ECHO SRC / Upsampling Guide

This guide is for anyone who wants to use the built-in upsampler in ECHO Next. Upsampling here means the on-device PCM sample-rate conversion under DSP -> ECHO SRC / Upsampling. It is not HQPlayer, and it is not PCM-to-DSD conversion.

One sentence to remember first:

Upsampling is advanced DSP, not sound-quality magic. Make sure your computer, DAC, drivers, and exclusive output are stable before you turn it on.

Confirm three things before enabling it

You need enough compute

Upsampling calculates new PCM samples in real time. Higher ratios, longer filters, and more aggressive quality policies mean more CPU / GPU load.

Setting	Hardware demand	Advice
2x PCM	Light	Fine starting point for laptops and ordinary desktops
4x PCM	Moderate	Recommended starting point; balances stability and effect
8x Ultra	High	Needs a stable CPU, driver, and DAC; do not make it your default at first
Advanced Filters / CUDA	High to very high	Only after basic upsampling is confirmed stable

If you hear pops, stutter, noticeably slower track changes, or repeated fallbacks in the playback state, lower the ratio first instead of stacking heavier filters.

You need a dedicated DAC

Upsampling is best suited to wired external DACs, USB decoders, sound cards, or dedicated audio devices with stable drivers. Do not try to validate it with Bluetooth headphones, Bluetooth speakers, monitor audio, or laptop speakers.

The reason is simple: the upsampled high sample rate ultimately has to be received by the device. Without a dedicated DAC or stable sound card, the system mixer, Bluetooth codecs, device firmware, or virtual sound cards can re-process the result — the numbers you see on screen may not be what actually reaches the DAC.

Exclusive output is required

To verify ECHO SRC is really taking effect, use:

• WASAPI Exclusive
• or the official ASIO driver from your DAC / sound-card vendor

Do not validate upsampling with WASAPI Shared or System. Shared output goes through the Windows mixing chain, the final sample rate may be decided by the system default format, and ECHO SRC may be bypassed. Shared mode is fine for everyday listening; for verifying upsampling, use exclusive or official ASIO.

Shortest recommendation

For a first run, set things up like this:

Output mode = WASAPI Exclusive
Output device = your dedicated DAC / sound card
ECHO SRC = 4x PCM
Quality policy = Balanced
Control mode = Normal
Advanced Filters / CUDA = leave alone
DSD / SDM / HQPlayer = off

Play for 30 seconds to a minute. If there are no pops, no stutter, and track changes work, then consider higher ratios or more complex filters.

Where to find it

1. Play a local FLAC / WAV / MP3 first and confirm basic playback works.
2. Open DSP in the left navigation.
3. In the Sample Rate group, enable ECHO SRC / Upsampling.
4. Keep the control mode on Normal for now.
5. Choose 4x PCM as the ratio.
6. Choose Balanced as the quality policy.
7. Confirm the output is already WASAPI Exclusive or official ASIO.

The top of the page shows the current path, source sample rate, target sample rate, engine, quality policy, and precision. Seeing “waiting for next playback plan” right after changing settings is normal — check the actual state after switching tracks or restarting playback.

Choosing a ratio

ECHO SRC converts upward within the 44.1 kHz and 48 kHz families. Typical results:

Source	2x PCM	4x PCM	8x Ultra
44.1 kHz	88.2 kHz	176.4 kHz	352.8 kHz
48 kHz	96 kHz	192 kHz	384 kHz
88.2 kHz	Kept, or handled per target	176.4 kHz	352.8 kHz
96 kHz	Kept, or handled per target	192 kHz	384 kHz
176.4 kHz	Kept	Kept	352.8 kHz
192 kHz	Kept	Kept	384 kHz

If the source already meets or exceeds the current target, ECHO may bypass processing. It will not re-process audio just so the UI can say “upsampling”.

Recommended progression

1. Off: troubleshooting, verifying bit-perfect candidates, confirming raw playback.
2. 2x PCM: lightweight test, good for weak CPUs or checking the chain first.
3. 4x PCM: recommended everyday starting point.
4. 8x Ultra: experimental tier; try only after confirming the DAC supports 352.8 / 384 kHz.

Do not jump straight to 8x Ultra. Higher is not “more advanced” — it is heavier on the computer and pickier about drivers and DACs.

Choosing a quality policy

Policy	Focus	Advice
Balanced	Stability and cost balanced	Recommended for the first run
Transparent	Higher precision, favors transparency and low distortion	Try after Balanced proves stable
Low latency	Lower SRC overhead	For low latency, weak CPUs, or diagnosing stutter

If you cannot hear a difference, that is normal. Upsampling is not an effects layer for every track. The settings worth keeping are the ones that stay stable over long playback, never break track changes, keep the device locked, and sound comfortable to you.

Confirming it is active

Check these places first:

1. Whether Path on the ECHO SRC page changed from waiting to processing.
2. Whether Source sample rate and Target sample rate differ.
3. Whether Engine shows the current SOXR / FIR / CUDA state.
4. Whether the Bit-perfect path indicator in the top right reports you have left bit-perfect.
5. Whether the DAC screen, driver panel, or sound-card control panel shows the target sample rate.

The DAC or driver panel display matters most. If ECHO requests 176.4 kHz but the DAC ends up showing 44.1 kHz or 48 kHz, there is usually shared mixing, driver resampling, or a device limitation still in the output path.

Why it is no longer bit-perfect when enabled

Bit-perfect means delivering the file’s digital samples as unchanged as possible. Upsampling recalculates samples, so once ECHO SRC is actually processing, playback is no longer strictly bit-perfect.

This is not an error — it is an honest indicator:

• To verify raw output: turn off ECHO SRC, EQ, FIR, ReplayGain, channel tools, speed change, and other DSP.
• To use upsampling: accept that it changes the samples, and use the output state to verify the chain.

Using advanced mode

Advanced mode exposes Filter / HQ-style, CPU/GPU Quality Ladder, Filter 1x, Filter Nx, Compute / CUDA, and PCM Dither / Noise Shaping options.

As a beginner, do not change these at the same time:

• Filter 1x
• Filter Nx
• Compute / CUDA
• CPU/GPU Quality Ladder
• PCM Dither / Noise Shaping

If you want to explore step by step:

1. Run 4x PCM + Balanced in normal mode until it is stable.
2. Change only the quality policy to Transparent and listen to the same track.
3. Then try 8x Ultra, confirming the DAC locks 352.8 / 384 kHz reliably.
4. Only then enter advanced mode and swap filters.
5. Even with an NVIDIA GPU, do not enable CUDA right away. Get the CPU path stable first, then see whether CUDA is available.

If the UI reports CUDA unavailable, no NVIDIA driver detected, missing worker, or a fallback to CPU, follow the hint first. Low GPU usage does not mean it is inactive — real-time audio is often small low-latency workloads. Watch the active / fallback indicators in the playback state instead.

Plain-language guide to the advanced options

Option	What it does	Beginner advice
Filter 1x	Handles base-rate sources like 44.1 / 48 kHz	Start with the default or a light option
Filter Nx	Handles sources at 88.2 / 96 kHz and above	Keep it close to 1x at first, for easier troubleshooting
Quality Ladder	Picks a whole quality/load profile in one click	Start from Realtime Safe or HiFi
Compute / CUDA	Whether to attempt GPU compute	Only after the CPU path is stable
Dither / Noise Shaping	Dithering and noise shaping for fixed bit-depth output	Keep the default if unsure; Float output usually does not need it urgently

Advanced filter names look impressive, but the troubleshooting principle stays the same: change one variable at a time.

How to back out when things break

Symptom	Do this first
No sound	Turn off ECHO SRC, switch output back to System or WASAPI Shared
Pops / stutter	Drop from 8x Ultra to 4x PCM, then 2x PCM
Track changes noticeably slower	Turn off advanced filters / CUDA, return to normal mode
DAC does not show the target rate	Confirm the output is WASAPI Exclusive or official ASIO
Less stable with CUDA on	Switch back to CPU; retry after updating the NVIDIA driver
Bit-perfect indicator gone	Normal; turn off ECHO SRC to get bit-perfect back
No upsampling with DSD / HQPlayer	Normal; ECHO SRC does not stack with those chains

Fastest rollback path:

Turn off ECHO SRC -> turn off other DSP -> back to WASAPI Shared/System -> play an ordinary FLAC

Don’t use it like this

Do not validate upsampling on Bluetooth devices. Bluetooth passes through the system stack, codecs, and a wireless link — unsuitable for sample-rate verification.

Do not crank the Windows default format to an extreme sample rate long-term and assume everything now sounds better. The system’s shared output and ECHO SRC are two different chains.

Do not enable ECHO SRC while diagnosing silence, pops, DSD, ASIO, HQPlayer, or remote library issues. Fix basic playback first, then play with upsampling.

Do not judge “better” by “louder”. When comparing before and after, keep volume matched and listen to the same familiar passage.

One-sentence summary

Most users should use it like this:

1. Keep it off day to day.
2. To try built-in upsampling, use a dedicated DAC + WASAPI Exclusive + 4x PCM + Balanced.
3. Once stable, try Transparent.
4. After confirming both the computer and the DAC can keep up, try 8x Ultra or advanced filters.
5. When things break, go back to Off. Do not push through.

The point of ECHO SRC is a local PCM upsampling path that is explainable and reversible. Whether it works well for you depends on your compute power, your dedicated DAC, driver stability, and whether you genuinely hear a difference you enjoy over time.