Okto Research
Member
Thank you for the review!
The filter acronyms are explained in the product manual: FRLP = fast roll-off, linear phase; FRMP = fast roll-off, minimum phase; BW = brickwall etc.
The external USB jumper is a necessity, because the standard format Raspberry Pi does not have USB routed to its onboard header. However, it is a plug-and-forget thing and a small price to pay for making it possible to accommodate a standard-format Raspberry Pi.
All the "DAC hats" use I2S, which is limited in capability (no DSD) and require a specific driver for the device (while we wanted to allow users to install streaming software of their choice). The I2S is routed to dac8 Stereo's XMOS microcontroller as well, but it is reserved for future use.
For those interested in what does it take to do such a measurement and what are the limitations of the best audio analyzers out there, I would like to share couple of lines and measurements we have done in-house. The topic has already been touched by @MC_RME.
The APx555 audio analyzer has several voltage ranges which differ in performance. Each of them has its own baseline noise (present independently of the signal provided by the device under test (DUT)). When you increase DUT's voltage to the upper limit of the analyzer range, you defeat that baseline noise, until you get to the point of overloading the specific range where distortion skyrockets. The "sweet spot" for APx555 is at around 5V RMS and 10V RMS. On the other hand, at 4V, it is not possible to measure more than 122dB SINAD, even with a theoretical noiseless and distortionless DUT/DAC. This is shown in the product datasheet:
The full document is available here: https://www.ap.com/download/apx555-...AkQkL4JvNmsD4Aq2yWtnuzPbvVmcX4KY-BCHOYMkW-4uQ
THD+N on the Y axis is just a negative value of SINAD, or in other words, sum of distortion components and noise. You can see that almost all the time the noise is the dominant component, not distortion, since the lines are flat, even though analyzer's distortion definitely rises a bit towards each range's top end.
How were they able to achieve that when there are no ADC chips available with full-scale distortion even remotely close to -135dB? The analyzer overcomes that by splitting the incoming signal into two paths, of which one measures the voltage value of the DUT's output, while the other suppresses the test tone(s) as much as possible with a sharp rejection filter and amplifies the rest, so it can be stitched together again in software. This method is far superior compared to a direct measurement.
Below are measurements of the dac8 Stereo we have done using APx555B, a slightly upgraded version of the leading-performance analyzer with somewhat reduced distortion:
We can see where the analyzer switched its range from 2.5V to the more noisy 5V, strongly impacting the measured value. The full-scale voltage of the DAC is slightly over 4V, so we end there. By extrapolating the measurement, we can guess that if the burden of analyzer's noise was lifted, the result would be 125-126dB SINAD.
Your trolling can be easily exposed. There is no commercial audio analyzer with 128dB SINAD capability. You have no idea about a Fourier transform, otherwise you would understand what would it take to measure such SINAD of 100kHz signal. Also, a mere look at the spectrum of the dongle you are promoting tells us it would spoil the performance of amplifiers like Purifi or Benchmark AHB-2. Finally, a well designed USB device should be agnostic to the noise at the USB host side - the dac8 Stereo is.
Have fun and enjoy music everyone,
Pavel from Okto Research
The filter acronyms are explained in the product manual: FRLP = fast roll-off, linear phase; FRMP = fast roll-off, minimum phase; BW = brickwall etc.
The external USB jumper is a necessity, because the standard format Raspberry Pi does not have USB routed to its onboard header. However, it is a plug-and-forget thing and a small price to pay for making it possible to accommodate a standard-format Raspberry Pi.
All the "DAC hats" use I2S, which is limited in capability (no DSD) and require a specific driver for the device (while we wanted to allow users to install streaming software of their choice). The I2S is routed to dac8 Stereo's XMOS microcontroller as well, but it is reserved for future use.
For those interested in what does it take to do such a measurement and what are the limitations of the best audio analyzers out there, I would like to share couple of lines and measurements we have done in-house. The topic has already been touched by @MC_RME.
The APx555 audio analyzer has several voltage ranges which differ in performance. Each of them has its own baseline noise (present independently of the signal provided by the device under test (DUT)). When you increase DUT's voltage to the upper limit of the analyzer range, you defeat that baseline noise, until you get to the point of overloading the specific range where distortion skyrockets. The "sweet spot" for APx555 is at around 5V RMS and 10V RMS. On the other hand, at 4V, it is not possible to measure more than 122dB SINAD, even with a theoretical noiseless and distortionless DUT/DAC. This is shown in the product datasheet:
The full document is available here: https://www.ap.com/download/apx555-...AkQkL4JvNmsD4Aq2yWtnuzPbvVmcX4KY-BCHOYMkW-4uQ
THD+N on the Y axis is just a negative value of SINAD, or in other words, sum of distortion components and noise. You can see that almost all the time the noise is the dominant component, not distortion, since the lines are flat, even though analyzer's distortion definitely rises a bit towards each range's top end.
How were they able to achieve that when there are no ADC chips available with full-scale distortion even remotely close to -135dB? The analyzer overcomes that by splitting the incoming signal into two paths, of which one measures the voltage value of the DUT's output, while the other suppresses the test tone(s) as much as possible with a sharp rejection filter and amplifies the rest, so it can be stitched together again in software. This method is far superior compared to a direct measurement.
Below are measurements of the dac8 Stereo we have done using APx555B, a slightly upgraded version of the leading-performance analyzer with somewhat reduced distortion:
We can see where the analyzer switched its range from 2.5V to the more noisy 5V, strongly impacting the measured value. The full-scale voltage of the DAC is slightly over 4V, so we end there. By extrapolating the measurement, we can guess that if the burden of analyzer's noise was lifted, the result would be 125-126dB SINAD.
If the Pi5 is going to have the same connector arrangement, then yes. I expect a 4B+ model to come out first, which should definitely be compatible with today's 4B.
We are getting better, answering most e-mails in about 2 working days and cutting down our lead time.
Your trolling can be easily exposed. There is no commercial audio analyzer with 128dB SINAD capability. You have no idea about a Fourier transform, otherwise you would understand what would it take to measure such SINAD of 100kHz signal. Also, a mere look at the spectrum of the dongle you are promoting tells us it would spoil the performance of amplifiers like Purifi or Benchmark AHB-2. Finally, a well designed USB device should be agnostic to the noise at the USB host side - the dac8 Stereo is.
I doubt you have heard the dac8 Stereo in your life. We are a small company that only sell directly, the first dac8 Stereos were shipped just couple of weeks ago, so we know our customers and I don't think you are one of them.
Why not to delete the misleading post then?
The dac8 Stereo does not integrate Dirac software. There are many software products that run on a computer and can do a room correction for you, like JRiver, Roon, Audiolense or Acourate. This of course applies only to content played from that computer. It is also possible to perform DSP on the Raspberry Pi, for example using BruteFIR with Volumio. Again - it would only apply to the content played by or streamed to the Raspberry Pi.
Have fun and enjoy music everyone,
Pavel from Okto Research
Last edited:
