The concept is that if you carefully synchronise in at least a number of areas then you can subtract two captures from each other of a channel, this being the DUT plus acquisition card, handling a particular source signal - and show differences occurred if some aspect of the setup was altered? But then how do you determine whether the variation was purely due to the DUT altering its behaviour - or whether it was at least in part due to variation elsewhere in the setup, say within the acquisition card?
Last edited:
Just acquire the digital output at least a couple of times without changing anything you can have control of and compare the two (or more signals).
If any differences between them is within the variation threshold you chose for your purposes, then you can surely state the whole channel (that is DUT + acquisition card) is enough time invariant for your purposes, and this implies that surely the acquisition card also behaves in this way.
At this point you can change whatever you wish in the DUT (cables, power cords, position, cover, bulb lights around you, etc...) and be reasonably sure that the differences you see (if any), one more time getting the output channel at least two times (just to allow each masure confirms each other), are due to how you operated over the DUT.
To help in this.the measure accepts a precision threshold which will set the exactness with which it will reconstruct both the reference and the output digital signals (see chapter 4 and chapter 6, par. 6.2).
As I'm a hobbyst, parameters like humidity, pressure and similar things were beyond my control and you could anyway state that there always be out of control parameters, no matter how hardly you try, but once again having got the output two or more time without changing anything under your control can help.
If you have a look at the graph at page 6 (66 as absolute value) of chapter 5, where two acquisitions were made without changing anything in the whole channel (DUT+acquisition card+cables+position+etc...) you see that the maximum channel variance, at roughly 6,5KHz (which is simply the distance between the two curves) is roughly 1/30 of the space from 0.06 and 0.08 dB, that is roughly 0.0007 dB.
I then infered that the acquisition card variance couldn't be higher than this value, being free to decide if this threshold was good enough for my purposes.
I hope this clarifies.
By the way, my name is Daniele (better than dabudabu
)
Okay, Daniele it is, !
As you say, time invariance is at least part of the controls that need to be imposed. Along those lines, what sort of results did you see, or have you in fact tried such, between a relatively cold start for the channel, say 5 minutes after turn-on; compared with, say, several hours later, after exercising the channel to a lesser or greater degree?
Personally, I wouldn't be happy to say I would "be reasonably sure that the differences you see" are due only to the nominal DUT, I would include the acquisition card in, as being part of the DUT, until I managed to convince myself, by one means or another, that its behaviour was relatively invariant after the particular change. I agree in most cases it probably will not be a factor, but Murphy's Law will probably catch you out - the really key measurement you make turns out to be corrupted, because the capturing device misbehaves in an unexpected way,.
!As you say, time invariance is at least part of the controls that need to be imposed. Along those lines, what sort of results did you see, or have you in fact tried such, between a relatively cold start for the channel, say 5 minutes after turn-on; compared with, say, several hours later, after exercising the channel to a lesser or greater degree?
Personally, I wouldn't be happy to say I would "be reasonably sure that the differences you see" are due only to the nominal DUT, I would include the acquisition card in, as being part of the DUT, until I managed to convince myself, by one means or another, that its behaviour was relatively invariant after the particular change. I agree in most cases it probably will not be a factor, but Murphy's Law will probably catch you out - the really key measurement you make turns out to be corrupted, because the capturing device misbehaves in an unexpected way,
.
Daniele, maybe a video demonstration of this would be a good thing to have a look at?
Any chance you could post something on Youtube?
Equations and graphs are all well and good as supporting documents, but visual demonstration might make this test approach less confusing to those here who are interested. Even some photos or an animation would be great.
I realize English is not your native language so just a suggestion......
You should use a translation program as some of your writing is hard to follow and creates confusion when using certain terminology and that can easily be misconstrued or misinterpreted.
I would really like to see any example or description of the test equipment used to "acquire" the multiple output captures as well as the material and test equipment being used to inject source.
Even a simple block diagram of the test setup with descriptions would be invaluable.
Any chance you could post something on Youtube?
Equations and graphs are all well and good as supporting documents, but visual demonstration might make this test approach less confusing to those here who are interested. Even some photos or an animation would be great.
I realize English is not your native language so just a suggestion......
You should use a translation program as some of your writing is hard to follow and creates confusion when using certain terminology and that can easily be misconstrued or misinterpreted.
I would really like to see any example or description of the test equipment used to "acquire" the multiple output captures as well as the material and test equipment being used to inject source.
Even a simple block diagram of the test setup with descriptions would be invaluable.
Thank you for your suggestions!Daniele, maybe a video demonstration of this would be a good thing to have a look at?
Any chance you could post something on Youtube?
Equations and graphs are all well and good as supporting documents, but visual demonstration might make this test approach less confusing to those here who are interested. Even some photos or an animation would be great.
I realize English is not your native language so just a suggestion......
You should use a translation program as some of your writing is hard to follow and creates confusion when using certain terminology and that can easily be misconstrued or misinterpreted.
I would really like to see any example or description of the test equipment used to "acquire" the multiple output captures as well as the material and test equipment being used to inject source.
Even a simple block diagram of the test setup with descriptions would be invaluable.
As at the moment I haven't anything similar, probably it will be easier to just draw a diagram with some explanations. Anyway this is not a 5 minutes tasks, so please give me some time for it.
In the mean time I'll be happy to better explain (trying to use a better English
)You are right and this is a test that surely worth's trying.Okay, Daniele it is,
As you say, time invariance is at least part of the controls that need to be imposed. Along those lines, what sort of results did you see, or have you in fact tried such, between a relatively cold start for the channel, say 5 minutes after turn-on; compared with, say, several hours later, after exercising the channel to a lesser or greater degree?
Personally, I wouldn't be happy to say I would "be reasonably sure that the differences you see" are due only to the nominal DUT, I would include the acquisition card in, as being part of the DUT, until I managed to convince myself, by one means or another, that its behaviour was relatively invariant after the particular change. I agree in most cases it probably will not be a factor, but Murphy's Law will probably catch you out - the really key measurement you make turns out to be corrupted, because the capturing device misbehaves in an unexpected way,
At the moment I'm not really focused on additional experiments with the measure (my family and my job come first...).
My hope is that, also sharing the Octave framework I developed to support it, some other people will also take part in experimenting this as well as other topics.
I know this will not be easy as the framework is no more than a collection of Octave scripts to be commented and explained one by one, but enthusiasm could make the difference
A bit of a coincidence, but I also just recently started exploring the using of Octave to dissect waveforms, I have a blog entry mentioning this. Note, I'm completely new to Octave and that category of software, so I'm feeling my way in at the moment; your efforts may not instantly make sense to me, , but I would certainly be interested in looking at them ...
, but I would certainly be interested in looking at them ... - Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.