Page 2 of 3 FirstFirst 123 LastLast
Results 11 to 20 of 22

Thread: 5V vs 3.3V

  1. #11
    Fusion Brain Creator 2k1Toaster's Avatar
    Join Date
    Mar 2006
    Location
    Colorado, but Canadian!
    Posts
    10,045
    Not in MDX. We report the digital value. Since the thermometer is linear at both 3.3v and 5v, if it says 4v with a 5v VCC, it is going to read in 4/5*1023 = 818 and report that to the software. If it has a 3.3v VCC, it will report around 2.64v which is 2.64v/3.3*1023 = 818.
    Fusion Brain Version 6 Released!
    1.9in x 2.9in -- 47mm x 73mm
    30 Digital Outputs -- Directly drive a relay
    15 Analogue Inputs -- Read sensors like temperature, light, distance, acceleration, and more
    Buy now in the MP3Car.com Store

  2. #12
    Newbie
    Join Date
    May 2005
    Posts
    4
    Sooo.. I am only getting into this for a nice customizable LCD gauge in my track car. Given that every single thing on my car is 5v and things I will be adding (oil pressure, etc) will be 5v, should I just stick with finding a used v4?

  3. #13
    Fusion Brain Creator 2k1Toaster's Avatar
    Join Date
    Mar 2006
    Location
    Colorado, but Canadian!
    Posts
    10,045
    Quote Originally Posted by night View Post
    Sooo.. I am only getting into this for a nice customizable LCD gauge in my track car. Given that every single thing on my car is 5v and things I will be adding (oil pressure, etc) will be 5v, should I just stick with finding a used v4?
    You sure about that? Most things in a car are 12v. Or they are resistive, and then will work on any supply voltage.
    Fusion Brain Version 6 Released!
    1.9in x 2.9in -- 47mm x 73mm
    30 Digital Outputs -- Directly drive a relay
    15 Analogue Inputs -- Read sensors like temperature, light, distance, acceleration, and more
    Buy now in the MP3Car.com Store

  4. #14
    Newbie
    Join Date
    May 2005
    Posts
    4
    unless i am misunderstanding how this works with software.
    My car is an OBD1 Honda with tuning software + data logging. That software has built in gauges but they are very basic. However, with that the idea was set to use a full time pc in the car for gauges (and multiple webcams later on for track lap recording). I happened on this and with the 5v fusion would be awesome.
    "Every single thing in the car" was referring to engine controls. IAT, ECT, TPS, MAP, Wideband, etc. All ECU inputs are 5v reference. Then add a couple 5v pulse converters and I can pass everything through for a custom gauge.
    Maybe even an accelerometer for some ghetto street tuning if I can figure out it's accuracy.

    I am frankly wired the hell out atm. I just finally finished a complete custom harness for the engine/car. So any extra work makes me cringe.

  5. #15
    Raw Wave
    Join Date
    Nov 2009
    Posts
    2,118
    5V interfaces are not the same as a 5V supply.

    Most inputs can EASILY be scaled (down) to 5V, or 3.3V etc. (Resistive aka voltage divider - usually two 2c resistors.) That's for both digital and analog inputs.

    And most digital outputs are ground (open collector) switching so the target's voltage does not matter.

    Only output inversion and analog output upscaling requires more complexity (eg amaybe resistors AND a transistor).

  6. #16
    Newbie
    Join Date
    May 2005
    Posts
    4
    well.. i mean 3 wire sensors with |Ground|5v reference|0-5v return signal| to ecu. But I guess what you said still applies.

  7. #17
    Raw Wave
    Join Date
    Nov 2009
    Posts
    2,118
    Well hopefully 2k1Toaster or others will correct if I'm wrong....
    Or unclear.

    But the sensor's supply voltage isn't an an issue for uPCs, PICs etc - unless you have to scale up - ie, expand a 0-1V signal to 0-5V for 5x better resolution.

    Hence FB's change to 3.3V should not be a problem other than adding 2 resistors.


    FWIW, the following from a blog somewhere:
    But I warn you - stop reading now. Later you will find resistance is futile. (To pun 111 of 1001 - any boobs not intended.)


    Even though resistors may have a tolerance of 1% hence a 10k marked resistor may be between 9990 & 1010 Ohm, only refers to their marked error - not that their values change from that (except for normal temperature drift etc).

    Consider too that the design may not be that accurate. EG - assume an 10k impedance for a 5V output sensor scaled for 10k. That means a 3.4k & 6.6k divider (see Wiki etc).
    In practice, using preferred values, you might choose 3.9k & 6.8k which gives 3.178V out for 5.000V in (less than FS 3.3V). That's 3.7% less than the expected 3.300V. Note the impedance is now 3.9+6.8 = 11.1k, but that may not matter - it depends on the sensor.

    Add to that the 2% error from TWO 1% resistors for an error of over 5% (ie - a max of 5.7% even though it could be in your favor - ie, 3.7% - 2% = 1.7%).


    Why does the above NOT matter?
    Because you recalibrate the FB or uPC etc - that corrects for component error.
    And the FB/PC has no idea what it is measuring - it merely sees a value of 1023 or 4095 for "full scale" analog inputs using 10-bit & 12-bit ADCs respectively.
    Hence if the FB is 10-bit, it reads "1023" for the original FB with a 5V sensor, AND for the new "3.3V FB" and the same 5V sensor thru 3.4k & 6.6k 1% resistors (possible 2% error) etc.

    If 2% additional error is ok, then fine.
    Otherwise calibration is required.

    If the 3.9k & 6.6k 1% resistors are used (5.7% error), the FB can be "rescaled" to reduce that error to the mere 2% resistor marking errors. IE - tell the FB that its reading of 3.178V (a value of 986/1024) is NOW full scale.

    But best is to calibrate. Simplest is a 5.000V reference into the voltage divider and measure the output. It might be 3.210V instead of the expected 3.178V (for 3.9k & 6.8k). Cool - so now [3.21/3.3 x 1024 = 996 "less 1") 995 represents full scale - for the input.


    I've probably stated the obvious to those in the know.
    And probably too little for those that don't...
    (And maybe the FB is not calibrated at that "hardware" level - ie, binary - if using a GUI etc.)

    But there are so many ways of tackling and designing these features, yet some are very simple.


    All designs should be similar in that they want maximum transmission voltage (ie, scale-down the signal at the receiving (FB) end), and that FS sensor output should be full scale ADC input or lower (but as close as possible for maximum resolution).
    Whether you use special resistors for a true 1023 FS value, or "common" values to obtain 986 for FS (ie, 37 less in 1024 = 3.6% error (should be 3.7% as above, but rounding errors...) is up to you. Does that "loss of scale" matter compared to the simplicity of common values.
    (FYI - I merely used the E12(?) scale for "preferred values" eg, resistors being 10,12,15,18,22,27,33,39,47,56,68,82,100,120.... But 1% resistors may have 24 preferred values eg, 10,11,12... )


    Then there are the original sensor errors and non-linearity....


    Don't get too carried away with accuracy & bit-values when calibration is required anyhow.
    And do NOT reverse horse & cart.... You use ADC bits to determine the (eg 1024) steps, not range & increments (eg to arrive at 1020 or 1030 as the FS ADC value. The increment size comes from FS divided by bits (eg 1024)!

  8. #18
    Fusion Brain Creator 2k1Toaster's Avatar
    Join Date
    Mar 2006
    Location
    Colorado, but Canadian!
    Posts
    10,045
    Quote Originally Posted by OldSpark View Post
    Well hopefully 2k1Toaster or others will correct if I'm wrong....
    Or unclear.

    But the sensor's supply voltage isn't an an issue for uPCs, PICs etc - unless you have to scale up - ie, expand a 0-1V signal to 0-5V for 5x better resolution.

    Hence FB's change to 3.3V should not be a problem other than adding 2 resistors.
    Well correct that it isn't a problem other than adding 2 resistors. It is a problem if the sensor spits out 5v into the 3.3v port. With a VCC+1.7v signal going in, it is not good. There are internal clamps that *should* just throw away everything over VCC, but they have a maximum current capability, and I wouldn't want to trust those for every scenario.

    Quote Originally Posted by OldSpark View Post
    FWIW, the following from a blog somewhere:
    But I warn you - stop reading now. Later you will find resistance is futile. (To pun 111 of 1001 - any boobs not intended.)
    That's so geeky in so many ways! 111 of 1001 will give you one of her sideways head turns and raised eyebrows for that one.


    Quote Originally Posted by OldSpark View Post
    Even though resistors may have a tolerance of 1% hence a 10k marked resistor may be between 9990 & 1010 Ohm, only refers to their marked error - not that their values change from that (except for normal temperature drift etc).

    Consider too that the design may not be that accurate. EG - assume an 10k impedance for a 5V output sensor scaled for 10k. That means a 3.4k & 6.6k divider (see Wiki etc).
    In practice, using preferred values, you might choose 3.9k & 6.8k which gives 3.178V out for 5.000V in (less than FS 3.3V). That's 3.7% less than the expected 3.300V. Note the impedance is now 3.9+6.8 = 11.1k, but that may not matter - it depends on the sensor.

    Add to that the 2% error from TWO 1% resistors for an error of over 5% (ie - a max of 5.7% even though it could be in your favor - ie, 3.7% - 2% = 1.7%).


    Why does the above NOT matter?
    Because you recalibrate the FB or uPC etc - that corrects for component error.
    And the FB/PC has no idea what it is measuring - it merely sees a value of 1023 or 4095 for "full scale" analog inputs using 10-bit & 12-bit ADCs respectively.
    Hence if the FB is 10-bit, it reads "1023" for the original FB with a 5V sensor, AND for the new "3.3V FB" and the same 5V sensor thru 3.4k & 6.6k 1% resistors (possible 2% error) etc.

    If 2% additional error is ok, then fine.
    Otherwise calibration is required.

    If the 3.9k & 6.6k 1% resistors are used (5.7% error), the FB can be "rescaled" to reduce that error to the mere 2% resistor marking errors. IE - tell the FB that its reading of 3.178V (a value of 986/1024) is NOW full scale.
    The FB can scaled it's reference ADC voltage, but we fix it. It is still a shared sample/hold cap into a single ADC essentially. So if we were to change or reconfigure the reference voltage for every channel, throughput would die. So the FB will still give you back a value that is scaled between 0 and 3.3, but you can take that number and then rescale it to 0-3.178v if you wanted. The error on the FB side as well as the quantization error however remain fixed.

    Quote Originally Posted by OldSpark View Post
    But best is to calibrate. Simplest is a 5.000V reference into the voltage divider and measure the output. It might be 3.210V instead of the expected 3.178V (for 3.9k & 6.8k). Cool - so now [3.21/3.3 x 1024 = 996 "less 1") 995 represents full scale - for the input.


    I've probably stated the obvious to those in the know.
    And probably too little for those that don't...
    (And maybe the FB is not calibrated at that "hardware" level - ie, binary - if using a GUI etc.)
    We try to provide as raw of information as possible. We do spit out the value is raw binary. We will not give you "3.210" but we will give you "995".

    Quote Originally Posted by OldSpark View Post
    But there are so many ways of tackling and designing these features, yet some are very simple.


    All designs should be similar in that they want maximum transmission voltage (ie, scale-down the signal at the receiving (FB) end), and that FS sensor output should be full scale ADC input or lower (but as close as possible for maximum resolution).
    Whether you use special resistors for a true 1023 FS value, or "common" values to obtain 986 for FS (ie, 37 less in 1024 = 3.6% error (should be 3.7% as above, but rounding errors...) is up to you. Does that "loss of scale" matter compared to the simplicity of common values.
    (FYI - I merely used the E12(?) scale for "preferred values" eg, resistors being 10,12,15,18,22,27,33,39,47,56,68,82,100,120.... But 1% resistors may have 24 preferred values eg, 10,11,12... )


    Then there are the original sensor errors and non-linearity....


    Don't get too carried away with accuracy & bit-values when calibration is required anyhow.
    And do NOT reverse horse & cart.... You use ADC bits to determine the (eg 1024) steps, not range & increments (eg to arrive at 1020 or 1030 as the FS ADC value. The increment size comes from FS divided by bits (eg 1024)!
    All good advice. The only thing I would add, is that companies make resistors much more tolerant than 1% if you want it. But like OldSpark said, if calibrated, it doesnt matter. You could use 10% resistors as long as the actual value produced results in the correct range.

    A 10% resistor with a marked value of Y and a real value of X will always be value X. (Well in normal temperature and blah blah semiconductor equations... Just assume that it is always X for normal situations...) The only thing is thing between a 10% marked Y value and a 1% marked Y value, is that for the 10%, 0.9*Y <= X <= 1.1Y and for the 1%, 0.99*Y <= X <= 1.01*Y.

    I have used 0.1% resistors in designs before. (0.999*Y <= X <= 1.001*Y). Also you can buy resistors in pretty much any value, and they add in series as well. So with a little shopping you can get exactly the value you want. If you get 10% tolerance resistors then definately calibrate it. If you use 0.1% or 0.05% resistors then I would still calibrate it. Remember your sensor isnt going to be 0.05% accurate right off the bat. Temperature, humidity, altitude, the type of music playing on the radio, all effect the output of a sensor. So what works for me here, now, today may or may not work the same for you, then, some time in the future.
    Fusion Brain Version 6 Released!
    1.9in x 2.9in -- 47mm x 73mm
    30 Digital Outputs -- Directly drive a relay
    15 Analogue Inputs -- Read sensors like temperature, light, distance, acceleration, and more
    Buy now in the MP3Car.com Store

  9. #19
    Raw Wave
    Join Date
    Nov 2009
    Posts
    2,118
    Phew - we do agree! (I am saying the same as you in case that is somehow misunderstood).
    And FB uses typical binary speak. (Not translated thru some GUI that shows 995 (or 3E3 or 1111100011 etc) instead of 3.21V etc. Yay! Real stuff!!!) Ooops - I meant shows 3.21V instead of....

    And agreed with resistors, but in so many way, that is WHY we use digital - "actual" analog component manufacture tolerance is trivial - the final unit is merely calibrated. (In nonvol memory etc.)

  10. #20
    MySQL Error scott_fx's Avatar
    Join Date
    Dec 2004
    Location
    Los Angeles Ca
    Posts
    3,998
    Quote Originally Posted by bigjimley View Post
    I was wondering the reasoning for the switch myself. I have 22 analogue inputs, over half are 5v sensors. I'm starting to think I may have to stick to the v4's as well. To bad, I was really looking to free up a USB port.
    sorry, this is off topic. but why would v6 free up a usb port?
    New System in progress:
    M10k
    Phaze TD1500 ~> Dynaudio MD130
    Phaze TD1500 ~> Seas g18rnx/p
    Zapco Ref 500.1 ~ 12" tc-9
    Behringer DCX2496 ~ Envision Electronics psu
    Transflective Xenarc

    My Car Pc Install
    My Boat Pc worklog

Page 2 of 3 FirstFirst 123 LastLast

Similar Threads

  1. M2 ATX no power to 5V and 3.3V rail?
    By CTdubbin in forum M2-ATX
    Replies: 1
    Last Post: 11-07-2009, 12:16 AM
  2. not getting +5v or +3.3v
    By mkkbb in forum Power Supplies
    Replies: 13
    Last Post: 04-09-2002, 12:41 PM
  3. Replies: 6
    Last Post: 01-18-2002, 03:39 PM
  4. My new Sproggy style power supply
    By Telek in forum Power Supplies
    Replies: 93
    Last Post: 12-12-2001, 05:52 AM

Bookmarks

Posting Permissions

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