Differentiator help..

[huricaine]

So a notch filter should work then to detect the distortion? Any idea on a design? Do you think active would be the way to go? Would be hz be feasable to do? I myself and many for car audio use midbass to play lower then 80hz because not all subwoofers play good upto 80hz due to a bit of inductance. If the circuit detected harmonic distortion created from 50hz, it would create a safe margin for people wanting to lower the LPF considerably below 80hz without worry for compatability. Would 50hz distortion be created around 100hz? It would be a safe bet that most users do not set their LPF upto 100hz in car...

 

[audioguru]

Symmetrical clipping distortion produces odd harmonics. So 50Hz harmonics would be 150Hz, 250Hz, 350Hz etc.
Un-symmetrical distortion when only the top or bottom is clipped produces even harmonics. So 50Hz produces 100Hz, 200Hz, 300Hz etc.


There are lots of active notch filter projects and circuits in Google.

 

[huricaine]

Great I will google for some rightnow...Do I need to look for a high Q, sharp rolloff filter? And how does the notching filter detect the distortion, by cutting the fundamental out with a very sharp rolloff? Since bandpass filters are said to be the "opposite" of notch filters, will a bandpass filter work too by making my distortion frequency(s) be the only thing that is allowed through? This way, I can use any fundamental frequency below 80hz to detect distortion right? Well actually the lower frequencies would create harmonics somewhere within 80hz and below huh? Would I have to choose a sort of limit like 50-80hz if I used a bandpass filter?

 

[audioguru]

A notch filter removes one frequency which is the fundamental frequency in your case. Then the remaining frequencies are distortion and noise.
The Q of a notch filter determines its bandwidth but you need a high Q so that the 2nd and 3rd harmonics are not reduced. Because it has a very narrow bandwidth (a high Q) the generator frequency must be tuned to the notch filter frequency.

A sub-woofer amplifier has a lowpass filter on its input so you should feed it with a 20Hz to 80Hz sinewave. When it is clipping then it will have many harmonics on its output but with symmetrical clipping the 3rd harmonic is the strongest then the remaining odd harmonics are reduced more and more as their frequency increases.

You can use a bandpass filter to detect the 3rd harmonic but I don't think you can get its Q high enough to remove the fundamental frequency.

 

[WTP Pepper]

It doesn't matter a jot if there's distortion above 20kHz. No one could hear it, even if the tweeter could reproduce it!

Not quite true. If there is enough power in the fundamental and harmonics it could heat and distort the voice coil which could distort the coil former and cause mechanical problems which result in you hearing their effect. Albeit not direct harmonics.

 

[huricaine]

I'm not sure if the average amp symetrically clips or not, it may be pretty close in symetry but isn't it safe to assume that no amp is perfect and would clip slightly un-even? Also, dosn't the peaks get distorted before the peaks clip?

When you say to use 20-80hz were you suggesting that when using a notch filter for a subwoofer amp because it has a low pass filter? Does the low pass somehow effect the filter?

What Q should I be looking to use, 10 maybe?

 

[huricaine]

nvm....

 

[huricaine]

Guys, I built an active differentiator based on this page: https://www.radio-electronics.com/info/circuits/opamp_notch_filter/opamp_notch_filter.php

I used their 50hz fnotch example. Well, lastnight I used it in my car connected to my amp with a scope attached to the output and the results weren't great. I used a voltage divider on the input btw. The little voltage I got was around a few hundred mV to a little over 1v at max volume(vs 17V originally). The voltage kept climbing as volume level did and when in the known clipping region with the volume knob, the voltage was proportional, not a big enough difference(it's not filtering at 50hz at all).

What could I be doing wrong? I checked my circuit(breadboard) and didn't find anything wrong. Also, the waveform seem distorted for sure it was a little loppy for lack of a better word. Also, when hard clipping, the peaks rounded off, they still went flat but the corners were rounded.

I also don't know what part of the circuit is responsible for the Q? Is it th evoltage divider negative feedback?

 

[ronv]

You can't just play music into it. It needs to be a sine wave at the notch frequency.

 

[unclejed613]

i think this is something like what you are looking for. it's a distortion analyzer that works with music as well as single tones... **broken link removed**

the device works by algebraic subtraction of the input and output waveforms.

 

[huricaine]

You can't just play music into it. It needs to be a sine wave at the notch frequency.

No I played a 50hz sinewave...

 

[huricaine]

i think this is something like what you are looking for. it's a distortion analyzer that works with music as well as single tones... **broken link removed**

the device works by algebraic subtraction of the input and output waveforms.

I'll def hae to take a look...

 

[audioguru]

The notch filter circuit uses resistors and capacitors that are matched extremely well. Its opamp is biased at 0V so it needs a positive and negative supply.

 

[huricaine]

A dual supply? I created a dual supply with a 12v dc source, two 4.7k resistors and two 470uF capacitors for a virtual GND...Or do you mean single sided?

I used polyester caps, they have J in them which I forgot what tolorance that is? But my resistors are 1% tolerance.

 

[audioguru]

A virtual supply will not work if it is not done correctly. Actually it is simple to bias an opamp at half the single supply voltage.
The bias resistors can be as high as 100k ohms since an opamp has a very low input bias current. Then the filter capacitor value can be low.
You do not need resistors as accurate as 1% to make a virtual supply, 5% is fine.

J is 5%, K is 10% and M is 20% tolerance for capacitors.

I re-arranged the circuit so that the wires do not cross and I added a virtual supply to bias the opamp at half the supply voltage.
R1 must be exactly the same value as R2. None of the signal resistor values are less than 2k ohms.
The source impedance must be less than 100 ohms and the output load must be greater than 2M ohms.

 

[huricaine]

I forgot about that page saying that the load cannot be less then 2M. I will eventually do something like add a peak detector and send it to a comparator or an opamp or something that saturates at a maximum allowable voltage to goto an LED. Can I utilize 2M ohms at the output of the notch? Before I use an LED and peak detector I want to simply measure the ouput to make sure it's working properly, I will use an oscope to see, should I attatch a 2M resistor on the opamp output to GND then put the scope probe common to the output?

Also, what do you think about the Twin T filter? Should I try and use that for a higher Q?

 

[huricaine]

I just tried the single sided circuit with the same notch network values I used before, I measured upto about 230mV at max volume(before clipping) at 50hz, and only 125mV at 80 and 100hz at the same volume(max before clipping). So does this mean I missed the 50hz mark by a bit? I don't understand, the values should atleast have gotten me close...

Oh and keep in mind that I am still using a voltage divider for the input for the 17Vrms max output from the amp so the little voltage is atleast partly due to that(linearly).

EDIT: I am now going to try a 30hz to get an idea...

 

[huricaine]

Sorry 200mV and 125mV is wrong I was able to get about 1V tops. Well anyway I re-did the test with a sweep(20hz-2khz) and I noticed it the amplitude started high at about 200mV(because the volume was not at max) and I saw it notch down to about 51 or 59mV(cannot remember exactly but it's close enough for the explaination) and it started to pick back up as expected. Well, I figured the frequency of the notch with my scope and the center was about 460hz well, I had about a few secs for it to go from the notch amplitude and rise which means that it probly had a notch stretch at the same amplitude for maybe 430 - 460hz is what I assume from what I recall. 460 was actually around the high side.

But anyway, my smart self forgot to use a 2M load! Can I go from output to gnd with a 2m resistor?

EDIT: I just aded 2M of resistance from output to GND and I got a notch at about 335hz.

EDIT: Dang for some reason I forgot the GND connection to the opamp...It's strange how it still gave a notch or even output? After adding the GND to the pin, I get a distorted waveform so i'm not sure what's wrong. I used polarised capacitors for the power supply, electrolytic. Is that bad?

 

[audioguru]

Please post the exact circuit that you used including resistor and capacitor values, opamp number, power supply voltage and source impedance.
The load is supposed to be MORE than 2 Meg ohms because it is parallel with R4 which must be the same value as R3.

Electrolytic capacitors for the virtual ground biasing are fine. Replace the 1uF capacitor with 10uF.

When the output is distorted then reduce the input level.

 

[huricaine]

I hope you do not mind but I used your schematic and wrote values in it.

A few notes: I have tried exactly 2M on the output, maybe it should change to a bit higher?

For power source I used a 12V remote turn on lead which is probably not he best idea as it cannot deal with too much current, do you know how much current this circuit may draw?

I'm not sure what the gainof the op-amp comes out to at given frequency but with the circuit before when I used a dual supply, I didn't get distortion from the opamp from having the volume as high as the audio amplifier would go before, and into clipping so I figure the opamp is probly not meeting it's limits.

I should probably move from the 12V REM lead to something like a simple 9V batt or something as far as the wave distortion problem goes.