I will take a listen to those tracks. Right now I'm preparing for company and have lots of cat fur to expel from the listening area. 🙂
It looks like the consensus is that time-coherent (Stereophile's phrase) speakers aren't as important to the DIY intelligentsia.
I thought for sure some one would say "How can you image without it?" 🙂
I get the simplicity and attractiveness of a full range, single driver speaker. I really do, but doubt I'd be willing to make the sacrifices.
Thanks everyone!
Best,
Erik
It looks like the consensus is that time-coherent (Stereophile's phrase) speakers aren't as important to the DIY intelligentsia.
I thought for sure some one would say "How can you image without it?" 🙂
I get the simplicity and attractiveness of a full range, single driver speaker. I really do, but doubt I'd be willing to make the sacrifices.
Thanks everyone!
Best,
Erik
Not sure what IR delay is, but if you mean driver delay, have that already, since I built the speakers. 🙂
Well I'm getting a miniDSP this week, or early next week. Maybe I'll attempt doing an active crossover with time delay as an experiment. Another DIYer who has been working on a 3-way active assures me the sound quality is not as good, but for a quick n dirty experiment it seems worthwhile.
Best,
Erik
Well I'm getting a miniDSP this week, or early next week. Maybe I'll attempt doing an active crossover with time delay as an experiment. Another DIYer who has been working on a 3-way active assures me the sound quality is not as good, but for a quick n dirty experiment it seems worthwhile.
Best,
Erik
How is the miniDSP going to help exactly? 😕 Time delay only isn't going to give you a fair shot here...
Pretty easy. Measure your frequency response including phase, with some smoothing (maybe 1/12 or 1/6th) and a well chosen window edge so you get reasonable looking low frequency response. Dial in the delay (in OmniMic) to get the phase the flattest you can get it without it tilting positive (other than at phase wrap vertical lines of course), then save the FRD file.
Load the FIR file into RePhase, work its sliders to adjust for flatness of both phase and dBs (note there are multiple banks you can use). Set the number of taps for RePhase to use (based on whoever much your FIR hardware can do) and the sample rate. Generate and save the correction impulse response file and import it into your hardware (very easy into MiniSharc or OpenDRC). Check the FR and PR again, to make sure it's good and you're done.
MiniDSP makes several pieces now that can do FIR filtering (phase and magnitude responses separately adjustable).
eriksquires said:
MiniDSP can be quite good with a little work. The SHARC DSP chip is about as good as there is, I think, so that leaves interface and analog issues. I use an OpenDRC-DA (as DAC, input selector and volume control), added a 500uH low-R choke between the DSP chip board inside it and the DAC board, which totally removed a slight whine noise that was getting through. I also changed the clock oscillator in the DAC section with a NDK NZ2520SD*, which seemed to make things sound better (though again, not exactly a blind test nor in any reasonable time frame). Though you'll never get some people to ever admit anything inexpensive could be good 😉
[* I should mention though, that that oscillator is a VERY teeny little dot of a part -- I could never have gotten it connected without a microscope and some very fine solid wires to tack onto it to make short connections!).
Last edited:
MiniSharc? Yes, MiniDSP? I don't think so... but I'd also advice against using the above method to find the phase you'd want to correct with RePhase. Figure out when you want the phase to be correct. Frequency dependent windows will help you get there, just not the one currently implemented in REW.
Bill, send me an IR from REW as measured at the listening position. I'll send you a FIR correction file made in DRC-FIR with my recipe to compare. You have well behaved phase with your speaker, lets see if we can make you a believer... 🙂 Make it at 44100 sample rate as I don't have templates for every sample rate a.t.m. Are you up for an experiment? I promise you don't have to like it. But you might...
Same thing I've experienced in less controlled home test. I also noticed a difference in physical sensation(chest) between 100-300hz with linear phase. A steep changing phase slope as found in a 4th order acoustic crossover (250hz) had much less kick and articulation in that range with the same frequency response and SPL.
All done outdoors with prosound woofers and pro coaxials.
Also find myself less likely to get bored or fatigued with music.
Last edited:
The DSP chip makes no difference. It's the interface that matters. MiniDSP can do FIR filters, but you have to enter the pole-zero form directly, not very effective.
The way I see it a time smeared system is a more polite version or rendering of the sound. The level of excitement goes down. Making it further removed from real sounds.
As I was getting my son from school yesterday I was cycling trough our city to pick him up. Someone dropped a 2 by 4 from a truck on the ground right next to me. The sheer power and broadband strike when it hit the floor was huge.
I remember thinking: you'll never get that kind of sound on a time smeared system. It will always be a more polite version of that same sound. It just doesn't have the same sense of impact. I saw the whole thing with the beam coming so I wasn't taken by surprise. But to me that's the difference. Along with the gain in coherency you get.
I bet a test like yours will be way more telling than headphones. There's more than one sensor at work when we hear something like this. You feel it as much as you hear it. True time coherency can make the midrange feel powerful, because the harmonic structure of the sound is in tact. The more low extension you've got, the bigger the impact. I don't care how many disregard the importance of time coherency, I'm going to miss it when it's not there.
I don't regret spending a lot of time on it to get it right. It was worth it to me.
Maybe I should add a more smeared version to my system for easy listening background music. It's pretty dominant with time coherency.
Last edited:
Interesting. You are suggesting to include some room influence in lower frequencies also in regards to phase? So it is not just the phase of the first wavefront but an average phase over a short time window? Would this ever happen in reality? In your example with the 2 by 4, u would change its phase relationship according to its surroundings/ reflections?
My stomach say phase should follow the anechoic response, but I will try your way to compare.
Wesayso, exactly! Bad time behaviour makes everything appear more quiet with less impact, averaging impulses, taking out the kick. If you want to cut something with an axe, you do not mount allpasses inbetween, you just hit it with the bare axe.
Most of us are in a room 😉. We will have room reflections and other determining factors around us. What I try and accomplish is getting rid of the room as much as possible. If I had a recording of that beam hitting the ground, that's what I want to hear in my room, including those reflections you mentioned. Preferably without hearing my room too much! I'd like to be transported to the recording, the "you are there" vs "they are here" argument comes to mind.
Solve everything you can in the room using acoustic treatment (like damping panels at first reflections etc) and use a supporting speaker design that has an even off axis response compared to the on axis signal.
Next step is to use frequency dependent windows at the listening position to correct (mostly) the speaker and only some parts of the room. At higher frequencies I'm only trying to correct the speaker. That first wave front to be exact.
At lower frequencies I use a shorter FDW (that leaves out the higher frequencies) to correct some of the room problems (excess phase window), but I'm not going to try to fix all of it. You need to spend a lot of time to learn what the room does at which frequency. Solving as much as you can with treatment is going to help more than trying to fix it digitally.
The APL_TDA graph I showed earlier shows the sum of my left and right speaker as measured at the listening position. The separate channels by themselves still show room influences that I do not fix. (that would sound horrible, I tried 🙂) I do exchange some energy from the left channel to the right and vise versa in trouble areas below 70 Hz. All to get an even response over a large area.
Together they sum in the low end to a coherent signal. Due to the sheer size (in height) of the speakers (line arrays) the bass out in the room is very even.
The frequency dependent window is fixing the early wave front at the listening position. You can see that if you look at the early waterfall plots. They are way cleaner when corrected this way.
Early waterfall plot of the midrange as measured at the listening position.
Without phase correction they show way more disturbance at 300-400 Hz.
Here's a later one, getting better:
(by limiting the correction window for phase to a lower frequency)
Look at a waterfall plot of a single full range speaker and compare. A good test would be to try and correct a full range speaker with decent output down low and correct it with a ~5 cycle window. Compare the waterfall plots you get. That would make an awesome near field system like gmad uses: http://www.diyaudio.com/forums/full-range/275730-convolution-based-alternative-electrical-loudspeaker-correction-networks.html
That simple system of him was a big inspiration to me, helping me to get a clear picture of what is going on. He started that thread to help others that are interested to give it a try.
The free DRC-FIR is also my goto tool for this. With my own template that deviates from the standard ones to solve my particular problems with my room and my speakers. I can't promise this works for every type of speaker, except full range and synergy/unity, basically every speaker that has reasonable off axis response compared to the on axis.
(even though full range speakers seem more time coherent from the start they have their quirks too, jut look up an early waterfall graph)
Lots more info in my thread as well.
Last edited:
Interesting, just to make sure, do you mean flat power response (period) or flat power response in something like a 45 degree horizontal window?
Or flat power response (period) but in a restricted frequency range?
FredrikC said:
I have adressed this earlier but the confusion seem to live on.
IMO it's unfortunate to call discrete reflections from edge diffraction "group delay" (which your study was about). This kind of phenomena easily results in combfiltering which is something totally different than a pure time/phase delay from crossovers for example.
The ear is very insensitive to phase distortion at high frequencies.
Power response is always defined over all space. Yes flat power response down to the lowest frequencies is not possible at anything but DI = 0 dB.
I'm not confused about what I did, but you seem to be. Our study did not look at edge diffraction, but was meant to simulate Higher Order Modes in waveguides. That it also results in comb filtering is beside the point. Group delay is the same issue regardless of how it is created. It can have comb filter effects in some cases and not in others, but those are all part of the phenomena's.
I don't know what you consider "high frequencies" and if you mean above say 7 kHz, then I would agree. But there is some evidence that 700 - 7 kHz does have some hearing perceptional significance.
Feeding REW with some perfect band-pass minimum phase devices and looking into waterfall plot seems that below 700Hz also makes sense to what some of us think we hear, as what ErnieM sense in below and also related in below post from wesayso. Know its acoustic speaker in a room we talk about and not perfect band-pass minimum phase devices so probably there is only a chance to get it right at a limited space/axis at listening position, but seems to be possible in real world when looking plots from wesayso's system.
In first picture its to show ringing tails that is added with low frq XO points compared to a XO free domain. First slices of wavefront seems distorted way up high into MF and HF even XO point frq is low.
In second picture all plots is XO free, but is to show the lower in frq the system stop-band the less ringing tails up in MF and HF area.
Attachments
Last edited:
- Status
- Not open for further replies.