Is M Scaler pass-through bit perfect, and why is the scaled output quieter?

Practical answer. On the Hugo M Scaler, Watts said pass-through is not bit perfect, and that the scaled output is quieter because of a deliberate 2.76 dB insertion loss that pass-through also applies Head-Fi, Dec 2018Head-Fi, Oct 2019Head-Fi, May 2021. M Scaler on versus pass-through is level matched by design Head-Fi, Nov 2018. Comparing either against Dave fed directly, he said in 2019 to turn Dave direct down by 3 dB, and that even then it is "still not exactly volume matched" Head-Fi, Oct 2019.

Is pass-through bit perfect? No. In December 2018 he wrote: "And of course pass-thru isn't bit perfect as it has to compensate for the M scaler filter attenuation" Head-Fi, Dec 2018. In October 2019 he repeated it: "Please remember that pass-thru is not bit perfect - it's intent is to allow one to easily hear the effect of the M scaler" Head-Fi, Oct 2019. In September 2018 he described the mode as having "filter gain equalised," selected with the OP SR button, where red is pass-through Head-Fi, Sep 2018. Inside, he said, the WTA filter becomes "a 1 tap filter," with the memory kept identical so the group delay is the same Head-Fi, Dec 2018. Pass-through follows the incoming sample rate rather than changing it Head-Fi, Nov 2018Head-Fi, Jan 2019. Speaking generally in 2021, he said "You can't have a volume control that is bit perfect, as it obviously changes the data" Head-Fi, Aug 2021.

Why is the scaled output quieter? In October 2019 he said the insertion loss "is essential to avoid clipping from inter-sample overloads and from Gibbs phenomena," and that pass-through has the same loss "to make the pass-thru comparison fair" Head-Fi, Oct 2019. In 2021 he said the -2.76 dB reduction applies "on all outputs including pass-through," so clipping behavior is the same in both modes Head-Fi, May 2021. Speaking of his WTA filters in general in 2022, he said all of them have a built-in 2.76 dB overload, for Gibbs phenomenon as well as inter-sample overloads, which he called "a bit of a red herring"; his measured worst case was 2.41 dB with square waves and 0.8 dB with random data, which he said is more like music Head-Fi, Apr 2022. In May 2023 he gave the setting as -2.73 dB, said about 2 dB of headroom covers Gibbs phenomenon, and called it a convenient number for the SPDIF de-emphasis filter built into his DACs Head-Fi, May 2023.

Level matching by product.

• M Scaler on versus pass-through: in 2018 he said pass-through has the same 2.75 dB loss and headroom, "so it's volume matched" Head-Fi, Sep 2018Head-Fi, Nov 2018.
• Blu MkII era: in May 2017, when the M Scaler being shown was in the Blu MkII Head-Fi, May 2017, he said its 2.76 dB loss meant raising Dave's volume by 3 dB to match, which left the M Scaler 0.234 dB louder Head-Fi, May 2017Head-Fi, May 2017. In November 2018 he said the Hugo M Scaler is "exactly the same as blu 2, with the 3db headroom" Head-Fi, Nov 2018.
• Hugo TT2: in December 2018 he said TT2 clips at +7 dB on the volume, and "Throw in an M scaler, then we are at +10 dB needed to clip it" Head-Fi, Dec 2018. In June 2020 he said -3 dB gives 3 V RMS, the maximum output with 0 dBFS input, and "If an M scaler is used, you need to set it to -4 as the M scaler attenuates by 2.75 dB" Head-Fi, Jun 2020.
• Dave: in April 2019 he gave +3 dB solo and +6 dB with the M Scaler before any possibility of clipping Head-Fi, Apr 2019; in January 2022, +4 solo and +7 with an M Scaler for 0 dBFS sources Head-Fi, Jan 2022. Both answers put the M Scaler setting 3 dB above solo.

Other pass-through behavior. The delay is the same in pass-through as in normal operation, he said in 2019 and 2020 (0.661 s with a 48k source, video mode off) Head-Fi, May 2019Head-Fi, Apr 2020. Only one BNC is active in pass-through, so connect the right cable or select the matching coax input Head-Fi, Aug 2019Head-Fi, Sep 2018. In 2021 he said 705.6/768 kHz files are treated like pass-through: "scaled in volume (so all inputs stay at the same volume level)" and noise shaped to 24 bits for the BNC outputs Head-Fi, Dec 2021. Separately, he told owners in July 2018 to feed the M Scaler bit perfect data from the source Head-Fi, Jul 2018Head-Fi, Jul 2018.

Gaps. The pass-through statements concern the Hugo M Scaler. At CanJam London in July 2018 he said it uses "exactly the same filter core that I've used in the Blu" c w 23:19, but this archive has no statement about a pass-through mode on the Blu MkII, and nothing on pass-through or output level for the Quartet M Scaler. In the same talk he says pass-through "won't change the gain, so that's exactly the same gain" and "You don't need to worry about the volume control" c w 24:08. That transcript is machine-made, so check the recording for exact wording.

Quotes

“Please remember that pass-thru is not bit perfect - it's intent is to allow one to easily hear the effect of the M scaler.”2019-10-18 Head-Fi, Oct 2019

“And of course pass-thru isn't bit perfect as it has to compensate for the M scaler filter attenuation.”2018-12-07 Head-Fi, Dec 2018

“The pass through mode also has the headroom, so it's volume matched.”2018-11-02 Head-Fi, Nov 2018

“this is done with the gain reduction of -2.76dB on all outputs including pass-through.”2021-05-10 Head-Fi, May 2021

Dated positions

Caveats and open points

• Pass-through statements concern the Hugo M Scaler; no statement was found on pass-through for the Quartet or BLU MkII, and nothing on Quartet output level.
• The 2017 insertion-loss posts (Head-Fi, May 2017, Head-Fi, May 2017) predate the Hugo M Scaler and do not name the device; they are read as the M Scaler in the Blu MkII because a May 2017 post places the M Scaler on Blu 2 units (Head-Fi, May 2017).
• He gives the margin as 2.73 dB (2017 and 2023), 2.75 dB, 2.76 dB and -2.766 dB in different posts; he did not say these are different quantities.
• An August 2017 post (Head-Fi, Aug 2017) says 'it' attenuates by 2.73 dB, that you can increase the volume to compensate, and that it won't change the clipping point or maximum power output; the post does not name the product, so it is not used in the answer.
• The 2022 and 2023 overload-margin posts speak of his WTA filters and DACs in general, not only the M Scaler.
• His two Dave volume answers (2019 and 2022) differ by 1 dB in absolute settings; he did not explain the difference. The 2020 TT2 answer (-3 to -4) is a 1 dB shift while the 2018 TT2 clip points shift by 3 dB; he did not reconcile them.
• Head-Fi, Dec 2021 does not name the product; it mentions BNC outputs and is read here as describing the Hugo M Scaler's 705.6/768 kHz handling.

Topics: M Scaler and Quartet settings and operation, Volume, gain and output level