I'm not speaking regarding "total line delay". I'm speaking regarding "Group delay" in opamp filters: you will receive some frequencies earlier than other!
Group delay is always bad, but withich a good threshold?
In fact Bessel filter has a flat group delay, but it is a disaster for frequency response flattness. I need to move more on Butterworth, with good response but bad group delay. So need a compromise between the two: then the question how much (in us) group delay is still ok?
Have a look at Blauert and Laws from 1978.
There are already numerous threads on this topic here at DIYaudio.com and elsewhere. It's an interesting topic, but "audibility" requires a subjective evaluation....and if doing that any conclusions get really interesting.
Dave.
Mmmm interesting.
It appears that a group delay of less than 400us is not audibile for the most and, thus, it is not harmful.
So we are speaking about hundreds of microseconds, and not a few.
Then I don't understand why it should be so important maintaining flat the group delay for a post DAC analogue filter... A Butterworth filter do not have milliseconds of group delay...
Also, it seems that loudspeakers have group delays many orders higher!
Nor do I. I've been using high order elliptic or chebyshev reconstruction filters for my DAC designs for many years now.
There are two things to consider besides audibility or inaudibility:
1. Amplitude roll-off can be compensated for in a linear-phase FIR oversampling filter. Compensating for phase errors would require a FIR filter with asymmetric impulse response, which is more expensive because it needs twice as many multiplications per second.
2. Due to a series of coincidences, the first Philips CD players already used a linear-phase FIR oversampling filter. Philips then also chose a nearly linear-phase analogue reconstruction filter, compensated for its roll-off with the FIR oversampling filter and used the phase linearity of the CD player as a selling point. They were quite successful with that, so competitors soon also went for nearly linear phase.
By the way, there is a fairly new thread about an allpass filter audibility test on this forum: Audibility of allpass phase distortion (test)
1. Amplitude roll-off can be compensated for in a linear-phase FIR oversampling filter. Compensating for phase errors would require a FIR filter with asymmetric impulse response, which is more expensive because it needs twice as many multiplications per second.
2. Due to a series of coincidences, the first Philips CD players already used a linear-phase FIR oversampling filter. Philips then also chose a nearly linear-phase analogue reconstruction filter, compensated for its roll-off with the FIR oversampling filter and used the phase linearity of the CD player as a selling point. They were quite successful with that, so competitors soon also went for nearly linear phase.
By the way, there is a fairly new thread about an allpass filter audibility test on this forum: Audibility of allpass phase distortion (test)
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.