I ran the test again last night on three powered speakers. A Mackie, a Roland and a JBL EON. All three are "pro" speakers in plastic cases. The Mackie and the Roland are both small single drivers, the JBL is a 1" horn and 15" woofer.
The tests were run several times each, trying for the best results.
Looking at the waveforms I can see that my marker beeps do not record the same from one run to another on the EON and the Roland. I don't know why. The Mackie is very consistent.
The tests were run several times each, trying for the best results.
- Mackie: Very good null with good extraction of hidden track. Some Buckeroo remaining
- Roland: Just noise. Can't get anything good at all.
Maybe defective? Maybe too loud. - JBL EON: Decent extraction, some Buckeroo left.
Looking at the waveforms I can see that my marker beeps do not record the same from one run to another on the EON and the Roland. I don't know why. The Mackie is very consistent.
Here are my results with the JBL EON and the Mackie monitor. I am not posting the Roland, as I can't seem to get good results. Perhaps a faulty recording.
My guess at the results of these first tests is that the louder signal (Buckeroo) is modulating the lower signal (Chuck E). This is almost completely absent in the DAC-ADC loop, so must be coming from the acoustic recording and associated electronics. Is most of it coming from the speaker itself? Likely, but further tests would be needed.
My guess at the results of these first tests is that the louder signal (Buckeroo) is modulating the lower signal (Chuck E). This is almost completely absent in the DAC-ADC loop, so must be coming from the acoustic recording and associated electronics. Is most of it coming from the speaker itself? Likely, but further tests would be needed.
from me, 
!!
Yes. In my case the speakers were run with a -16dB FS (-13dB RMS) pink noise giving 96dBA at the microphone. A-weighting was used so as to not favor larger speakers with extended bass response.
The small speakers I tested very close, about 30cm from the driver. The larger speaker was tested at 1 meter from the speaker face.
I do not know if these levels are ideal, but that is what I used. A Shure SM 81 mic was used to record and measure. This is a very flat cardioid microphone.
The small speakers I tested very close, about 30cm from the driver. The larger speaker was tested at 1 meter from the speaker face.
I do not know if these levels are ideal, but that is what I used. A Shure SM 81 mic was used to record and measure. This is a very flat cardioid microphone.
I never had much success with doing deep nulls using loudspeakers. If you look at the kind of precisions needed in amplitude and phase to get a null that makes the "main" signal negligible, it isn't hard to see why. Electronics are pretty repeatable, you can get pretty much the same result if you play the same thing through the same device twice. With speakers, things are always changing. Compression, heating, any memory effects in suspensions, background noises, air currents, tiny physical position movements, reflections off stuff in the room -- all make deep nulling of an ACOUSTIC audio signal a pain to attempt. The main attraction of the signal nulling method is that it is exceedingly sensitive. But it is also the main failing of the method, it is so sensitive that it becomes difficult to create any kind of control situation that is stable enough to allow it to be able to cancel signals when subtracted.
Maybe for your low-level detail tests it won't be so necessary to cancel so deeply. If so, maybe you can get close enough, I hope so.
When making any kind of speaker-played recordings you have to be downright compulsive about keeping things "the same". That means that speaker and microphone position cannot be moved, not even a millimeter, don't even bump their cables. Nothing in the room should change between recordings, including humans in there doing the recording - stay out, and preferable don't go back in between recordings. Reflected signal levels off of anything much in the room are a lot more audible than most people think and changes in them will show up like sore thumbs in a difference recording.
BTW, maybe not exactly for this thread, but I think some mention should be about "hidden detail" and "perception of hidden detail", as they aren't likely to be the same. Different things might make music seem more detailed, and some won't have anything to do with more accuracy or S/N. So something that shows more perceived detail might even have less "detail" that might be determined mathematically. A good example is SET tube amps and my system speakers. The SETs distort like hell (all that "clean first watt" stuff, when measured, is still miles from even approaching a chip amp's distortion on any harmonic), and their response changes with speaker load. But either the distortion or the change to eq gives more perceived detail, small things just get noticed more.
BTW, what would happen if all the reviewers (the ones who "heard things" they'd "never heard before" on their old records, after using a new component), would they hear those same things now if they listened again with the system restored to how it had been before? Would those thing in the recordings be inaudible, unheard, again? I doubt it. I think with ears, it's a matter of getting things noticed as much as it is about getting noise and distortion down.
I'll stay tuned...
Bill
Maybe for your low-level detail tests it won't be so necessary to cancel so deeply. If so, maybe you can get close enough, I hope so.
When making any kind of speaker-played recordings you have to be downright compulsive about keeping things "the same". That means that speaker and microphone position cannot be moved, not even a millimeter, don't even bump their cables. Nothing in the room should change between recordings, including humans in there doing the recording - stay out, and preferable don't go back in between recordings. Reflected signal levels off of anything much in the room are a lot more audible than most people think and changes in them will show up like sore thumbs in a difference recording.
BTW, maybe not exactly for this thread, but I think some mention should be about "hidden detail" and "perception of hidden detail", as they aren't likely to be the same. Different things might make music seem more detailed, and some won't have anything to do with more accuracy or S/N. So something that shows more perceived detail might even have less "detail" that might be determined mathematically. A good example is SET tube amps and my system speakers. The SETs distort like hell (all that "clean first watt" stuff, when measured, is still miles from even approaching a chip amp's distortion on any harmonic), and their response changes with speaker load. But either the distortion or the change to eq gives more perceived detail, small things just get noticed more.
BTW, what would happen if all the reviewers (the ones who "heard things" they'd "never heard before" on their old records, after using a new component), would they hear those same things now if they listened again with the system restored to how it had been before? Would those thing in the recordings be inaudible, unheard, again? I doubt it. I think with ears, it's a matter of getting things noticed as much as it is about getting noise and distortion down.
I'll stay tuned...
Bill
Thanks Bill for the advice and info. Yes, it's a pain. I was very careful not to change anything and quickly do the two recordings back to back. But even that may not be enough. Background noise is hard to avoid, too.
True that a lot of different things can lead to a perception of more detail. Frequency response being a big one! Just trying to find thru testing if anything else can be detected, and if it even matters.
True that a lot of different things can lead to a perception of more detail. Frequency response being a big one! Just trying to find thru testing if anything else can be detected, and if it even matters.
RESA
I just found out about this thread. Very interesting - I plan on trying it.
I have a related topic here: http://www.diyaudio.com/forums/full-range/273348-resonance-enhanced-selective-amplification-resa-aka-ddr.html
I just found out about this thread. Very interesting - I plan on trying it.
I have a related topic here: http://www.diyaudio.com/forums/full-range/273348-resonance-enhanced-selective-amplification-resa-aka-ddr.html
What I would do is make a single test track of the clean and contaminated clips, repeated over and over again - so: clean, contaminated, clean, contaminated, clean, contaminated, etc. Careful splicing and joining would ensure that the click markers occur exactly equally spaced along the track; and then record that in one go, say at different volumes for the DUT. This means that DAW editing will needed to pull out the recorded contaminateds, to align with the cleans - but the big upside is that conditions will have barely varied between the two versions being played, for best matching.
Thanks for the link! Drivers are such imperfect animals that it's hard to balance the faults of one vs the virtues of another. Faults in one area may actually make up (subjectively) for faults in another area. Psycho-acoustics are complex.
But I think with some work, sweat and thought, we should be able to tease a lot more useful info out of our tests and measures.
A little OT, but the JBL EON measures very flat. I was surprised, because it's an awful sounding speaker. Tubby, thick and heavy describe it for me. What is it doing wrong that is not showing up in the on-axis F.R.?
Could be ringing in the LF due to plastic panel resonances, or a very narrow power response at high-frequencies.
Also, if its voiced for stand-mounting, putting it on the floor will gain you ~6dB of lower midrange immediately.
I haven't found a JBL Eon that I like. (NB, only played with 15" ones)
The earlier ones with the exposed horn are a compact way of making a lot of noise, but the later ones (full grille) I've had running into limiting in a small pub with dude + guitar. Nowhere near the oomph they ought to have.
Chris
Also, if its voiced for stand-mounting, putting it on the floor will gain you ~6dB of lower midrange immediately.
I haven't found a JBL Eon that I like. (NB, only played with 15" ones)
The earlier ones with the exposed horn are a compact way of making a lot of noise, but the later ones (full grille) I've had running into limiting in a small pub with dude + guitar. Nowhere near the oomph they ought to have.
Chris
Yes, I don't think the plastic case does them any good. I bought a dozen of the 15s back in '96 when they first came out. Never liked them. Wanted the 10s, but they were having production problems. The 10s were much better for my use.
The QSC K8 is a great sounding powered speaker in a plastic case - best I've ever heard. Why so much better than the EON? I don't know.
If we can actually get this test to work, it might tell of something new about what gives a speaker its distinct sound.
The QSC K8 is a great sounding powered speaker in a plastic case - best I've ever heard. Why so much better than the EON? I don't know.
If we can actually get this test to work, it might tell of something new about what gives a speaker its distinct sound.
Now this is interesting! Could you explain exactly where it went wrong, what the problems were - I would appreciate that, thanks!
Yes, getting 100% correct alignment, or as close to it as computationally possible is key, and quite a bit of fiddling may be required to achieve that - if the clock of the ADC has slight drift for any reason there'll be a problem, but this can be compensated for. Also, to get precise alignment of the offset samples may take some doing, one may need to go to decent levels of upsampling to achieve optimum alignment, a 1 sample offset can make all the difference(!!) - I needed to use almost ridiculous levels of that sort of adjustment to get good nulls in some experiments.
Wasn't there some "magic" audio file comparison program ... yes, here's an old thread on it, "DiffMaker" apparently does all those difficult resampling-and-sample-matching things automagically:
http://www.diyaudio.com/forums/digital-source/207436-software-tool-comparing-wav-files.html
I'm agnostic, even a bit skeptical about all this DDR stuff, but if you guys actually find something new and useful, great.
http://www.diyaudio.com/forums/digital-source/207436-software-tool-comparing-wav-files.html
I'm agnostic, even a bit skeptical about all this DDR stuff, but if you guys actually find something new and useful, great.
"Automagically" is the word - simply put, it doesn't work! I played with it on a couple of occasions, and gave up, in frustration each time - I fed it with a couple of self-contaminated files, done in an exceedingly straightforward way, and it completely failed to make sense of what was done. IOW, it can handle only a very small range of problems, and in many cases will crash or output a nonsense result, absolutely miles away from the real issue - again, a poorly executed, untrustworthy tool is worse than no tool at all ...
At the moment the best approach is very careful manual manipulation, and analysis of the files ... speaking of which, Pano, would you still have any of the files which were recordings of a test signal playback which combined clean and contaminated in a single pass? If you do, I would be interested in having a look at the data, to see if I could extract something meaningful from there - if you were willing to upload one to be available?
Thanks,
At the moment the best approach is very careful manual manipulation, and analysis of the files ... speaking of which, Pano, would you still have any of the files which were recordings of a test signal playback which combined clean and contaminated in a single pass? If you do, I would be interested in having a look at the data, to see if I could extract something meaningful from there - if you were willing to upload one to be available?
Thanks,
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.