Measuring Frequency Response - Where to connect o-scope leads?

[Deke609]

Hi all -

I am having a devil of a time figuring out how to use my oscilloscope - specifically, I can't figure out how to set up the leads to measure frequency response (at least I think that's my problem). 

The scope (Picoscope 2204A) is a usb scope with 2 channels (A and B), plus an AWG out (Arbitrary Waveform Generator) through which I can send a frequency sweep signal.

I have 4 available BNC leads - 2 traditional with points/clips on the hot and an alligator clip on the other, and 2 with stackable banana plug connectors.

I am able to send the frequency sweep signal through my Stereomour II - I know this b/c I can hear the sweep through my LCD4s.  I achieved this by connecting the AWG out to the right channel rca input of the amp.  In case this is part of the problem, I'll mention how I made this connection. I made a banana plug to rca adaptor by soldering a red 20 ga wire to the center pin on the rca connector, and a black 20 ga wire to the shield of the connector, and then soldered the same color banana binding posts to the other ends of the wires.  I connected a banana plug lead to the AWG out and plugged its red plug into the post connected to the rca center pin and the black to the one connected to the shield.  Result: frequency sweep plays through right headphone.

Where I think I'm screwing up is with the other connections.  I used just one lead, A, attached to the right speaker outs on the amp: red to red, black to black, with my headphones also plugged into both right and left outs (just in case someone is worried: the headphones can handle the output of the amp).  But when I ran the frequency sweep, 5HZ to 20kHz, 2V amplitude, I got the inverse of what I was expecting: instead of the usual frequency response graph showing dB roll-off at lower frequencies ( like a y = log x graph), I got a graph showing higher (but negative) dBs at very low frequencies that precipitously drops off and levels out at a much lower constant across the remaining frequencies (like a y = log 1/2x graph, but with a much steeper initial drop-off). 

I'd be very appreciative if someone could instruct me on how/where to connect the o-scope leads to the amp! Or any other pointers. I am completely bewildered!

Many thanks in advance,

Derek

[Paul Birkeland]

The first thing you want to do is to load the amp with a resistor.  This will just take out a variable that you have, even if it doesn't appear to be the issue.

Can you post a photo of what the scope returned to you? 

When I do FR graphs, I do periodic noise and an average over at least a few seconds. 

The other thing that I do every time I do FFT work is to plug the output of my device to the input, then sweep the device itself to make sure everything is working properly.

[Deke609]

Thanks again, Paul.

The first thing you want to do is to load the amp with a resistor.  This will just take out a variable that you have, even if it doesn't appear to be the issue.

I will be putting together some pairs of 50W switchable dummy loads tomorrow evening (2/4 ohms and 8/16 ohms). I will try again tomorrow with the 2 ohm load to match the present 2 ohm OT configuration.

Can you post a photo of what the scope returned to you? 

Attached is one of the latest I ran - this one shows a more gradual negative slope, with some odd periodic peaks (first time I've seen that). The frequency scale is off - I am able to zoom-in with the software, but this is how the software saved the image of the sweep.

But what's up with the -ve dBs?  Is the Y-axis scale inverted?

(And just to be clear, for anyone else who might read this, I am not problem-solving an issue with the amp. The amp is stellar - I am just trying to figure out how the scope works so I can play around with loading, damping, distortion, and other fun stuff once the amp is more fully burned in.)

When I do FR graphs, I do periodic noise and an average over at least a few seconds. 

I will look up "periodic noise".  I am taking the average values - when I switch to magnitude or peak, the line becomes ragged.  By "over a few seconds" do you mean the entire sweep over a few seconds, or each step (e.g., 1 Hz increment) for a few seconds? I've set the sweep to do 100mHz per ms.  Is that too fast? I figured it was slow, since the default increment of the scope was 10 Hz or more per ms.

The other thing that I do every time I do FFT work is to plug the output of my device to the input, then sweep the device itself to make sure everything is working properly.

I will try this tomorrow.

I am also going to look into my soundcard settings -- maybe there's a bass-boost setting that's been turned on or I've disabled something.

thanks again!

Derek

[Paul Birkeland]

The graph you posted is essentially a graph of nothing.

-82dBV is like 80uV or so RMS.  This is probably the noise floor of the scope.

[Deke609]

Thanks Paul - this at least confirms that I am not measuring what the amp is actually doing (It is definitely amplifying the sweep).  I suspect I'm not hooking up the leads properly.  I've tried to find a tutorial online dealing with amps - but all the ones I've found presume that one already knows how to use the leads.  Having used a multimeter, I assumed that I did too, until I tried to use the scope.  I will try some simpler measurements tomorrow to see if I can figure it out.

[Paul Joppa]

I'll expand on the negative dBs a bit. Decibels are a measure of a ratio, not a voltage or a current or a sound pressure by itself. If you look at the scope display you posted you will see that it is displaying "dBV" which is a widely-used shorthand for dB relative to 1.0 volt. (It's probably 1.0 volt rms, though that's not explicit.) Since dB is logarithmic, any ratio less than one produces negative dBs, 1.0v is zero dB, and voltages greater than 1.0 are positive dBs. Thus for example, -60dBV is 0.001 volt or 1.0 millivolt.

[Deke609]

Many thanks Paul J. It never occurred to me that a 1:1 ratio would be reflected as 0, and that anything less would be -ve. That clears up the -ve dBs for me.  It also points me in the direction of the problem.  If I'm seeing -80 dBs I'm effectively not measuring any voltage -- literally, as PB wrote, "a graph of nothing" -- in a situation where I know there is measurable voltage being outputted. 

I tracked down an "Oscilliscope Basics" guide by TEK that Doc had linked to in an older post, and it has some helpful pointers about using the probes.  First and foremost, it emphasized the importance of good grounding.  It also suggested to me that I should be using the traditional style and not banana plug terminated probes for taking the measurements b/c they are likely more accurate (although accuracy is not the main problem I am having - any measurement, however wildly inaccurate, would be an improvement over measuring nothing )

On the upside: there is nothing like hands-on, trial-and-error troubleshooting for learning. I don't known when it happened (I'm only pushing 50), but it seems at some point in the aging process I lost the ability to learn by any means other than making mistakes.  And come to think of it, in my 20s I was pretty much incapable of learning from my mistakes.  So I'll chalk up this change as an improvement.

[Paul Birkeland]

When I was first dinking around with scopes, it was really helpful to have someone who was familiar with their operation standing next to me to help out. 

A lot of the newer scopes have calibration lugs that put out something like 1V at 5kHz for properly compensating probes.  If you have these, you should see the waveform while adjusting the probe compensation.

Otherwise, get your phone out and download a signal generator app, generate a 60Hz sine wave at maximum level, then use an 1/8" mini to RCA cable to make that waveform accessible to your scope.  If you have a a scope probe that's two alligator clips, the black one to the shell of RCA jack and the red one to the center pin should get you your signal.  Your scope should have a button to automatically acquire the signal (not in frequency/amplitude mode, but rather in amplitude/time).

[Doc B.]

Yes, go at this methodically. First, as PB says, just put in a sine wave and measure voltage amplitude over time. That's the easiest thing to measure. Don't worry about a frequency sweep yet. Second, as PB suggested, take the amp out of the equation. Just get the tone from the output of the scope to the input of the scope through the test leads.

The issue can be the setting of the scope rather than the connections. Generate a 1KHz tone at 1V pk-pk output. That should give you a textbook looking sine wave display if your X and Y scales are set correctly.

[Deke609]

Thanks PB and Doc

I will take the patient and methodical approach.

Derek

[Deke609]

Just a quick update.

The advice to take the amp out of the equation and just measure directly from the scope's signal generator output was a great call.  It has helped me to further narrow down the problem.

No problem generating and viewing sine waves, square waves etc.  Generated a 1 kHz square wave, set the trigger to repeating increasing slope, adjusted the frequency axis to display the complete square and was able to adjust the compensation trimmer to get a perfect square. Sine waves look perfectly good too after trimming.

Then for kicks, I generated a sine wave frequency sweep, still in "scope view".  No problem. Goes from long waves to increasingly short waves.  ... BUT, as soon as I switch to "spectrum view", the same -80ish dB plot shows up. 

I'm going to call the manufacturer tomorrow for help.

Thanks to all for your assistance.  I'll figure this thing out eventually!

The nice thing about taking the amp out the equation is that I can listen to it while I alternately plead with and curse the scope 

Derek

[troplin]

Derek I think there might be a misunderstanding on what the spectrum view actually is.
Usually, the spectrum view on a scope shows the current spectrum (measured over a short period), i.e. while the sweep is ongoing, there should be a spike wandering from low to high frequency or vice versa.
This is not the same as the frequency response, which you might be expecting.

[Deke609]

Derek I think there might be a misunderstanding on what the spectrum view actually is.
Usually, the spectrum view on a scope shows the current spectrum (measured over a short period), i.e. while the sweep is ongoing, there should be a spike wandering from low to high frequency or vice versa.
This is not the same as the frequency response, which you might be expecting.

Many thanks Tobias - This is much appreciated insight. I do see a moving spike when the measurement is set to "magnitude". My assumption was that this moving spike was capturing the info I wanted: dBs x frequency, and that I could generate a linear frequency response plot by selecting "hold peak" instead of "magnitude".  I'm guessing that I'm wrong about this?

Derek

[troplin]

I'm not an expert with scopes, but I think you're probably right that the "hold peak" function should do that.
Maybe you are measuring the minimum peak instead of maximum?

IMHO a scope is not quite the best tool for this job. Just use your PC sound-card and a audio measurement application (I'm using "Room EQ Wizard"). With that setup the application is in control of both, signal generation and measuring at the same time.
This has the advantage that the application knows the currently generated signal at any time and this enables a whole bunch of additional measurements like phase shift, distortion, decay and so on.
Additionally you can just use normal audio cables and don't have to fiddle with leads.

[Deke609]

Thanks Tobias -

I think the "hold peak" is in fact the way to generate a frequency response plot with the scope I have. It looks like the scope/software is acting up.  I discovered that repeated clicks on the "auto setup" button eventually persuades the scope/software to plot something that looks like a frequency response curve.  PB was correct - the plot I previously attached was simply capturing the frequency noise floor of the scope itself.  I verified this by leaving the probes unconnected to the scope and selecting "spectrum view": exact same plot.

I an starting to doubt whether this $140 do-it-all scope can really do the work of what usually costs upwards of $2K (minimum) in equipment.  Go figure! 

Problems I've identified so far:

(1) channel "cross-talk" (I think that's what it is called?): signals incoming or outgoing from the AWG or a channel are picked up by the other channel -- example: when channel A is directly connected to the AWG out, but nothing is connected to channel B, B shows a noisy signal plot similar to and slightly weaker than the AWG Out and channel A -- I haven't popped open the case of the scope yet, but I suspect that the bnc ports and any associated PCB processors are not (sufficiently) shielded; and

(2) when connecting the AWG Out directly to the input of the scope and running a 5Hz to 20kHz frequency sweep, the resulting plot using "peak hold" is wonky -- showing higher dB's from 0 Hz to about 20 Hz (the first five Hz of which are not even generated by the sweep I've chosen), then leveling to a flat plot until 20 kHz when the plot drops off as expected -- the dB uptick at super low frequencies only disappears if I let the sweep run for more than 1/2 hour;

Because of the cross-talk issue, I can't be certain how much of the signal I'm measuring is from the target versus the source.

I am going to go back to the drawing board and ask Doc, PB and PJ about what measurements I need to look at distortion, loading/damping, etc. and the desired accuracy levels of same, and then figure out my best course of action in terms of equipment.  I am as much interested in the theory and playing around with the measurements and scope as I am with fine tuning the SII to match my headphones -- and so am willing to invest a bit in getting some equipment that is up to the task so that I can continue learning about the behavior of analog audio circuits. I am slowly (very slowly) starting to work my way through Morgan Jones' text, "Valve Amplifiers" and a few others -- with frequent internet search excursions to re-learn lots of math stuff that I forgot decades ago. 

Thanks for the software recommendation. I looked into Room EQ Wizard - impressive set of functions.  But I do not have a loudspeaker system - I'm using the SII with Audeze LCD4 headphones, so I don't think the mic approach will work.  Not 100% sure of that though.  Maybe I could directly plug the output of the amp into my laptop.  But I think I'd like to go the traditional testing equipment route, provided it's not prohibitively expensive, so that in the future I can poke around with different components and sub-assemblies to see what they are doing -- e.g., low pass and high pass filters (something I've just been reading about and am fascinated by).

For now I am going to take a break from trying to scope the amp and turn my attention to finishing the base: I am going to try to clad it in 30 ga copper sheet. Wish me luck!

cheers and thanks,

Derek

[edited for clarity]