Problem with FFT on Hantek DSO1200

Thread Starter

skip.ele

Joined Nov 27, 2011
53
I am having trouble believing or understanding what I am seeing on this scope with regards to the FFT feature. I started off trying to measure the noise on my home built amplifier and I have been playing with it for several days now but I'm new to FFT and while I may not fully understand what I am seeing I surely don't believe it. Long story short, I'm now looking at the noise level of the scope itself meaning I'm connecting the probes input to its own ground and looking at the FFT trace on the scope. Right now it is showing a noise floor (average) of -12 dbVrms. Does that sound accurate for a probe shorted to ground? Seems it should be much lower than that. I measured the noise on my amplifier output and the average there is -8 dbVrms. This does not seem nearly low enough either. I was expecting something around -45 dbVrms. The amp may be a little noisier that normal because it has a high gain of x1000 but the dbVrms readings don't match up with my peak to peak signal on the scope channel which is about 4mv peak to peak. I'm using AC coupling on the scope which is supposed to provide better FFT performance. I also noticed that every time I adjust the scope channels volts/division the average dbVrms value of the FFT plot changes drastically. Basically if I increase the volts/division for CH1 from 10mv to 1v the dbVrms reading goes up from -8 dbVrms to .6 dbVrms. Should this be happening?

thanks
skip
 

THE_RB

Joined Feb 11, 2008
5,438
Yeah that's pretty normal.

The scope FFT is math generated from whatever samples the ADC can pick up.

If you input an extremely low signal (like a grounded scope probe) the ADC values are very small, and similar to each other. That gives them a very poor resolution and range, so the math process for detecting the FFT frequencies has a lot of error.

If you connect a sine or other decent strength signal, that gives a large ADC range (difference between hi and lo ADC values) and the math process has much more information to work with. So by the time it has scaled the large signal it can more accurately scale the harmonics and other frequencies on the display.

So basically to get a decent FFT display and dB scaling etc you need a decent sized, real signal.
 

Thread Starter

skip.ele

Joined Nov 27, 2011
53
While I can certainly see what you are saying as true I was hoping to use the FFT to analyze the noise level of the amp. Do I need to have an input signal to do that? What would be the proper approach for noise analysis in this case. I have a single channel amp with a gain of x1000 from two stages. I also have 3rd order high and low pass filters (300 - 10khz) and AGC. The intended audio source is an electret microphone. From input to output to scope configuration what should I be doing?

Thanks for the help.
Skip
 

THE_RB

Joined Feb 11, 2008
5,438
Maybe someone else can speak up? I've never used the FFT to try to measure really small noise from an amp.

I assume you are grounding the amp input, and measuring the noise at the amp output?

If the signal out is only very small (some uV?) you need to crank the gain up on the scope front end (best Vdiv) so you have a good amplitude signal to measure, then hope the FFT can be accurate with that.

Why do you need to find the frequency components of the amp noise?
 

Thread Starter

skip.ele

Joined Nov 27, 2011
53
Well I wanted to get some kind of db comparison of the noise of my amp compared to others. I wanted to get some idea of how good or bad it is. I need to determine if it can be made better, quieter. I am mainly trying to learn as much about this as I can. This project was intended to be a learning experience from the beginning. I had never used FFT before and thought it would be a good opportunity to do that. Plus I wanted to test the frequency response of the amp too especially since I have high and low pass filters. I know you don't have to use FFT to do that but again I thought it would be interesting to see it in the frequency domain.
But also now that I am into this I want to try to determine the quality of my scope for performing FFT. Is there a benchmark that I can use to test the quality of my scopes FFT?

And thanks for sharing your thoughts on this.
skip
 

Thread Starter

skip.ele

Joined Nov 27, 2011
53
First let me say that my scope is a wind-up toy compared to yours. If you dont need it for the next couple of years I'd love to keep it for you. :)

I had already looked at your video a few days ago and it's funny but the one thing I need more information on is the one thing you didnt talk about and that is the actual dbVrms values you were presenting onscreen. What is the reference for the FFT and what was the actual dbVrms noise floor value of the signal presented without the speaker attached. Just looking at the scope with nothing attached. I'm trying to figure out how to interpret the actual values.

thanks for the video and the help.
skip
 

THE_RB

Joined Feb 11, 2008
5,438
Well I wanted to get some kind of db comparison of the noise of my amp compared to others. I wanted to get some idea of how good or bad it is. I need to determine if it can be made better, quieter.
...
The normal way is to just measure the amplitude of the noise signal out of the amp, you can do that with a standard 'scope trace.

So with grounded inputs, if your amp makes 10mV noise on the output it's better than an amp that makes 20mV noise on the output.

There should not be any significant frequency components in the noise apart from some mains frequency, so you can set the 'scope timebase to get general HF noise, and at a slower timebase to see the 50/60Hz component.
 
Top