I've been playing around with 'high-definition' digital audio files. I've also been reading people's opinions of same. It seems playback of DSD64 from (for instance) DSF files is accompanied by ultrasonic noise in the 40kHz range.
I have a Topping D10 DAC (the first version), which uses the ES9018K2M DAC chip with an XMOS XU208 USB interface, so it should be able to play up to 384kHz sample rate. I've installed the D10 Theselon driver from the Topping website, and the ASIO wrapper for foobar2000 from hydrogenaudio. For foobar2000's output I selected the Topping D10 ASIO device.
I installed a freebie spectrum analyzer component into foobar2000 and played a hi-res 24-bit/192kHz music file, and saw this:
My DAC is displaying that it's playing 192kHz sample rate, but I only see up to 24kHz displayed in the plugin. The high frequency upper limit of the display doesn't change if I switch to the Default (speakers): Topping D10 output, which is set to run at 96kHz in the Windows Sound settings. The D10 display does change to 96k when I make this switch, so I know the DAC is responding. Is foobar2000 resampling everything to 48kHz sample rate internally?
Also, I notice a rising level of high frequency noise from about -75dB at 10kHz up through -65dB at 24kHz. I wish I could see beyond 24kHz to see how high up in frequency that noise extends.
I also installed Spek and loaded in the same file. I did get a display up to 96kHz (1/2 of 192k, as hoped for). I see this:
I'm not sure what to make of that. There's a pink area extending up a little past 40kHz. That's got to be noise all the way up there. Above that it gets darker, which I believe means the noise decreases to down around -95dB up there.
What is all that ultrasonic noise up there between 22kHz and 40kHz?
Here's a 24-bit/44.1kHz WAV music track loaded into Spek:
The upper frequency limit is 22kHz, as expected for a 44.1kHz sample rate. The file itself has no information above 21kHz, indicating a sharp filter. It looks like Spek is reporting what's there in the audio file, which is what I was hoping for.
I see there is a wider variety of software spectrum analyzers designed to measure incoming signal from an AD converter, so it can measure the performance of audio devices (amplifiers, etc.). Can REW or ARTA be used as a spectrum analyzer of digital audio files? Is there anything that can display a spectrum analysis of DSF files or SACD ISOs?
Is there a better tool than Spek for this?
Thanks for any advice on this.
I have a Topping D10 DAC (the first version), which uses the ES9018K2M DAC chip with an XMOS XU208 USB interface, so it should be able to play up to 384kHz sample rate. I've installed the D10 Theselon driver from the Topping website, and the ASIO wrapper for foobar2000 from hydrogenaudio. For foobar2000's output I selected the Topping D10 ASIO device.
I installed a freebie spectrum analyzer component into foobar2000 and played a hi-res 24-bit/192kHz music file, and saw this:
My DAC is displaying that it's playing 192kHz sample rate, but I only see up to 24kHz displayed in the plugin. The high frequency upper limit of the display doesn't change if I switch to the Default (speakers): Topping D10 output, which is set to run at 96kHz in the Windows Sound settings. The D10 display does change to 96k when I make this switch, so I know the DAC is responding. Is foobar2000 resampling everything to 48kHz sample rate internally?
Also, I notice a rising level of high frequency noise from about -75dB at 10kHz up through -65dB at 24kHz. I wish I could see beyond 24kHz to see how high up in frequency that noise extends.
I also installed Spek and loaded in the same file. I did get a display up to 96kHz (1/2 of 192k, as hoped for). I see this:
I'm not sure what to make of that. There's a pink area extending up a little past 40kHz. That's got to be noise all the way up there. Above that it gets darker, which I believe means the noise decreases to down around -95dB up there.
What is all that ultrasonic noise up there between 22kHz and 40kHz?
Here's a 24-bit/44.1kHz WAV music track loaded into Spek:
The upper frequency limit is 22kHz, as expected for a 44.1kHz sample rate. The file itself has no information above 21kHz, indicating a sharp filter. It looks like Spek is reporting what's there in the audio file, which is what I was hoping for.
I see there is a wider variety of software spectrum analyzers designed to measure incoming signal from an AD converter, so it can measure the performance of audio devices (amplifiers, etc.). Can REW or ARTA be used as a spectrum analyzer of digital audio files? Is there anything that can display a spectrum analysis of DSF files or SACD ISOs?
Is there a better tool than Spek for this?
Thanks for any advice on this.
I found an old demo version of MusicScope, which looks like it was a great program. Oh well, the company closed its doors last year and their free 2.1.0 version is nowhere to be found. But the demo version 2.0.7 has enough functionality for me to capture this:
As you can see on the left side of the graph, most of the recorded information is centered around 400Hz. That makes sense.
What's alarming are the three ultrasonic spikes, which can't have anything to do with musical information.
The lowest one is at 38.4kHz and is at -49dB level. The next one up (to the right) is at 57.6kHz and is at about -70dB. Then there's the big whopper of a spike at 76.9kHz and is down only -34dB.
Is this ultrasonic noise generated by noise shaping? Or is that quantization noise recorded into the track? Does the DAC filter that out? (I suspect it does, with all those FIR choices I see in the PCM5322 chipset options in Moode Audio.
With all these whistling oscillations going on, and the switching frequencies of all my wall warts and class D amps, not to mention the clock frequencies of my laptop and my Raspberry Pi, there must be a lot of ultrasonic energy bouncing around my living room.
As you can see on the left side of the graph, most of the recorded information is centered around 400Hz. That makes sense.
What's alarming are the three ultrasonic spikes, which can't have anything to do with musical information.
The lowest one is at 38.4kHz and is at -49dB level. The next one up (to the right) is at 57.6kHz and is at about -70dB. Then there's the big whopper of a spike at 76.9kHz and is down only -34dB.
Is this ultrasonic noise generated by noise shaping? Or is that quantization noise recorded into the track? Does the DAC filter that out? (I suspect it does, with all those FIR choices I see in the PCM5322 chipset options in Moode Audio.
With all these whistling oscillations going on, and the switching frequencies of all my wall warts and class D amps, not to mention the clock frequencies of my laptop and my Raspberry Pi, there must be a lot of ultrasonic energy bouncing around my living room.
Many DACs use noise shaping which lowers the noise in the audio band and pushes it up above there. So that may be a reason you see the hf noise.
BTW it is logical that you see only audio signal up to 20+ kHz or so, even when sampling at 384kHz.
Sampling doesn't create additional audio. It better not!
Jan
BTW it is logical that you see only audio signal up to 20+ kHz or so, even when sampling at 384kHz.
Sampling doesn't create additional audio. It better not!
Jan
The spectral analysis was of the audio file, not of the output from the DAC. My concern is that those ultrasonic noise spikes are in the file. The DAC would try to play this like it was wanted signal, no?
For comparison, a 24/96k file, also from HDtracks. Note that there are no ultrasonic spikes as there are in the 24/192k file in the first post.
DSD is a 1-bit encoding. It has to use noise (especially ultrasonic) to encode an analog level.
The noise is mostly removed by the filters in the DAC.
16-bit PCM encodes analog levels to one part in 65,536. The amount of noise required to avoid quantization error is much smaller.
Ed
The noise is mostly removed by the filters in the DAC.
16-bit PCM encodes analog levels to one part in 65,536. The amount of noise required to avoid quantization error is much smaller.
Ed
But what happens if a 24/192k PCM file transcoded from a DSD master encodes the ultrasonic noise present in the original DSD file?
Do DACs put a low pass filter on the output to suppress that ultrasonic noise? (I hope so, but I don't know so.)
Do DACs put a low pass filter on the output to suppress that ultrasonic noise? (I hope so, but I don't know so.)
Hello.
I read on your graphs that the music file you tested is Coltrane's "A Love Supreme".
It must have been recorded on tape.
Though I am not versed in analogue recording techniques, I wonder if the suspect 38.4 and 76.9 kHz tones you have observed are not simply the tape bias signal or related to it.