PGGB supports four levels of internal calculation precision. offers a Processing Precision setting; uses Adaptive Precision. Available options depend on your license and processing mode.
PGGB / The processing explained
PGGB, a perfect upsampler
for your music.
means increasing the by calculating extra between the existing ones as accurately as possible.
PGGB does this on your computer, with and output processing chosen for your . Here is what happens along the way, and why each part matters.
Click an underlined term for a plain-English explanation.
HOW IT WORKS / 01
The signal path through PGGB.
convert your digital tracks to an . Many DACs perform , or before converting into an analog signal. PGGB lets your computer do more of that preparation.
The signal path inside PGGB
DecodeDecode the music track at high precisionExplore
Decode samples at high precision.
PGGB reads the source audio and represents each sample as a number between -1 and 1 using its internal calculation , up to 256 bits. input is also converted into high-precision values for processing.
Keeping samples at high precision lets subsequent calculations retain the smallest changes.
Crossfeed & EQOptional adjustments to your soundExplore
Apply optional crossfeed and EQ.
makes headphone listening more natural; adjusts the tonal balance or applies room correction for your speakers. These optional adjustments happen before .
Crossfeed and EQ use the same high-precision arithmetic to keep numerical errors extremely small. With them off, the signal goes straight to the next stage.
Resample & filterUpsample or downsample to the output rateExplore
Calculate the output samples and filter high frequencies.
The Precision Optimized Engine matches the filter to the accuracy of its internal arithmetic. It calculates the new samples at the while removing unwanted content in the upper band.
GainSet the level and leave headroomExplore
Adjust the level for reconstructed peaks and EQ.
After upsampling, some new peaks can exceed the -1 to 1 range and cause . PGGB applies your setting. leaves the level unchanged unless is needed to avoid clipping. Reducing the signal level can also help avoid overloading later stages of the playback chain. Fixed gain and other modes let you choose how the final level is set.
DSD / PCMConvert to your selected output formatExplore
Convert the result to DSD or PCM for your DAC.
DSD.
The modulator converts the high-precision, high-sample-rate upsampled signal into a , shaping its to reach an precision equivalent to about 133 bits.
This is made possible by PGGB’s ultra-low-noise 25th-order modulators.
DSD modulationPCM.
Adaptive noise shaping converts the high-precision upsampled signal to the selected , producing the final samples.
PGGB uses dithered noise shapers up to 41st order to reach an audible-band precision equivalent to about 125 bits within 32-bit PCM output.
PCM noise shaping and bit depthOffline / LiveSave the result to file or send it to your DACExplore
Save or stream the same upsampled DSD or PCM audio.
Offline.
Write a file in the format you selected, keeping the and artwork where available.
File writing and metadataLive.
Send the stream through or on Windows, on Mac, or a supported network endpoint.
Live output connections
Select a card to explore what happens at that stage.
HOW IT WORKS / 02
Decode the track for high-precision processing.
The music arrives as a file or from a Live source. PGGB it into and represents them using arithmetic. Each value is between -1 and 1, ready for the next processing stage.
That precision determines how much numerical detail the calculations can retain. More precision reduces accumulated as the signal passes through the processing stages.
Higher precision does not add information missing from the recording. It lets PGGB calculate accurately from the information present while keeping processing noise extremely low.
A track can be processed at 256-bit precision and then written as 24-bit PCM or converted to one-bit DSD. The processing precision does not affect the size of the output file.
HOW IT WORKS / 03
Apply crossfeed and EQ before resampling.
Your optional adjustments come before . When they are off, the signal moves straight to the next stage.
Crossfeed and ARC.
blends a little of each channel into the other for headphone listening. can tailor this processing to your measurements, with optional room reflections available Offline.
PGGB’s implementation adds no tonal coloration and remains true to the source.
Equalization.
adjusts the tonal balance or applies a headphone, speaker or room correction. PGGB supports and .
Because EQ runs at PGGB’s , it is transparent as it adds no rounding errors of its own.
HOW IT WORKS / 04
Calculate samples at the output rate.
After any or , PGGB calculates the new at your chosen . It calculates these values accurately and reduces unwanted distortion associated with the highest .
Interpolation: calculating new samples.
For example, converting a 44.1 kHz track to 705.6 kHz gives your 16 times as many samples per second. PGGB calculates these additional samples to follow the described by the recording, as if that waveform had been sampled at the higher rate.
That calculation is called . describes how closely the calculated values match the ideal waveform and thus how accurately PGGB upsamples.
How does the curve come from the samples?
With , each original sample contributes a small curve. Adding all those curves gives the . PGGB calculates values along it at the output sample rate you choose.
For downsampling, it calculates fewer samples per second and filters out frequencies that the lower rate cannot represent.
The V9 engine
Filter length and arithmetic precision.
The Optimized Engine chooses a filter length that keeps reconstruction error below arithmetic error in the frequency band being preserved. At that point, arithmetic precision becomes the main limit on reconstruction accuracy.
PGGB uses up to 256-bit arithmetic to retain very small differences with extremely little . V9 reaches that accuracy with shorter filters, making this high-precision processing faster and much less demanding on RAM.
Why not keep making the filter longer?
Making the filter longer brings little practical benefit in that band once its error is below the calculations’ rounding error. The new engine chooses the length needed to reach that point and processes the upper frequency band separately.
In PGGB, leave Precision optimized selected under Settings → Processing to use this engine.
High-frequency filtering
Reduce high-frequency distortion.
The highest frequencies a recording can hold are harder to reconstruct accurately. PGGB handles this upper band separately, preserving the while reducing unwanted high-frequency distortion.
For a 44.1 kHz recording, the blog examines the 20–22 kHz region, just above the usual hearing range, and explains why its treatment matters.
Is this the same as HF Noise Filters?
They are different controls. The Precision optimized engine also handles source-noise filtering internally.
The separate HF Noise Filters controls reduce high-frequency noise already in a recording. They apply when you select the Legacy engine under Settings → Processing.
HOW IT WORKS / 05
Set the gain and avoid clipping.
After and , PGGB applies the selected setting. This sets the signal level before the final or conversion.
Auto: avoid clipping.
leaves the gain unchanged unless is needed to avoid . A can peak above the original , so checking only those original values would miss .
Fixed gain and other modes.
Fixed gain applies your chosen gain. Other gain modes can set a target peak or use existing album-peak information. Choose the approach that suits your files and playback setup.
Headroom, normalization and loudness
is the space between the signal’s highest peak and the maximum level the output can handle. Attenuation makes room for peaks that would otherwise clip.
Peak Normalize sets each track’s highest peak to a chosen level. It does not make different recordings equally loud: tracks with different dynamics can still have different average levels.
HOW IT WORKS / 06
PCM: noise shaping and final bit depth.
When you choose output, the enters the noise-shaping and final bit-depth stage.
PCM output to your has a finite set of determined by the . Converting a high-precision value to one of these available levels introduces a small difference called .
changes the distribution of that error. reduces its contribution in the while allowing more at higher frequencies. At high , there is more space above the audible band for this noise.
PGGB’s adaptive PCM noise shaping can reach an audible-band noise level equivalent to about 125 bits within a 32-bit output bit depth. The result is a sequence of high-quality PCM ready to be saved to a file or streamed. The DAC and the rest of the output path still need to handle its .
Where does dither fit?
When reducing a result to its output bit depth, can follow the signal predictably and become audible, signal-correlated distortion. is a small, deliberately added random signal that breaks up this relationship. It trades signal-dependent distortion for a .
Dither and noise shaping have different jobs and can be used together. PGGB uses dithered noise shaping. Dither breaks up the error’s relationship with the music; noise shaping controls where more of that error sits in frequency. Neither removes noise already present in a recording.
Background: oversampling and noise shaping (Analog Devices, PDF) ↗
HOW IT WORKS / 07
DSD: modulation to a one-bit stream.
Choose output and the result goes to the selected instead of the output stage.
A has only two levels. The proportion of high and low values over a short interval represents the signal’s . An analog filter recovers the varying signal from that rapid sequence.
A delta-sigma modulator uses and to produce this one-bit stream. The selected modulator and DSD determine how its is distributed.
PGGB’s own modulator designs go up to 25th order, the highest published order we know of. Its six are grouped as Low Noise, Very Low Noise and Ultra Low Noise.
PGGB’s ultra-low-noise modulators can reach an noise level equivalent to about 133 bits within the one-bit stream.
Compare the modulatorsHow does the modulator order compare?
For context, Weiss Saracon documents eighth- and tenth-order DSD modulators. DpdoEngine’s parameter reference lists orders up to eleven, while its product page advertises fifteenth order. PGGB offers designs up to 25th order.
This compares published modulator orders. Noise performance also depends on the design, output rate and measurement conditions; order alone does not establish a noise level or listening result.
HOW IT WORKS / 08
Save a file or send audio to your DAC.
The prepared or signal now takes the route you chose: a new file for Offline listening, or a stream for Live playback.
Offline: write the file.
PGGB writes the result in your selected file format. PCM can use WAV, RF64, FLAC or WavPack; DSD can use DSF or WavPack. Available choices depend on the output settings.
, including cover art, is transferred where possible. Your regular player can then play the prepared files without having to repeat PGGB’s processing.
Live: send the stream.
PGGB sends the prepared audio to the chosen : or on Windows, on Mac, or a supported network endpoint.
For network output, use PGGB-Audio-Link (PAL) when the endpoint can run it; use UPnP when PAL is not an option. Connect a PAL endpoint.
The output format and rate must suit the and connection. Calibrate your chosen settings to confirm the computer can process them in real time.
Match the output to your DAC’s conversion path
A mode can avoid the DAC’s own oversampling. Other PCM DACs use their internal processing differently at high input rates. A direct DSD path and a DAC that converts DSD to PCM are different cases too.
Choose the format, rate and for your DAC’s actual input and conversion path, rather than assuming one format is always preferable.