Why are DAC maunufacturers make DSD output 6dB lower than PCM?

[gdpr]

A number of DAC manufacturer's are limitng the output for DSD with -6dB, compared to output for PCM.

I have personal experience with Holo Audio Spring DAC and iFi micto DAC's.  Other models from these manufacturers do the same, as specified in their manuals.

 

I understand that theoretically DSD can go up to 0dB, and even uo to +3dB for short peaks.

 

But in practice, this turns out to be counterproductive, at least when upsampling both PCM and DSD (DSF) to DSD256/DSD512.

 

From the same SACD album, upsampling the SACD layer (DSF file) to DSD256 results in  (+/-) 6dB lower output level, compared with the redbook layer (AIFF file) .

This means I have to crank up the volume by 6dB (on my preamp) to keep same sound level.

 

Using a DAC from one of the forementioned brands,  the output level is pushed down another 6dB, as the DAC receives DSD data.

As a result, I am receiving a -12dB level with DSD, compared with PCM.

 

So, why are these (and probably other) DAC manufacturer's do this? 

 

I find this rather annoying to say the least (one of the reasons why I have sold my Holo Audio Spring dac) 

 

Dirk

[audiventory]

I don't see point decrease level in DAC. Because it is low frequency filter.

 

When DSD recorded or processed overload can cause broken stability of sigma-delta modulator.

In our conversion software we reduce output level at 0.4 dB (headroom). Because we use special stability keeper inside the modulator. It allow keep low noise level and decrease headroom.
But common reducing is 6dB.

 

Read details ("Noise, Maximal Level and Stability Issues" part) https://samplerateconverter.com/educational/what-is-dsd-audio

[xx3stksm]

It's reasonable to have attenuation by 6dB in 1bitDSM as long as your modulator is more than 3rd order. In case of 1st and 2nd order, you don't have such restriction, but low order DSM isn't enough for audio application. High order DSM(5th or 7th) is mandatory to decrease quantization noise. However, 1bitDSM has only two values 1 and 0. If you want to have maximum amplitude, the easiest way is a succession of 1, which isn't allowed in high order DSM. Because high order DSM must have a high-frequency for sufficient noise-shaping.

 

If your DSM outputs a succession of 1, It means low frequency. High order DSM forces you to have a mixture of 1 and 0 even if you want maximum amplitude. That's why what you can have is not 11111111 but 11011011. Average value of 11011011 is 6/8. The same restriction occurs in a minimum amplitude. What you can have isn't 00000000 but 00100100. The average value is 2/8. In 1bitDSM, what you can have is usually not from 1 to 0 but from 6/8 to 2/8. Maximum swing in 1bit DSM with high order ends up 6/8-2/8=4/8. The standard 1bitDSM can't use 100% amplitude. If your DSM is multibit(5bit or 6bit), the restriction disappears even if in high order one.    

[firedog]

I agree with the OP. Roon also does this, which is annoying if I play a mixed PCM/DSD set of songs. Sometimes I convert everything to PCM just to avoid this. 

[Miska]

On 12/10/2018 at 5:27 PM, ddetaey said:

A number of DAC manufacturer's are limitng the output for DSD with -6dB, compared to output for PCM.

I have personal experience with Holo Audio Spring DAC and iFi micto DAC's.  Other models from these manufacturers do the same, as specified in their manuals.

 

I understand that theoretically DSD can go up to 0dB, and even uo to +3dB for short peaks.

 

But in practice, this turns out to be counterproductive, at least when upsampling both PCM and DSD (DSF) to DSD256/DSD512.

 

From the same SACD album, upsampling the SACD layer (DSF file) to DSD256 results in  (+/-) 6dB lower output level, compared with the redbook layer (AIFF file) .

This means I have to crank up the volume by 6dB (on my preamp) to keep same sound level.

 

DSD 0 dB is specified as 50% modulation index that translates to -6 dB PCM level in direct conversion. And short term peaks are specified to be max +3 dB which is 75% modulation index and translates to -3 dB PCM level in direct conversion.

 

Some DACs like ESS apply 6 dB gain to DSD content and as result risk clipping on DSD content that exceeds 0 dB DSD level. They may simply run out of their voltage rails when encountering DSD content with +3 dB peaks.

 

So for better result, if you switch between PCM and DSD for output, turn down the PCM level. Most DACs have lowest THD+N level at about -10 dBFS level.

 

[Miska]

On 12/25/2018 at 8:07 PM, firedog said:

I agree with the OP. Roon also does this, which is annoying if I play a mixed PCM/DSD set of songs. Sometimes I convert everything to PCM just to avoid this. 

 

There's no such problem if you stick to keeping output at either PCM or DSD only. For example I'm not ever seeing such problems. Just don't switch between the two. And if you do, turn down the PCM level. Most PCM content is anyway pushed too high and have pile of digital clipping.

 

Good thing in such is also that you likely end up having less attenuation in your preamp. If you push the digital level to max and then as a result end up keeping preamp set to for example -40 dB volume setting, you first push the DAC analog stages to maximum distortion and then attenuate everything a lot, just to be amplified again up by typical power amp gain of +36 dB... In optimal case your preamp gain would be 0 dB or something on the positive side, instead of attenuation.

 

[jamesg11]

So using roon as front end with hqp, what pcm is optimal (idsd bl)?

/& with a passive pre?