Frequency Response Linearity

[Eugene101]

Hi,

I'm trying to measure frequency response of an audio amp using AD2 and Audio Analyser Suite.

In order to check the equipment I've measured FR diectly connecting W1 to Scope 1 & Scope 2. the picture I've receive is as follows:

Looks like there's some slight non-linearity of the AD2 itself on frequencies above 3 KHz. Can it be so, or am I doing something wrong?

Also, measuring a square wave at the 20KHz shows some imperfections.

Is it the limitation of the scopes or the wavegen?

Thank you!

[attila]

Hi @Eugene101

Is the wiring to the two scope channels completely identical and symmetric, splitter in middle and identical cables ?
If you are using probes at 10X you could try adjusting the trimmers.
The device has two trimmers for each channel for positive and negative inputs. These may need a bit of adjustment for high range, above 5Vpk2pk, 500mV/div. The 2nd and 3rd trimmers are for +, and 1st and 4th for - input.

With short BNC cables + T splitter I see ~0.005dB difference at low and ~0.02dB high range between the channel readings. This may also come from the BNC adapter capacitance difference.

[Eugene101]

Thank you for the reply!

The connection is done through the special board that I've made. It's ok. The difference betwen channels is explainable.

But I see on your screen the same curve that was exactly the subject of my question: why it has a form like this? Shouldn't it be flat? Or the level off this fluctuation is neglectable?

 

Edited April 10 by Eugene101

[attila]

Hi @Eugene101

Ideally it should be flat but nothing is perfect. I think such 0.2dB deviation is acceptable, unnoticeable with common usage.
https://www.ni.com/en/shop/electronic-test-instrumentation/oscilloscopes/what-are-oscilloscopes/3-hidden-oscilloscope-specs-that-really-matter.html
 

If you take relative measurement, like it is done by default in Network Analyzer, the different between channels is lower 0.002-0.05dB, see C2 is relative to C1
For more accurate measurement a close compensation can be used, an initial close loop reference curve R1, see M1 is C1 relative to R1 (earlier saved C1 curve)

[Eugene101]

13 hours ago, attila said:

For more accurate measurement a close compensation can be used, an initial close loop reference curve R1, see M1 is C1 relative to R1 (earlier saved C1 curve)

Thank yo very much, this is what I was looking for.