To a certain amount. But with a smart process you get more resolution the lower the frequency.
Controlled DBT. Your choice of the format (e.g., ABX, triangle, sorting...). Recordings must have identical mastering, the biggest reason why SACD and CD often actually do sound different. I think the basic idea of M&M was a good one, using a high res source and interposing a 16/44 "bottleneck." There are a lot of other ways to do it as well- the important thing is same master, level-matched, controlled for non-auditory cuing. If you're seriously interested in doing it, contact me via email and I'll be happy to help you set it up.
I convinced myself that at least I can't hear a difference at normal listening levels by making 16/44 copies of 24/96 recordings (my home-made ones) and using foobar for an ABX, but that's not well-controlled.
I convinced myself that at least I can't hear a difference at normal listening levels by making 16/44 copies of 24/96 recordings (my home-made ones) and using foobar for an ABX, but that's not well-controlled.
Identical mastering ...?
It's going to be very difficult to observe such. I will have to use a true 24/192 of which 16/44 exist for the comparison. I have such with 16/44 and then hi-res download, how would i confirm the mastering ..?
we did a similar DBX test 6 months ago and i could tell 12 out of 12 times the 24 bit recording from the 16. 16 bit recordings are more forward and in your face and has more noise, 24 bit is softer , a bit laid back and has more of an natural presence.
Maybe you could provide me with 2 recordings done your way for the test...? It must be true 24 bit not up sampled.
It's going to be very difficult to observe such. I will have to use a true 24/192 of which 16/44 exist for the comparison. I have such with 16/44 and then hi-res download, how would i confirm the mastering ..?
we did a similar DBX test 6 months ago and i could tell 12 out of 12 times the 24 bit recording from the 16. 16 bit recordings are more forward and in your face and has more noise, 24 bit is softer , a bit laid back and has more of an natural presence.
Maybe you could provide me with 2 recordings done your way for the test...? It must be true 24 bit not up sampled.
@Ed, I can tell the difference without even listening, so that's the problem I referred to with "control." 
My experiment was fine for my own education, but if I thought I heard differences, a properly skeptical third party would want me to control things better before accepting my results.
You can download a couple of the high res versions of my recordings (I've posted links to the Soundcloud site) and save 16/44 versions if you have software like Goldwave (what I used).
@a.wayne, these aren't upsampled, I recorded tham in 24/96 format directly. As I told Ed, if you have high res material, you can easily convert it to 16/44 for your own comparison and amusement. I use Goldwave (cheap and versatile), but there's lots of other good audio editing software out there.
My experiment was fine for my own education, but if I thought I heard differences, a properly skeptical third party would want me to control things better before accepting my results.You can download a couple of the high res versions of my recordings (I've posted links to the Soundcloud site) and save 16/44 versions if you have software like Goldwave (what I used).
@a.wayne, these aren't upsampled, I recorded tham in 24/96 format directly. As I told Ed, if you have high res material, you can easily convert it to 16/44 for your own comparison and amusement. I use Goldwave (cheap and versatile), but there's lots of other good audio editing software out there.
Huge:
1. Of exceedingly great size, extent, or quantity.
2. Of exceedingly great scope or nature
(thefreedictionary.com)
That is what I mean by 'huge'. Like the huge difference between night and day, or black and white, or wrong and right. A huge difference would be still noticeable using rubbish ancillary equipment during the most 'stressful' double-blind test. A 2kHz low pass filter would be huge. Half-wave rectified audio would be huge. Any other use of 'huge' could suggest male bragging.
1. Of exceedingly great size, extent, or quantity.
2. Of exceedingly great scope or nature
(thefreedictionary.com)
That is what I mean by 'huge'. Like the huge difference between night and day, or black and white, or wrong and right. A huge difference would be still noticeable using rubbish ancillary equipment during the most 'stressful' double-blind test. A 2kHz low pass filter would be huge. Half-wave rectified audio would be huge. Any other use of 'huge' could suggest male bragging.
I would suggest going from 24/96k to 16/48k. This is easy to do with out error. (drop the last 8 bits and take every second sample).
Unlikely. And if you'll recall, I didn't hear a difference between native high res and down-converted.
no that isn't "legal" down conversion
decimation requires digital filtering to prevent aliasing of content above the target sample rate Nyquist
there is no problem with rational fraction decimation either as long as intermediate word length resolution is adequate - which is required for the filtering anyway
dither with noise shaping would give better perceptual S/N with the word length reduction - is "fair" in that virtually all commercial music CD today use some version of dither with noise shaping
Last edited:
Only if you actually hear a difference. The key is to use a well regarded SRC and redither. I suggest searching the mastering section over on The Womb (http://www.http://thewombforums.com) and see what "the professionals" use.
In my experience, mastering engineers put audiophiles to shame when it comes to obsessing over subtle differences. Where an audiophile may optimise their system to be "musical", MEs generally optimise their systems to be "analytical". How many audiophiles obsess over the audibility of the different types of noise-shaped dither, and which is best for different musical genres? The other main difference is that where an audiophile may consider tweaks such as Shakti stones and unobtanium cored power cables, MEs tend to keep their tweaks in the realm of sound engineering practice.
Thought the same ...
Are you kidding me , do you know any ..? all the ones i have ever met ( highly rated) Do not obsess in the way you describe . Some Producers and Studio engineers , Yes ..
As soon as you do that, you're placing a lot of faith in the resampling algorithm used by that particular piece of software. Have you looked at the source code for Goldwave ? Since it's not available I'm guessing no.
(Linear interpolation ? Cubic spline ? Other ? Floating point or integer based ?)
We can only cross our fingers and hope that the author has done a good job of it, (rather than taking shortcuts for speed or through lack of understanding of signal processing theory) and doing a good job of downsampling and up-sampling is NOT guaranteed by any means, many pieces of software have messed this up badly in the past and continue to do so. Doing it right is non trivial.
Even things such as digital volume scaling are handled incorrectly by many pieces of software, for example older versions of iTunes (thankfully not the current versions, which have a very good scaling / resampling engine) and Winamp were culprits - if you turned the digital volume control down below maximum they simply scaled the 16 bit amplitude values by a ratio and rounded to the nearest integer with no re-dithering. It doesn't take a strong grasp of sampling theory to know that this is the wrong way to do it and will introduce artefacts...
To truly compare a high res sample and a down sampled version of it fairly would require the resampling algorithm to be fully disclosed and scrutinised to make sure it is not introducing digital artefacts that shouldn't be there and could be heard.
Last edited:
Probably true, but like I said, it sounded identical to me, and when it comes to digital artifacts, I'm not exactly deaf. I suspect the errors (if any) are buried below where other noise sources cover them up.
edit: You're right, it's not open source, but they claim that "Resampling is done using a high quality polyphase algorithm with a filter length of 192."
edit: You're right, it's not open source, but they claim that "Resampling is done using a high quality polyphase algorithm with a filter length of 192."
Err, no.
Converting from 24 bits to 16 bits by simply throwing away the 8 least significant bits (truncation) is exactly the wrong way to do it, and will result in digital artefacts. (Assuming that the analog noise floor of the original 24 bit recording was below the 16th bit)
This was a mistake that even some professional mastering equipment made in the early days, and there is bound to be poorly written PC based audio editing software which still does this.
The 24 bit signal has to have 16 bit digitally added dither applied before truncation to 16 bits, which means that the LSB of the final 16 bit version is no longer the same as the "same" bit in the original 24 bit sample.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.