lordpil's blog


if the hum is so bad its clipping the output of the amp on the ad597, the lpf wont average right and youll have error
kbbl lunch'
its hum and noise
also your ground return for the uc is through the tc amp grounds
which is maybe good or bad
what did you do?
bigass cap across the tc input?
how big is that cap?
stacked ceramic?
as long as the leakage is low enough
and it might add a bunch of latency

Posted by renesis at 05:34 | permalink | 0 comments


1uF will have too much leakage
use like 1nF of ceramic max
i would try 100pF across the tc probe, then do an RC after the AD597 out
like use the lowest thing you have
for the RC use like 10K and 10uF
wtf is 18pf for, random
oh maybe the eth stuff
use a bunch of those
but yeah, for the RC, cut trace to adc, put 10K in series, and 10uF from adc input to ground
and you measurements will have like 1 sec of latency, but 60hz hum should be reduce pretty significantly

Posted by renesis at 05:29 | permalink | 0 comments


see if TC7 has the least noise issues
if it does, then you prob need to rework the grounds

Posted by renesis at 05:19 | permalink | 0 comments


probably, unfortunately this means nothing in terms of electronics
but if you like the ui, thats a pretty good reason to use something
ttmustang: you might want to take a PCB and cut up the ground trace
and bring them all back to your regulator ground
like, bunch of ~20awg wires going back to one ground point, from your AD597, the uC, the logic stuff, the connectors

Posted by renesis at 05:14 | permalink | 0 comments


yeah you need better grounding habits
dont talk shit, you use eda for kids
haha notice how you can see both layers in eagle
yeah orcad is way annoying
just is, all the spreadsheet based config, didnt really like the cad ui, felt like some legacy dos shit
im not korean, i dont play starcraft

Posted by renesis at 05:09 | permalink | 0 comments


but yeah you can try averaging in software, but for it to work nearly as well as an analog filter you have to sample real fast
and have huge average sets
i was using filters with -3db set to like 1Hz, 4Hz, sometimes less
so you can put an rc on the output
to get rid of hum
but something that simple needs to be set at really low freq to get rid of meaningful amounts of hum
like, prob between 0.1Hz and 10Hz
yes
then to get rid of high frequency stuff, try putting like 100pF across the thermocouple input
and make sure the probe is grounded somewhere
at the system ground, on the earthed chassis
hopefully system ground
hmm
then it could be ground loops, maybe
depends on your supply, and how your system is ultimately grounded, and whats up with how the probe is mounted

Posted by renesis at 05:04 | permalink | 0 comments


put the scope probe in 10x mode, set voltage divisions to like 50mV, set time to like 10mS
put it on the ad597 output
trace should be flat, if you see noise, its that
the adc will prob give random values around the actual measured value
if you sample fast enough and/or can wait long enough, you can software average
or you can RC LPF
same compromises, the more you reduce the noise, the more latency you get from actual measurements
its not the number of samples
its the latency
like, all your readings will be delayed, maybe half a second to a few seconds
dont do that
do 8 samples and divide by 8, or do 16 samples
do 2^x numbers, its just shift operations, it takes much less time

Posted by renesis at 04:59 | permalink | 0 comments


or its noise
or it was a bad connection to the plug

Posted by renesis at 04:51 | permalink | 0 comments


heh

Posted by renesis at 04:14 | permalink | 0 comments


pasta sauce doesnt really need to have fat
its just the oil, dont really need that
i must not have chrons

Posted by renesis at 04:08 | permalink | 0 comments


analog oscopes will usually show all the peaks
sometimes digi ones will leave them out because of sampling or display res

Posted by renesis at 03:57 | permalink | 0 comments


that shit has an spl meter?
it says it measures sound
and it has dB in its mode selector
im guessing thats dBspl
also light
ha you can see the robber grip overmold peeling off

Posted by renesis at 03:32 | permalink | 0 comments


on a diff adc input?

Posted by renesis at 03:16 | permalink | 0 comments


its 1/8" threaded or what
extension wire responds faster than normal probes, too
well if you need it to be hardcore like that

Posted by renesis at 03:07 | permalink | 0 comments


are your adc readings stable?
oh
isnt 597 for type j?
oh, no thats for k nm
you dont have any filtering on the adc inputs so if theres noise being picked up by the tc cable itll show up in the adc readings
across the inputs?
buy extension wire from mcmaster, and weld the ends
crimp/weld them into some ring terminals if you need to mount them with a screw
its like $.33/ft
why
you cant just immerse the wires?

Posted by renesis at 03:02 | permalink | 0 comments


fuck that
put it in boiling water, it should says 1V
put it in between some ice cubes, it should go down to 0V
ttmustang: 21.9 degrees Celsius = 71.42 degrees Fahrenheit
google knows
oh the ic?

Posted by renesis at 02:51 | permalink | 0 comments


(base resistors)

Posted by renesis at 02:45 | permalink | 0 comments


youre going to use four 555 instead of 4 pins on a uc?
at that point i would just dedicate a whole soic8 uC to it or something
then use a dip8 uc
man now you need an even bigger uC
_unreal_: i give up use an army of 555s

Posted by renesis at 02:40 | permalink | 0 comments


yeah but this $5 chip does it automatically
has the input amps, CJC, error amps
for just a sensor amp, at $5its a bit much, but itll do a whole thermostat system by itself
thats worth it
sculptor: yes
like i said, decoupling cap, done
_unreal_: that shit needs base resistors
for HA and HB
LA and LB maybe you get away without them
the way it is now, its using the current clamping in the 74 series
he did
weird
yeah that part seems neat
why a 555?
just use a timer output on a uC

Posted by renesis at 02:35 | permalink | 0 comments


yeah thats where youre fucking up
just get j thermocouple extension wire
and twist the ends
its faster response than any commercial probe
because the shit has no mass
how long do you need, ttmustang?
$2
http://www.mcmaster.com/#type-j-thermocouple-wire/=em8ryy
i made a weed vaporizer with digital feedback loop!
and i guess you would have to buy connectors
male and female
anyway, redo comparison with TC probe as $2+plug instead of $30

Posted by renesis at 02:30 | permalink | 0 comments


to what temp tho?
im thinking like, 250F or so
_unreal_: the transistors need base resistors
hes doing some sort of furnace thing, tho
so i dont know if hell be able to do the whole system with coolant sensors
corn husk furnace, i guess
oh i didnt see your first response
ttmustang: isnt that more than the ad597?
thermocouple probes are a ripoff
just get extension wire and twist the ends, weld somehow
soldering even works, just not nearly as reliable
and its a bitch to solder

Posted by renesis at 02:25 | permalink | 0 comments


ad597?
sculptor: you can set it up like, standalone sensor, or use it as a thermostat
but in its simplest config, you just need a decoupling cap, and itll output 10mV/C absolute
attach DMM, instant wide range temp sensor
just ignore decimal, maybe a zero
ttmustang: you usually want to ground the thermocouples somehow
or theyre going to feed the amps mad hum
if you dont ground it at the sensor IC, you want it to be grounded by what you mount the probe to
im not looking at the datasheet
and yeah, car temp sensor might not be very accurate or have a wide range, but if its enough and its cheaper, prob ok, prob easier

Posted by renesis at 02:20 | permalink | 0 comments


Top | Add to Technorati Favorites


© 2007 lordpil.   XHTML 1.0! CSS! Site design by GNAA  Blog Engine by pbx | MULTI2 | ian hanschen | lolwat