DC coupling vs AC coupling in old oscilloscope.

Thread Starter

FDanielAS

Joined Oct 17, 2022
1
I've been measuring the output of an old ramp generaton with an not so old oscilloscope.
As far as I know, AC coupling "removes" DC components of the signal measured so you only see transients, and the signal I measured had a pretty small DC component, so I tried to use this coupling to "remove it", but then I got what you can see in the pictures attached.
I read the manual so I know that its AC coupling doesn't work at frequencies lower than 20KHz, so I know now why it didn't work, but still curious, why did this happen? or what shall I interpret from this readings?

311186351_988473198766090_5437681919132414727_n.jpg
262848219_824381638687062_8770839048755982766_n.jpg
 

WBahn

Joined Mar 31, 2012
27,933
I've been measuring the output of an old ramp generaton with an not so old oscilloscope.
As far as I know, AC coupling "removes" DC components of the signal measured so you only see transients, and the signal I measured had a pretty small DC component, so I tried to use this coupling to "remove it", but then I got what you can see in the pictures attached.
I read the manual so I know that its AC coupling doesn't work at frequencies lower than 20KHz, so I know now why it didn't work, but still curious, why did this happen? or what shall I interpret from this readings?
I'm not sure I'm following. You say that you know why it didn't work and then ask why it didn't work.

The scope's front end can't just eliminate a DC component. Think about this. How would it be possible to eliminate the 0 Hz component but not the 0.000000001 Hz component?

What the scope does in capacitively couple the input, which results in the input signal being put through a low-pass filter. That filter has a particular cutoff frequency such that frequency components in a signal significantly about the cutoff frequency pass through unchanged and frequency components significantly below it are heavily attenuated. I don't know the details of that Tek's filter, so I don't know if 20 kHz is the cutoff frequency, or if it is a frequency above which they are saying that there is no significant attenuation (meaning that the actual cutoff frequency might be quite a bit lower, perhaps somewhere between 2 kHz and 10 kHz).

In your waveform, the only portion of it that has significant high-frequency content is the falling edge, that is why you get the sharp downward spike at that point.
 

MrChips

Joined Oct 2, 2009
27,711
The input is coupled via a series capacitor which creates a CR high pass filter. This is also a differentiator, i.e. it produces the derivate of the input signal. That is what you observe in the second screen shot.
 

tautech

Joined Oct 8, 2019
307
I've been measuring the output of an old ramp generaton with an not so old oscilloscope.
As far as I know, AC coupling "removes" DC components of the signal measured so you only see transients, and the signal I measured had a pretty small DC component, so I tried to use this coupling to "remove it", but then I got what you can see in the pictures attached.
I read the manual so I know that its AC coupling doesn't work at frequencies lower than 20KHz, so I know now why it didn't work, but still curious, why did this happen? or what shall I interpret from this readings?
Hardly, AC input coupling removes the DC component and places the 0V reference position midway on the signals amplitude.
We can still take many meaningful measurements excepting absolute amplitude WRT 0V.
IMO your scope is broken based on the attached screenshots from a cheap 100 MHz DSO.
 

Attachments

sarahMCML

Joined May 11, 2019
201
Hardly, AC input coupling removes the DC component and places the 0V reference position midway on the signals amplitude.
We can still take many meaningful measurements excepting absolute amplitude WRT 0V.
IMO your scope is broken based on the attached screenshots from a cheap 100 MHz DSO.
He's looking at a 1 second ramp on AC coupling, It's pretty much bound to do that!
 

MrChips

Joined Oct 2, 2009
27,711
I fired up my Tektronix TDS220 to find out.
Yes, it shows a proper square wave down to 0.1Hz (the lowest on my function generator).

Thus, the AC signal coupling capacitor must be bad.

Edit: I forgot to switch the input coupling to AC. Ignore the above comments.
 

crutschow

Joined Mar 14, 2008
31,144
I agree that the AC capacitor seems to be bad.

Can you check with a signal generator to see at what sinewave frequency the display starts to show a drop in amplitude?
 

MrChips

Joined Oct 2, 2009
27,711
I made a mistake with my measurements. I forgot to switch the scope input to AC. duh!
The -3dB point on the AC input is about 10Hz.

Use a sine wave input signal and measure the low frequency limit on AC coupling.
 

crutschow

Joined Mar 14, 2008
31,144
Keith is right.
The AC output does look correct for such a low frequency (1Hz) ramp.
Below is the simulation for a 10Hz rolloff AC coupling as determined by MrChips, which is very close that observed by the TS:

1666110274132.png
 

WBahn

Joined Mar 31, 2012
27,933
Hardly, AC input coupling removes the DC component and places the 0V reference position midway on the signals amplitude.
We can still take many meaningful measurements excepting absolute amplitude WRT 0V.
IMO your scope is broken based on the attached screenshots from a cheap 100 MHz DSO.
With a 1 Hz waveform, the only way that I can see to only remove the DC component would be to digitally capture the DC-coupled waveform and then arithmetically compute the average and subtract it. But that would limit your resolution to what you can capture at the full signal of the waveform, and often the reason for AC coupling is to be able to get the resolution appropriate to the small signal riding on top of a large DC offset, which brings it back to AC-coupling via a capacitor, which means that there is going to be a finite cutoff frequency and information content below that will be lost in the filter.
 

crutschow

Joined Mar 14, 2008
31,144
You can get a close rendition of the 1Hz ramp signal with capacitive coupling if it has a 10mHz roll-off, but that would have a 17s time-constant for settling to the average DC on the input (taking about 85s to reach steady-state), which would be normally unacceptable.
 

WBahn

Joined Mar 31, 2012
27,933
I read the manual so I know that its AC coupling doesn't work at frequencies lower than 20KHz, so I know now why it didn't work, but still curious, why did this happen? or what shall I interpret from this readings?
@FDanielAS: Could you go back and confirm that 20 kHz figure? That's not making much sense.

Scopes are routinely used to AC couple line frequencies which are 50 Hz to 60 Hz. That would imply a cutoff frequency no higher than about 10 Hz. These measurements would be made with the high-impedance input, which is normally 1 MΩ. That would put the cutoff frequency when using the 50 Ω input at 20,000 times that, or about 200 kHz.

So check and make sure that what you were looking at didn't say that you won't get accurate waveforms for signals below 200 kHz when using the 50 Ω input.

It might be worth posting a link to the manual and indicating on what page you found the information.
 

WBahn

Joined Mar 31, 2012
27,933
You can get a close rendition of the 1Hz ramp signal with capacitive coupling if it has a 10mHz roll-off, but that would have a 17s time-constant for settling to the average DC on the input (taking about 85s to reach steady-state), which would be normally unacceptable.
Sure, but so what? We could get a close rendition of a 0.001 Hz waveform with capacitive coupling if we have a 10 µHz roll-off. No scope (at least not a commercially-produced scope) is going to do either of these. The smallest cutoff frequency that makes sense is going to be in the 1 Hz to 10 Hz range. Even the 1 Hz range is likely going to result in objectionable settling times, so most are going to be ~10 Hz. If you go above 10 Hz, then you start impacting captures of powerline singles (or, more to the point, powerline noise riding on other signals). These are a very common point of interest, so you want to be five time constants lower than 50 Hz, or 10 Hz at the the most.

I'm willing to bet that the cutoff on his scope is 10 Hz for 1 MΩ input and 200 kHz for 50 Ω input. I have no idea where the 20 kHz number is coming from, but I'm taking it with a grain of salt until I see it in a spec sheet or user manual.
 

Kjeldgaard

Joined Apr 7, 2016
475
I have a Tektronix TDS 2014, so after some searching I found:

TDS1000B/2000B Series Oscilloscope User Manual:
  • Lower Frequency Limit, AC Coupled:
    • ≤ 10 Hz at BNC
    • ≤ 1 Hz when using a 10X passive probe

I also think it fits with what I experience with AC coupled measurements with the instrument.
 

crutschow

Joined Mar 14, 2008
31,144
Sure, but so what?
No need to get snippy.

My "what" was just to make the point that the TS's waveform could be reproduced by passing through a high-pass filter with a low enough corner, even though impractical to do in a scope, based upon your statement below, nothing more.
With a 1 Hz waveform, the only way that I can see to only remove the DC component would be to digitally capture the DC-coupled waveform and then arithmetically compute the average and subtract it.
But if you happened to have a small-amplitude, low-frequency waveform ridding on a large value DC that you want to see, I think that using a simple external RC circuit would be a simpler way to do it than the digital approach you mentioned.
 

crutschow

Joined Mar 14, 2008
31,144
TDS1000B/2000B Series Oscilloscope User Manual:
  • Lower Frequency Limit, AC Coupled:
    • ≤ 10 Hz at BNC
    • ≤ 1 Hz when using a 10X passive probe
That reduction in the corner frequency with the 10X probe makes sense, since it increases the resistance the blocking capacitor sees from 1meg to 10meg.
 

WBahn

Joined Mar 31, 2012
27,933
No need to get snippy.

My "what" was just to make the point that the TS's waveform could be reproduced by passing through a high-pass filter with a low enough corner, even though impractical to do in a scope, based upon your statement below, nothing more.
But if you happened to have a small-amplitude, low-frequency waveform ridding on a large value DC that you want to see, I think that using a simple external RC circuit would be a simpler way to do it than the digital approach you mentioned.
The context of the discussion is about what is going on inside an oscilloscope, not what can be done with some external circuit.
 

WBahn

Joined Mar 31, 2012
27,933
I have a Tektronix TDS 2014, so after some searching I found:

TDS1000B/2000B Series Oscilloscope User Manual:
  • Lower Frequency Limit, AC Coupled:
    • ≤ 10 Hz at BNC
    • ≤ 1 Hz when using a 10X passive probe

I also think it fits with what I experience with AC coupled measurements with the instrument.
That jives with I've said earlier -- 10 Hz with the 1 MΩ input. So that would be 1 Hz with a 10 MΩ input and 200 kHz with the 50 Ω input. I hope the TS will confirm what we found (or thinks he found) in the documentation.
 
Top