Hi all,
I find it not so simple to remove a DC component between op-amp stages. What am I missing?
The standard solution is a simple RC filter, say 1 uF/47k from op amp out to op amp in. But I found that regardless of what type cap I use or what combination of R and C to get similar turnover frequency, it seems to reduce dynamics by a wide margin: drum and especially bass slam is greatly reduced. This applies even if I put an op amp buffer before or after the filter, or both. The only solution I found that preserves the dynamics is to correct for the DC through a separate op amp stage. Here too I am confused because there are many topologies to choose from but I couldn't find a definive reference to the ins and outs of DC servos.
I have more caps in the signal path downstream so the goal is not necessarily to eliminate any cap in the signal path... although if it's elegant, then, sure... but just to do it right...
Specific questions:
1) RC filters:
1.1: what's the problem? Does the source component have trouble charging the cap instantaneously (current output limitations)? ... but it will not change much even is the source is a buffer with no other load.
1.2: I saw various derived topologies to "simple" interstage RC filters: resistors in the signal path before or after the cap, and/or large resistor to gnd before the cap. What's the purpose of this and is there a good application note detailing ins and outs?
2) DC servo:
2.1: any source for a good application note on this subject? I found various topologies that were economical with explanations about the trade-offs.
2.2: is there a way to make a self-adjusting DC servo (no offset drift)?
Thank you...
MBK
I find it not so simple to remove a DC component between op-amp stages. What am I missing?
The standard solution is a simple RC filter, say 1 uF/47k from op amp out to op amp in. But I found that regardless of what type cap I use or what combination of R and C to get similar turnover frequency, it seems to reduce dynamics by a wide margin: drum and especially bass slam is greatly reduced. This applies even if I put an op amp buffer before or after the filter, or both. The only solution I found that preserves the dynamics is to correct for the DC through a separate op amp stage. Here too I am confused because there are many topologies to choose from but I couldn't find a definive reference to the ins and outs of DC servos.
I have more caps in the signal path downstream so the goal is not necessarily to eliminate any cap in the signal path... although if it's elegant, then, sure... but just to do it right...
Specific questions:
1) RC filters:
1.1: what's the problem? Does the source component have trouble charging the cap instantaneously (current output limitations)? ... but it will not change much even is the source is a buffer with no other load.
1.2: I saw various derived topologies to "simple" interstage RC filters: resistors in the signal path before or after the cap, and/or large resistor to gnd before the cap. What's the purpose of this and is there a good application note detailing ins and outs?
2) DC servo:
2.1: any source for a good application note on this subject? I found various topologies that were economical with explanations about the trade-offs.
2.2: is there a way to make a self-adjusting DC servo (no offset drift)?
Thank you...
MBK
What you are generally looking at is a resistor at the output of the source op-amp, a capacitor, then a resistor at the input of receiving op-amp to ground.
There could of course be more depending on inverting\non-inverting configuration of your input op-amp.
1) If you have a resistor and capacitor in a circuit, you have a filter, either high pass or low pass, determined by f3db=1/2piRC.
You will want that corner frequency well out of the audio band.
2) Remember that the receiver side of that cap is essentially floating. Leakage current from an op-amp will charge that cap up to the rails if you do not have a resistor to ground to blead the charge. The leakage current through the resistor will cause off-set and you will need to chose the resistor accordingly, taking into account that is also part of a low-pass filter.
There you have it.... and
3) Make sure you use a really good cap! Elna Cerafine for lytics (and the other hundred suggestions - or one), or a high quality film cap would be even better.
Oh... If you are running the output op-amp in a non-inverting configuration, and it has FET inputs, you are looking at non-linear CV issues if your source impedance is high. If you look at op-amp specs for FET input amps, they ususually test the non-inverting configuration with low source impedance or they use the inverting configuration for the distortion test.
Good Luck!
There could of course be more depending on inverting\non-inverting configuration of your input op-amp.
1) If you have a resistor and capacitor in a circuit, you have a filter, either high pass or low pass, determined by f3db=1/2piRC.
You will want that corner frequency well out of the audio band.
2) Remember that the receiver side of that cap is essentially floating. Leakage current from an op-amp will charge that cap up to the rails if you do not have a resistor to ground to blead the charge. The leakage current through the resistor will cause off-set and you will need to chose the resistor accordingly, taking into account that is also part of a low-pass filter.
There you have it.... and
3) Make sure you use a really good cap! Elna Cerafine for lytics (and the other hundred suggestions - or one), or a high quality film cap would be even better.
Oh... If you are running the output op-amp in a non-inverting configuration, and it has FET inputs, you are looking at non-linear CV issues if your source impedance is high. If you look at op-amp specs for FET input amps, they ususually test the non-inverting configuration with low source impedance or they use the inverting configuration for the distortion test.
Good Luck!
Thanks...
Actually I am aware of the basics... and what I tried was all sorts of DC blocking variations:
1) source=low z op amp output - 10k volume pot (wiper is output; R is to gnd) - series cap - R to gnd - line driver in with 10k input z;
2) source=low z op amp output - 10k volume pot - non inverting op amp buffer- series cap - R to gnd - line driver in with 10k input z;
3) source=low z op amp output - 10k volume pot - non inverting op amp buffer - series cap - R to gnd - non inverting op amp buffer - line driver in with 10k input z;
4) same as 1-3 but with 200R to 1k series resistance after the source.
The series caps were 0.47u to 10u, from film caps (10 u, 4.7 u and 2.2 u speaker x-o grade, 2.2 u line voltage AC grade, 0.47 and 1 u regular line level grade red polyprop caps) to electrolytics (those were the cheap kind, nonpolarized of course), if using the larger ones I tried bypassing too; all with matched R to gnd to get f(c)=1/2(PI)RC=3.5 to 7 Hz approximately. Op amps are OPA 2134 with up to 40 mA drive capability at the 0-4V swing of the signal, rails at 12 V.
The result was always a very noticeable loss in "punch".
2 Solutions kept the "punch" intact:
1) when I got rid of the caps and removed DC by a biased op amp stage
2) when the source was cascoded, so the output was a transistor, not an op amp.
So... do larger cap values need a "cap driver" as source
?
MBK
Actually I am aware of the basics... and what I tried was all sorts of DC blocking variations:
1) source=low z op amp output - 10k volume pot (wiper is output; R is to gnd) - series cap - R to gnd - line driver in with 10k input z;
2) source=low z op amp output - 10k volume pot - non inverting op amp buffer- series cap - R to gnd - line driver in with 10k input z;
3) source=low z op amp output - 10k volume pot - non inverting op amp buffer - series cap - R to gnd - non inverting op amp buffer - line driver in with 10k input z;
4) same as 1-3 but with 200R to 1k series resistance after the source.
The series caps were 0.47u to 10u, from film caps (10 u, 4.7 u and 2.2 u speaker x-o grade, 2.2 u line voltage AC grade, 0.47 and 1 u regular line level grade red polyprop caps) to electrolytics (those were the cheap kind, nonpolarized of course), if using the larger ones I tried bypassing too; all with matched R to gnd to get f(c)=1/2(PI)RC=3.5 to 7 Hz approximately. Op amps are OPA 2134 with up to 40 mA drive capability at the 0-4V swing of the signal, rails at 12 V.
The result was always a very noticeable loss in "punch".
2 Solutions kept the "punch" intact:
1) when I got rid of the caps and removed DC by a biased op amp stage
2) when the source was cascoded, so the output was a transistor, not an op amp.
So... do larger cap values need a "cap driver" as source
?MBK
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.