72 software Sample Rate Converters put to the test

[bibo01]

This is a test by my friend Tom Gefrusti. The procedure employed to test 72 SRCs is this:

1) original native 24/88.2 kHz files (pink noise)

2) downsampling from 24/88.2 kHz to 16/44.1 kHz (with all listed software)

3) oversampling and re-quantization of the scaled files ... 16/44.1 kHz files are brought back to 24/88.2 kHz

4) specular comparison (NullTestDifference) between the native original file and the processed file.
--------------------------------------------------------------------------------
GREEN = Acoustica v6.0 - No dithering
RED = Acoustica v6.0 - Noise shaping and dithering
PURPLE = Acoustica v6.0 - Noise shaping only
YELLOW = AIMP v4.01 - Dither
BLUE = AIMP v4.01 - no Dither
WHITE = Ashampoo Music Studio v6.1 (no choice of dither and noise shaping)

--------------------------------------------------------------------------------
GREEN = AudioMove v1.2 - best
RED = AudioMove v1.2 - better
PURPLE = AudioMove v1.2 - good
YELLOW = Audiomuxer 0.9.6.0 - r8brain resampler
BLUE = Audiomuxer 0.9.6.0 - SSRC resampler
WHITE = AUI Converter free v5.6.3 - With dither

--------------------------------------------------------------------------------
GREEN = AUI Converter v5.6.3 free - No Dither
RED = AVS Audio Converter v8.1
PURPLE = AVS Audio Editor v8.1 - colorized dither, noise shaping A
YELLOW = AVS Audio Editor v8.1 - colorized dither, noise shaping B
BLUE = AVS Audio Editor v8.1 - gaussian dither, noise shaping A
WHITE = AVS Audio Editor v8.1 - gaussian dither, noise shaping B

P.S.: unseen colors are superimposed

--------------------------------------------------------------------------------
GREEN = AVS Audio Editor v8.1 - Triangular dither, noise shaping A
RED = AVS Audio Editor v8.1 - Triangular dither, noise shaping B
PURPLE = AVS Audio Editor v8.1 - uniform dither, noise shaping A
YELLOW = AVS Audio Editor v8.1 - uniform dither, noise shaping B
BLUE = AVS Audio Editor v8.1 - no dither
WHITE = Awave Studio 10.6 - FIR filter

--------------------------------------------------------------------------------
GREEN = Awave Studio 10.6 - Linear filter
RED = Awave Studio 10.6 - only downsample
Cockos REAPER PURPLE = V5.17 - Better (384pt Sync)
YELLOW = Cockos REAPER V5.17 - Extreme HQ (768pt HQ Sync)
BLUE = Cockos REAPER V5.17 - Fast (16pt Sync)
WHITE = Cockos REAPER V5.17 - Fast (IIR + Linear Interpolation)

--------------------------------------------------------------------------------
GREEN = Cockos REAPER V5.17 - Fast (II x2 + Linear Interpolation)
RED = Cockos REAPER V5.17 - Goode (192 pt Sync)
PURPLE = Cockos REAPER V5.17 - HQ (512pt Sync)
YELLOW = Cockos REAPER V5.17 - Low (Linear Interpolation)
BLUE = Cockos REAPER V5.17 - Lowest (Sampling Point)
WHITE = Cockos REAPER V5.17 - Medium (64pt Sync)

--------------------------------------------------------------------------------
GREEN = dBpoweramp v15.1 - SSRC resampler, no dither
RED = dBpoweramp v15.1 - SSRC resampler, RPDF
PURPLE = dBpoweramp v15.1 - SSRC resampler, TPDF
YELLOW = dBpoweramp v15.1, resampler undeclared (no choice of dither and noise shaping)
BLUE = Dexster Audio Editor v4.4 (no choice of dither and noise shaping)
WHITE = EZ CD Audio Converter v3.1.7 (no choice of dither and noise shaping)

--------------------------------------------------------------------------------
GREEN = GoldWave v6.21 (no choice of dither and noise shaping)
RED = Izotope RX4 -CD preset [filter steepness 32.0, 0.96 Cutoff shift, Pre-ringing 1.00 (ie linear phase)], MBIT
PURPLE = Izotope RX4 -CD preset [filter steepness 32.0, 0.96 Cutoff shift, Pre-ringing 1.00 (ie linear phase)] no Dither
YELLOW = Izotope RX4 - preset CD [filter steepness 32.0, 0.96 Cutoff shift, Pre-ringing 1.00 (ie linear fasTPDF
BLUE = Izotope RX4 - Custom [filter steepness 32.0, 0.96 Cutoff shift, Pre-ringing 0.00 (ie minimum phase)], MBIT
WHITE = Izotope RX4 -Custom [filter steepness 32.0, 0.96 Cutoff shift, Pre-ringing 0.00 (ie minimum phase)] no Dither

--------------------------------------------------------------------------------
GREEN = Izotope RX4 Custom [filter steepness 32.0, 0.96 Cutoff shift, Pre-ringing 0.00 (ie minimum phase)], TPDF
RED = Izotope RX4 - Custom [filter steepness 172.0, Cutoff shift 0.97, Pre-ringing 0.1] MBIT
PURPLE = Izotope RX4 - Custom [filter steepness 172.0, 0.97 Cutoff shift, ringing Pre-0.1], no Dither
YELLOW = Izotope RX4 - Custom [filter steepness 172.0, 0.97 Cutoff shift, ringing Pre-0.1], TPDF
BLUE = Izotope RX4 - Custom [filter steepness 434.0, Cutoff shift 1.00, Pre-ringing 1.00 (ie linear phase)], MBIT
WHITE = Izotope RX4 Custom [filter steepness 434.0, Cutoff shift 1.00, Pre-ringing 1.00 (ie linear phase)] no Dither

--------------------------------------------------------------------------------
GREEN = Izotope RX4 - Custom [filter steepness 434.0, Cutoff shift 1.00, Pre-ringing 1.00 (ie linear phase)], TPDF
RED = MP3 Stream Editor v3.4.4.3163 (no choice of dither and noise shaping)
PURPLE = Nero Recode 2016 (no choice of dither and noise shaping)
YELLOW = Nero WaveEditor v14 (no choice of dither and noise shaping)
BLUE = Nero WaveEditor v14, medium anti-aliasing filter
WHITE = Nero WaveEditor v14, accurate anti aliasing filter steepness 434.0, Cutoff shift 1.00, Pre-ringing 1.00 (ie linear phase)] no Dither

--------------------------------------------------------------------------------
GREEN = Nero Wave Editor v14, anti aliasing filter fast
RED = OcenAudio v3.0.7 (no choice of dither and noise shaping)
PURPLE = Saracon v01.61 - no dither
YELLOW = Saracon v01.61 - POWr1
BLUE = Saracon v01.61 - POWr2
WHITE = Saracon v01.61 - POWr3

--------------------------------------------------------------------------------
GREEN = Saracon v01.61 - PDF
RED = TAudioConverter v0.9.9 (no choice of dither and noise shaping)
PURPLE = Waveshop v1.0.14.001 - best (no choice of dither and noise shaping)
YELLOW = Waveshop v1.0.14.001 - fastest (no choice of dither and noise shaping)
BLUE = Waveshop v1.0.14.001 - medium (no choice of dither and noise shaping)
WHITE = XRecode II v1.0.0.223 (no choice of dither and noise shaping)

--------------------------------------------------------------------------------
The best software + setting is = Izotope RX4 - Custom [filter steepness 434.0, Cutoff shift 1.00, Pre-ringing 1.00 (ie linear phase)], no Dither

Among the 72 software settings there are others that stand out: dBpoweramp is a great example, also some settings of Saracon. You just need to look out at the maximum bandwidth and the least possible noise which means maximum congruence with native master file.

 

[John Kenny]

Thanks bibo, interesting results
I haven't checked but how well do they correlate with the SRC analysis of samplerate converters?

 

[bibo01]

Re. large discrepancies within the same software, let's take for example the action by:
1) Awave Studio 10.6 - Linear filter
2) Awave Studio 10.6 - only downsample

There is a very noticeable difference (generalized by NTD). Here, sifting and digging into these differences, it comes out that:

a) phase incoherence in the waveform compared to the original

b) phase shift in the waveform compared to the original

c) Spectral inconsistency with the original spectrum (also in the global response linearity level)

In practice within some software there are very marked differences on almost all parameters of the waveform.

On the other hand, if we take the best software, in comparison with the original waveform you get a perfect overlap and consistency.

 

[bibo01]

Thanks bibo, interesting results
I haven't checked but how well do they correlate with the SRC analysis of samplerate converters?

Well, here it's been used a different signal rather than the usual sweep, but there should be a high correlation if they did things right.

 

[John Kenny]

Well, here it's been used a different signal rather than the usual sweep, but there should be a high correlation if they did things right.

In the brief check I did - it seems there is correlation - well done

 

[amirm]

This is great info and clearly labor of love! Please express our thanks to your friend Tom.

 

[bibo01]

I will.

 

[amirm]

FYI, I moved the thread to Audio Library.

 

[Mivera]

Wheres Hqplayer?

 

[bibo01]

The SRCs tested are all offline ones, capable of saving their down/oversampling to a file. HQPlayer is not.

 

[Mivera]

The SRCs tested are all offline ones, capable of saving their down/oversampling to a file. HQPlayer is not.

Unless you get Jussi to offline convert some for testing like I did

 

[bibo01]

I did that too... multiple times

 

[Mivera]

I did that too... multiple times

With pink noise, or music?

 

[bibo01]

Pink noise, silence... PCM, DSF...

 

[Mivera]

Pink noise, silence... PCM, DSF...

Where's the measured data? I know you have the dynamic data on the Italian forum, do you have anything else? I wish there was more test results. I'm quite blown away with the DSD7 256+ fs modulator. I'm wondering what it does different to sound so good?

 

[March Audio]

An additional resource

http://src.infinitewave.ca/

 

[John Kenny]

You could use something like Totalrecorder to route the output from HQPlayer to disk instead of lineout?
You could then analyse it with your techniques

 

[amirm]

You could use something like Totalrecorder to route the output from HQPlayer to disk instead of lineout?
You could then analyse it with your techniques

I am not sure totalrcorder works with ASIO/WASAPI for bit-accurate capture. Does it?

 

[John Kenny]

I am not sure totalrcorder works with ASIO/WASAPI for bit-accurate capture. Does it?

Totalrecorder may be incompatible with HQplayer, I don't know it would require trying it but AFAIK, there are two methods for recording a soundstream using Totalrecorder:
- direct mode where it uses one of its own drivers (kernel stream) to convert the digital stream from the output of HQPlayer into Totalrecorder
- or virtual device capture mode where "Sound is captured exactly in the same format that is used by an audio player to pass audio data to a sound board."
How appropriate these methods are & whether they introduce some additional signal differences I don't know. I guess it could be tested with an existing SRC for which results have already been generated & use Totalrecorder to capture this stream, do the processing & compare to previous result.

See here for more info on capture modes/drivers

 

[amirm]

Mike, is there a way you can facilitate getting captured outputs of HQPlayer for this type of evaluation? I think it would be a very useful thing and I would imagine the developer would be in favor of it.