Author Topic: undersampling waveform  (Read 3365 times)

0 Members and 1 Guest are viewing this topic.

Offline kokodinTopic starter

  • Regular Contributor
  • *
  • Posts: 82
  • Country: pl
undersampling waveform
« on: April 25, 2017, 07:06:55 am »
Hello. Since i have a problem with measuring one particular waveform rms value, the waveform is out of range of my meter, i was thinking about a way to cheat.
Since i know the exact frequency of my waveform and i know it is constant in shape, there might be a way around it, by sampling the waveform at 10 or 20 times slower frequency than the waveform cycle.
In theory it would work like calipers, by downsampling the waveform. But to make this work for measurment it would also need output this sampled waveform in real time.
So i was thinking about some kind of analog circuit with fast switching input and output stages at .
Repeater that would operate at a given frequency.
Signal is about 15.625 kHz, my meter have limit of aboit 1kHz at rms measurment. From the frequency you can probably tell it is a pal signal thing, to be more specyfic i am still fighting my tv heater voltage with no proper equipment.
I am dubble cautious now, because i measured with a scope heater voltage of my other tv with the same crt model. And the waveform was a little bit sharper and different, also non rms meter shown a bout 1 Volt difference between them. I did not mess around with this one.
They have the same crt so they should be driven by the same voltages. but who knows, one have color amps powered with 215V dc the other have 240V there, and both tvs are in specs with that. (the 240V cathode amp one shows lower heater voltage though)

edit
It could probably also work with buffered analog meter, but i don't have one. What i have is a cheap analog meter though so i could use its scale for something. Either as a whole unit with a buffer stage with a low impedance output added or as a scale for something new.
« Last Edit: April 25, 2017, 07:16:00 am by kokodin »
 

Offline Benta

  • Super Contributor
  • ***
  • Posts: 5840
  • Country: de
Re: undersampling waveform
« Reply #1 on: April 25, 2017, 09:35:34 am »
You could think about playing with an AD8436 and spin your own RMS -> DC converter.

 

Online David Hess

  • Super Contributor
  • ***
  • Posts: 16548
  • Country: us
  • DavidH
Re: undersampling waveform
« Reply #2 on: April 25, 2017, 10:00:51 pm »
You do not even need to know the shape or frequency of the signal to make an undersampled measurement of the RMS value.  The standard deviation of the signal is equal to the RMS value and the standard deviation does not change when the signal is undersampled.

So take the output of your sampling circuit and feed it into your low frequency RMS meter.  This is how RF sampling RMS voltmeters work.  If I took the output of my Tektronix 7S11 vertical sampling amplifier and fed it into my low frequency RMS voltmeter, then I could make RMS voltage measurements into the GHz range.  Lookup the Racal Dana 9301A and HP 3406A for instruments that work this way.

If you have a way to sample into the digital domain, just calculate the standard deviation.
 

Offline kokodinTopic starter

  • Regular Contributor
  • *
  • Posts: 82
  • Country: pl
Re: undersampling waveform
« Reply #3 on: April 26, 2017, 05:43:38 pm »
Well i did not expect it actually make sense, but aparently it does  ;D

Thank you for reasuring me.

I find something simpler than digging trough full device schematics and separating the sampler circuit
http://webpages.charter.net/dawill/tmoranwms/Elec_Sample.html
Basically it is what i was thinking about , it is even explained :] but i have a question.
http://webpages.charter.net/dawill/tmoranwms/Elec_Sampler5_lg.gif this schematic looks simple enough to make, a little bit more complicated than a regular guitar stompbox, but my input signal is about 30-35Vpp and this sampler is powered from 12V
I could either recalculate entire schematic to 40v power suply (unlikely) or push my signal trough high impedance voltage divider, let's say 1:4.
So the question is, would it work better if it was recalculated for bigger signal, or would it make no difference and i only would have to multiply it back by 4? Because i am assuming that this circuit is what i need. (correct me if i am wrong)

I can probably use my scope to compare slower waveform to its oryginal and find it out after i make it, but i would rather know if it is in fact correct circuit to use.
 

Online David Hess

  • Super Contributor
  • ***
  • Posts: 16548
  • Country: us
  • DavidH
Re: undersampling waveform
« Reply #4 on: April 26, 2017, 08:42:53 pm »
Samplers usually rely on a low impedance signal but there are some exceptions.  The example you show is a closed loop sample and hold which will deliver very good accuracy limited by the Vbe mismatch of the 2N4401 differential pair and has a high input impedance but is much slower than most samplers.  It will still be more than fast enough for your purposes and could be modified for high voltage operation.
 

Offline kokodinTopic starter

  • Regular Contributor
  • *
  • Posts: 82
  • Country: pl
Re: undersampling waveform
« Reply #5 on: April 27, 2017, 12:19:22 am »
well i was thinking about high input impedance circuit, because it is not a signal itself that interest me, but rather the rms voltage of the signal under load, in this case it is a crt heater filement. I do not wish to load the filement with another low impedance , because this would throw off my measurment.
Power source for the filement in this tv is a crude extra winding in flyback transformer,feeding of return lines from horizontal deflrction. There is no much way of controlling the current or voltage of that thing, the only adjustment i have is an extra choking coil with an adjustable core. Pretty much closed loop
As an added bonus, if i change screen geomatry or base voltage of the set, the heater voltahe also changes.
 

Online David Hess

  • Super Contributor
  • ***
  • Posts: 16548
  • Country: us
  • DavidH
Re: undersampling waveform
« Reply #6 on: April 27, 2017, 04:05:02 pm »
There has got to be an easier way to measure that than building a sampler to feed your RMS measuring meter.  I think Benta's suggestion about using an RMS converter IC would save you a lot of time.  Linear Technology and Analog Devices make several which would be suitable.

http://www.analog.com/en/products/linear-products/rms-to-dc-converters/ad8436.html#product-overview
http://www.linear.com/product/LTC1968

 

Offline kokodinTopic starter

  • Regular Contributor
  • *
  • Posts: 82
  • Country: pl
Re: undersampling waveform
« Reply #7 on: April 27, 2017, 06:04:16 pm »
to be honest i looked up the AD8436  in electronic price comparing engine and it come up only as a premade board of about half a price of a new scope so i got confused and droped it. this ltc1968 chip is around 10 usd in TME shop so it might be worth the effort, but i would have to dig into details of its datasheet , because i would rather not blow it up on a first try. (input voltage of about 1 V pp and all that stuff), The converter itself look rather simple in its aplication , but it require a meter front end with a range switch and some point of reference . Front end attenuation must be known to multiply the output back to its proper value.and then measure , so a lot of points where it can go wrong :]
 

Online David Hess

  • Super Contributor
  • ***
  • Posts: 16548
  • Country: us
  • DavidH
Re: undersampling waveform
« Reply #8 on: April 27, 2017, 06:58:48 pm »
Calibration is pretty simple for these; the RMS value of a DC signal is just the value of the DC signal.

Attenuate the input signal to the minimum value compatible with the accuracy you want so as to maximize the supported crest factor.
 

Offline kokodinTopic starter

  • Regular Contributor
  • *
  • Posts: 82
  • Country: pl
Re: undersampling waveform
« Reply #9 on: April 28, 2017, 01:05:59 pm »
Well yes, if we talking dc it should just show dc value and input shold equal ottput.
However datasheet say that input is a diferential input where one of the leads is ac coupled, or should be, and the other dc coupled.
are you saying that for calibration i should feed it with dc while both input terminals being dc coupled?

From what i understand  to use this thing relatively easily i should make high impedance voltahe devider of around 35-40 times To match peak to peak voltage of my tv to the 1V pp input, after that i might want to put back to back zener diodes as a overvoltage protection, or even 4 fast diodes (2 in series paralell with two in reverse ) then use standard aplication for let's say for symetrical power suply and buffer the output with  opamp of 35to 40 times gain, also powered from symetrical power suply. that should be able to be used with standard dc multimeter. Shown dc voltage should be correct without more calculations.
well that is a theory, it might work it might not.
 

Online David Hess

  • Super Contributor
  • ***
  • Posts: 16548
  • Country: us
  • DavidH
Re: undersampling waveform
« Reply #10 on: April 29, 2017, 09:53:47 pm »
Well yes, if we talking dc it should just show dc value and input shold equal ottput.
However datasheet say that input is a diferential input where one of the leads is ac coupled, or should be, and the other dc coupled.
are you saying that for calibration i should feed it with dc while both input terminals being dc coupled?

I will assume you are talking about the LTC1968.

Check the application examples in the datasheet again.  Most users are interested in AC RMS but AC+DC is done by not using any coupling capacitor.  For a single ended AC+DC input, one input is grounded.

Quote
From what i understand  to use this thing relatively easily i should make high impedance voltahe devider of around 35-40 times To match peak to peak voltage of my tv to the 1V pp input, after that i might want to put back to back zener diodes as a overvoltage protection, or even 4 fast diodes (2 in series paralell with two in reverse ) then use standard aplication for let's say for symetrical power suply and buffer the output with  opamp of 35to 40 times gain, also powered from symetrical power suply. that should be able to be used with standard dc multimeter. Shown dc voltage should be correct without more calculations.
well that is a theory, it might work it might not.

The input impedance of the LTC1968 is pretty high and the heater for a CRT very low so this should not be a problem.
 

Offline kokodinTopic starter

  • Regular Contributor
  • *
  • Posts: 82
  • Country: pl
Re: undersampling waveform
« Reply #11 on: May 01, 2017, 11:16:58 am »
i feel kind of dumb , because i forget about ac+dc voltages at all :] oh well.

I was browsing TME site for avalable rms-dc converters and i find another one i might be interested
while ltc1968 have pretty good performance and bandwitch, not to mention pretty simple aplication. It look more like a dedicated part for being build into a multimeter, besides the shop have a really small quantiti of those, under 10 pieces
so i did a search for rms-dc converters and AD736JNZ poped up AD736JNZ
, they have more tnan 700 of those and the price is also a bit smaller.
It have smaller bandwidth , but still enough i thinh,  can be powered from 16,5V+- and accept input voltages up to 12V+- p-p (1Vrms for power suply 5 to 16V+-) so it would work great from two 9v batteries as a stand alone unit , there is even a bunch of different aplications to use it , like for example adustable gain of the output voltage and premade input voltage devider. Not to mention it is in dip package so it wuld be easier to not destroy while soldering :] Not if i would destroy an smd package.

with input devider of 10 i would have 500-800mVrms at 3,5V p-p so i think a pretty resonable parameters for this converter
« Last Edit: May 01, 2017, 11:19:36 am by kokodin »
 

Offline kokodinTopic starter

  • Regular Contributor
  • *
  • Posts: 82
  • Country: pl
Re: undersampling waveform
« Reply #12 on: May 02, 2017, 01:56:59 pm »
i read trough that ad736 datasheet and i made aproximate circuit baced on internal schematic from datasheet , it kind of work :]
i have absolutly no faith in it in high frequency operations but it reads low ac voltages on multimeter milivolt scale just fine
since i probably missed somethng internaly (like that constant current source) my mockup circuit is overbiased and shows 1V on ac coupled and 2V on dc coupled modes without signal , dc measurment is also biased to one side, so the mockup circuit is unusabel with more distorted wavefoms, but at least i know how would it work with oryginal ic without buying or breaking one :]
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf