Thank you George. Even with only 4 hours of sleep I actually understand what you are saying. A nice little calculus problem. Actually simpler than that it appears. You only have so many steps that your ADC can slice the sample frequency in time dependent on the peak voltage you have to divide or the smallest fraction of that you can use as a multiple. Surely I'm not saying that well but I do understand. I think I can imagine how this also works depending on whether that sin wave is at 20hz or at 20Khz.
It's always nice to have a good teacher George.
So back to my other question. Now if you used a CDP at what I'll call the normal fixed output of 2v rather than using the adjustable output that the CDP can truncate and you then do your attenuation somewhere else after the conversion to analog is this a different kind of distortion we will see in the end or will you see similar distortion if measured at the speaker input terminal? So rather than having the nv output pp you have full voltage swing at the output of the CDP and you reduce that gain in the analog side of things with some form of attenuator. Which would be the preferred method?
It's always nice to have a good teacher George.
So back to my other question. Now if you used a CDP at what I'll call the normal fixed output of 2v rather than using the adjustable output that the CDP can truncate and you then do your attenuation somewhere else after the conversion to analog is this a different kind of distortion we will see in the end or will you see similar distortion if measured at the speaker input terminal? So rather than having the nv output pp you have full voltage swing at the output of the CDP and you reduce that gain in the analog side of things with some form of attenuator. Which would be the preferred method?
If on the other hand Richard is talking about the distortion created by the digitisation of this weak signal, here is what I think:
The precision in height of slicing (which is around ½ LSB and which is close to what is called quantisation error) is -intuitively calculated- 30nanoV and when mathematically derived, it is even less.
So this 30nV is the 1/3 of what the 0.1% of the 1mVpp (which is 1nV) that Richard is talking about.
I hope the numbers are correct
George
Good Morning all!
Yes. this is what I was getting at.
Now throw in some numbers from jitter, analog noise and harmonics and various other artifacts and we have a hard time In Practice reaching and keeping below those numbers in a system of components.
So, this brings me back to the conclusion that more bits to throw away for the cause to get the range of low distortion required to be truely transparent in practice. Which is starting to happen for other reasons as well (clipping headroom, etc) BUT, The masters which are recorded at 24/96 or better now but can only be accessed via those HD/HiRes file downloads. When you get rid of all the processing to make and playback a spinning LP or CD (16bits) AND it is 24/96+ the improvement is - I dare to say - dramatic.
Thanks for the fun, guys. Now back over to JC.
THx-RNMarsh
anybody capturing analog today for audio in a studio is using 24/96
the limit is the room, mic, preamp's, other analog noise floors for best in class audio ADC having >120 dB S/N (nobody's delivering 144 dB S/N regardless of the 24 bit math)
and you could imagine sound engineers actually have some concept of system gain structure, don't regularly burn 60 dB of level
for the dominant multibit delta sigma monolithic converters today the -60 dB re fs THD is way below the noise floor
even 16 bit RedBook CD with noise shaped dither is delivering spot noise in our sensitive 3 KHz +/- region of >120 dB
the 1st half of http://downloads.izotope.com/guides/izotope-dithering-with-ozone.pdf looks very readable, shows noise floor plots of real dithers used in modern audio tools
the limit is the room, mic, preamp's, other analog noise floors for best in class audio ADC having >120 dB S/N (nobody's delivering 144 dB S/N regardless of the 24 bit math)
and you could imagine sound engineers actually have some concept of system gain structure, don't regularly burn 60 dB of level
for the dominant multibit delta sigma monolithic converters today the -60 dB re fs THD is way below the noise floor
even 16 bit RedBook CD with noise shaped dither is delivering spot noise in our sensitive 3 KHz +/- region of >120 dB
the 1st half of http://downloads.izotope.com/guides/izotope-dithering-with-ozone.pdf looks very readable, shows noise floor plots of real dithers used in modern audio tools
Last edited:
Not heard of DVD-A or SACD then?
Of course much HD material is DSD (so effectively PWM) and we haven't even considered multibit vs single bit DACs yet.
Dvv, I think that your questions are difficult to answer properly.
First, there have been many articles on minimum distortion, BUT THD is one of the worst predictors without weighing the harmonics STRONGLY. Putting any kind of number down, without qualifying the harmonics is almost pointless.
Speaker sensitivity is interesting in that there are limits on physical drivers that keep the dynamic range to something 'managable' no matter what the sensitivity. On one hand you have the problem with clean power and heat problems with inefficient speakers (85dB), and cost, excursion limitations, and even Doppler distortion with efficient speakers with limited cone area.
Besides, what is important is what we can hear with quality audio playback equipment, rather than any single number that we could state. Yet, I do know what you are talking about, so keep on truckin' !
First, there have been many articles on minimum distortion, BUT THD is one of the worst predictors without weighing the harmonics STRONGLY. Putting any kind of number down, without qualifying the harmonics is almost pointless.
Speaker sensitivity is interesting in that there are limits on physical drivers that keep the dynamic range to something 'managable' no matter what the sensitivity. On one hand you have the problem with clean power and heat problems with inefficient speakers (85dB), and cost, excursion limitations, and even Doppler distortion with efficient speakers with limited cone area.
Besides, what is important is what we can hear with quality audio playback equipment, rather than any single number that we could state. Yet, I do know what you are talking about, so keep on truckin' !
Well a sound bar is just a shrunk radiogram with new problems and costco sell a lot of those.
Richard, what PMA is showing is the SIGNAL AVERAGED distortion after dithering, I'm pretty sure. I think that is where the essential problem lies. Dithering, WHEN averaged, hides the essential nonlinearity that has to be there by the finite number of bits used at low levels. It hides it pretty well, in fact, BUT the instantaneous nonlinearity is still there. OK, SY prove me wrong, I would like to believe that 'digital is perfect' in principle.
I would suggest you read the carefully supervised DBLT of 16 bit system results found wanting and why by F. Toole.
What impressions do you get out of the listening from direct downloads of 24/96+ master files? In fact, anyone. A lot of downloads have occured and the Big Boys are also into it big time. HiRes. Great news for the HiEnd goals of more realistic sound.
Its isnt too far fetched to say it is closer to what you would get with a first gen copy of an analog master in your home for play back. Now that would make an interesting comparison. HiRes vs 1st gen analog master.
-RNM
Last edited:
Hi Fas, Kind
I don’t mix recordings for a living but dynamics and stereo image have been primary interests for a good chunk of my life.
I don’t see the problem as being mostly the medium but what is recording industry custom now and what is possibly the primary thing limiting dynamics in a home hifi.
If one takes an oscilloscope and looks at the amplifier output, shockingly often one finds instantaneous voltage clipping on peaks.
This may be the amplifier but can also be from any part of the signal chain upstream.
About 15 years ago, QSC was going around demonstrating their amplifiers with an A/B/X switcher.
They came to the company I worked at and we set up a pair of Unity horns and about a dozen different amplifiers, including my Threshold stasis II. I had my favorite test CD’s and spent a while listening back and forth and eventually had a couple tracks where I thought best brought out a difference between the amp’s.
Then we went on to the ABX switcher. The amplifiers tended to fall into two different sounds which I could readily discern but these differences were small AND not at all what I can explain. The difference was on the decay side of transients. With the “good” amplifiers, the decay was smooth like an asymptotic curve but the “other” sound was where the decay became granular or like it has steps in the amplitude or something???.
The really unexpected thing was relevant to the post though, with my stasis compared to the good pro amps, at a level not higher than I had used in my living room, my stasis began to sound “less dynamic”. On the LED indicator the peaks were in the -20 to -10dB range so this was weird and weird they were undetectable as different amps up to that point.
An oscilloscope on the Threshold output showed that around the level where the dynamics were altered, what I saw was instantaneous voltage clipping.
Everyone knows what “clipping” sounds like BUT this form of clipping you can’t hear as a “flaw” because it may only exist for one or two cycles, VERY SHORT, not like icky clipping and I could only hear it by switching between “with and without” and the “sound was simply being less dynamic.
Since then, in larger systems, I have found dynamic limiting due to gain structure being off is also a common issue and the best, maybe only easy way to find out is with an oscilloscope playing dynamic music.
Another little appreciated thing is that the “VU” meter represents more or less “how loud it sounds” while how loud it IS can be very different. It is another view one can have looking at the microphone voltage with an oscilloscope or having a “peak hold” sound level meter. Two tests I did at home were eye openers. I have a tile floor in the kitchen, tossing a table spoon on the floor, from a distance of about 5 feet to the SLM, produced a peak in the mid 130’s. Sitting in my car, with the windows rolled up and then slamming the door closed normally, produced a peak of nearly 140dB. Now, if you record these sounds and limit those peaks, it still sounds like a spoon hitting the floor and a car door closing, it just doesn’t sound real, in part because some of the peak is missing.
Due to the peaks theoretically required to re-produce the sounds one see’s as mic Voltage, it is not often done.
I have a couple recordings that might be fun for you guys interested in dynamics.
Using speakers from work at home, realistic peaks are no problem but the recordings were. The dream I have had most of my life has been to capture a live event converted into realistic stereo image and reproduced it such that one can be fooled when visual clues are removed.. These were done using a microphone invention in the works at work, these are the two front channels, more or less your visual field but the system covers a hemisphere with 8 channels. As they aren’t compressed, the average level (more like subjective loudness)is very low compared to “normal” recordings and as a result, you need to turn the gain up to reach a subjective “normal” level. Also do try these with headphones FIRST!! As these are difficult to do well with loudspeakers.
The first is a long parked freight train starting up. First they bunch up all the slack in the couplers and then go forward so that they start each car rolling one at a time. These are 200 ton coal cars which made a satisfying sound, recorded about 150 feet from the tracks.
https://www.dropbox.com/s/jq5n4gj4mpptjpn/TrainStart.wav?dl=0
This was the 4th of July parade in Deerfield IL 2013, taken about 75 feet from the road lined with people and several large trees. Like before, turn the gain up to a realistic level.
https://www.dropbox.com/s/8208qvei00qxzxz/parade section3.wav?dl=0
This is an old recording, version 1 at a friends outdoor bbq. His kids were in an Irish folk music group and a number of them played. I left the beginning on the long side as it was fun listening to the sounds of the kids, summer bugs and the neighbors AC unit that came on. The mic array was about 20 feet from the kids, maybe should have been closer.
https://www.dropbox.com/s/c3c0si0r7giud2w/Johns bbqTrack 04.wav?dl=0
Enjoy
Tom
Think about the sampling theorem for a minute or two. I am getting the feeling that maybe this time you're not huckstering and you truly don't understand it.
While you're thinking, relisten to the demos in the Werner Ogiers article on dither I've linked to about a zillion times. Averaging has zero to do with this.
Well, that is just another what i call 'artifacts' in the system which prevents the full dynamic range being explored. BTW -- dynamic range is NOT measured (Standards) by any averaged numbers. it is the peak values measured which in digital systems defines the dynamic range.
I bought this for you guys:
View attachment 1860 AES.PDF
THx-RNMarsh
John you and Richard can really make engineers cringe - dither decorrelates quantization error - don't you understand the meaning of decorrelate? after all you do brag of grad level EE course when it suits you to claim EE expertise that you never actually display in these discussions
If Richard has a clear psychoacoustically controlled demonstration of 24/96 actually scoring better vs 16/44 with competent production in a representative home music playback scenario he should be publishing his own AES papers
Well John, the market I am after isn't made up of people who sit fixated in the "sweet spot" masturbating over soundstage (when they're not busy masturbating over being able to hear the saliva gurgling at the back of their female jazz vocalist du jour's throat).
Or people who want their living spaces occupied by an ugly and disparate collection of boxes connected with a tangle of cables propped up on myrtlewood blocks.
The market I'm after are people who actually enjoy listening to music rather than being neurotically obsessed with "sounds" and who would like something much better than a "cute" three piece affair with two inch speakers and an eight inch fart box "subwoofer" or crappy *** "soundbar."
se
I'm really naive about digital distortion. I just know what I hear, and I am never completely satisfied for some reason.
Of course, I go back a long way, in fact I paid a Phd (Engineering) to do some advanced aliasing filter design, over 40 years ago, and he also gave me the computed harmonics for any number of bits used at any instant. It did not look like PMA's measurement, so what gives? Where did the non-linearity go? I think that is has to have been there, just from the simple model of not using all the bits, all the time.
For the record, I am always just stupid, and I am not marketing anything. I hate the very idea of marketing, and that is why I never advertised Vendetta Research products during their product life. My business partner always held that against me, because we COULD have sold more, IF I would have marketed the products. So just call me stupid and uneducated, but I still have questions. '-)
Of course, I go back a long way, in fact I paid a Phd (Engineering) to do some advanced aliasing filter design, over 40 years ago, and he also gave me the computed harmonics for any number of bits used at any instant. It did not look like PMA's measurement, so what gives? Where did the non-linearity go? I think that is has to have been there, just from the simple model of not using all the bits, all the time.
For the record, I am always just stupid, and I am not marketing anything. I hate the very idea of marketing, and that is why I never advertised Vendetta Research products during their product life. My business partner always held that against me, because we COULD have sold more, IF I would have marketed the products. So just call me stupid and uneducated, but I still have questions. '-)
In order:
Yes, and you might want to catch up on modern technology basics.
Yes, because you peek.
Perhaps you didn't understand enough to ask the right questions? PMA's measurement is what you get when you actually bother to look.
- Status
- Not open for further replies.