Any statement with "always" is likely to be incorrect.😀 If you are testing, for example, two reasonably good-measuring blocks of gain and use music as the test signal, following basic protocols can make cheating extremely difficult; not impossible, but very difficult and uncertain. You as the experimenter just have to be willing to do the basic protocols and to have the DUTs with which to do them.
Could you be more specific? If you mean measuring protocols, it was done and published. Or you mean test (subjective test) protocols?
I had some time so I did Mooly's new listening tests after all. PMs sent to Mooly.
I suspect these tests might show up some interesting results and (after sending the PMs) I have now checked the spectrums and can confirm that, as far as my limited understanding goes, there's nothing to worry about from the point of view of hearing damage when listening to these so there's no reason not to try.
I suspect these tests might show up some interesting results and (after sending the PMs) I have now checked the spectrums and can confirm that, as far as my limited understanding goes, there's nothing to worry about from the point of view of hearing damage when listening to these so there's no reason not to try.
I have to wonder how much difference the play-back system makes. My basic choices are: 1) Internal Realtek stuff in PC and cheap PC speakers, 2) As #1 but with 'phones at the output of the cheap PC speakers, 3) Emu0202 and it's headphone output, 4) Schiit DAC and 'phone amp (this was what I used), 5) Burning to CD and playing that on a dedicated player.
Then there's the software. One could use Audacity, Windows Media Player or a multitude of other players. And then there's the choice of 'phones or speakers, but that discussion could take years. I chose 'phones because my current room and speakers leave much to be desired.
Nothing I own is TOTL, and the Emu and Schiit are probably best, but what's the general opinion of how to get the best resolution on these files?
Then there's the software. One could use Audacity, Windows Media Player or a multitude of other players. And then there's the choice of 'phones or speakers, but that discussion could take years. I chose 'phones because my current room and speakers leave much to be desired.
Nothing I own is TOTL, and the Emu and Schiit are probably best, but what's the general opinion of how to get the best resolution on these files?
For highest resolution, I use, PC, external DAC, my own design of headamp (8Vrms/32ohm, THD<0.001% anywhere, Rout = 2.7ohm, S/N>110dB/1Vrms) and Sennheiser HD 598. In Mooly's new listenning tests no.1 and 2 I have clearly described (in PM) the background accompanying more or less all of the files and sent Mooly the right vote, together with description. Only the best and most transparent hardware allows the listener to find the differences, to make a choice of the best recording and to forget about effects that make the mistakes of the hardware more acceptable (re votes for cassette and mp3).
Oooh i could write pages in this quistion seen from experience, and as always others could have different. Cooking down: Renesas/Nec chipset PCIex1 cheap card dedicated for the EMU USB connection, try find slot going direct to CPU northbrige this is better than via chipset southbrige, and if possible to get uniqe IRQ for this slot is also preferable. I also seen RealTek perform nice but it depends on motherboard manufacture care and can be discovered by running freware Rightmark Audio Analizer. Regarding onboard audio there is light if choosing Intel brand, cause i read next generation will route HD-audio (Where codec sits) bus direct to CPU North bridge, but lets see. Try evaluate if the written can be worth for you and use Rightmark freeware to test for best performance. Last thing installed memory and timing settings can have huge influence on sound picture (timing can be trimmed in BIOS).
Regards Ricky
I've found my in ear headphones B&W C5 plugged directly into a stock Creative Titanium HD soundcard to be OK.
I'm running under Ubuntu Linux and pulseaudio and for these tests I have assumed that people are interested in an opinion on the original files and have been checking using pactl list whilst the sample is running that there is no resampling going on.
This has required reconfiguring the system to use the test files' native sample rate which is a bit of a hassle:
The driver for the soundcard requires module parameters in /etc/modprobe.d/alsa-base.conf to set multiple and reference_rate.
For example:
options snd-ctxfi multiple=1 reference_rate=44100
multiple can be 1, 2 or 4 reference_rate 44100 or 48000
After that you also need settings in /etc/pulse/daemon.conf
resample-method = speex-float-10
default-sample-format = float32le
default-sample-rate = 44100
I also use:
default-fragments = 32
default-fragment-size-msec = 25
Before the start of these tests, most of my music was from CD and I had come to the conclusion that the best mode of operation was to upsample to 88.2kHz. This gives a smoother sound without the noise of 176.4 and because it's a multiple of the original rate doesn't seem to dull the detail as far as I can tell.
So my normal settings are:
options snd-ctxfi multiple=2 reference_rate=44100
resample-method = speex-float-10
default-sample-format = float32le
default-sample-rate = 88200
Which uses the speex-float-10 filter to upsample to 88.2kHz
speex-float-10 takes a lot of CPU but is OK going from 44.1 up to 88.2. My system can't cope with speex-float-10 when downsampling 192kHz files though so I have to change the settings if I want to listen to them.
I haven't noticed a difference in sound between audio players on Linux. They either play the files or they don't. All the sound difference comes about from the choice of pulse audio's default-sample-rate, the resample-method and the rate the sound card is running at. The default-fragments and default-fragment-size-msec affect the buffering. When you use CPU intensive resampling the audio can get choppy and I found the buffer settings above fixed that.
Last time I looked, there are no Linux drivers for the later Creative cards and that one only works because I bought it by accident thinking it was going to work and then had to fix the driver myself. So don't consider this a recommendation.
I haven't tried doing these listening tests with upsampling. From past experience, I would expect it to make quite a big difference.
I've recently migrated my tube amp to run from a Windows PC and it's working but I'm going to install Linux dual boot and use that in the future as I have a better understanding of what's going on under the covers on Linux.
I gave Pavels Bach files a good listen, both yesterday and today. To play them through my main speakers I have to burn them to CDRW first.
Anyhow, I preferred the Bach1, the WAV-MP3-WAV file. The difference was very small though with speaker listening.
Today I tried them again, this time played on the PC listening with Sony MDR-V7 headphones and my choice was the same, I preferred the WAV-MP3-WAV file although the headphones showed a less focused image... not as "together" and more splashy. The impression of more hf detail remained though, I felt I could "see" the rosin coming of the bowstring with this one, and that's the one that got my vote 🙂
Very interesting results for all of these tests I am finding.
Anyhow, I preferred the Bach1, the WAV-MP3-WAV file. The difference was very small though with speaker listening.
Today I tried them again, this time played on the PC listening with Sony MDR-V7 headphones and my choice was the same, I preferred the WAV-MP3-WAV file although the headphones showed a less focused image... not as "together" and more splashy. The impression of more hf detail remained though, I felt I could "see" the rosin coming of the bowstring with this one, and that's the one that got my vote 🙂
Very interesting results for all of these tests I am finding.
One assumes the "right" answer is always the cleanest flat-response unprocessed file, but I'll play devil's advocate and suggest that live music in your living room might be fatiguing after a while. Some adjustments are probably beneficial to the home listening experience and the question is what and where to do them. It doesn't surprise me at all that a more technically flawed signal can sound better than the original under some circumstances.
I tried doing a quick ABX test on Pavel's Bach files. To my surprise, I couldn't reliably tell the WAV-MP3-WAV from the original. (I couldn't hear any obvious differences and scored exactly 50% in 10 trials.)
In a sense I'm not surprised. Years ago I ran tests with different bit rates of MP3 to see when I couldn't tell the difference any more, and it was somewhere between 192 and 256k. I notice the MP3 artifacts in cymbals and hi-hats first, and this recording has none, so maybe it can squeeze by with 160.
I did the tests using the setup I use for music production, a Macbook Pro with a M-Audio Firewire Audiophile interface, hooked up to a home-made solid-state power amp driving Dynaudio nearfield monitors.
Pavel, was the "bach difference" file unity gain or did you amplify it?
In a sense I'm not surprised. Years ago I ran tests with different bit rates of MP3 to see when I couldn't tell the difference any more, and it was somewhere between 192 and 256k. I notice the MP3 artifacts in cymbals and hi-hats first, and this recording has none, so maybe it can squeeze by with 160.
I did the tests using the setup I use for music production, a Macbook Pro with a M-Audio Firewire Audiophile interface, hooked up to a home-made solid-state power amp driving Dynaudio nearfield monitors.
Pavel, was the "bach difference" file unity gain or did you amplify it?
Try with headphones. Listen for the background wind in the opening section. One of them sounds windy, the other one sounds more dynamic.
Guess: your monitors can perhaps not quite manage the bass required to expose this???
Last edited:
Have you looked into DRC DRC: Digital Room Correction ?
I spent a while testing this. It can make a significant improvement, particularly in fixing low frequency phase issues, integrating subwoofers and eliminating room resonances. It can also fix the frequency response from your tube amplifier. So you can get the benefit of tube clarity with a flat frequency response.
The basic goal is that it preprocesses the audio signal to get a flat frequency response at head height at the location in the room where you normally sit 😀
Believe it or not, I had to modify the DRC code and rebuild it to use 64 bit floating point arithmentic rather than 32bit because I could hear the 32 bit limitation in the signal processing chain 🙂
DRC does quite a lot of maths. I guess the errors were accumulating.
I have been redoing my setup from the beginning to convert it from prototype to final deployment. I haven't got as far as re-enabling DRC yet but I do intend to turn it back on because it does help a lot.
You're probably right, my monitors are only rated down to 55Hz according to Dynaudio.
Listening on headphones, there is a kind of low frequency rumbling that I didn't notice on the speakers. It sounds like the reverb of a large hall. However I'm struggling to detect any difference in it between the two files.
Edit: I love DRC for fixing room resonances, and 32 bits is often not enough for the kind of digital filter you might use to notch out a room resonance. When you make an IIR filter with a low centre frequency and high Q, the arithmetic errors are amplified massively.
Listening on headphones, there is a kind of low frequency rumbling that I didn't notice on the speakers. It sounds like the reverb of a large hall. However I'm struggling to detect any difference in it between the two files.
Edit: I love DRC for fixing room resonances, and 32 bits is often not enough for the kind of digital filter you might use to notch out a room resonance. When you make an IIR filter with a low centre frequency and high Q, the arithmetic errors are amplified massively.
Hi nice meeting one with experience in DRC. I am not at the point yet have other setup fixture's first, but have quistion about DRC that is rumbling. Can it do the nice FIR correction where you only correct group delay not freq response in lows caused by speakerdriver/cabinet highpass filter, so it is possible to do square waves by speakers ?
Regards Ricky
Last edited:
ABC as good as foobar2000 ABX?
Also those versed in the arcane art of doing DBLT are also those best equipped to cheat, even without 21st century tools like file viewere 😱
Especially dem Beach Bums (like me) who want to appear a Pseudo Gurs and Golden Pinnae 🙂
Of course, everyone else here would NEVER be so evil 😀
_________________
I downloaded foobar2000 ABX and am very impressed with it as a straight player as well as an ABX tester.
Anyone know of a similarly good programme to do ABC?
Or could we persuade Pawlowski to do a version for ABC? Or even something simple that allows instant comparison between 3 synchronized files?
I do know of such a programme from the R&D dept. of at least 1 important audio company but alas, its not available to beach bums 😡
If not, we may have to ask for a Foobar2000ABX screen shot from each Listener along with his comments 🙂
It's impossible to foil the cheaters 100%. We can only try & make it more difficult for them.
Also those versed in the arcane art of doing DBLT are also those best equipped to cheat, even without 21st century tools like file viewere 😱
Especially dem Beach Bums (like me) who want to appear a Pseudo Gurs and Golden Pinnae 🙂
Of course, everyone else here would NEVER be so evil 😀
_________________
I downloaded foobar2000 ABX and am very impressed with it as a straight player as well as an ABX tester.
Anyone know of a similarly good programme to do ABC?
Or could we persuade Pawlowski to do a version for ABC? Or even something simple that allows instant comparison between 3 synchronized files?
I do know of such a programme from the R&D dept. of at least 1 important audio company but alas, its not available to beach bums 😡
If not, we may have to ask for a Foobar2000ABX screen shot from each Listener along with his comments 🙂
Last edited:
Just info if not already knowing. Also they have component "WASAPI output support" which for most all sounddevices can do bitperfect as ASIO drivers can and bypass VISTA/Win7/8 way to lock wordlengt and samplerate.
Hehehe you are all so crazy 😀
Imagine how Foobar knows which is which. Its because every files are represented by binary numbers. A simple programming will do it.
- Status
- Not open for further replies.