Curse of the Fourier / Combinatorial Explosion

[Mr Boochie]

That's the thing about progress, it's not linear. Sometimes I wonder how many stone ages we've been through.

Yeah, "progress" or knowledge seems to go in quantum leaps. Sometimes it seems like the right combination of many things, at the right time.

Almost a coincidence.

 

[AudioAvarice]

I personally think some of those 'quantum leaps' are far from coincidence, let's not forget that Nikola Tesla's papers have not seen the light of day since seized by the government after his death in the 1940's.

 

[Olle]

Way back when before digital there was always a debate on how much band width we really needed beyond 20 to 20,000. Mac amps were 16 to 40,000 HZ +/- .1 db with less than 10% phase shift. Phase shift is just another way of expressing frequency response. 360 degrees is 3 db. Band width is +/- 3 db or a 6 db variation. Ampex suggested 5 times, but did that mean 10 to 50,000 HZ or 20 to 100,000 HZ. I have for gotten. Another thing was how much signal to noise ratio was needed. If you have two pieces of electronic with the same signal yo noise then when a signal passes through them you loose 6 db of ratio or gain 6 db more noise than you had before. So If you have a room with say with 24 db of noise in the midrange where the ear is most sensitive and you play your music with peaks of 104 db do you really need electronics with more than 80 or 86 or 92 db signal to noise depending where you are using separates or not. If you have 80 db signal to noise then your distortion has to be less than .01% or you can hear the distortion as noise. .001% requires a signal to noise of 100 db. But if you have a sound 15 to 30 db louder than another sound in a similar spectrum the louder sound is said to mask the quieter sound so you can't really make an intelligent decision as to whether you can identy the quiter sound or distortion. So that means most of the time you can't hear distortion less than 2 per cent if its in the same spectrum as the fundamental sound. So if frequency response defines rise time, signal to noise defines noise and distortion, what left is there to measure. HArmonic and IM distortion are just forms of noise so they can be described by signal to noise it seems. But how to separate the primary sound from the noise to be measred. Thats why we have THD and IM tests. The problem is the IM tests only describe what happens between two frequencies. What happens between 10 or 100 frequencies react with each other. Thats the kind of testing we really need I suggest.

I would take my dual trace 100 MHZ oscilloscope and plug the input to my scope and then sample the out put going to the speaker of the units under test. I would use a 700 HZ tone to calibrate the two channels with a gain of the test unit set about 40 db and line up to two wave forms so the looked identical then it the differential function to fine tune the adjustments so the difference became as close as possible to a straight line ever increasing the sensitivity of the display looking for differences between the two traces. Which be distortion between the input and out put. Then I would remove the 700 HZ calbration tone and insert music or pink noise. Some units under test did a great jon some not so great. I tried using pink noise but the results were almost just to hard for my mind to decide which unit was best. All I could do was look at the envelope of distorted noise and decide which units had the smallest height of the envelope of noise. I will say older tube units had the widest variations in performance. SS units the best especially as more time passed during the late 60's 70's and early 80's. Units using digital stages were horrible at first during the early 80's and early 90's.

The question is or was what I had seen correlate to what I heard. Most of the time I would say yes. But there are always exceptions to the rule. So my method wasn't perfect for sure. It was just my way of looking at things at the time. I would think with todays powerful computers the answers could be much more accurate and real numbers could be assigned to the changes as the signal went through each piece or groups of electronics. Then we would know the difference. Then how do we interpret that difference so we can eliminate it? Thats way beyond me today and was back then in the 60's and 70's.

It was just and idea I had on how to see the distortion electronics were adding to the sound we heard. It was the total distortion I was seeing. I didn't know what was frequency errors, noise or harmonic or intermodulation or time shifts, etc etc. Maybe If I had used different test frequency to calibrate the signals of input versus output I would have gotten better results. But I didn't.

Maybe a spectrum analyzer on the distortion product would have been useful

 

[onanysunday]

This may just be review for some, but one way of skipping ahead to what matters (with all unseen factors combined) is running a real-time analysis (Audyssey or otherwise) within the acoustic listening environment from a listening position using a microphone to see the sum total / actual (aggregated) frequency response curve with everything added together. This can be achieved through software or EQ /mastering processors. I use an upgraded Behringer DEQ2496. It gives the final, actual FR curve of everything combined together with its unintended dips and peaks displayed at various frequencies. It explains "what/where", but not "when/why/how" -which we still don't have as precise a handle on (curse of the fourier) within all the unseen factors effecting sound; combined inputs/ stages, circuits/ components used to implement them, environment, system synergy and its overall effects to final sound and measured frequency response from a theoretical or mental perspective as heard through an imperfect, acoustic environment beforehand. Measuring the final stage of perception where everything has already been combined together, and adjusting for it afterwards, gives us more control over the sound and gets us closer to what we're trying to achieve. There will always be more uncertainty within combined inputs (and their overall effect on final sound that is heard) when compared to the output of everything already combined together, or what can be measured at the very end (or last stage) as the sum total frequency response curve of all unseen factors combined.