martyn: What would you prefer to use, Digital or Analog sensors?
    [18:42] <Martyn> I prefer digital sensors
    [18:42] <Martyn> analog sensors are really terribly subject to noise
    [18:42] <Martyn> digital sensors can be places at the FAR periphery of a board, using i2c or SPI
    [18:42] <Kenny9999> AHHHHHH
    [18:42] <Martyn> or even placed -off- board, which is key
    [18:42] <Martyn> Plus, analog sensors are subject to A/D conversion noise and errors
    [18:43] <Martyn> So people keep telling me. I'll make an account on AQ forums then
    [18:43] <Martyn> also, the faster the processor, the more important it is to use digital vs/ D/A
    [18:43] <Kenny9999> Ah nice, continue, I like it...
    [18:43] <@kha> Martyn: not true...
    [18:44] <Martyn> because you can buy digital sensors of ALL kinds of resolutions, and using I2C or SPI means you can have -arbitrary- precision on your sensor
    [18:44] <Martyn> but on analog, you're stuck with whatever the hell D/A converter the manufacturer decided to give you, plus you're limited in the number of sensor channels
    [18:44] <Martyn> one SPI bus beats the hell out of 8 A/D channels
    [18:44] <Martyn> kha : I'd like to see your reasoning on this one
    [18:44] <@kha> Martyn: a digital sensor just has the A/D converter in it
    [18:45] <Martyn> kha : Yes, and it's a DIFFERENT a/d converter per device. You're not limited by number -or- precision.
    [18:45] <@kha> and in many cases it's a far more limited A/ converter than you can have externally
    [18:46] <Martyn> kha : For example, I have a photomultiplier with IR sensitivity and full 16 bit A/D, capable of taking 20k samples/sec
    [18:46] <@kha> Martyn: read the data sheets... and check some quality A/D:s out there
    [18:46] <Kenny9999> some argument already!
    [18:47] <Martyn> kha : But I also have a sonar sensor array that can return anything from 10bit to 16bit depending on how it's programmed
    [18:47] <Martyn> kha : I'm sorry, but there is no way in hell having one particular A/D converter on-chip is going to beat the flexibility of what I can buy in the market with SPI/I2C
    [18:55] <@kha> Martyn: who said a on chip A/D (they are crap)[18:55] <@kha> (most of the time)

    So, what you thing is the best, with my monkey brain, I vote for digital sensors!!!
    Truly superior pilots are those who use their superior judgment to avoid those situations where they might have to use their superior skills"- Author unknown

  2. #2
    Senior Pilot xtrmtrk's Avatar
    Join Date
    Jan 2012
    Bellevue, Washington USA
    Blog Entries
    Reputation Points (Add)
    Can you really make blanket statements like that? Don't you have to look at each device, especially the digital ones, to see what kind of internal a/d they are actually doing? I'd guess it's different for different price points and design targets.

    It seems like with analog you are assuming that you can do better a/d than the sensor's manufacturer can (or at least has chosen to for a particular configuration). With digital you're allowing for the possibility that the designer can create the best conversion hardware/firmware.

    You're also subject to the quality and resolution of your a/d hardware. If you only have a couple bits resolution, with modern analog devices you're probably leaving precision on the table.

    I'm a software guy so I'm probably biased, but I really think you have to examine each device's specs individually.

    I think a much better question would be what's the best possible accelerometer configuration we could get that costs less than $10 at the component level and is usable in the AQ environment. Could be analog or digital. And I know that question has been asked many times and we need to keep asking it. I'm sure the answer changes as new products are available.
    2013 Carnage (so far!): 26 propellers, 1 frame, 5 arms, 1 motor mount, 2 motors, 2 ESCs, 1 GPS

  3. #3
    Moderator AeroQuad Core Software Developer kha's Avatar
    Join Date
    Apr 2011
    Espoo, Finland
    Blog Entries
    Reputation Points (Add)
    Yes, there is no simple answer for this question... but one aspect is clear:

    Digital sensors are far easier to work with

    Going the analog road needs one to o the homework carefully and also design the system so that the analog signals are not disturbed.

  4. #4
    Senior Pilot jsmcms's Avatar
    Join Date
    Mar 2010
    Detroit, Michigan USA
    Reputation Points (Add)
    Well said.
    Work hard, but play harder!

  5. #5
    Senior Pilot
    Join Date
    Jan 2011
    Suffolk County NY , USA
    Reputation Points (Add)
    Most ( Dont want to say ALL .. because someone will find one .. ) sensors are analogue at the core. A 'digital' one just has an onboard A/D.

    Also the concept of *better* is subjective. Better at what ? If we are talking about *potential* raw sensor performance , then yes analogue will almost always be better.

    I think the real question isn't "is digital better then analogue" .. its "Should I design my own A/D section , or use a pre-designed one"

    And to be honest , there are very few of us on this forum that could design a *better* a/d section then the geeks at ivensense.

  6. #6
    Senior Pilot wooden's Avatar
    Join Date
    Dec 2011
    Pacific Northwest
    Blog Entries
    Reputation Points (Add)
    I doubt any of us could design a better ADC than the people who are paid to do it for a living.

    Analog signals are touchy, all sorts of noise can creep in from all sorts of sources. One of the senior design teams at my school discovered this during their project - they spent several weeks just testing different filters and casings until they finally got their system quiet enough that they could actually do their project. Trying to do this on a flying, vibrating object with several different wireless communication systems going simultaneously is no simple task.

    Besides, the ADCs on most of these chips today are quite powerful and give more than enough resolution for a stable copter flight.

  7. #7
    Moderator AeroQuad Technologist Honk's Avatar
    Join Date
    Apr 2010
    Blog Entries
    Reputation Points (Add)
    DAMN!!! This discussion is old and I have the answer: people here don't wanna spend $50-100 extra to have a good acquisition system and that's it. That's why MultiWii and AQ and all other Ardu projects are gonna keep using mainstream underspeced mobile phone sensors. It flies and people are happy because it's cheap.

    I doubt any of us could design a better ADC than the people who are paid to do it for a living.
    I do it for a living now. But it's not even about "better"! It's about application and MONEY! Invensense makes CHEAP sensors for NON-VIBRATION ENVIRONMENT use while we wanna use it for a vibration FILLED environment. Of course they're not suited to that, it's more expensive and mobile phones can't integrate the accels for linear speed/movement and can not integrate the rate gyro's for very long to have an absolute angle. And even less of that in a vibration environment. Invensense are GOOD at designing devices for low prices to handheld devices. I don't think they make anything with more performance than that.

    Of course you can squeeze out MUCH more of a sensor if you have access to its analog output and can choose your ADC unit on your own. Every time I've talked to people who's been in the industry for long, they say that you can't combine clock domains and stuff on the same silicon as analog stuff. It's simply degrading signal quality. What has always been the way is to use an analog output sensor, separate ADC chip and separate microcontroller or equal.

    This is a board I'm gonna get made as an analog shield for Baloo:
    * 1x 16 bit 500kSPS 8 ch ADC for gyr/acc with mux auto cycling
    * 1x 24 bit 2kSPS 1 channel
    * separate voltage reference IC's

    This is what I made the other day as a standalone gimbal stabilizer using 2 ADXRS620 and a Pro Mini integrating them @ 15/30/60kHz each (by fiddling with adc clock prescaler) and digital tail servos refreshed @ 560Hz:

    And today I maidened it, although with a lousy expendable compact cam, broken roll servo gears and not correct gains:

    Sorry for all the ranting, but I stand firm that analog is the way to go for our crafts BECAUSE: no finished digital commercial sensor chip in the low budget price range is made to be on a vibrating copter. That's it. With the analog gyro's and the Atmega sampling it I could easily hold an absolute angle (by integrating them) for an entire flight! No accelerometer involved = no leans/linear acceleration/centripetal force messing with attitude. For example. Or if we wanna utilize accelerometer for linear dampening up or to the sides, we do NEED a good analog-to-digital accelerometer acquisition system. I'm 100% convinced of that.
    Smart people don't need a signature.



Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts