Case Study: Achieving 93.77dB Flat SNR with Yamaha Montage M—Pushing Practical Limits at Home(Z-weighted/Flat)

[waiwai]

[Introduction] Hello ASR community. I’ve been working on a project to push the measurement limits of a home environment, focusing on how far consumer gear can go when the electrical environment is fully optimized. I wanted to share a milestone achieved using my Yamaha Montage M as the source, connected to a Focusrite Scarlett 16i16 (4th Gen).

[The Achievement] To establish a pure source signal, I utilized the Montage M's onboard ANX (Analog Physical Modeling) engine to generate a 1kHz sine wave. This allowed me to test the instrument's output stages at their peak potential. To ensure measurement integrity, both the Montage M and the Scarlett 16i16 were operated at a matched 48kHz sampling rate. Furthermore, by using the OCXO-stabilized Scarlett as the recording master, I aimed to capture the Montage’s analog output with the highest possible timing precision and minimal aliasing.

• SNR (Z-weighted / Flat): 93.77 dB
• THD+N: 0.00052%
• Noise Floor: Stable below -140dB across the audible spectrum.

[Practical Context & Optimization] What makes this 93.77dB (Flat) significant is that it was measured with the actual instrument fully connected and active, rather than a shorted input or a disconnected state. This represents a real-world recording/monitoring scenario.

I achieved this through two main pillars:

1. Optimization of power supply and grounding: A systematic overhaul of the power delivery and grounding topology across the entire chain to eliminate common-mode noise and ground loops.
2. Master clock synchronization: Integration of an 8663 OCXO as the master clock for the Scarlett 16i16. While the Montage M does not accept external clock input, stabilizing the recording interface's time domain with a high-precision reference was crucial for capturing the instrument's analog output with maximum phase integrity.

[Subjective Impact: The "Brain Glitch"] While the numbers are satisfying, the subjective result was startling. During a recent ASMR monitoring session, the transparency reached a level where the hardware "disappeared." A faint stomach growl from the performer was reproduced with such phase accuracy and lack of floor noise that I momentarily mistook the sound for my own body. This "brain glitch" confirms that reducing the noise floor to this level has a profound impact on spatial localization and realism.

[The Road to >100dB] I am currently finalizing a formal academic paper that details the specific schematics and grounding methodologies used here. To maintain the integrity of my upcoming publication, I cannot share the full technical details just yet, but I wanted to demonstrate what is possible even with a consumer interface.

My next goal is to solve current challenges involving interconnect shielding and further power isolation to reach >100dB SNR (Flat/Z-weighted) before the final paper submission. I look forward to sharing the full methodology with this community once published!

I will try to answer as many technical questions as possible, but please understand some details are reserved for my upcoming paper. Happy to discuss the results!

 

[waiwai]

"Finally reached the milestone! Previously with my Montage M (Synthesizer), I was hitting a ceiling around 93.77dB, but by switching the source to a dedicated Topping DX7s, I’ve successfully broken the 100dB SNR barrier in a typical home environment.

Measured result: 101.55dB SNR (THD 0.00016%).

Through meticulous grounding (utilizing a ham radio antenna earth) and ferrite core optimization, I managed to extract the full potential of the DX7s into the Scarlett 16i16. Since this is a raw, unweighted flat measurement, I believe this rivals the signal purity of high-end mastering studios. What do you guys think?"