DSP Measurements and Rising Noise Floor

[tw99]

As a result of this thread Tony from miniDSP reached out to me unprompted and gave me an updated firmware for the 2x4HD to test out.

This firmware implemented two improvements: 1) changed the gain structure so volume control acts before ASRC meaning you can use attenuation in the miniDSP to eliminate intersample over clipping and 2) modified IIR assembly code to improve performance with low frequency EQ. As I understand similar firmware improvements are available for the Flex/SHD/DDRC-24 platforms.
...

Overall some really nice improvements in performance and love that they reached out and fixed this issue unprompted.

Michael

There doesn't yet seem to have been a MiniDSP software release that includes this fix. At least there's nothing in the latest release notes from March 2023 that suggests it was included in the last release.

 

[Piere]

Encountered the same issue last year. Looked around also, no fix at the software release pages.

 

[mdsimon2]

There doesn't yet seem to have been a MiniDSP software release that includes this fix. At least there's nothing in the latest release notes from March 2023 that suggests it was included in the last release.

They reached after March 2023 so I wouldn't expected it to be in the release notes. It is disappointing that they are dragging their feet on the update.

Michael

 

[Piere]

Here my findings a year ago of a 24dB/okt LR X-Over at 100Hz. The measurements were done completely in the digital domain through USB, in as well as out.

Flat (blue) versus 24dB/oct LR (yellow), test tones were 30Hz and 50Hz, both at -10dB:

Same test at different signal levels:

Looks noise level is a steady 120dB below signal level. Although at levels regarded inaudible it generates still low level noise modulation.

 

[tw99]

They reached after March 2023 so I wouldn't expected it to be in the release notes. It is disappointing that they are dragging their feet on the update.

Michael

There's a new Device Control release 1.12 which contains both the fixes that you mention above.

 

[Stoutblock]

There's a new Device Control release 1.12 which contains both the fixes that you mention above.

Ironically the new manual uses -6 db as the example for setting the input attenuation

 

[Hayabusa]

Here my findings a year ago of a 24dB/okt LR X-Over at 100Hz. The measurements were done completely in the digital domain through USB, in as well as out.

Flat (blue) versus 24dB/oct LR (yellow), test tones were 30Hz and 50Hz, both at -10dB:

View attachment 301516


Same test at different signal levels:

View attachment 301517


Looks noise level is a steady 120dB below signal level. Although at levels regarded inaudible it generates still low level noise modulation.

To me this looks exactly like the noise floor of single precision floating point in combination with a certain FIR filter length.

for instance 512 taps:

float = 24 bits -> -144dB + 10log(512) = -117dB

 

[Sokel]

Just a quick one to see how media players like foobar loaded with a software filter looks like.

Test rig is:
Foobar 2000 with Mathaudio plug-in
Multitone Analyzer recording with External Source setting
E-MU 0204 ADC
Khadas tone1 DAC
Test signal is 30Hz,96Khz,64-bit FP
Analog I/O


No filter



80Hz L-R 24db filter (aprox)


Both for comparison





Can't see a difference.

 

[ITJ]

I used the TOSLINK output from a 2015 Macbook Pro running at 96 kHz / 24 bit for playback and a Hifime UR23 TOSLINK to USB converter for capture.

Does this also mean that the Hifime UR23 is a digital transparent capture device, or least doesn't generate a big degradation of the incoming signal.

 

[mdsimon2]

Does this also mean that the Hifime UR23 is a digital transparent capture device, or least doesn't generate a big degradation of the incoming signal.

Yes, as far as I can tell it is completely transparent. However, it has some usability drawbacks:

1) Only works up to 96 kHz
2) Sample rate set by computer needs to match incoming sample rate, it will not adapt to changing sample rates

For a single sample rate system, it is a very good solution.

Michael

 

[somebodyelse]

1) Only works up to 96 kHz

Their S2 Digi seems to go up to 192kHz and includes an input. I suspect the setting of sample rate limitation still applies, and probably input and output rates must match.

 

[mdsimon2]

Their S2 Digi seems to go up to 192kHz and includes an input. I suspect the setting of sample rate limitation still applies, and probably input and output rates must match.

I have one of those as well, it definitely works with inputs up to 192 kHz.

The S2 digi is an odd device as it shows up as separate input and output devices in MacOS. In this way it seems like you can actually set separate sample rates for input and output.

The output doesn't seem to work well. I first noticed this when trying to use it as a TOSLINK input / output device with CamillaDSP, I would get audible dropouts that were not CamillaDSP buffer over/under runs.

I then tried it as a pure USB to TOSLINK device and noticed the same issues. This was on my MOTU UL Mk5 setup, it may be that other DACs do not have issues with the S2 Digi but it is the only source I've ever had an issue with the MOTU.

Michael

 

[glc650]

The output doesn't seem to work well. I first noticed this when trying to use it as a TOSLINK input / output device with CamillaDSP, I would get audible dropouts that were not CamillaDSP buffer over/under runs.

So what do you use instead of the S2 digi to avoid dropouts?

 

[mdsimon2]

So what do you use instead of the S2 digi to avoid dropouts?

To clarify, it is only the output that has dropout issues. I don't really have a need for the output as I have a lot of devices that can generate TOSLINK output.

I have several older Mac Minis and an older MacBook pro which have TOSLINK outputs, I mostly use these when testing as it avoids another connected device. For more complicated setups where I need multiple digital outputs / inputs I'll use a MOTU Ultralite Mk5 or a RME Fireface 800.

Michael

 

[ITJ]

Yes, as far as I can tell it is completely transparent. However, it has some usability drawbacks:

1) Only works up to 96 kHz
2) Sample rate set by computer needs to match incoming sample rate, it will not adapt to changing sample rates

For a single sample rate system, it is a very good solution.

Michael

What does a single sample rate source look like?

I am planning to DIY a 2ch AV Receiver. Using a HDMI switch with integrated toslink audio extractor, Camilla DSP as room correction using a raspberry pi.

Would different HDMI sources produce different sample rates? As I don't think I have any control over the output.
I can set the EDID to 2/5/7 channel audio for the HDMI inputs.

 

[mdsimon2]

Many HDMI sources only output 48 kHz, Apple TV falls in this category. Many TVs with HDMI input and optical output will resample to deliver 48 kHz to the optical output.

All of my CamillaDSP systems use an Apple TV as a source.

Michael

 

[ITJ]

Many HDMI sources only output 48 kHz, Apple TV falls in this category. Many TVs with HDMI input and optical output will resample to deliver 48 kHz to the optical output.

All of my CamillaDSP systems use an Apple TV as a source.

Michael

Many thanks for your input.

I did a quick test using the display function of a Topping E30 with the following HDMI sources.
Raspberry Pi: 48kHz
Blu Ray player: 48kHz
Wii HDMI: 48kHz
Chromecast: 48kHz
Windows Laptop: 48kHz

On the laptop I could change the sample rate between 32, 44, 48kHz and 16 or 24 bits, which seems the be the accepted audio settings by the HDMI switch. By default Windows will choose 16bit 48kHz

It seems like this HDMI switch will be pretty good as a single sample rate source.

 

[Nisse10000]

Many thanks for your input.

I did a quick test using the display function of a Topping E30 with the following HDMI sources.
Raspberry Pi: 48kHz
Blu Ray player: 48kHz
Wii HDMI: 48kHz
Chromecast: 48kHz
Windows Laptop: 48kHz

On the laptop I could change the sample rate between 32, 44, 48kHz and 16 or 24 bits, which seems the be the accepted audio settings by the HDMI switch. By default Windows will choose 16bit 48kHz

It seems like this HDMI switch will be pretty good as a single sample rate source.

You might want to have a look at the follwing thread since ta least some are having issues with a similar set-up: https://www.audiosciencereview.com/...ill-get-a-whole-lot-easier-and-cheaper.48233/

 

[Ken Tajalli]

Interesting thread. Thanks to @Sokel for providing a link in another thread & of course to @mdsimon2 for starting it.
I wonder how a Mojo2 would behave in a similar test. After all, one of Mojo2's claims to fame is its internal Transparent tone control system.
A lot of claims by designer Rob Watts, regarding how advanced (good?!) the implementation is, so a test may shine some light on the matter, one way or another.
Just saying . . .

 

[danstahl]

I've put together a brief pros/cons list. Is this an accurate assessment/summary? I'm tempted to go with MiniDSP Flex HT and try it as a pure DAC with Camilla; and if Camilla proves too difficult to work with HT audio (resampling+complex pipelines/filters+ processing speed for audio/video sync), I can fall back to the internal filters.

	MiniDSP Flex HT (USB+Camilla)	MiniDSP Flex HT (HDMI)	Motu (Camilla)
SINAD	Best in class noise and distortion	With internal DSP, noise and distortion rises	Very good noise and distortion
Processing	Variable processing time	Fixed processing time (4.2 ms)	Variable processing time
Tweakability	Highly customizable	IIR/PEQ only, limited filters	Highly customizable
Sample Rate	Fixed (upstream resamples)	Variable (no upstream resampling), internally resampled to 48KHZ	Fixed (upstream resamples)
Input attenuation	Yes, with Camilla	No, digital clipping possible if input is clipped (should be rare)	Yes, with Camilla OR with Motu volume knob
Looks	1MU, sleek	1MU, sleek	1MU, rack mountable with extra ears, not WAF friendly
Cost	950	1000 with wifi dongle	700 with rack mounts