I'm trying to use a minimalist approach to design a line level crossover for a bi-amp system.
I need to drive a midrange+tweeter (with traditional passive crossover), line level crossed with a woofer, let say at 400 Hz, 12dB/oct. The 2 amplifier modules are located inside each loudspeaker cabinet.
I will use a simple volume pot (may be followed by a unity gain buffer) as preamp unit, distant about 6 feet from the loudspeakers, and I'm tempted to avoid the additional line stages required for the crossover filters. Using a passive line level crossover could be possible, but I don't like the non optimal impedance load.
So I'm working on the following idea: why don't use the amp itself to implement LP-HP filters ?
One first 400Hz 6dB RC section can be used at the very input of the amplifiers, and a second 400Hz 6dB cut can be implemented into the feedback network. May be also the miller cap of the woofer amp should be increased.
Am I missing something? There are negative effects limiting the bandwith of the amps ? I'm using a classic single input diff + vas + driver + EF topology, class AB, NFB.
Let me have your comments, and thanks in advance.
Gianluca
I need to drive a midrange+tweeter (with traditional passive crossover), line level crossed with a woofer, let say at 400 Hz, 12dB/oct. The 2 amplifier modules are located inside each loudspeaker cabinet.
I will use a simple volume pot (may be followed by a unity gain buffer) as preamp unit, distant about 6 feet from the loudspeakers, and I'm tempted to avoid the additional line stages required for the crossover filters. Using a passive line level crossover could be possible, but I don't like the non optimal impedance load.
So I'm working on the following idea: why don't use the amp itself to implement LP-HP filters ?
One first 400Hz 6dB RC section can be used at the very input of the amplifiers, and a second 400Hz 6dB cut can be implemented into the feedback network. May be also the miller cap of the woofer amp should be increased.
Am I missing something? There are negative effects limiting the bandwith of the amps ? I'm using a classic single input diff + vas + driver + EF topology, class AB, NFB.
Let me have your comments, and thanks in advance.
Gianluca
Hmmm......
The high pass cascaded 1st order would not be a problem.
The low pass input is not a problem either but the feedback loop
is. You need to increase the compensation for stability into the
lowest gain, for a simple first order filter = unity gain.
This will likely crush the slew rate of the power amplifier,
but with a 400Hz c/o frequency things should be fine.
Note as described these are not 2nd order filters with adjustable Q.
sreten.
The high pass cascaded 1st order would not be a problem.
The low pass input is not a problem either but the feedback loop
is. You need to increase the compensation for stability into the
lowest gain, for a simple first order filter = unity gain.
This will likely crush the slew rate of the power amplifier,
but with a 400Hz c/o frequency things should be fine.
Note as described these are not 2nd order filters with adjustable Q.
sreten.
Assuming stability is not an issue, i think the following question should be asked (I and I've no idea what the answer is):
In the bands outside the XO region, which scheme (line level XO preceeding the amp vs. an XO inserted in the feedback loop) produces less noise and non-linearities? I would want to confirm this after the fact with actual measurement.
A lesser but important question would be: Can you sucessfully implement the desired XO slopes when inserted in the FB loop? That is: supposse the XO you desire is a 24dB L-R XO -- is that what you actually get or are actual results different from predicted?
The reason I bring this up, is that I tried vaguely analogous six months ago with a headphone amp and did not like the results. The roll-off of curves were not what I wanted/predicted and THD+N was unacceptably elevated across frequency range. No doubt, someone brighter than I could have made it work - but it was MY project not theirs. I concluded that even if I persevered until it worked, I was going to end up with something that was far lest minimal and elegant than simply keeping the two functions (filter vs. amplification) separate.
On the otherhand, if the amplifier you use resembles a discrete opamp maybe the whole thing is more straight forward -- just a random thought.
In the bands outside the XO region, which scheme (line level XO preceeding the amp vs. an XO inserted in the feedback loop) produces less noise and non-linearities? I would want to confirm this after the fact with actual measurement.
A lesser but important question would be: Can you sucessfully implement the desired XO slopes when inserted in the FB loop? That is: supposse the XO you desire is a 24dB L-R XO -- is that what you actually get or are actual results different from predicted?
The reason I bring this up, is that I tried vaguely analogous six months ago with a headphone amp and did not like the results. The roll-off of curves were not what I wanted/predicted and THD+N was unacceptably elevated across frequency range. No doubt, someone brighter than I could have made it work - but it was MY project not theirs. I concluded that even if I persevered until it worked, I was going to end up with something that was far lest minimal and elegant than simply keeping the two functions (filter vs. amplification) separate.
On the otherhand, if the amplifier you use resembles a discrete opamp maybe the whole thing is more straight forward -- just a random thought.
sam9 said:
a) for low pass functions stability is a main issue, and fundamentally
affects the re-compensated amplifiers performance.
b) by definition the XO in the feedback loop should have lower
noise and distortion in the stopband.
c) getting complicated -
you cannot insert a filter function (except 1st order) in the feedback loop.
(Edit : you can insert a 2nd order bandpass or notch filter but
you can't insert anything with higher than 1st order stopbands)
Using feedback topologies you can implement 2nd order functions with a gain
stage and adding a first order filter to the input means the maximum order
you can implement with a single gain stage is 3rd order.
A problem with the 2nd order topologies is gain is restricted.
d) Find this confusing - why would a headphone amplifier need filters ?
sreten.
") Find this confusing - why would a headphone amplifier need filters "
1- Equalization with regard to a specific headphone
2- Same as above but for sharp anomolies appearently the result of the interaction of specific headphones (even very good ones) and the ear canal. You can find details at the Linkwitz Lab websete.
3 (and most important to me)- Compensation for hearing loss of an elderly relative. I did this quite easily with an LM386, but that hasn't the greatest sonics. (This is non-trivial - my mother can now hear TV and radio again for the first time in 15 years! Sometimes even the best hearing aids don't work well in these situations.) I wanted to so something similar for someone else I know who misses high quality listening.
"b) by definition the XO in the feedback loop should have lower
noise and distortion in the stopband."
Won't argue with the principal, I just like to see such things confirmed with real harware.
1- Equalization with regard to a specific headphone
2- Same as above but for sharp anomolies appearently the result of the interaction of specific headphones (even very good ones) and the ear canal. You can find details at the Linkwitz Lab websete.
3 (and most important to me)- Compensation for hearing loss of an elderly relative. I did this quite easily with an LM386, but that hasn't the greatest sonics. (This is non-trivial - my mother can now hear TV and radio again for the first time in 15 years! Sometimes even the best hearing aids don't work well in these situations.) I wanted to so something similar for someone else I know who misses high quality listening.
"b) by definition the XO in the feedback loop should have lower
noise and distortion in the stopband."
Won't argue with the principal, I just like to see such things confirmed with real harware.
A basic problem with doing this type of thing in the feedback loop is that the gain can never become less than 1 (100% feedback). Suppose you have an amp with 20dB gain, you can then roll it off say at 20dB per octave for just one octave, then the gain is 1 and stays there. Normally in a xover you want the output to continue to decrease "forever".
All this is of course separate from the mentioned potential stability problems.
Jan Didden
All this is of course separate from the mentioned potential stability problems.
Jan Didden
"gain can never become less than 1 "
I think the sudden realization of this when I was looking into to it was one of the "complications". I deal better with many things (not just electronic) by breaking them down into small parts and addressing them separately. Building filters into a FB loop goes counter to this even it were the ideal approach.
I think the sudden realization of this when I was looking into to it was one of the "complications". I deal better with many things (not just electronic) by breaking them down into small parts and addressing them separately. Building filters into a FB loop goes counter to this even it were the ideal approach.
Hi,
Have seen a commercial Denon amp that had the Baxandal tone control in the feedback loop of the power amp to save cost. Not a good sounding amp btw.
It can be done provided the amp is unity gain stable. Here you can run into problems because most power amps aren’t. Look what a Sallen & Key 2nd order filter is … It is basically build around a unity gain amp (buffer). If you implement it that way you need to drive it with a large swing at the input and I have seen no power amps that can deal with that.
The way it possible can be done it to attenuate first the output of the amp so you have unity gain and then build a S & K filter with it. Thus if your power amp has a nominal gain (with it’s own feedback loop) of say 26 dB then attenuate the output with 26dB and then configure it as a S & K filter. Just a suggestion.
Cheers
Have seen a commercial Denon amp that had the Baxandal tone control in the feedback loop of the power amp to save cost. Not a good sounding amp btw.
It can be done provided the amp is unity gain stable. Here you can run into problems because most power amps aren’t. Look what a Sallen & Key 2nd order filter is … It is basically build around a unity gain amp (buffer). If you implement it that way you need to drive it with a large swing at the input and I have seen no power amps that can deal with that.
The way it possible can be done it to attenuate first the output of the amp so you have unity gain and then build a S & K filter with it. Thus if your power amp has a nominal gain (with it’s own feedback loop) of say 26 dB then attenuate the output with 26dB and then configure it as a S & K filter. Just a suggestion.
Cheers
There are lots of commercial amplifiers with passive preamps
and active tone controls built into the poweramps feedback loop.
The compromises involved in compensating the power amplifier
for useless tone control settings (full treble cut) are depressing.
This also applies to a lesser extent to pre-amp line stages.
sreten.
and active tone controls built into the poweramps feedback loop.
The compromises involved in compensating the power amplifier
for useless tone control settings (full treble cut) are depressing.
This also applies to a lesser extent to pre-amp line stages.
sreten.First of all, thanks to everybody for the more than interesting comments.
Actually I would like to focus the discussion over the following simplified scenario:
- highpass amp: I cannot figure out any problem in simulation limiting the lower frequency corner at 400Hz.
- lowpass amp: things are more subtle here, the desired amp bandwidth is only 20 - 400 Hz. Again I cannot see problems in simulation (I also increased the miller cap from 68pF to 300pF).
It's true that the NFB filter is 6dB/oct, but only until the unity gain is reached (after that the response remains at unity gain). The 6dB input filter (remember I want to reach 12dB/oct) will operate normally, instead. Considering the amp's overall gain is about 30dB, I would say that for a woofer - satellite crossing at 400 Hz, it could somehow work... extending the concept to a general filter implementation via NFB is totally different matter.
Different opinions, before I start building ?
Thanks again for the excellent contributions,
Gianluca
Actually I would like to focus the discussion over the following simplified scenario:
- highpass amp: I cannot figure out any problem in simulation limiting the lower frequency corner at 400Hz.
- lowpass amp: things are more subtle here, the desired amp bandwidth is only 20 - 400 Hz. Again I cannot see problems in simulation (I also increased the miller cap from 68pF to 300pF).
It's true that the NFB filter is 6dB/oct, but only until the unity gain is reached (after that the response remains at unity gain). The 6dB input filter (remember I want to reach 12dB/oct) will operate normally, instead. Considering the amp's overall gain is about 30dB, I would say that for a woofer - satellite crossing at 400 Hz, it could somehow work... extending the concept to a general filter implementation via NFB is totally different matter.
Different opinions, before I start building ?
Thanks again for the excellent contributions,
Gianluca
IMO you have bigger fish to fry than the series feedback
not being able to go lower than 0dB gain, the combination
of input roll-off and feedback roll-off will be fine.
I'd be more concerned about baffle step compensation.
At 400Hz your likely to be near the middle of it. One approach
would be to use 1st order at 400Hz for the feedback loops
and use gain between the two amplifiers to allow for BSC.
In this scheme the input rolloffs for bass would be 100Hz
to 200 Hz, and for the midrange 800Hz to 1.6KHz.
sreten.
not being able to go lower than 0dB gain, the combination
of input roll-off and feedback roll-off will be fine.
I'd be more concerned about baffle step compensation.
At 400Hz your likely to be near the middle of it. One approach
would be to use 1st order at 400Hz for the feedback loops
and use gain between the two amplifiers to allow for BSC.
In this scheme the input rolloffs for bass would be 100Hz
to 200 Hz, and for the midrange 800Hz to 1.6KHz.
sreten.sreten,
you are absolutely right.
My theorical BSC freq is 500Hz, so I better move both NFB roll-offs to 500Hz. Lowpass input roll-off will be something between 125Hz and 250Hz. Highpass input roll-off between 1kHz and 2KHz.
And after I can play with amps gain to balance BSC.
Thanks a lot for the suggestion !
Gianluca
you are absolutely right.
My theorical BSC freq is 500Hz, so I better move both NFB roll-offs to 500Hz. Lowpass input roll-off will be something between 125Hz and 250Hz. Highpass input roll-off between 1kHz and 2KHz.
And after I can play with amps gain to balance BSC.
Thanks a lot for the suggestion !
Gianluca
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.