I tried upping the gain for my LM3886 based amp, so that it could be used to power my bass driver and could compensate for the dipole rolloff,
I used a 10k input resistor and a 1M feedback resistor in parallel with a 4.7nF cap. The result is a very loud hum and 10V DC offset (I tested on a speaker and used a cap on the output, just in case there was too much DC).
What is the cause do you think? Could it be the cap in parallel? Over a certain frequency the gain is under 10 and maybe it causes it to be unstable, I'm not sure, just thinking.
Is there any other way to incorperate this 1st order low pass filter easily (minimal component count)? I'm trying to avoid an extra active device or a PLLXO (loss of gain, which is already being pushed).
I used a 10k input resistor and a 1M feedback resistor in parallel with a 4.7nF cap. The result is a very loud hum and 10V DC offset (I tested on a speaker and used a cap on the output, just in case there was too much DC).
What is the cause do you think? Could it be the cap in parallel? Over a certain frequency the gain is under 10 and maybe it causes it to be unstable, I'm not sure, just thinking.
Is there any other way to incorperate this 1st order low pass filter easily (minimal component count)? I'm trying to avoid an extra active device or a PLLXO (loss of gain, which is already being pushed).
Basically, it is the attenuation of the feedback network at the frequency where the gain around the loop drops to unity that matters for the stability. With the values you use, the attenuation factor of the feedback network has dropped to unity long before the loopgain drops to unity, and high frequency oscillation will almost certainly occur. As the oscillation amplitude tends to grow until some non-linear limiting effect limits it, it may well result in large DC offset, loud hum, and hearing damage for any bats living nearby.
If it is an inverting amplifier, you could split the 10kohm resistor in two equal parts and add a capacitor to ground at the node where they are connected (with a value C=1/(2*pi*f*Rpar), where Rpar is the parallel value of the two parts (2.5kohm)).
If it is an inverting amplifier, you could split the 10kohm resistor in two equal parts and add a capacitor to ground at the node where they are connected (with a value C=1/(2*pi*f*Rpar), where Rpar is the parallel value of the two parts (2.5kohm)).
I was thinking of doing just that (almost), but why would it be Rpar? Wouldn't I just use 5k? Also, I had trouble understanding why it wouldn't work to leave the 10k resistor and place a capacitor to ground after it, I guess I should read the 464 pages I printed out of that opamp design reference. Thanks, I'll do what you said, and while I do that I'll try to understand why it's the paralleled value, I'm sure it'll come to me.
Due to the negative feedback, the signal voltage at the inverting input of an inverting amplifier is very small, at least at low signal frequencies where the loop gain is high. Therefore, the inverting input has an almost zero impedance to ground (it is a so-called virtual ground). A capacitor between the virtual ground and the real ground has either virtually no influence because there is no signal voltage across it anyway, or it causes high-frequency oscillations by changing the phase shift at the frequency where loop gain drops to unity too much.
As for the parallel value, if you write out the equations for a voltage source driving an R-C-R network into a zero ohm load, you find that the cut-off frequency is determined by the parallel value.
Anyway, I'm sure you'll find most of these things explained in detail in your 464 pages op-amp reference.
As for the parallel value, if you write out the equations for a voltage source driving an R-C-R network into a zero ohm load, you find that the cut-off frequency is determined by the parallel value.
Anyway, I'm sure you'll find most of these things explained in detail in your 464 pages op-amp reference.
Ah, I see now, thanks for the explanation.
That 464 page referance was the Op Amps For Everyone from TI that everyone was talking about, time to read it I guess.
Thanks again.
That 464 page referance was the Op Amps For Everyone from TI that everyone was talking about, time to read it I guess.
Thanks again.
Splitting the input resistor won't work if there is a pot on the input (which there is) now will it? The pot would change the frequency response of the filter wouldn't it? Hmmm... I have to find another way of doing this then...
Yes, unfortunately you are right...
Maybe one way to reduce the influence of the potmeter would be to connect an RC series network consisting of the 4.7nF you were intending to use and a 47 kohm series resistor across the 1Mohm feedback resistor. Then split the 10kohm in two times 5kohm and put a 82nF capacitor from the midpoint to ground (ideally Ctoground=47kohm*4.7nF/2.5kohm=88.36nF).
Most of the filtering is now done by the 4.7nF, which is not affected by the potmeter. The 82nF which is affected by the potmeter only determines the stop band response for -26dB and more attenuation. For high frequencies, the feedback network attenuation is determined by 47kohm//1Mohm and 5kohm, giving an attenuation greater than ten. Therefore, stability should be no problem.
Maybe one way to reduce the influence of the potmeter would be to connect an RC series network consisting of the 4.7nF you were intending to use and a 47 kohm series resistor across the 1Mohm feedback resistor. Then split the 10kohm in two times 5kohm and put a 82nF capacitor from the midpoint to ground (ideally Ctoground=47kohm*4.7nF/2.5kohm=88.36nF).
Most of the filtering is now done by the 4.7nF, which is not affected by the potmeter. The 82nF which is affected by the potmeter only determines the stop band response for -26dB and more attenuation. For high frequencies, the feedback network attenuation is determined by 47kohm//1Mohm and 5kohm, giving an attenuation greater than ten. Therefore, stability should be no problem.
MarcelvdG I really should thank you for all your help so far.
I think that -26dB would be quite alright and I could do without the extra resistor and capacitor on the input (at least give it a try and see first). But how did you come up with the 47k value? I ask because I understand how it would work, the gain would never hit unity, and would always stay above a certain amount because of the fixed resistor in series with the cap. But if I wanted to keep a 10k input resistor, and use a 1.1M feedback resistor, what resistor/cap combination would be required to have say a -3dB point of 30Hz while not resulting in a gain so low as to induce instability?
If you or anyone know of a website explaining such a type of circuit it would be greatly apreciated, thanks again.
I think that -26dB would be quite alright and I could do without the extra resistor and capacitor on the input (at least give it a try and see first). But how did you come up with the 47k value? I ask because I understand how it would work, the gain would never hit unity, and would always stay above a certain amount because of the fixed resistor in series with the cap. But if I wanted to keep a 10k input resistor, and use a 1.1M feedback resistor, what resistor/cap combination would be required to have say a -3dB point of 30Hz while not resulting in a gain so low as to induce instability?
If you or anyone know of a website explaining such a type of circuit it would be greatly apreciated, thanks again.
Because of your remark about the gain being less than ten in the first post, I assumed that the LM3886 is compensated for gains greater than or equal to ten. This means that at high frequencies, the feedback network should attenuate about ten times or more (the attenuation can be less at low frequencies). With 5kohm and a relatively big capacitor from the negative input to ground, you then need 45kohm from the output to the negative input. With 47kohm and 1Mohm (both E12 values), you are pretty close to 45kohm.
By the way, a more accurate response can be obtained if you split the 10kohm resistor into unequal parts. For example:
8.87kohm from potmeter to the capacitor to ground,
47nF capacitor to ground,
1.13 kohm from the capacitor to the negative input,
between the negative input and output 10kohm in series with 4.7nF in parallel with 1Mohm.
Now the parallel resistance determining the cut-off frequency of the 47nF capacitor varies between about 1kohm and 1.13kohm when the potmeter output resistance varies between zero and infinity.
By the way, a more accurate response can be obtained if you split the 10kohm resistor into unequal parts. For example:
8.87kohm from potmeter to the capacitor to ground,
47nF capacitor to ground,
1.13 kohm from the capacitor to the negative input,
between the negative input and output 10kohm in series with 4.7nF in parallel with 1Mohm.
Now the parallel resistance determining the cut-off frequency of the 47nF capacitor varies between about 1kohm and 1.13kohm when the potmeter output resistance varies between zero and infinity.
- Status
- Not open for further replies.