Author Topic: DS2072 int(ch1*ch2) math error?  (Read 2113 times)

0 Members and 1 Guest are viewing this topic.

Offline Circlotron

  • Frequent Contributor
  • **
  • Posts: 809
  • Country: au
DS2072 int(ch1*ch2) math error?
« on: February 08, 2016, 12:20:28 PM »
Okay, so I'm looking at the blue voltage line and the yellow current line and wanting the integral of the two so I can see the total energy (milliJoules) in the event.

First pic looks pretty good. After things finish the purple integral line stays at a steady final level.

Second pic, scope setttings unchanged but make things a bit more energetic. Now that the yellow current returns to zero but the blue voltage does not, this seems to cause the purple integral line to start drifting downward. It should not because non-zero blue is being multiplied by *zero* yellow, so purple should stay put.

Third pic is the same trace slowed down a little and this causes purple to be offset downward (maybe ok) but it also shows how far the the purple can drift down while blue is supposedly being multiplied by zero.

Anyone else had this kind of issue?


 

Offline markone

  • Frequent Contributor
  • **
  • Posts: 367
  • Country: it
Re: DS2072 int(ch1*ch2) math error?
« Reply #1 on: February 08, 2016, 12:43:38 PM »
Actually seems weird, but if you look closely the first screen you can also notice the integral line stopping a little before than CH1 and CH2 reach zero, like the math was limited in precision or out of phase.

For the rest it's like CH1 is never exactly equal to zero.

You could try to make the integral of CH1 and CH2 separately to sort out that.

Okay, so I'm looking at the blue voltage line and the yellow current line and wanting the integral of the two so I can see the total energy (milliJoules) in the event.

First pic looks pretty good. After things finish the purple integral line stays at a steady final level.

Second pic, scope setttings unchanged but make things a bit more energetic. Now that the yellow current returns to zero but the blue voltage does not, this seems to cause the purple integral line to start drifting downward. It should not because non-zero blue is being multiplied by *zero* yellow, so purple should stay put.

Third pic is the same trace slowed down a little and this causes purple to be offset downward (maybe ok) but it also shows how far the the purple can drift down while blue is supposedly being multiplied by zero.

Anyone else had this kind of issue?
 

Offline markone

  • Frequent Contributor
  • **
  • Posts: 367
  • Country: it
Re: DS2072 int(ch1*ch2) math error?
« Reply #2 on: February 08, 2016, 12:46:18 PM »
Also in the second screen the integral start to drift a little already from  the trace start.
 

Offline Circlotron

  • Frequent Contributor
  • **
  • Posts: 809
  • Country: au
Re: DS2072 int(ch1*ch2) math error?
« Reply #3 on: February 08, 2016, 12:54:12 PM »
Also in the second screen the integral start to drift a little already from  the trace start.
Yes, while yellow is still at zero.
Maybe there is some kind of small DC offset in the scope?
 

Offline Circlotron

  • Frequent Contributor
  • **
  • Posts: 809
  • Country: au
Re: DS2072 int(ch1*ch2) math error?
« Reply #4 on: February 08, 2016, 12:56:35 PM »
For the rest it's like CH1 is never exactly equal to zero.

You could try to make the integral of CH1 and CH2 separately to sort out that.
How do you think I could do that? Say int(ch1*10)  ?

Now I have to leave to go to work. All excited but can't try this.  :scared:

Will keep watching this thread at work.
 

Offline markone

  • Frequent Contributor
  • **
  • Posts: 367
  • Country: it
Re: DS2072 int(ch1*ch2) math error?
« Reply #5 on: February 08, 2016, 12:57:15 PM »
Also in the second screen the integral start to drift a little already from  the trace start.
Yes, while yellow is still at zero.
Maybe there is some kind of small DC offset in the scope?

Exactly,

try to integrate only CH1 then CH2 separately.
 

Offline markone

  • Frequent Contributor
  • **
  • Posts: 367
  • Country: it
Re: DS2072 int(ch1*ch2) math error?
« Reply #6 on: February 08, 2016, 12:58:47 PM »
Ok, do not have this scope, i cannot replicate, anyway i'm interested to.
 

Offline rich

  • Regular Contributor
  • *
  • Posts: 245
  • Country: gb
Re: DS2072 int(ch1*ch2) math error?
« Reply #7 on: February 08, 2016, 01:18:40 PM »
Is anything in practice truly zero? Ch2 scale is 500V/div, so even 100uV on ch1 would give: -1250V * 100uV = -0.125 and the math scale is viewed at 5/div.
 

Offline Circlotron

  • Frequent Contributor
  • **
  • Posts: 809
  • Country: au
Re: DS2072 int(ch1*ch2) math error?
« Reply #8 on: February 08, 2016, 02:43:41 PM »
Okay, so ran a self calibration after running about 1/2 hour. No improvement.

Try this:
Ch1 at 5v/div and x10.
Ch2 at 5v/div and x10. Leave Ch2 off.
2mS div.
Math = Intg(Ch1*1000)
Math scale = 2v/div.

As you shift Ch1 trace above and below centre the purple integral trace slopes up or down depending on the exact Ch1 offset from centre. Slope is *NOT* proportional to offset.

Not only that, if you turn Ch2 on it changes the integral slope of Ch1 and you can see Ch1 trace move up about 2 pixels. I made a video about that a while back. -->
 

Offline Circlotron

  • Frequent Contributor
  • **
  • Posts: 809
  • Country: au
Re: DS2072 int(ch1*ch2) math error?
« Reply #9 on: February 08, 2016, 03:07:44 PM »

Try this:
Ch1 at 5v/div and x10.
Ch2 at 5v/div and x10. Leave Ch2 off.
2mS div.
Math = Intg(Ch1*1000)
Math scale = 2v/div.

As you shift Ch1 trace above and below centre the purple integral trace slopes up or down depending on the exact Ch1 offset from centre. Slope is *NOT* proportional to offset.

Also just noticed that if you go to average mode the integral slope gets 3 or 4 times steeper.  :-//

Edit -> Going to average mode just moves the position of the line up or down a whisker to the /average/ position. Manually moving the non-averaged line with the shift knob has much the same effect on the integral slope line.
« Last Edit: February 08, 2016, 03:33:34 PM by Circlotron »
 

Offline Circlotron

  • Frequent Contributor
  • **
  • Posts: 809
  • Country: au
Re: DS2072 int(ch1*ch2) math error?
« Reply #10 on: February 08, 2016, 05:02:52 PM »
The more I think about this whole thing the more I reckon it might be just rounding errors - trying to do too much with 8-bit data. Multiply it out too far and it just gets too gritty.
« Last Edit: February 08, 2016, 06:00:54 PM by Circlotron »
 

Offline markone

  • Frequent Contributor
  • **
  • Posts: 367
  • Country: it
Re: DS2072 int(ch1*ch2) math error?
« Reply #11 on: February 08, 2016, 09:01:03 PM »
The more I think about this whole thing the more I reckon it might be just rounding errors - trying to do too much with 8-bit data. Multiply it out too far and it just gets too gritty.

This would explain also what i said before "you can also notice the integral line stopping a little before than CH1 and CH2 reach zero, like the math was limited in precision or out of phase."

Af far as  "As you shift Ch1 trace above and below centre the purple integral trace slopes up or down depending on the exact Ch1 offset from centre. Slope is *NOT* proportional to offset." have you made a video ?
 

Offline alsetalokin4017

  • Super Contributor
  • ***
  • Posts: 1563
  • Country: us
Re: DS2072 int(ch1*ch2) math error?
« Reply #12 on: February 08, 2016, 10:11:44 PM »
Yeah, I've had these worries myself on the 1054z. A pity too because one of the main uses I had intended for this scope is power and energy balance measurements.

But if you want to "prove" that your apparatus under test is a "free energy" device, with a continuously decreasing energy integral... it's ideal.
The easiest person to fool is yourself. -- Richard Feynman
 

Offline markone

  • Frequent Contributor
  • **
  • Posts: 367
  • Country: it
Re: DS2072 int(ch1*ch2) math error?
« Reply #13 on: February 09, 2016, 08:37:48 AM »
Yeah, I've had these worries myself on the 1054z. A pity too because one of the main uses I had intended for this scope is power and energy balance measurements.

But if you want to "prove" that your apparatus under test is a "free energy" device, with a continuously decreasing energy integral... it's ideal.

One thing that i have learned in the DSO era is that if you have to make decent power measurement on power line, power inverters and so on, a digital 8 bit scope (forget virtual additional bits) plus differential probes and current clamps, anyway expensive, is not the way to go, a dedicated instrument is a way better solution. 

Fo instance years ago i had an experience with a quite expensive Lecroy setup, a Wavesufer 8bit (plus virtual) 4CH 500Mhz DSO (13K euro) +  Lecroy CP150  clamp current probes (4000 Euro each )+ Lecroy ADP305 voltage differential (1600 Euro each), after lots of  inconsistent measurement on a 2 channel power inverter (a device working up to 50Khz), we switched  to a 4CH Zes Zimmer power analyzer @ 15K euro that succeeded in its job at first try.

Then the Lecroy kit was used efficiently for some other things, i have a nice memorys for its big touch screen  ;)
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf

 

http://opalkelly.com/