CHANGSTAR: Audiophile Headphone Reviews and Early 90s Style BBS
Lobby => Headphone, IEM, and Other Audio Related Discussion => Topic started by: xnor on June 22, 2013, 07:13:00 PM
-
ahoy
time for some little comparative listening. Here are two 3 sec snippets of Firefly:
a.flac (https://www.dropbox.com/s/oyhhawm30vcq3u1/a.flac)
b.flac (https://www.dropbox.com/s/4tvag7cp7u631kr/b.flac)
in 44.1/16 compressed with FLAC.
Can you pick out a difference between the two, if so do you prefer one of them?
Hint: one of the files was processed differently. I will reveal which one and how after some guys had a chance taking a listen.
(this might not be the right place for this, so please don't walk the plank2)
-
Permission denied. MediaFire detects the material as copyrighted.
Maybe use a different hosting solution?
-
Woah woah woah. When did we get another sailor with a Guybrush avatar?!?
-
Oh lol, I'm new here but been using that avatar for a few years.
Will fix the links.. edit: fixed.
edit2: Does it work?
-
My ears cannot detect those differences (reliably) on cheap PC speakers at a quick listen and to me border on placebo realm but they are certainly measurable and very quantifiable.
-
lol, I actually thought zerodeefex was xnor from Head-Fi for the longest time because of that.
-
:-DD nope
-
I am not sure if it is placebo or not since the differences don't jump out at me but I thought I heard the snap in the beginning of the song ever so slightly louder on B. I'll listen some more.
Edit:Nevermind I think its placebo. I'll listen some more
-
My placebo feeling says I like A more and feel like it is less annoying than B but I say placebo because the difference is quite small on my hd280pro.
Without over analyzing, and based on my comparisons without trying to scientifically eliminate placebo (I havent done an ABX yet) I feel the lower frequencies of the vocals are ever so slightly richer on A. This is the opportunity to shoot me down hehehe but yah like I said I myself have a feeling its just placebo since I don't hear anything jump out at me like when comparing oggs vs mp3s vs the original flac file.
EDIT: ahh nevermind... I discovered that its in the recording where sometimes the voice gets slightly thinner at times. Its both present on A & B
First AB Test here
foo_abx 1.3.4 report
foobar2000 v1.2.8
2013/06/23 08:30:13
File A: C:\Users\donunus\Downloads\a.flac
File B: C:\Users\donunus\Downloads\b.flac
08:30:13 : Test started.
08:35:58 : 01/01 50.0%
08:36:35 : 02/02 25.0%
08:37:36 : 03/03 12.5%
08:38:10 : 04/04 6.3%
08:38:24 : 05/05 3.1%
08:38:56 : 05/06 10.9%
08:39:17 : 06/07 6.3%
08:39:39 : 06/08 14.5%
08:40:10 : 06/09 25.4%
08:40:40 : 07/10 17.2%
08:41:29 : 07/11 27.4%
08:41:45 : 07/12 38.7%
08:42:31 : 08/13 29.1%
08:43:19 : 08/14 39.5%
08:43:49 : 09/15 30.4%
08:44:06 : Test finished.
----------
Total: 9/15 (30.4%)
I got a perfect 5/5 at first since I could tell the difference of the flow of the sound when switching from track to track then fatigue sets in after that I was becoming less sure the higher the trials went. Anyway, this was with an hd280pro out of a realtek soundcard and an O2 amp. When a dac arrives here I'll test it again.
-
No no, I'm not :boom: you down. I know how hard such tests are when you basically have no friggin clue what to look for.
ABX is nice, but in order to start you should be able to hear (or at least think* you can hear) a difference between A and B.
*) I've confidently ABX'd files and it turned out I was completely wrong. I think everyone who's done a couple of ABX tests has experienced that, especially if the results are hidden during the test.
Btw, did I mention that I really like some of those smilies? :D <- Well, not particularly that one.
-
A sounded like it had better treble energy, thus better clarity/sharpness, while B sounded a bit shaved off at the upper reaches for the first ten seconds. I noticed it after one listen on my mediocre 10 yr old JBL loudspeakers.
-
can we get a hint of what to listen for and I can test using ABX if I can hear the things that you talk about?
-
Testing if foobar cheats... I converted A to an mp3 then transcoded back to flac LOL
foo_abx 1.3.4 report
foobar2000 v1.2.8
2013/06/23 09:49:43
File A: C:\Users\donunus\Downloads\a.flac
File B: C:\Users\donunus\Downloads\b.flac
09:49:43 : Test started.
09:49:52 : Trial reset.
09:50:02 : 01/01 50.0%
09:50:13 : 02/02 25.0%
09:50:22 : 03/03 12.5%
09:50:29 : 04/04 6.3%
09:50:34 : 05/05 3.1%
09:50:38 : 06/06 1.6%
09:50:43 : 07/07 0.8%
09:50:47 : 08/08 0.4%
09:50:51 : 09/09 0.2%
09:50:57 : 10/10 0.1%
09:51:02 : 11/11 0.0%
09:51:07 : 12/12 0.0%
09:51:10 : 13/13 0.0%
09:51:14 : 14/14 0.0%
09:51:18 : 15/15 0.0%
09:51:21 : 16/16 0.0%
09:51:27 : 17/17 0.0%
09:51:33 : 18/18 0.0%
09:51:37 : 19/19 0.0%
09:51:39 : 20/20 0.0%
09:51:42 : 21/21 0.0%
09:51:45 : 22/22 0.0%
09:51:48 : 23/23 0.0%
09:51:51 : 24/24 0.0%
09:51:55 : 25/25 0.0%
09:51:58 : 26/26 0.0%
09:52:04 : 27/27 0.0%
09:52:08 : 28/28 0.0%
09:52:12 : 29/29 0.0%
09:52:15 : 30/30 0.0%
09:52:18 : Test finished.
----------
Total: 30/30 (0.0%)
At least we know that I wasn't cheated the first time around hehehe
-
I believe you and the results. It wasn't that hard a test, unlike 320 vs. Flac...
-
A sounded like it had better treble energy, thus better clarity/sharpness, while B sounded a bit shaved off at the upper reaches for the first ten seconds.
I thought the same thing at first. Sample A seemed to have a bit more energy up top, but after a few more listens the difference seemed to fade away. Only have my Etymotics plugged in right now.
-
The more you listen to it, the more your senses will dull and desensitize. Then, it's time for a break. I was relying on my first instincts before that stage set in. There's probably more going on that I didn't notice. I didn't bother to assess it past the first half.
-
Yah thats what happened with the first five that I got right 5/5 then it went downhill from there.
Like xnor said though, if we knew exactly what to look for, it would be easier to consistently get which one is which. Sometimes I think I hear a difference but I can't grasp what the difference exactly was so I just did the abx test to see if it was imagined. Then the straight 5/5 proved to me that I was actually hearing a difference but it all faded away after that.
-
A sounded cleaner, more refined w/ clearer more precise dynamics, imaging, micro dynamics, air, and natural timbre. B sounded rougher/grainier around the edges and more smeared, almost like a MP3. Cymbals on the percussion set sounded harsh and mechanical, instruments were softer and less natural sounding feeling more claustrophobic w/ less air to breathe. This hurt pankton found in the trailing decays and settling of notes where A was superior.
Only listened to them twice because running those two 30 sec clips back to back induces fatigue extremely quick. If I were to design a A/B or DBT, those truncated 30 sec snippets would be the quickest way to induce a 50/50 split over extended listening. I had to take a 2-3 minute break just between those two listens as the first set started to boggle my head and ears after it recycled on repeat before I had time to think about it.
B is more fucked up than A, you can tell me otherwise but I won't believe you based on what I heard and tell you to verify your labeling methodology because it would be wrong.
Edit - Although A is better, it's not even as good as my hi-rez rip of Nighfly so your version is either a worse master or has been processed, or perhaps even a higher rez mp3 than B even.
-
Ok so we both like "A" better and I still did get a 5/5 before I forced myself to do more trials. I just can't say much about the plankton though since I am only using the hd280pros but "A" did seem more relaxed and less etched. Still If I listen closely they do sound like they are from the same master since I can hear the tape's wow and flutter.
-
The version is based on the Japanese SACD released 2011. Maybe it's the remastering or the downmixing that makes it sound worse to you.
-
So what is exactly the difference between both files? The way it was ripped?
Did it go through an expander?
-
The version is based on the Japanese SACD released 2011. Maybe it's the remastering or the downmixing that makes it sound worse to you.
The original pressing is less V-shaped compared to this pressing
-
So what is exactly the difference between both files? The way it was ripped?
*** SPOILER ***: I think A is the original, and B has been processed with a filter that has flat or almost flat frequency response, but rings (there is apparently both pre- and post-ringing) at 4-5 kHz. I have an impulse response extracted, although it may not be 100% accurate.
Edit: because of the pre-ringing, it is easy to ABX the difference using only the first second of the samples. The filtered version has an audibly less "sharp" attack.
-
thanks xnor. these tests are always fun. after a quick A/B/A/B, A sample had better definition on note attack and sounded a little more dynamic and detailed. sample B was a little grainier sounding and tonally a little more chalky. subjectively, i preferred A in my rig.
-
Gee don't analyze the files, just tell us what you hear. :P
-
A sounds ever so slightly CLEANER and B has some degree of grain to some of the high hats and cymbals, as most have mentioned. You'd be hard-pressed to hear it on crappy gear. I could barely hear the differences on a TH900 but when I took out the ear-shattering HD700, it was there. I went back to the TH900s and heard the same. If you processed the original copy to be cleaner, then I'd think you have A, and B is the original.
That'd be my hypothesis based on my observations. I could be wrong because I just don't do ABX tests well. :p
-
Crapware:
Player: Foobar2000
DAC/Amp: Focusrite 2i2 (I know xnor loves this redlicious piece of equipment, but he doesn't know it.)
Phones: HD558 (... yup ... and a used one)
Cables: The one that came with the Focusrite, but modded with holy water.
Conditions: Subject was definitively tired (long day at Disney with the kids)
My results:
foo_abx 1.3.4 report
foobar2000 v1.2.6
2013/06/24 01:23:24
File A: C:\Users\osito\Downloads\a.flac
File B: C:\Users\osito\Downloads\b.flac
01:23:24 : Test started.
01:25:20 : 01/01 50.0%
01:25:44 : 01/02 75.0%
01:26:48 : 01/03 87.5%
01:27:10 : 02/04 68.8%
01:29:06 : 03/05 50.0%
01:29:36 : 03/06 65.6%
01:30:14 : 03/07 77.3%
01:30:52 : 04/08 63.7%
01:31:45 : 05/09 50.0%
01:32:07 : 05/10 62.3%
01:32:55 : 06/11 50.0%
01:33:10 : 06/12 61.3%
01:35:16 : 07/13 50.0%
01:35:49 : 07/14 60.5%
01:36:04 : 07/15 69.6%
01:36:41 : 08/16 59.8%
01:37:52 : 09/17 50.0%
01:38:47 : 09/18 59.3%
01:39:55 : 09/19 67.6%
01:40:38 : 10/20 58.8%
01:41:20 : 10/21 66.8%
01:41:45 : 11/22 58.4%
01:42:11 : 12/23 50.0%
01:42:34 : 13/24 41.9%
01:43:04 : 14/25 34.5%
01:43:55 : 14/26 42.3%
01:44:34 : 15/27 35.1%
01:45:06 : 15/28 42.5%
01:46:07 : 16/29 35.6%
01:47:57 : 17/30 29.2%
01:48:06 : Test finished.
----------
Total: 17/30 (29.2%)
I also felt A was a little better. It was hard for me to tell the files apart.
-
Total: 17/30 (29.2%)
I also felt A was a little better. It was hard for me to tell the files apart.
Maybe investing in a horribly distorted Eddie Current tube amp will help improve the results? :)) ;) I shudder at the thought of getting around 50% using a HD650 and O2 amp like some MIA folk in hiding. It truly is amazing how a non linear, noise and distortion machine sold as snake oil could provide a listener the illusion of repeatedly and accurately discerning variances in mastering and playback technique. Powerful Jedi mind trick this SET tube stuff! :&
-
Using crappy cheap headphones (although bright headphones that may make the difference more audible) from a sound card:
foo_abx 1.3.4 report
foobar2000 v1.2.2
2013/06/24 14:56:01
File A: F:\a.flac
File B: F:\b.flac
14:56:01 : Test started.
14:58:08 : 01/01 50.0%
14:58:16 : 01/02 75.0%
14:58:42 : 02/03 50.0%
14:59:18 : 03/04 31.3%
14:59:33 : 04/05 18.8%
14:59:48 : 05/06 10.9%
15:00:14 : 06/07 6.3%
15:00:35 : 07/08 3.5%
15:00:56 : 08/09 2.0%
15:01:15 : 09/10 1.1%
15:01:45 : 10/11 0.6%
15:02:47 : 11/12 0.3%
15:03:06 : 12/13 0.2%
15:03:41 : 13/14 0.1%
15:04:08 : 14/15 0.0%
15:04:24 : 15/16 0.0%
15:04:52 : 16/17 0.0%
15:05:08 : 17/18 0.0%
15:05:47 : 18/19 0.0%
15:05:54 : 19/20 0.0%
15:06:01 : 20/21 0.0%
15:06:22 : 21/22 0.0%
15:06:33 : 22/23 0.0%
15:06:43 : 23/24 0.0%
15:06:51 : 24/25 0.0%
15:06:58 : Test finished.
----------
Total: 24/25 (0.0%)
Also, I did a quick test yesterday with a 7/7 result listening only to the beginning of the files. This time, I chose a point somewhere in the middle of the sample. An easily noticeable difference is that the attack of the snare drums is smeared in a way similar to MP3 compression. See also my other post above for an explanation of how I think the sample was processed.
-
Just goes to show that such tests are often just as much about the experience, skill and consistency of methodology implemented by the user. That many such tests can sometimes, or often be a reflection upon the user rather than the gear involved. It's no wonder that it may sometimes be difficult for the casual or even potentially biased to distinguish between correlation and causation when interpreting data.
-
Thanks for the logs and thoughts.
Should I post how one of the files was processed?
-
*** SPOILER ALERT ***
.
.
.
.
.
.
So the filter used to process B has pre- and post-ringing but its magnitude of the FR is flat.
The pre-ringing is what some of you recognized as similarity to MP3 compression. Pre-ringing is known to be (a lot) more offensive than post-ringing, so successful ABX logs were expected.
Here's what the CSD of the pre-ringing looks like:
(http://xserv.mooo.com/xnor/audio/images/ringing1-rev.png)
and post-ringing (http://xserv.mooo.com/xnor/audio/images/ringing1.png).
.
.
.
.
.
More listening tests to come, if you like.
-
So I have a theoretical question that's been bugging me for awhile. Maybe for another thread, but I can move it later if anyone thinks it's worth exploring. The general common consensus is that the digital domain gives us perfectly accurate audio reproduction. Bits are bits and all that. Yet, digital appliances like DACs seem to often employ either a specific selection or choice of digital filters. So the question is, if digital audio reproduction is perfectly accurate, why do we need the existence of different filters and which filter represents the 'correct' one?
-
So I have a theoretical question that's been bugging me for awhile. Maybe for another thread, but I can move it later if anyone thinks it's worth exploring. The general common consensus is that the digital domain gives us perfectly accurate audio reproduction. Bits are bits and all that. Yet, digital appliances like DACs seem to often employ either a specific selection or choice of digital filters. So the question is, if digital audio reproduction is perfectly accurate, why do we need the existence of different filters and which filter represents the 'correct' one?
Timely question. An interesting discussion starting up over at AudioStream.com, involving Charlie Hansen from Ayre and Gordon Rankin from Wavelength. Very much worth a look.
-
Filters in the DAC are needed to create a smooth transition from one sample voltage to the next sample voltage.
As DAC's must be capable of handling various bit-rates this 'smoothing' can not be done correctly with a fixed analog filter as that would need to change it's cut-off point depending on the sample frequency which is not practical/feasible in the analog plane but rather easy to do in the digital domain as these DAC's 'work' on much higher frequencies anyway. As there are a few ways to Rome and people like to have choices different filters are easily made and selectable.
But in essence you are right in assuming that even when the audio-data enters all DAC circuits in a similar way (otherwise it wouldn't be bit perfect) the analog waveform is NOT exactly the same when it comes to the frequencies that come near the nyquist frequency of the signal.
-
Thx burn and solder! I'll take a look at the article, hope it saves me the trouble of writing the piece I was thinking about. This stuff is such a time suck for people that don't work in the industry or don't make money off it..
-
The general common consensus is that the digital domain gives us perfectly accurate audio reproduction. Bits are bits and all that. Yet, digital appliances like DACs seem to often employ either a specific selection or choice of digital filters. So the question is, if digital audio reproduction is perfectly accurate, why do we need the existence of different filters and which filter represents the 'correct' one?
You need an anti-aliasing filter. That's basically a steep low pass filter that removes (in practice: attenuates) frequencies above Nyquist.
Now there are some parameters to this filter.
One is phase. If you go min phase you will have quite a bit of phase shift in the audible range, because the filter needs to be steep.
In most DACs you'll therefore find linear phase FIR (finite impulse response) filters. No matter how steep the filter, the phase shift will be linear / the group delay will be constant.
Then we have slope/steepness which depends on the order of the filter. A steeper filter does therefore not only increase computational cost, but also means more ringing in the time domain. This ringing is usually not in the audible range.
There are some audiophiles that tell you reducing the ringing will improve sound quality, but all it will do is cause a more aliasing - into the audible range. And I can tell you, aliasing sounds really bad.
The cutoff frequency has to be high enough to not attenuate audible frequencies, but low enough to achieve high enough attenuation at Nyquist.
There's also stopband rejection. This is designed to be in line with the performance of the DAC. No point in using a 180 dB filter if the noise floor of the DAC is at -100 dB.
If you want to play around with those parameters I can recommend SoX. There's a SoX resampler plugin for foobar2000 too.
(http://upload.wikimedia.org/wikipedia/commons/thumb/6/60/Butterworth_response.svg/512px-Butterworth_response.svg.png)
Imagine cutoff frequency = Nyquist. Everything after the green vertical line would be mirrored (aliased) back to below Nyquist.
This would be a damn bad anti-aliasing filter.
-
There's also a good discussion of the trade-offs inherent in filter design here.
http://resonessencelabs.com/digital-filters/
-
^ I may have skipped something but I didn't see them mentioning that all files were already filtered at creation, either in the ADC or later using a resampler, so you have no influence on that. No matter what playback filter you choose, you cannot undo the initial filter.
If the signal is already properly bandlimited applying another filter doesn't do much. You can however degrade the signal by using an apodizing filter that doesn't remove frequencies above Nyquist, i.e. doesn't remove all aliased frequencies.
-
Ok. Now I used my computer headphone out and my KSC75...
foo_abx 1.3.4 report
foobar2000 v1.2.6
2013/06/25 09:12:22
File A: C:\Users\osito\Downloads\a.flac
File B: C:\Users\osito\Downloads\b.flac
09:12:22 : Test started.
09:12:57 : 01/01 50.0%
09:13:11 : 02/02 25.0%
09:13:33 : 03/03 12.5%
09:13:45 : 04/04 6.3%
09:13:59 : 05/05 3.1%
09:15:10 : 06/06 1.6%
09:16:23 : 07/07 0.8%
09:16:47 : 08/08 0.4%
09:17:25 : 09/09 0.2%
09:17:56 : 10/10 0.1%
09:18:52 : 11/11 0.0%
09:19:14 : 11/12 0.3%
09:19:49 : 12/13 0.2%
09:20:06 : 12/14 0.6%
09:20:33 : 13/15 0.4%
09:21:24 : 14/16 0.2%
09:22:03 : 14/17 0.6%
09:22:25 : 15/18 0.4%
09:22:43 : 15/19 1.0%
09:23:07 : 16/20 0.6%
09:23:11 : Test finished.
----------
Total: 16/20 (0.6%)
Using crappy cheap headphones (although bright headphones that may make the difference more audible) from a sound card:
Total: 24/25 (0.0%)
Nice!... :)p7 But...
(http://cdn0.meme.li/instances/300x300/39115159.jpg)
Linky (http://www.changstar.com/index.php/topic,302.0.html)
-
Appreciating the dialog here, thanks xnor.
I did have a quick listen to the files but didn't do any testing. I though they were such poor masters that they both sucked.
:)
-
I've read the SACD is the same as the DVD-A so all I can think of is that you don't like the 5.1 to 2.0 downmix or you just dislike the recording itself. Will compare the downmix to the CD-layer version.
What version do you recommend, if any?
-
Appreciating the dialog here, thanks xnor.
I did have a quick listen to the files but didn't do any testing. I though they were such poor masters that they both sucked.
:)
Personally I find that album overrated for mastering quality. I never user it for testing.
-
Personally I find that album overrated for mastering quality. I never user it for testing.
Exactly. I do use "Walk Between the Raindrops" sometimes when I want to hear something with shitty compressed sound.
I like Donald Fagen a lot, but that album just pissed me off.
-
You need an anti-aliasing filter. That's basically a steep low pass filter that removes (in practice: attenuates) frequencies above Nyquist.
Because ultra-high frequencies at the sampling rate, whether NOS or over-sampled or up-sampled, may destroy tweeters, cause amps to oscillate - they make the rest of the chain unnecessarily amplify and attempt to reproduce extra unneeded signals which the amps/tweeters may have never designed to do so. It's an engineering thing, which just as much about about margins and tolerances "the oh-shit factor" as making science tangible.
(This is why I feel Harold's O2 design with the pot between the voltage multiplier section and the current buffer is retarded - and why I've had to answer more than a few PMs here and over at HF on how to make the O2 work with even not so "hot" sources. I've even made the O2 clip myself with slightly less DAC voltage output that I thought would happen in theory. The assumption that the O2 would be always used with wimpy output DACs, or that end-users would possess sufficient technical knowledge about their DAC's voltage output, appropriate op-amp gain, etc. , is not one that would ever be made by a credible engineer meaning to offer the product to the public. At the very least, a warning label regarding DAC/source output and internal gain settings should have been silk-screened near the input jack.)
Some DAC chips produce more high-frequency junk than others. For example, I've used the output straight from AKM voltage-out DAC chips, but generally this is still not advised. I've heard of some folks rely on tubes (amps of output section of DACs) when not using a filter because of tubes' inherent bandwidth limitations to usually that of human hearing (when tubes are used for audio applications.)
The farther the ultra-high frequency junk is from 20kHz (~limit of human hearing), the better because more gentle, less steep filters (which have less phase shift and probably other more desirable behavior - this is ultrabike's area of expertise) can be used. The very early CD Players were NOS, and there was some audiophile debate back in the early 80s whether this harshness was because of the steep brickwall analog filters used to kill junk at 44.1kHz. Thus oversampling DAC were born. But please, let's not debate that. That was the 80s.
-
Anti-aliasing filters are used in ADCs, not DACs [correction by ultrabike: DACs can implement decimation filters which are a form of digital anti-aliasing].
An anti-aliasing filter is a steep lowpass filter which tries to eliminate frequencies above Nyquist, which is half the sample rate.
For instance, if you are converting an analog signal to digital at 48kHz, your ADC should have a lowpass that ideally rejects everything above 24kHz. This is because any frequency above the halfway mark (24kHz) could be falsely interpreted by the DAC stage as a corresponding lower frequency.
Look at image d in this figure:
(http://i.imgur.com/bfA6BZT.png)
The analog signal's frequency is 0.95 of the sample rate...in the case of 48kHz, we have a 45.6kHz sine wave. The black squares are the actual data points captured by the ADC stage. Now, when a DAC tries to reconstruct the sine wave, you can see that the data points form a nice super-imposed sine at 0.05 of the sample rate or 2.4kHz. There's just no way for the DAC to reconstruct anything above the Nyquist frequency, so we must get rid of those frequencies. You can imagine how nasty this could sound with actual music instead of sine waves. Ultrasonic harmonics/noise/garbage being interpreted as lower frequencies, mixed into the audible spectrum. Nasty.
These filters can't be theoretically perfect in practice because we're still dealing with the limitations of analog. The filter cannot be infinitely steep, as would be most ideal. The best ADCs/DACs typically are those with painstakingly-designed analog stages.
There are software tricks that can be done to reduce the complexity of the analog anti-aliasing filters (on the ADC side) and reconstruction filters (on the DAC side) required, BUT you still need great analog design to get quality results from digital...whether converting from analog or to analog. The design and quality of these analog filters is one big reason why DACs and ADCs can sound so different. It's not just ones and zeroes.
-
Digital filtering at the recording stage is called anti-aliasing and is indeed meant to remove signals above half of the sample frequency as they can be 'mirrored' back in the audible range. In this case the sample clock is the mirror. All signals NEAR that sample frequency that are present in the original signal, that is to be recorded, thus have to be removed in an analog way.
The higher the sample frequency the higher the analog content may still be or the less 'steep' the filter can be.
As Purrin already mentioned, the steeper the filter the more post ringing (for analog) is present and the more phase and amplitude (ripple in the upper frequency range) problems exist.
In the Digital to Analog conversion part a filter is essential to create smooth transitions from one output voltage step to the other.
The posted diagram already shows the 'steps' get bigger the higher the signal frequency is.
Those 'errors' need to be removed.
This can be done by an analog (low pass) filter OR by upsampling (creating sample points between the actual samples) and calculating where they might have been through algorithms thereby reducing the output steps in size.
This way the amplitude of the 'steps' is made smaller and shifted to a higher frequency which is EASIER to filter with a NOT steep analog filter to remove the high frequency 'errors'.
So in the DAC a filter is needed to 'smooth' the analog output so the steps are not there anymore.
Both digital and analog ways exist to reach the same goal (an analog signal without 'steps') and both have advantages and disadvantages depending on the TYPE of filter used.
Both in the analog as well as in the digital realm different types of filters exist with different behaviour in the AUDIBLE band and outside of it.
-
Anti-aliasing filters are used in ADCs, not DACs.
Depends...
In the case of an up-sampling DAC, if the up-convert rate is an integer multiple of the original rate, then an anti-imaging (or interpolation) filter is used. However, if the up-convert rate is not an integer multiple of the original rate then down-conversion might be required as well. Down-conversion might imply a decimation filter, which is kind of a digital anti-aliasing filter...
---
On a different note, I find interesting that the ringing in the test track was applied in the same (roughly) region that my KSC75 and HD558 exhibit ringing... around 4 and 5 kHz... Maybe something like a modded HD800 with less nasties in that range will be more revealing in an ABX.
-
Digital filtering at the recording stage is called anti-aliasing and is indeed meant to remove signals above half of the sample frequency as they can be 'mirrored' back in the audible range. In this case the sample clock is the mirror. All signals NEAR that sample frequency that are present in the original signal, that is to be recorded, thus have to be removed in an analog way.
The higher the sample frequency the higher the analog content may still be or the less 'steep' the filter can be.
As Purrin already mentioned, the steeper the filter the more post ringing (for analog) is present and the more phase and amplitude (ripple in the upper frequency range) problems exist.
Good points. We can't have it both ways...speaking in terms of theoretical ideals in the time domain versus freq domain and vice versa.
In the Digital to Analog conversion part a filter is essential to create smooth transitions from one output voltage step to the other.
The posted diagram already shows the 'steps' get bigger the higher the signal frequency is.
Those 'errors' need to be removed.
This can be done by an analog (low pass) filter...
I thought a low-pass would only work in this application if the DAC could somehow generate a true impulse train (which is impossible in the real-world since the impulses are infinitely narrow).
So, if your DAC uses zero-order hold (sounds like what you're describing by "voltage steps"), then you need a filter that looks like this:
(http://i.imgur.com/RDxnxt0.png)
OR a multirate filter (downsampling, upsampling).
But a low-pass will not reconstruct the effects of zero-order hold. Am I wrong here?
Depends...
a decimation filter, which is kind of a digital anti-aliasing filter...
Thanks for the correction.
-
I thought a low-pass would only work in this application if the DAC could somehow generate a true impulse train (which is impossible in the real-world since the impulses are infinitely narrow).
So, if your DAC uses zero-order hold (sounds like what you're describing by "voltage steps"), then you need a filter that looks like this:
(http://i.imgur.com/RDxnxt0.png)
OR a multirate filter (downsampling, upsampling).
But a low-pass will not reconstruct the effects of zero-order hold. Am I wrong here?
I think you are correct. Those are compensation filters which can be implemented in the digital domain...
---
So... Did you give the audio files comparo a shot?
popcorn
-
Haha! Yes.
Don't have the log in front of me, but I scored 8/10 correctly identified with K701s.
Also, I don't care for Fagen or the production of this particular album. Just my opinion.