And that is essentially what you do get when a system, even poor ol' miserable, badly maligned, dead straight stereo, gets it right. The illusion is that the end wall or whatever, past, behind the speakers disappears, and a space as large as the recorded acoustic opens up, appears in its place. A comparison would be that it is as if the listening room was transported to the exterior wall of the concert hall, say, an opening punched into that space and your listening room attached at that point.
How this can occur is that enough acoustic clues, low level details, get through cleanly to the hearing system, enabling the mind to effortlessly construct this "mirage" ...
Frank
Check out how much electrical interference in the home is dragging down the quality. What I do is shut down the whole house electrically except for the one circuit feeding the audio; and this means pulling every fuse and circuit breaker at the power meter box. Every one: lights, stove, water heater ... except the circuit feeding the audio. Also, make sure that all power cords not needed to run the audio test are pulled from that remaining live circuit. Power down totally every cell or mobile phone, battery operated laptop, you get the idea.
Now, does that make a difference, particularly in the area of treble clarity and cleanness? I've learnt a lot from this simple exercise ...
Frank
Probably well meant advices, but pls check
Marsh Sound Design ~ Experience the Sound with Marsh Sound Design Products !!!
to realize whom you advise.
Marsh Sound Design ~ Experience the Sound with Marsh Sound Design Products !!!
to realize whom you advise.
I've found that the tiniest "imperfection" can make the difference between live, and not live, and only by going through everything with a fine tooth comb, not assuming anything, do I make real progress. More than anything else it's an attitude, an approach that works for me; unless I do what I try with total rigour I'll get misleading results. The line between success and "not so good" is, oh so fine, and that's the point I'm trying to make ...
Frank
Frank
Fas42, you are on to something. My lab is situated in one bedroom of my apartment. In order to do any serious measurements, I have to turn off everything that I can. Then, I still get residuals from nearby apartments. This is a known, IBM did a paper on it, and it can cycle with people's appliance use habits. It seems to me, that we have a dual opportunity to both consider EMI 'proofing' our electronics, as good as possible, and removing the sources of the EMI from our power lines, by either turning everything off (somewhat extreme) to adding an external AC power conditioner to 'purify' the AC waveform. I am up for both, so long as the solutions are 'practical' and somewhat affordable. Sometimes they are not affordable, but still effective.
Last edited:
What you are describing, fas42, is exactly what I go after. I can achieve the MOST success where my close associates and I control both the design and construction. The Blowtorch was made this way. I can get only relatively perfect success when I have my designs made offshore, by others who may be compliant, but not as aggressive at 'getting it right'.
It is sort of what you get when you buy a Honda product rather than an expensive Mercedes, you might say.
It is sort of what you get when you buy a Honda product rather than an expensive Mercedes, you might say.
Last edited:
I hate to stick on this point but little has been done to confirm what is heard or not heard. I for one am not yet convinced. Counter examples no matter how carefully done continue to be dismissed out of hand.
Why aren't they even "metadata".
FFT does what for most everyone in science? It tells us what frequency components are present. You are beginning to sound like those who accuse others of 'believing in' THD.RNMarsh said:
Ask Richard Marsh, for it was he who first used the term. In the context he may mean someone who recognises good sound when they hear it, but doesn't know enough maths or science to properly understand the circuits producing that sound.john curl said:
Last edited:
Almost any microphone will have noise increasing at low frequencies. A fun experiment you might want to try is to put a horn on to a GE Novasensor pressure guage. The lowest range one is not too sensitive (15uV/Pa) but it does go to DC. You will find that barometric pressure has a 1/f spectrum like most things and eventually will swamp out the music.
I might ask why is this relevant?
Last edited:
Ah, DF96, describing engineer Ed Simon as a 'lay listener' was more an implied insult, than anything accurately describing Simon, himself.
For the record, I strongly depend on 'lay listeners' to give me the feedback necessary to make better audio electronics. In fact, pleasing these 'lay listeners' is what I do, both as a personal interest, as well as to earn an income. May there always be happy 'lay listeners'.
For the record, I strongly depend on 'lay listeners' to give me the feedback necessary to make better audio electronics. In fact, pleasing these 'lay listeners' is what I do, both as a personal interest, as well as to earn an income. May there always be happy 'lay listeners'.
SY & Scott
I had wanted to avoid sampling theory but it has clearly raised its head.
If it takes a "Golden Ears" multiple hours of listening to "hear" these "Differences" then what data would we have to capture of real music to do a Fourier analysis of it.
To be as accurate as possible since there are music components to 100 khz we should look that high. We really can't set the whole record length as a few hours but let's cheat and do 10 second samples.
So that would give us 1,000,000 points for a transform. If we use FFT we need n log
operations so that would be 6,000,000. We would want 200,000 samples (or more) per second for 60 seconds per minute and 60 minutes for one hour. That would be 4.32e15 operations.
Now how many instructions will it take per operation?
In 76-77 I built my own Fourier based analyzer as I could not afford a commercial unit. Even 1/3 octave analyzers were a bit steep. But someone may have the ability to do it in fewer instructions than I used so let them answer.
I later built a room simulator based on the TRW multiplier accumulator chip when that became available. It was a fun and insightful toy.
Of course if we want to capture music including direction we will need six channels for mono. With the bandwidth limits of microphones we just might need to use three different types so bring the final count up to 36 channels!
Now amazingly enough there are computers large enough top process this transform! (But then how do you analyze all of it?)
DF
The issue is that you have a definition of distortion that is limited. If you feel better about yourself by denigrating others, so be it.
Scott,
Yes noise increase as the frequency goes down not just for microphones, but for air motion etc. The classic tale is of the folks who built mic preamps that were RF condensor based and good to DC. Their recording meters were always clipping from stuff they often couldn't hear. One time they really went crazy and a bit later they understood why when the thunderstorm got close enough for them to hear the lightning strikes.
I have had to measure the noise impact of steel wheels on steel tracks. The flat spots from braking create giant levels of LF energy. Some of which is perceived by other than one's ears.
I had wanted to avoid sampling theory but it has clearly raised its head.
If it takes a "Golden Ears" multiple hours of listening to "hear" these "Differences" then what data would we have to capture of real music to do a Fourier analysis of it.
To be as accurate as possible since there are music components to 100 khz we should look that high. We really can't set the whole record length as a few hours but let's cheat and do 10 second samples.
So that would give us 1,000,000 points for a transform. If we use FFT we need n log
operations so that would be 6,000,000. We would want 200,000 samples (or more) per second for 60 seconds per minute and 60 minutes for one hour. That would be 4.32e15 operations.Now how many instructions will it take per operation?
In 76-77 I built my own Fourier based analyzer as I could not afford a commercial unit. Even 1/3 octave analyzers were a bit steep. But someone may have the ability to do it in fewer instructions than I used so let them answer.
I later built a room simulator based on the TRW multiplier accumulator chip when that became available. It was a fun and insightful toy.
Of course if we want to capture music including direction we will need six channels for mono. With the bandwidth limits of microphones we just might need to use three different types so bring the final count up to 36 channels!
Now amazingly enough there are computers large enough top process this transform! (But then how do you analyze all of it?)
DF
The issue is that you have a definition of distortion that is limited. If you feel better about yourself by denigrating others, so be it.
Scott,
Yes noise increase as the frequency goes down not just for microphones, but for air motion etc. The classic tale is of the folks who built mic preamps that were RF condensor based and good to DC. Their recording meters were always clipping from stuff they often couldn't hear. One time they really went crazy and a bit later they understood why when the thunderstorm got close enough for them to hear the lightning strikes.
I have had to measure the noise impact of steel wheels on steel tracks. The flat spots from braking create giant levels of LF energy. Some of which is perceived by other than one's ears.
SY & Scott
So that would give us 1,000,000 points for a transform. If we use FFT we need n log
Ed, I already told you student Matlab on my PC does a 4 million point double precision FFT in a few seconds. This is all beside the point no one claimed music or noise is periodic you did.
Take another signal processing example. Take the impulse response of an RIAA filter (out as far as you want, < 24bit LSB?), then convolve that with your signal to apply the RIAA filter. This process is enormously speeded up by using overlapped FFT techniques (convolution is simply multiplication in the conjugate domian). The answer can be proven EXACTLY the same, no "infinities" required.
Opposite, Ed claims that you can't formally treat an acquisition of music or noise as periodic. We've now apparently moved the goalposts, minimally miked "purist" recordings don't count, and you have to look at an hour.
Yes, many new tests should be done, for the reason that much of "confirmed" was done 50 and more years ago on a bearded equipment.
It's the new meme. Get with it, Gramps. You'll see it next week in Urban Dictionary.
- Status
- Not open for further replies.