Print

Print


Hi Dave:

Have you done controlled tests (ie same source, same playback chain, perhaps same ADC if it does 
both formats, perhaps same DAC if it does both formats) comparing 192/24 PCM and 1x or 2x DSD? I 
have heard such comparisons in a good mastering suite, using the newest Mytek DAC, although I'm not 
sure the ADC was the same exact unit (or even made by the same companies, since there are only a 
relative few ADCs that do both PCM and DSD, but the number is growing). I also did a controlled 
comparison with my Tascam DA-3000, which to my ears sounds quite good and sounds amazing for $800 
street. The DA-3000 is capable of 192/24 and 2x DSD. In both cases, my ears prefer PCM because I 
think DSD does something to "haze out" the top end a little bit. I think the LP crowd likes DSD 
because the top end definitely sounds un-bright and un-metallic, with normal-sounding source 
material. I found that if you feed DSD over-bright, harsh-sounding source material, it will 
faithfully reproduce the over-brightness and harshness. I do however think there is a subtle "shave 
off" going on at the very top with DSD digital and SACD sound. It's not always a bad thing, but I 
tend to prefer high-resolution PCM because to my ears it's more output=input (for better or worse). 
These differences I hear could well be related to ADC or DAC behavior, not format behavior. But I 
have been surprised that I tend to hear them almost universally when listening closely to familiar 
material.

The general word I get in the world of commercial record companies and mastering engineers is that 
there's a niche market of audiophiles who vocally prefer DSD, but it's a niche market and the cost 
of dealing with trying to work in a pure DSD environment has to be justified by the potential sales 
of the end product. The Japanese market, for example, is still quite viable for certain titles in 
SACD physical media and pure DSD downloads. But, going forward,  the number of titles where the 
extra workflow and expense can be justified is likely to be small. On the other hand, PCM workflow 
is pretty much PCM workflow, regardless of the resolution (although some mastering places really 
don't want to work at more than 96/24 because all of their tools are optimised for 96/24). So I'm a 
big advocate for moving forward and standardizing on 192/24 -- get updated tools and move forward. 
The reason I say this isn't because I hear differences on the same file between 192 and 96 -- I 
can't say I'd pass any blindfolded tests, even if I was working on the file. The reason I want the 
higher sample rate is because I want the transfer to capture things I can't hear that maybe someone 
like Jamie Howarth can use to fix problems that I can hear. For instance, some modern mag-film 
electronics will pass signal up to where the bias was on old recordings made on Westrex dubbers, for 
instance. If you transfer at 192/24 and send the file to Plangent, they can fix the wow and flutter 
caused by the sprocket transport. If you put a line in the sand and refused to transfer beyond 
96/24, you wouldn't recover the bias and, if that piece of mag film rots from vinegar syndrome in 
the vault, what you did at 96/24 is all you've ever got. And by the way, this precludes DSD 
transfers for me because they noise-shape and roll off frequencies up where bias would be, at any 
DSD resolution.

-- Tom Fine


----- Original Message ----- 
From: "Dave Burnham" <[log in to unmask]>
To: <[log in to unmask]>
Sent: Thursday, March 19, 2015 3:37 PM
Subject: Re: [ARSCLIST] Downsample


> Well, Tom, you are absolutely correct and the differences you describe are exactly the differences 
> between CD and SACD, I'd say for the same reasons, but I expect the contrast wouldn't be as 
> dramatic as between 96/24 and MP3.
>
> db
>
> Sent from my iPhone
>
>> On Mar 19, 2015, at 2:57 PM, Tom Fine <[log in to unmask]> wrote:
>>
>> PS -- John Chester is right. If you take something from 96/24 to 44.1/16, there will be some 
>> level of audible difference, depending on the source material. I know there are plenty of 
>> quasi-religious arguments out in webland about "CD resolution is perfect" and there's a flawed 
>> "ABX" test reported in the AES Journal, but trained listeners who listen for their profession, 
>> such as members of ARSC, should hear differences. To take a crack at describing the differences I 
>> generally hear, you hear less of a fade out on a reverb tail, or less "air" around an instrument 
>> being played in a space. On the human voice, I hear a bit less of the low-level stuff that often 
>> happens in the throat and that sensitive mics can pick up, also less breath in a quiet vibrato. 
>> The net result, again to my ears, tends to be a flattening of the stereo image as heard through 
>> monitor speakers, it's more conforming to the dimensions and spread of the speakers rather than 
>> seeming deeper, wider and higher than the speakers. The audible differences shouldn't be drastic, 
>> but tend to be audible to a careful listener. Obviously I'm talking about differences in 
>> well-recorded content with subtle tonal and volume-level contrasts and changes. Something full-on 
>> slamming loud and harsh may sound equally good or bad at many resolution levels, including some 
>> lossy. But, it's worth noting that lossy CODECs use perceptual encoding and often change the 
>> frequency spectrum as heard over good speakers in a full-sized room at normal listening levels.
>>
>> -- Tom Fine
>>
>>
>> ----- Original Message ----- From: "Tom Fine" <[log in to unmask]>
>> To: <[log in to unmask]>
>> Sent: Thursday, March 19, 2015 2:32 PM
>> Subject: Re: [ARSCLIST] Downsample
>>
>>
>>> Hi Steve:
>>>
>>> Since it's free to try, definitely try dBpoweramp:
>>> https://www.dbpoweramp.com/
>>>
>>> The batch converter uses very good SRC and dither-down software, to my ears. By the way, I don't 
>>> know why this is, but their batch converter, using the LAME engine, makes better-sounding MP3 at 
>>> the same bitrates than iTunes using licensed Frauhoffer (sp?) software. Other programs I've 
>>> tried that used the LAME engine sounded worse.
>>>
>>> You can get super-fancy in the area of "downsampling" as you call it. I don't hear any 
>>> difference that I can describe between software that uses similar SRC methods and similar dither 
>>> profiles. I think the differences were more radical back in the day of hardware converters built 
>>> into digital buss racks feeding Sony 1630 machines.
>>>
>>> -- Tom Fine
>>>
>>> ----- Original Message ----- From: "Steven Smolian" <[log in to unmask]>
>>> To: <[log in to unmask]>
>>> Sent: Thursday, March 19, 2015 2:09 PM
>>> Subject: [ARSCLIST] Downsample
>>>
>>>
>>>> What is a reasonabley priced program that will downsample from 96/24 to
>>>> 44.1/16 with no audible sonic alteration?
>>>>
>>>> Steve Smolian
>>>
>
>