After having experimented with both kinds of feedback, I cannot say that one of them sounds better than other.
However, we chose to implement post-filter feedback in our coldamp amplifiers mainly in order to avoid that frequency response dependency on the load connected. IMHO that limits the applications of the amplifier.
I am sure that the ones that use pre-filter feedback, such as Lars, have tuned their amps so that they work well with the vast majority of typical loads, but that (correct me if I am wrong) forces the designer to have a higher than necessary filter cutoff frequency (in order to ensure full bandwidth with low impedance loads, for example). That has some drawbacks, in my opinion. (carrier frequency attenuation, etc).
However, we chose to implement post-filter feedback in our coldamp amplifiers mainly in order to avoid that frequency response dependency on the load connected. IMHO that limits the applications of the amplifier.
I am sure that the ones that use pre-filter feedback, such as Lars, have tuned their amps so that they work well with the vast majority of typical loads, but that (correct me if I am wrong) forces the designer to have a higher than necessary filter cutoff frequency (in order to ensure full bandwidth with low impedance loads, for example). That has some drawbacks, in my opinion. (carrier frequency attenuation, etc).
ssanmor: It is true that you get a lower carrier attenuation, but i think it is widely accepted that this carrier residue has no significance at all, and is only a 'cosmetic flaw'. In my op there is no reason to chose the whole working principle based on solving a 'cosmetic flaw'.
About load dependent freq response, this is a theoretical discussion we have had many times here on diyaudio.com. Those who are in favor of this problem are right, as long as you measure with true ohmic loads, such as induction free resistors, and very low impedance such as 2 Ohm or less. Then you will get a few dB dip at 20 kHz.
In real life however the tweeter usually has a self inductance 10 times higher than a typical output choke (12uH) , and so there is - in most cases - no real problem with load dependent freq response.
That is why i dont really see any drawbacks of pre-filter NFB, as you all probably know
And the good thing in a free world we are all entitled to our own opinion
ssanmor, let me ask you a question, even if you can not establish which of the two is better, can you at least say if there is a difference in the sound or not?
Nice stuff you make, BTW
All the best from
Lars
About load dependent freq response, this is a theoretical discussion we have had many times here on diyaudio.com. Those who are in favor of this problem are right, as long as you measure with true ohmic loads, such as induction free resistors, and very low impedance such as 2 Ohm or less. Then you will get a few dB dip at 20 kHz.
In real life however the tweeter usually has a self inductance 10 times higher than a typical output choke (12uH) , and so there is - in most cases - no real problem with load dependent freq response.
That is why i dont really see any drawbacks of pre-filter NFB, as you all probably know
And the good thing in a free world we are all entitled to our own opinion ssanmor, let me ask you a question, even if you can not establish which of the two is better, can you at least say if there is a difference in the sound or not?
Nice stuff you make, BTW
All the best from
Lars
Thanks for your thoughts, Lars.
We don't also think that a very small switching frequency residue is of utmost importance, in that aspect I agree with you, although a better attenuation there also implies a better attenuation at the hundreds of MHz range, where it IS important to have good rejection.
However, we like to have a flat freq. response up to 20KHz for any load, and our amps have found applications from 2 ohm (subwoofers in parallel) to 100 ohm nominal (megaphony installations). For this reason we prefer to have this very controlled, although we respect the other point of view, of course.
About sound, well, to be sincere, my listening sessions with the amp in pre-filter mode have been very limited, but I heard no significant difference between two modes with 4 ohm speakers (where freq. response dependency was not a problem in any of the two cases).
Thanks for the compliments, Lars. Your stuff is (and has always been) very nice also and above all, innovative. I saw your modules years ago when they were the only ones and got impressed. They motivated me a lot to investigate and develop my own modules. Thanks for that from all the audio community.
Best regards,
Sergio
We don't also think that a very small switching frequency residue is of utmost importance, in that aspect I agree with you, although a better attenuation there also implies a better attenuation at the hundreds of MHz range, where it IS important to have good rejection.
However, we like to have a flat freq. response up to 20KHz for any load, and our amps have found applications from 2 ohm (subwoofers in parallel) to 100 ohm nominal (megaphony installations). For this reason we prefer to have this very controlled, although we respect the other point of view, of course.
About sound, well, to be sincere, my listening sessions with the amp in pre-filter mode have been very limited, but I heard no significant difference between two modes with 4 ohm speakers (where freq. response dependency was not a problem in any of the two cases).
Thanks for the compliments, Lars. Your stuff is (and has always been) very nice also and above all, innovative. I saw your modules years ago when they were the only ones and got impressed. They motivated me a lot to investigate and develop my own modules. Thanks for that from all the audio community.
Best regards,
Sergio
Greetings,
I use both post and pre feedback on one of my designs. The pre is mainly used for oscillation and basic stability, the post i use to mop up any non-linearities in the output filter and provide accurate DC feedback, i also amplifiy and invert the local ground signal, injecting this into the amp massivly reduces the noise floor and removes all audible hiss. This technique works very well for me, you may find different uses for using both, or simply decide it is not needed. Thats my view anyway.
Take care
Mad.P
I use both post and pre feedback on one of my designs. The pre is mainly used for oscillation and basic stability, the post i use to mop up any non-linearities in the output filter and provide accurate DC feedback, i also amplifiy and invert the local ground signal, injecting this into the amp massivly reduces the noise floor and removes all audible hiss. This technique works very well for me, you may find different uses for using both, or simply decide it is not needed. Thats my view anyway.
Take care
Mad.P
Oh really how do you know about that ?
http://www.diyaudio.com/forums/clas...class-d-vs-ffa-audio-class-d-sample-test.html
http://www.diyaudio.com/forums/clas...class-d-vs-ffa-audio-class-d-sample-test.html
Last edited:
Well my real life work (not simulators) show that post filter feedback will not actually do much to improve THD.
But on the other hand you can also get a good low THD with pre filter THD (in real life i mean).
So i guess i just dont agree with you on this one.
It's not a matter of 'not going throgh the trouble' of implementing post filter feedback. My experience just tells me that post filter feedback has some inherent limitations that i don't want my amplifiers to have. Limitations that i can't see any way to work around without damaging the sound. As simple as that.
I know it can also be said that pre filter amps have inherent problems, like loss of gain at low impedances at 10-20 kHz. However i find it easy to get around these limitations without damaging the sound. (With a simple RC compensation filter).
UcD is by far not the only post filter feed back modules, there are many of them out there. And they all feature the same characteristics. And none of them are better in sound tests than a good pre filter feedback setup. Sorry.....
Anyway since there is so much prejustice around post filter feedback, and it's so easy to implement, i think my next generation of Class D amplifiers might have both options on board.
Postfilter feedback is only for getting a low impedance on the output and a higher control of the speaker system, the analog filters in such speaker system can be a factor of trouble when using pre filter feedback, but as people tell here, a high quality coil, high switching frequency does wonders, I think also that the differences in thd are very low because a coil is a liniair element of itselfs, a multilevel class d on 2 Mhz (multiplication of carriers) lose the need of post filter feedback, and I think prefilter feedback does sound better, (kind of open loop effect coil).
regards
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.