PowerSerge
Member
Im sure they sound amazing but they look diy
Right, whatever it's called I would definitely not want something like that for monitoring. 45ms is pretty hardcore long latency as well, for studio work it means it's impossible to play on any electric/electronic instrument while listening through them, that's the reason there's low latency mode and you have to manually switch between them
Why would that pink noise pulse ramping up effect be? Is this true for electrical devices as well, and is this effect only audible in the lower bass, or would any driver need to have an associated group delay with it's LF roll off to avoid this? Would the effect disappear or get shifted to below audible if a subwoofer was added to this speaker and integrated so that it went down to 5 Hz with the phase still flat? I'm really curious about how and why this effect would happen. Must it always happen? Is it intrinsic to the nature of a signal with flat phase through the high pass zone?Erin reviewed the smaller Pulsar which yielded the ff. step response:
View attachment 306215
Ex Machina Pulsar MKII 3-way 8-inch Active Studio Monitor Review
Ex Machina Pulsar MKII Reviewwww.erinsaudiocorner.com
The low end might not exactly audibly "pre-ring"...
@KSTR briefly had this to say across at DIYAudio:
"... this appears to be a speaker where the natural phase response from the woofer highpass was corrected to linear phase which creates the point-symmetric step response.
That's a neat trick for speakers to have "fast bass" down low. The price you pay is that some signals (like certain kick drum sounds) may be rendered with audible artifacts. The rising slope in the time-domain is not ringing and does not sound like ringing, rather it sounds like pink noise pulse ramping up (very interesting acoustical illusion)."
Okay, so TLDR: it's because correcting phase response to linear for the bass doesn't make sense. The linear phase is not worth the bodies that it buries.Why would that pink noise pulse ramping up effect be? Is this true for electrical devices as well, and is this effect only audible in the lower bass, or would any driver need to have an associated group delay with it's LF roll off to avoid this? Would the effect disappear or get shifted to below audible if a subwoofer was added to this speaker and integrated so that it went down to 5 Hz with the phase still flat? I'm really curious about how and why this effect would happen. Must it always happen? Is it intrinsic to the nature of a signal with flat phase through the high pass zone?
Why would that pink noise pulse ramping up effect be? Is this true for electrical devices as well, and is this effect only audible in the lower bass, or would any driver need to have an associated group delay with it's LF roll off to avoid this? Would the effect disappear or get shifted to below audible if a subwoofer was added to this speaker and integrated so that it went down to 5 Hz with the phase still flat? I'm really curious about how and why this effect would happen. Must it always happen? Is it intrinsic to the nature of a signal with flat phase through the high pass zone?
Some Adam subwoofers have even longer processing delays for phase correction, but they do not linearize the phase of the speakers to the extreme extent Ex Machina does. Yeah, the low latency mode would be the default for most live monitoring workflows.
So the gist is we can't really hear the group delay down there, but we might hear the ramp up effect some times if we try to force the phase flat. I wish I understood how that works better. I should just make linear phase filters with steep cutoff rates at different frequencies and listen to music through them on headphones. See if I can hear that effect. Headphones should do it was well as speakers. In any case, it seems the phase can be flattened to a significant degree without noticeable effect, and without too much effort.Okay, so TLDR: it's because correcting phase response to linear for the bass doesn't make sense. The linear phase is not worth the bodies that it buries.
The reason it works for crossovers and not for the low end roll off is because a HPF and LPF will "correct" for one another's phase response. A typical minimum phase (read: passives, or actives sans phase correction) crossover has phase wraps around the crossover point(s) from the pass filters. They will each get to a point where you'll see, basically, inverse phase rotation for each filter - and that can be corrected for pretty easily via all-pass filters (analog, a la PSI) or in DSP with FIR filters. This doesn't cause pre-ringing, even with super steep filters (8th order, even, is OK).
But it doesn't work for LF roll off because there's no LPF to go along with the HPF of the system, so instead you just get that weird ramping effect.
If you add a sub, you now have a LPF to work with.
Effectively, yes - you're hearing the pre-ring from a linear phase HPF. And yeah, headphones make it more obvious but any decent speaker should demonstrate it clearly enough.So the gist is we can't really hear the group delay down there, but we might hear the ramp up effect some times if we try to force the phase flat. I wish I understood how that works better. I should just make linear phase filters with steep cutoff rates at different frequencies and listen to music through them on headphones. See if I can hear that effect. Headphones should do it was well as speakers. In any case, it seems the phase can be flattened to a significant degree without noticeable effect, and without too much effort.
I dunno ....about not being able to hear bottom-end group delay.So the gist is we can't really hear the group delay down there, but we might hear the ramp up effect some times if we try to force the phase flat. I wish I understood how that works better. I should just make linear phase filters with steep cutoff rates at different frequencies and listen to music through them on headphones. See if I can hear that effect. Headphones should do it was well as speakers. In any case, it seems the phase can be flattened to a significant degree without noticeable effect, and without too much effort.
How do you calculate low frequency resolultion based on taps and sample rate? Also, why do more taps make the pre-echo more audible? I would normally expect that more is better.Either way, it will take a good number of taps. I think maybe 16k taps @ 48kHz sample rate is probably a good minimum choice. Frequency resolution for that is 6Hz.
A decent rule of thumb seems to be have freq resolution at least 1/3 of lowest freq intended to correct. So 16k @ 48 is probably good for work down to 18Hz.
Oh, and as a reminder, if the sample rate used is 96kHz, it takes twice the taps for equivalence.
The only times I've felt confident I can consistently hear pre-echo, has been with more taps and also using then to effect a linear-phase system high-pass.
Like 65k taps at 48kHz. You may want to play with that too.
Good luck with experiments...look forward to hearing your impressions.
Is that different than smoothing and windowing the response before applying correction? I've been shooting for a smooth response at the listening position with about 7 cycle windowing lately. The reason is that when I finally looked at the windowed response at the listening position I was seeing big drops in the frequency response at the crossovers, which didn't appear with longer windowing. I ruled out interference problems by measuring each driver separately. The nearfield equalization in my case wasn't holding out to the listening position, with more droop in the low end as I moved away from my tweeters. This has made a huge improvement in my perception of the sound quality.I see less taps as a form of "correction smoothing" sometimes.
I don't know how to determine how many taps my filters have. If I export an impulse from REW, is there an export setting that tells me the number of taps that will be employed? Or is that a setting on the convolution engine? I see I can choose the sample count and the sample rate on the impulse export.Either way, it will take a good number of taps. I think maybe 16k taps @ 48kHz sample rate is probably a good minimum choice. Frequency resolution for that is 6Hz.
A decent rule of thumb seems to be have freq resolution at least 1/3 of lowest freq intended to correct. So 16k @ 48 is probably good for work down to 18Hz.
Oh, and as a reminder, if the sample rate used is 96kHz, it takes twice the taps for equivalence.
Good luck with experiments...look forward to hearing your impressions.
Screw up as in fail to fully optimize, or screw up as in create something really horrible?But dunno, and don't really want to explore using REW for FIR anymore.....too easy to screw up, i think.
Let's just say I'd make it a mandatory requirement to run an electrical transfer function of the FIR filter, before I'd send it to an amp & speaker.or screw up as in create something really horrible?
I use REW to generate .wav convolution files to tune to the Harman curve, but with the bass boost reduced by about 4dB. I don't know whether it is perfect, but I haven't had any issues - it definitely improves the sound of my speakers' in-room response.Let's just say I'd make it a mandatory requirement to run an electrical transfer function of the FIR filter, before I'd send it to an amp & speaker.
Levels can be tricky using trace arithmetic and impulse inversions......but that may just be my lack of practice with REW.
When exporting to a .wav file, you can specify the number of taps, or you can leave that box unchecked and let REW decide. On some convolution files I generated a few months ago I had specified 64k. Yesterday I unchecked the box and let REW decide the number of taps, and it seems to be working great.I don't know how to determine how many taps my filters have. If I export an impulse from REW, is there an export setting that tells me the number of taps that will be employed? Or is that a setting on the convolution engine? I see I can choose the sample count and the sample rate on the impulse export.
So 64K samples means 64K taps? That's a lot more than I usually see specified. I've been leaving that box checked. Latency doesn't seem too bad. I'm going to have to watch OCA's video again. That video is what started me down this whole FIR filter path.When exporting to a .wav file, you can specify the number of taps, or you can leave that box unchecked and let REW decide. On some convolution files I generated a few months ago I had specified 64k. Yesterday I unchecked the box and let REW decide the number of taps, and it seems to be working great.
So 64K samples means 64K taps?