Mossca Audio Technology

Sample Rate Upsampling

Convert the original PCM audio to a higher sample rate through high-quality interpolation before sending it to the DAC for playback.

Sample rate upsampling is also commonly referred to as Upsampling or Sample Rate Conversion. For example:

44.1 kHz → 88.2 / 176.4 / 352.8 kHz

or:

48 kHz → 96 / 192 / 384 kHz

It does not turn the original music into “higher-resolution” audio.

What it actually does is:

  • Calculate new sample points between the original samples
  • Push spectral images to higher frequencies
  • Change the operating conditions of subsequent digital filters
  • Change the data rate received by the DAC

It is therefore a form of active digital processing.

Why Upsample the Sample Rate?

Digital audio is not a continuous waveform. It consists of a sequence of discrete sample points.

CD audio, for example, contains 44,100 samples per second. The player and DAC ultimately need to reconstruct a continuous analog waveform from those discrete samples.

That process inherently requires:

Interpolation + Reconstruction

In other words, interpolation and reconstruction.

The real question is therefore not “the original file already contains samples, so why add more?” but rather: how should the signal between those samples be reconstructed?

What Does Sample Rate Upsampling Actually Do?

Consider a simple example:

44.1 kHz → 176.4 kHz

This means the number of samples per second becomes four times greater.

The system does not simply copy each original sample four times. A proper process normally looks like:

Original PCM → Digital Interpolation Filter → Intermediate Sample Calculation → Higher-rate PCM

The additional sample points are mathematically reconstructed from the original signal.

Does It Add New Musical Information?

No.

If the original file is converted from:

44.1 kHz / 16-bit → 176.4 kHz / 24-bit

information already lost during recording cannot be recovered.

Upsampling cannot create:

  • New instrument detail
  • High-frequency information that never existed
  • Higher true recording resolution

Upsampling is not the same as restoring higher resolution. It changes how the signal is processed afterward, not the original musical content.

Why Can It Still Affect the Sound?

A DAC does not simply “read digital values and output them directly.”

A typical modern digital playback chain looks like:

PCM → Digital Filter → Oversampling → DAC Core → Analog Reconstruction Filter → Analog Output

If the player performs part of the sample rate conversion beforehand:

PCM → Mossca SRC → High-rate PCM → DAC

the digital processing subsequently performed inside the DAC may change.

Different DACs do not necessarily use the same internal path for every input sample rate. This is the main reason software upsampling can potentially produce audible differences.

Nyquist Frequency and Spectral Images

For a 44.1 kHz sample rate, the Nyquist frequency is 22.05 kHz.

In theory, useful audio information can be represented below this frequency. However, sampled digital signals also create periodic spectral images.

The DAC must suppress these images through digital and analog filtering.

What Happens After Upsampling?

If 44.1 kHz is upsampled to 176.4 kHz, the new Nyquist frequency becomes 88.2 kHz.

The spectral images are moved farther away from the audible frequency range.

Originally:

20 kHz → 22.05 kHz Nyquist

the filter has only a narrow transition region.

After upsampling:

20 kHz → 88.2 kHz Nyquist

the filter has substantially more transition bandwidth, making filter design less demanding.

Why Are Filters So Important?

Ideal interpolation theoretically requires a sinc filter, but an ideal sinc is infinitely long. Real systems can only use finite approximations.

Different filters therefore make different trade-offs between:

  • Passband flatness
  • Stopband rejection
  • Transition bandwidth
  • Phase response
  • Ringing
  • Latency
  • CPU cost

As a result, two different SRC implementations converting 44.1 kHz to 176.4 kHz may not produce exactly the same result.

How Can the Interpolation Filter Affect Sound?

Commonly discussed factors include:

  • Frequency response
  • Phase response
  • Pre-ringing
  • Post-ringing
  • Transition band
  • Alias suppression

A very steep linear-phase filter may offer excellent frequency response and strong Stopband rejection, but it may also produce more symmetrical Ringing.

Other filters may trade a small amount of frequency-domain performance for different time-domain behavior.

Therefore, “how many kHz” is only part of the story. The interpolation algorithm is equally important.

What Are Pre-ringing and Post-ringing?

When a digital filter processes a very sudden transient, such as an idealized impulse, small oscillations may appear before and after the transient.

Linear-phase FIR filters generally produce Ringing on both sides. Minimum Phase filters tend to move more of the Ringing to after the transient.

These are mathematical filter characteristics. They do not necessarily mean the effect is directly audible, but they affect how the waveform is reconstructed in the time domain.

Why Does the Player Need a High-quality SRC?

Sample Rate Conversion is not simply a “multiply by 2” operation.

It requires:

1. Interpolation 2. Filtering 3. Suppression of spectral images 4. Preservation of precision 5. Control of numerical error

Low-quality SRC can introduce:

  • Aliasing
  • Passband attenuation
  • Imaging residue
  • Phase anomalies
  • High-frequency distortion
  • Numerical noise

The goal of a high-quality SRC is to keep the converted high-rate PCM as close as possible to the original signal within the audible band while minimizing conversion artifacts.

Why Are Integer-multiple Sample Rates Common?

These are integer-multiple conversions:

44.1 → 88.2 → 176.4 → 352.8
48 → 96 → 192 → 384
By contrast, 44.1 → 192 is a non-integer-ratio SRC.

It is mathematically valid, but the resampling ratio and filter design are more complex.

For this reason, many high-quality playback systems prefer to preserve the 44.1 kHz family and the 48 kHz family, upsampling within each family.

The 44.1 kHz and 48 kHz Families

Common relationships are:

44.1 → 88.2 → 176.4 → 352.8 → 705.6
48 → 96 → 192 → 384 → 768

This avoids unnecessary conversion between the two sample-rate families.

Doesn’t the DAC Already Perform Oversampling Internally?

Many modern DACs do.

A typical Sigma-Delta DAC path is:

PCM Input → Digital Filter → Oversampling → Sigma-Delta Modulator → DAC Core

Therefore, even if the player outputs 44.1 kHz PCM, the DAC may internally convert it to a much higher processing rate.

One purpose of software Upsampling is to move part of this interpolation and filtering process from the DAC’s fixed internal algorithms into software running on the player.

What Is the Difference Between Software Upsampling and DAC Oversampling?

Direct playback:

44.1 kHz PCM → DAC Internal Filter → DAC Internal Oversampling

Software upsampling:

44.1 kHz PCM → Mossca SRC → 176.4 / 352.8 kHz PCM → DAC

Potential advantages of the latter include:

  • Software can use more complex filters
  • CPU resources are usually greater than those available inside a DAC chip
  • Algorithms can be updated more easily
  • Users can choose different output sample rates

However, this does not automatically mean it will sound better.

Why Is It Not Necessarily Better?

Different DACs handle high-rate PCM differently.

When receiving high-sample-rate PCM, some DACs may:

  • Reduce their internal Oversampling ratio
  • Use a different Filter
  • Change Modulator operating mode

Other DACs may still send 352.8 kHz input through a full internal digital processing chain.

Whether software SRC has a meaningful effect therefore depends on the DAC architecture.

Differences Between DAC Architectures

Sigma-Delta DAC

Modern Sigma-Delta DACs usually depend heavily on internal Oversampling, Filtering, and Modulation.

Software upsampling may alter the proportion of work performed by their internal Filter / Oversampling stages.

R-2R DAC

R-2R / Ladder DAC behavior varies much more.

Some designs use:

PCM → Digital Filter → Ladder

while others may provide:

  • NOS mode
  • OS mode
  • Custom FPGA Oversampling

These DACs may respond more noticeably to different input sample rates.

NOS DAC

NOS means Non-Oversampling and refers to DACs that do not perform conventional digital Oversampling internally.

PCM therefore reaches the Conversion stage more directly, making software Upsampling potentially more significant.

One way to understand it is: if the DAC itself does not perform Oversampling, the player’s SRC may directly determine how digital interpolation is performed.

NOS DACs minimize digital filtering, but there are trade-offs:

  • High-frequency images remain closer to the audible band
  • The analog filter has more work to do
  • High-frequency response may roll off
  • Imaging spectrum becomes more pronounced

A common configuration is:

44.1 kHz → Software High-quality SRC → 352.8 kHz → NOS DAC

This allows the player to perform digital interpolation while minimizing digital filtering inside the DAC.

What Might It Do to the Sound?

Depending on the system, listeners may perceive differences in:

  • High-frequency texture
  • Transient edges
  • Spatial presentation
  • Sound density
  • Overall smoothness or directness

These differences may come from:

  • SRC filter
  • DAC internal processing
  • Clocking architecture
  • Analog filter
  • Output stage

Upsampling itself does not have a fixed “sound signature.”

Is a Higher Sample Rate Always Better?

No.

44.1 → 88.2 may already provide a much wider filter Transition band.

Increasing further to 352.8 kHz can theoretically provide even more filtering margin, but it also increases:

  • CPU workload
  • USB data throughput
  • Memory bandwidth
  • DAC input workload

How the DAC ultimately processes the signal still determines the practical result.

A higher number does not automatically mean better sound.

Upsampling and Bit-Perfect Playback

If:

44.1 kHz PCM → SRC → 176.4 kHz PCM

the digital sample data has changed.

Upsampling is therefore not Bit-Perfect relative to the original PCM file.

Bit-Perfect playback aims for:

Source PCM → Unchanged Sample Data → DAC

Upsampling instead follows:

Source PCM → DSP / SRC → New PCM Stream

They are two different playback strategies.

What About Increasing Bit Depth?

You may also see:

16-bit / 44.1 kHz → 32-bit / 176.4 kHz

Sample Rate and Bit Depth should be considered separately.

Increasing Bit Depth alone does not create new original information.

However, using 32-bit float or 64-bit float internally during DSP can improve:

  • Calculation precision
  • Headroom
  • Numerical stability across multiple DSP stages

High-precision internal processing is therefore not the same thing as making the source file “higher resolution.”

Why Does Real-time SRC Require Significant Processing Power?

High-quality SRC generally involves:

Input PCM → FIR Filter → Large Number of Multiply-Accumulate Operations → High-rate PCM

As the number of filter Taps increases:

  • Stopband rejection can improve
  • Transition band can become narrower
  • CPU usage increases
  • Latency may also increase

Real-time high-quality SRC is fundamentally a DSP computation task.

Mossca’s Sample Rate Upsampling Path

PCM Source → Decode → High Precision PCM → Sample Rate Converter → Higher-rate PCM → Audio Output → DAC

The Sample Rate Converter is developed by Mossca.

Mossca builds the appropriate PCM output path according to:

  • Original sample rate
  • Output device
  • DAC-supported sample rates
  • Current playback mode
How Is It Different from PCM → DSD?

Sample Rate Upsampling

PCM → SRC → High-rate PCM → DAC

The data remains multi-bit PCM.

PCM → DSD
PCM → Oversampling → Sigma-Delta Modulation → 1-bit DSD → DAC

The final data format becomes DSD.

Both involve High-rate processing, but they have different goals.

Upsampling changes the PCM sample rate. PCM → DSD additionally changes the digital encoding format and the type of data delivered to the DAC.

When Is Sample Rate Upsampling Useful?

Suitable when:

  • You want to use a high-quality software SRC
  • Your DAC handles high-rate PCM well
  • You use a NOS / R-2R type DAC
  • You want to compare DAC internal filtering with software filtering
  • The system has sufficient CPU and bandwidth

It may not be necessary when:

  • You want to preserve the original file Bit-Perfect
  • Power consumption is more important on a mobile device
  • You want a strictly Bit-Perfect playback path

A Simple Way to Understand It

With native PCM playback, the original sample rate is sent to the DAC, and the DAC decides how interpolation and reconstruction are performed.

With sample rate upsampling, the player first performs high-quality digital interpolation and then sends higher-rate PCM to the DAC.

What changes is not the original musical information, but who performs part of the digital reconstruction process.

FAQ

Does converting 44.1 kHz to 352.8 kHz add real detail? No. SRC does not automatically increase the valid information contained in the original recording.

Why can it still sound different? Because the SRC algorithm, DAC input path, internal filtering, and analog output stage may change.

Is a higher sample rate always better? No. Beyond a certain point, the benefit depends on the DAC architecture and system implementation.

Does Upsampling introduce distortion? Every SRC performs a mathematical transformation. A high-quality SRC aims to keep audible-band error, Aliasing, and frequency-response changes extremely low.

Is Upsampling Bit-Perfect? No. Once the sample rate changes, the output digital data is no longer identical to the original file.

Sample Rate Upsampling

Convert the original PCM audio to a higher sample rate through high-quality interpolation before sending it to the DAC for playback.