Thanks, I found that searching yesterday, there doesn't appear to be much info on it on line that I could find.
It's an interesting analysis. I need to read 3 or 4 times more to get my head around some of it. Certainly harmonics out as far as the accelerometer could measure.
From what I gather, the point JN makes is that if a sine wave close to Fs/2 is amplitude modulated by another sine wave below Fs/2, there may be one side band beyond Fs/2. If this side band gets filtered out by a brick wall filter, the original signal can no longer be faithfully recontructed.
This situation might occur with high pitched instruments with a sharp onset (attack), such as cymbals.
My take on this was at the time JN posted this, that if a sine wave is amplitude modulated so that one side band is beyond Fs/2, the modulated signal violates Nyquist. So after an AD-DA loop, appropriately brick wall filtered and reconstructed, it should look different from the original.
Oh, c'mon, there is plenty of exclusivity with CDs - Japan releases, XRCD, alternative transfers, limited editions and gold discs by Mobile Fidelity and DCC Compact Classics. Then there are companies that put out collector sets in limited quantities, like Mosaic records. Now, try to get your hands on any of these:
Mosaic Records - Out-Of-Print Limited Edition Collections
Is that exclusive enough? 😀
Look at figure 1 and 16. In both cases, the measured acceleration is off the scale in the first 100 to 250 uSec. It is hard to tell the timeframe given the time scales, they did not look closely at the pluck.
Jn
How? Does not happen in regular recorded music.
Modulation is unlikely to occur on acoustic instruments, the controversial hypothetical wave envelope or the likes may only be produced using electronic source which is OT.
Apparently not enough for those complaining bunch. 😀
The example of sine wave envelope modulation of a sine was used because it is the simplest to understand.
The attack and decay of an instrument is actually envelope modulation, but the spectra of the envelope is more complex, and changes with time, rending normal FFT measurement far more challenging.
Jn
Last edited:
Speaking for myself - NO, NO and NO. Call me thick, I have no problem with that 🙂
Last edited by a moderator:
100uS doesn't violate nyquist though? It's the first 50uS that is of interest in this disucssion and their logging doesn't really allow that.
Derfy,
High frequency loss depends greatly on the humidity in the air. I suspect it is less than 2 dB in most home reproduction systems.
What I have mentioned a few times before is that the original CD A/D converters were 9 bit linear with a 7 bit tail. The idea was to avoid zero crossing issues.
BTY I was at the AES convention when the CD standard was passed. I didn't run into anyone who thought the high frequency limit was wonderful. The attitude was we got a standard passed that folks could actually build.
Virtually all the original converters used multistage LC filters. I tried looking up the data sheet for the single can multipole filters that were mass produced for the early CD players. No luck, maybe the amazing George can do it.
Also to confuse the issue there are small children who respond to dog whistles!
Why waste time on folks with fixed ideas?
High frequency loss depends greatly on the humidity in the air. I suspect it is less than 2 dB in most home reproduction systems.
What I have mentioned a few times before is that the original CD A/D converters were 9 bit linear with a 7 bit tail. The idea was to avoid zero crossing issues.
BTY I was at the AES convention when the CD standard was passed. I didn't run into anyone who thought the high frequency limit was wonderful. The attitude was we got a standard passed that folks could actually build.
Virtually all the original converters used multistage LC filters. I tried looking up the data sheet for the single can multipole filters that were mass produced for the early CD players. No luck, maybe the amazing George can do it.
Also to confuse the issue there are small children who respond to dog whistles!
Why waste time on folks with fixed ideas?
Last edited:
I've tried on and off to find evidence of sidebands being created by acoustic instruments. It appears if and when sidebands are created they are due to system nonlinearity
Agreed, and agreed.
Problem is that it provides the possibility that the sound showed up that fast.
Jn
There is certainly the possibility of an acceleration that fast. I haven't digested enough to see if they managed to work out the envelope of the sound caused by that acceleration.
Fascinating stuff though. Shows just a small number of the variables a Luthier has to balance to get the required sound.
Fascinating stuff though. Shows just a small number of the variables a Luthier has to balance to get the required sound.
This is at last a well formulated thesis, thank you very much.
Since JN is not going to come with a prove, maybe together we can bring this to some conclusion because it seems easy to check.
So the product of two frequencies is needed, both below FS/2 (in my case 20Khz), so let's say one frequency is at 19Khz and the other at 4Khz.
The combination gives 15Khz and 23Khz in the 10msec time window that I have used.
I have multiplied this time signal with a raised cosine to avoid abrupt starting and endings.
Image below at the left shows the modulated signal first, then the 20Khz brick wall filtered version in the middle and the difference between the two at the bottom.
The second image shows the spectra of the 3 time domain images, the original, the filtered and the delta between the two.
So from the original product of 19Khz and 4Khz, being the sum of 15Khz and 23Khz, only the 15Khz signal is kept as expected.
2*Sin(19Khz)*Cos(4Khz)=Sin(15kHz)+cos(23Khz)
However one should not regard the confusing product 19/4, but the 15/23 additition, because this 15/23 is the acoustical signal.
So yes, the 15Khz we could hear before filtering is still there after filtering.
I see nothing alarming.
In case I'm overlooking something, just let me know.
Hans
P.S. It is a complete mystery to me why it took so many postings, to get to this simple result.
Last edited:
Matt, please don't forget it all started out as a thought experiment. The amplitude modulated sine wave etc. is just a halfway house on the road to reality. Which may, or may not, be there, untill we open the box.
I went one step further in my attempt to find a replacement circuit for the ESL63, this time with both transformers in place.
Both transformers have a gain of 75, giving 150 for the total gain.
I guessed the inductance from other similar transformers.
But since transformers are not at all my world, this may be incomplete, so help is appreciated.
I started with the impedance curve as seen from the PA, see image top left.
Then I rearranged the secondary components in such a way, that I got 3 sections.
The image at the right shows in green the signal coming from the transformers.
Blue looks to be the signal responsible for the bass, red for the middle section and cyan for the high end.
Again, any comment is welcome.
Hans
Both transformers have a gain of 75, giving 150 for the total gain.
I guessed the inductance from other similar transformers.
But since transformers are not at all my world, this may be incomplete, so help is appreciated.
I started with the impedance curve as seen from the PA, see image top left.
Then I rearranged the secondary components in such a way, that I got 3 sections.
The image at the right shows in green the signal coming from the transformers.
Blue looks to be the signal responsible for the bass, red for the middle section and cyan for the high end.
Again, any comment is welcome.
Hans