[theaudiologist1]

Can anyone here actually notice the distortion on DSD64 recordings? For those who are not aware, I am referring to the noise at the higher frequencies that is apparent on DSD64 that is not present on DSD128+ as the noise floor is increased.I see a lot of people on the internet saying DSD64 sounds "horrible" because of this flaw.

Has anyone here actually witnessed DSD64 sounding bad due to the noise issue? Most of my DSD files are DSD64 and I never had a problem with them.

 

[bigshot]

You are listening to the wrong people I suspect. None of this matters for the average consumer. It doesn’t even matter for critical listeners. For some people too much is never enough.

 

[71 dB]

Most of my DSD files are DSD64 and I never had a problem with them.

There is your answer.

 

[Davesrose]

I see a lot of people on the internet saying DSD64 sounds "horrible" because of this flaw.

How many are a lot? I doubt there are any saying that on internet sites that aren't audio forums. On audio forums, there might be a number. There's also number of folks on audio forums that claim there's an audible difference with power cords. Most my DSD sources are SACDs, which is DSD64 and was the height of "native" DSD albums. It's not limited within any confine of human hearing (be it noise or loud dynamic range). The people saying it's horrible are probably only looking at frequency graphs seeing a meaningless difference that one won't hear.

 

[Tjj226 Angel]

Can anyone here actually notice the distortion on DSD64 recordings? For those who are not aware, I am referring to the noise at the higher frequencies that is apparent on DSD64 that is not present on DSD128+ as the noise floor is increased.I see a lot of people on the internet saying DSD64 sounds "horrible" because of this flaw.

Has anyone here actually witnessed DSD64 sounding bad due to the noise issue? Most of my DSD files are DSD64 and I never had a problem with them.

Wellllll the situation is actually more complicated than what people have been pointing out.

With DSD64, the noise will quickly rise after 20Khz. To combat this sudden jump in noise, designers will try to filter out all audio above 20K. This filter has to be exceedingly steep at 20K. In fact in a perfect world, the frequency response of your dac would look like a perfect flat line out to 20K, followed by an immediate drop to 0. This kind of filter is called a brick wall filter.

In a perfect world, with a perfect brick wall filter, you probably would not notice the difference between DSD64 and DSDXXXX.

Sadly, there is no such thing as a perfect brick wall filter. In fact most brick wall filters actually kinda suck.

What you will often times see is a filter that actually might start rolling off your audio at 18khz so that the filter can achieve a 20db roll off at 20Khz. Different manufacturers will do different things in order to tune this filter so that it does the least amount of damage, but it's pretty much a crap shoot. Of course that isn't to say that filters can't be done well, but it is exceedingly difficult.

With DSD128 and 256, the sudden jump in noise occurs muuuuch further out in the frequency response. This means that the filter can have a slower roll off and be placed well out of the audio band. Obviously, this means the filters are much easier to design well and are virtually inaudible.

Long story short, it's most likely the DAC's fault.

 

[bigshot]

Good luck hearing noise above 20kHz at all.

 

[Davesrose]

Wellllll the situation is actually more complicated than what people have been pointing out.

With DSD64, the noise will quickly rise after 20Khz. To combat this sudden jump in noise, designers will try to filter out all audio above 20K. This filter has to be exceedingly steep at 20K. In fact in a perfect world, the frequency response of your dac would look like a perfect flat line out to 20K, followed by an immediate drop to 0. This kind of filter is called a brick wall filter.

In a perfect world, with a perfect brick wall filter, you probably would not notice the difference between DSD64 and DSDXXXX.

Sadly, there is no such thing as a perfect brick wall filter. In fact most brick wall filters actually kinda suck.

What you will often times see is a filter that actually might start rolling off your audio at 18khz so that the filter can achieve a 20db roll off at 20Khz. Different manufacturers will do different things in order to tune this filter so that it does the least amount of damage, but it's pretty much a crap shoot. Of course that isn't to say that filters can't be done well, but it is exceedingly difficult.

With DSD128 and 256, the sudden jump in noise occurs muuuuch further out in the frequency response. This means that the filter can have a slower roll off and be placed well out of the audio band. Obviously, this means the filters are much easier to design well and are virtually inaudible.

Long story short, it's most likely the DAC's fault.

With DSD64, it's still above 20K in which there's appreciable noise shaping. There might be more debate about effects of DSD to PCM (or with also mastering of PCM to DSD). I do have a high end stereo SACD player from when the format was at its height. I do prefer listening to it with all SACDs except one (and one can debate about whether it's the native DSD DAC or better analog stages/etc). With the one exception, it's a multichannel SACD: which I can play on my Oppo BD player (that converts DSD to PCM). Sound wise, if I do directly compare: my dedicated player sounds fuller....but if I'm not directly comparing, the Oppo still sounds good (and I wouldn't notice if I wasn't comparing). So if I'm just listening to music, I enjoy that disc with the Oppo as distinct organs are in front and behind me (and I have a good subwoofer to keep the lowest sub-bass).

DSD – The New Addiction by Andreas Koch

 

[Tjj226 Angel]

Good luck hearing noise above 20kHz at all.

Shrug

I have read that it is absolutely vital to filter that noise. However, I have never found a good explanation as to why that is, or even a visual graph to show what the damage would be.

 

[bigshot]

Ultrasonic noise can be loud enough to damage your hearing without you being able to hear it. But if SACDs could make people deaf, there would be a lot of deaf people out there.

The truth is that if you are dealing with DSD, you have plenty of room to roll off. You don't have to roll off with a brick wall filter. It's like oversampling. When you get up above 20kHz, you can apply a normal rolloff and it will work fine. Humans can't hear that high anyway, so if you have a small amount of spill above 20kHz, no one is going to hear it anyway.

 

[71 dB]

With PCM, especially with 44.1 kHz and 48 kHz, filtering ultrasonics is done because the sampling theorem requires to do so. Not using anti-alias filter would cause ultrasonics aliasing into lower frequencies that can be heard.

DSD has significantly more bandwidth, but the dynamic range dies out toward the higher frequencies. With DSD ultrasonics are filtered mostly to prevent IM-distortion produced by analog audio gear. Lower sampling rate PCM is a protection against this kind of distortion. Higher sampling rate PCM can suffer from this, but typically the amount of ultrasonics is significantly lower than with DSD.

So, ultrasonics are dealt with for different reasons in PCM and DSD world.

 

[Tjj226 Angel]

With PCM, especially with 44.1 kHz and 48 kHz, filtering ultrasonics is done because the sampling theorem requires to do so. Not using anti-alias filter would cause ultrasonics aliasing into lower frequencies that can be heard.

DSD has significantly more bandwidth, but the dynamic range dies out toward the higher frequencies. With DSD ultrasonics are filtered mostly to prevent IM-distortion produced by analog audio gear. Lower sampling rate PCM is a protection against this kind of distortion. Higher sampling rate PCM can suffer from this, but typically the amount of ultrasonics is significantly lower than with DSD.

So, ultrasonics are dealt with for different reasons in PCM and DSD world.

See this is what I have read time and time again, but IDK what that really means. How does an ultrasonic frequency cause audible issues in lower frequency signals? I get that it causes IM distortion, but wouldn't ultrasonic IM distortion still be relatively unnoticeable?

I feel like there is a bigger picture I am missing here.

 

[71 dB]

See this is what I have read time and time again, but IDK what that really means. How does an ultrasonic frequency cause audible issues in lower frequency signals? I get that it causes IM distortion, but wouldn't ultrasonic IM distortion still be relatively unnoticeable?

I feel like there is a bigger picture I am missing here.

It all come down to non-linearities in the ultrasonic frequencies, where analog gear isn't necessarily very linear. Non-linearities create sum and difference frequencies. So, for example 30 kHz and 33 kHz are inaudible, but their difference frequency 3 kHz is in a very sensitive frequency range of human hearing. Noise in the ultrasonic range can generate audible noise for this reason.

Why isn't audio gear very linear at ultrasonics? It is because trade-offs in engineering. Allowing ultrasonic range be less linear can help making the performance better in the audible range and vice versa. Non-linearities at ultrasonics is only a problem if excessive ultrasonic content is present in the recording.

 

[Tjj226 Angel]

It all come down to non-linearities in the ultrasonic frequencies, where analog gear isn't necessarily very linear. Non-linearities create sum and difference frequencies. So, for example 30 kHz and 33 kHz are inaudible, but their difference frequency 3 kHz is in a very sensitive frequency range of human hearing. Noise in the ultrasonic range can generate audible noise for this reason.

Why isn't audio gear very linear at ultrasonics? It is because trade-offs in engineering. Allowing ultrasonic range be less linear can help making the performance better in the audible range and vice versa. Non-linearities at ultrasonics is only a problem if excessive ultrasonic content is present in the recording.

Thank you for the explanation, now it all makes sense.

 

[bigshot]

Super audible frequencies can never make music sound any better, because we can't hear them when they are properly reproduced. But they can mess up sound when they are reproduced incorrectly. It's easier to just no deal with them at all.

 

[audiokangaroo]

Can anyone here actually notice the distortion on DSD64 recordings? For those who are not aware, I am referring to the noise at the higher frequencies that is apparent on DSD64 that is not present on DSD128+ as the noise floor is increased.I see a lot of people on the internet saying DSD64 sounds "horrible" because of this flaw.

Has anyone here actually witnessed DSD64 sounding bad due to the noise issue? Most of my DSD files are DSD64 and I never had a problem with them.

DSD64 sounds more opaque than DSD128. DSD64 has a lot of noise above 30 kHz. This ultrasonic noise is potentially audible and it could be the reason why DSD128 sounds better.