A listener's caveat

What I'm hearing in your post is to cast aside preconceived notions or audiophile propaganda. Reading that something digital or a processor was used to make an cd or Lp, that OP amps were used in an amplifier design, whether a cd is a remaster, whether a cd is'nt a remaster, whether an Lp is an early or later pressing. I have been guilty of these automatic predispositions and suspect other folks here have too. Or what you know or dont know wont hurt you, If it sounds good it sounds good right. Ever hear something and think it sounds pretty good until you find out x y or z and then inexplicably change your opinion of it?

 

Yes I am doing primarily measurements right now. Which I hope will not taint my listening tests later. But I have done some listening tests before and I just am not sure in some of the cases which is the most accurate. Because of mitigating factors with DACs.

I did that last night. It was my first attempt and the stuff you are seeing past the nyquist frequency might be easily fixable with an antialias filter.

What I think the site is missing and you are leaving out is the last step in most chains is to convert to 16-bit with dither. They really should put up an ideal sweep at 44.1 16-bit and let you see how noisy that is. When you take that last step into consideration those low level distortions in the first example become negligible. BUT the audible ringing that might be in the second one which is close to the same operation that the advanced SRCs perform I do not consider negligible.

Some of the SRCs that had a lower benchmark on that site I have heard reports of sounding the best, with the least audible ringing and a very small amount of distortion - a healthy compromise.

 

Why try to patch a less-than-good situation if you can get it right from the start with another algorithm? Or this just an intellectual exercise for you?

 

Hi Barry,

Interesting post and thread.

There are several things, though, that come to my mind upon reading it:

- There are levels and levels of critical listening, and levels upon levels of demand depending on the user, his equipment, his budget, and his intentions when he/she decides on what gear sounds best at a certain moment.

- I think it is a good idea to address the generic comments/opinions about audio as you have done here, but this also creates its issues: On the one hand you are mentioning that one should listen attentively rather than go by the marketing, I agree; but when you mention something like this, you suddenly see people asking you what sounds better than what, which would seem to have missed the point you made originally which was, 'listen for yourself, don't just take any opinion or data as right.'

- Each person, you included, arrives at his/her conclusions depending on their own experience with gear, software, and critical (or not) listening. Some time ago I was defending what I heard when I listened to the recordings I made at 192/24 with my current sound card as compared to lower resolutions. At that time you were totally centered on 96/24.

Now you are clearly behind the idea of recording at 192/24. You say that it is because you hadn't heard a transparent enough A/D that rendered that resolution correctly.

Well, I do not know how good or less good my current sound card is as compared to the one you are using because I haven't made comparisons. So, I do not think that I should make any statements in that regard. What I do know is that my ears tell me that my sound card (which I am sure is not the best one out there) does deliver better sounding recordings at 192/24. So, in my experience in my own context, and in yours within yours we agree on this. If I ever get to listen to your present A/D converter of choice I might discover that the differences between its recordings at 192/24 and lower resolutions are even more noticeable, but this - to me - in no way discards or lessen what my ears tell me with my current one.

- I believe that often certain statements like 192/24 recording is better than 44.1/16 are made in threads based on a person's experience and because it is easier/more straightforward to make such a comment than to go into the large amount of grey areas that delving into your otherwise logic and valuable post opens.

For example: When the target for an audio product is 44.1/16 and the person has assumed that he/she will get better results by recording originally at 192/24 (without going into the A/D converter quality variable) there is another 'grey' zone immediately present: will the downsampling + dithering algorithms used are not transparent enough some or all of the qualities that could be gained by recording at 192/24 will be lost in the process.

- Then, there is the gear/software synergy to consider (to me another 'grey' spot). Working with a great SRC, but not having the right gear to be able to hear what it is doing might not even let the person register the advantage. And this is only one of the many variables involved.

What I mean to say by this is that, as I see it, as one goes towards more exacting heights in one's recordings one must not forget that things become less cut and clear. We are being more exacting, yes, but there are different variables involved in every move that make for a 'grayer' panorama in that everything is not just black or white. The ability to notice and decide between the variables involved in each of these instances is not something that a novice in audio can usually perceive or understand first hand. It takes a process that involves a lot of listening. In these cases, it is much easier to transmit generic ideas which are used, as the post I cite below says, as shorthand so as not to confuse/overwhelm the reader.

This does not mean that your post is not valuable, but that it can open a world of possibilities as well as create a lot of confusion and, when not, make people take the 'easy way', which is usually the, "OK, then tell me which are the good products for this." If this is your intent, then it is OK, if your intent is to have people make decisions with common sense then make sure that is clearly defended.

This is my take not only for reverb, but for sound cards.

All the above is just my 2-cents worth.

[Excuse me if I have rambled during this post, I've been writing it while having to do something else].

 

Hi Key,

These are two distinct processes and each can introduce its own artifacts.
There is a good dither comparison here.

I'm not sure where such reports originated, who made them and under what circumstances. In my own tests (and in beta testing a number of SRC algorithms over the years), I've found the results on the infintewave page are very close to what is heard in listening tests. Closer, in correspondence to what is heard, in my experience, than any other set of measurements I've ever seen.

Also, I'm not sure of what is attained from measurements in the absence of listening tests. To my mind, their value comes in making the correlation, without which they have no value since the hardware/software is designed for no other purpose than to be listened to.

Again, just my perspective.

Best regards,
Barry
www.soundkeeperrecordings.com
www.barrydiamentaudio.com

 

Hi Metoo,

I think there is a difference between someone asking me about what I like and them taking this to automatically mean it will be what they will like.

The point of this thread however, is not simply about listening for one's self.
It is about not making general conclusions based on specific circumstances.

This is not correct. I was not "totally centered" on 24/96. It was what I used because at the time I had not heard it bettered. I prefer the converters I'm using now at 24/96 (heck, even at 16/44!) to most other converters I've heard (but certainly not all) at higher rates.

Again, in keeping with the theme of this thread, I never said "24/192 is no good and 24/96 is the best". I did not generalize from specific experiences. Just as I wouldn't say or support the idea that 24/192 is better than 16/44.
Sometimes yes. Sometimes no.

I don't understand why it is "easier" to make sweeping generalizations, which are logically unsupportable than to qualify a statement with one's personal experience, which is.

In some situations, with some converters, 24/192 will be better. In other situations, with other converters it will be worse. A generalization in either direction would be equally wrong.

Perhaps we just see this differently. It sounds to me like you are saying it is less confusing for the novice with sweeping generalizations, even if they are not accurate. Forgive me if I misunderstand but if this is what you are saying, I am 100% at odds with this. People don't learn by absorbing erroneous generalizations. They learn by practicing the avoidance of these.

I must also ask if you find it upsetting that folks ask me for recommendations on things like EQ (or soundcards, speakers, etc.). If so, I'm sorry about that but I will always answer these honestly and based on my own experience. The smart folks who ask me these questions, will use my answers as a basis for their own investigations - to see if they agree or not. The ones who don't want to think for themselves... well, they might like the generalizations this thread warns against.

In conclusions, you may feel differently and I completely respect how you feel. But when someone says something like "digital EQ is bad" or "192 is better than 44.1" my inner flag goes up with a warning that they are at best mistaken and are more than likely talking through their hats.

And if someone asks me what my favorite EQ is, I'll tell them what my favorite EQ is. This, with the understanding that I can't tell them what their favorite EQ is. ;-}

Best regards,
Barry
www.soundkeeperrecordings.com
www.barrydiamentaudio.com

 

People do it on this forum all the time!

 

Just a proof of concept really. I mean since SoX is free it does make it a weird excercise. I also just wanted to test out the idea of upsampling to something that both are multiples of.

And who knows maybe if I use the right lesser SRC and use that technique I may be able to actually beat SoX or iZotopes.

 

Actually I think we are talking about three different things not two. I am considering the entire chain in a normal work flow. Those other two tests are compartmentalizing the steps. My point is that after you convert to 16-bit with dither it brings a lot of those SRCs to about equal in mesurable performance.

When did I say I was not going to do listening tests?

 

I agree. I, myself, have asked you before... and then gone on to try things out on my own (as, it seems, Key is doing).

The problem, when it comes up, and it does, would be the full responsibility of the person getting the message.

Thanks for making this clear (at least to me). I hope it carries out in what I imagine to be an interesting many sided reflection.

When I said that you were "totally centered" on 96/24 I meant that you were doing your professional work at that resolution. I was just pointing out that in my own context I had my attention on 192/24 and was noticing positive differences. This meaning that everyone has his own experience within his own context. And that the 'advantages' of recording at higher resolutions can be also achieved (although perhaps less exactingly) with lesser sound cards. So, again, this is another 'grey' area that reinforces the 'it depends' context you are wanting to stress here.

We agree. If you understood that I was saying that you thought that 192/24 was no good I must have expressed myself with clarity.

But, let's take a look at how one as an amateur audio buff would look at this. One would probably say, "So, 192/24 can be good or not good, 96/24 can be good or not good, 44.1/24 can be good or not good. So, what should I do?" In this moment 'expert' advice (the same one you warn against) is needed or one is just stuck in a conundrum.

Now, if you say a higher resolution with a better A/D converter should deliver better results, you are setting a generic truth that can be achieved by using the right hardware. The next thing is, "How does my current hardware perform under this situation? What other other hardware out there that I can upgrade to within my budget to achieve this 'higher resolution recording is better' axiom?" In this case, the person has a line to follow, IMMO it is easier to connect the dots or be compelled to do it within this mindset. The other context to me is too diffused. This is were my point of view is different from yours.

Yet, the personal experience is, in itself, a specific circumstance in that it is your own personal experience within your own experiential (critical listening, previous audio experiences, how you are wired), hardware, and software context.

We agree, but I don't see a clear premise to follow in this statement, because the person is essetially being told that 'it depends', anything is possible depending on the hardware, but they are not informed/made aware (or might not even be able to notice) the specific features/qualities that would allow them to tell one case from the other, especially if they lack the experience. This is where I see that, although we agree on the basics, this invitation to a relative take on the issue might result either in confusion or in just adopting the 'expert' advice.

Yes, but they need a clear cut guideline to follow. A 'sweeping generalization' is not of use to you given your experience, but the opposite is not a good starting point for whomever is trying to make some sense of this.

On a side note, let me point out - in case it is not clear by now - that we do not disagree on the bottom line, just on what way it should be communicated. This is much like what happens when one learns music harmony. You are first taught all the rules and how to keep to them. Then, once you arrive at a certain knowledge of the fundamentals, you are told that you can do whatever you want. If you had been told this from the start you would not have learnt music harmony. If you want I can go more philosophical and tell you an example within a Buddhist philosophy context).

No, I am not. As I mentioned at the beginning of this post I am one of those who have previously asked you for recommendations and will in the future.

As to what folks do with your recommendations, I already answered that above and on a previous thread.

IIRC, I mentioned in my post above 192/24 vs. 44.1/16 (not 44.1/24 recording). There are many out there doing their recording at 44.1/16 and then tweaking the files. I guess that we can both agree that we can make a 'sweeping statement' that recording at 16 bits and tweaking a file (another thing is to leave it as is) will not deliver the best results one can get with his set up. Stressing on things like these, IMHO, goes a long way in clearing certain preconceptions that are more dangerous (for their popularity) than if 192/24 is better than 96/24 and, if so, when.

 

Hi Key,

My experience in this regard has been different. I have not found a situation where dithering alters my perception of an SRC algorithm.

I didn't suggest you were not going to listen or that you said that.
You clearly mentioned you'd be listening later.
I asked what the value of the measurements was in the absence of listening.
Once you've listened, I believe any value in the measurements will reveal itself.

Don't get me wrong. I hope what you're experimenting with works out.
I hope you don't merely equal the best SRC, I hope you exceed it.
Then I'll ask your permission to use what you've found in my work.

Best regards,
Barry
www.soundkeeperrecordings.com
www.barrydiamentaudio.com

 

This brought this up in my mind:
I was starting to look into a couple of things about SoX the other day. It started while looking into the different graphs at the SRC Comparisons website. I have chosen to use intermediate phase in my SoX settings until now (due to their lower distortion in the 1kHz tone graph), but my limited knowledge of the technical aspects is not letting me tell if its results - as seen in the phase response graph - are worse than those obtained by choosing linear phase. I have still do to a listening test because I haven't had the time.

Given that you seem to be quite aware of the phase element, what is your take on the differences shown by those graphs?

 

Hi Metoo,

I don't think our points differ so much after all.

However, what I would say to the novice in answer to your first paragraph is to try working with what they have and use it at different sample rates to see how it performs.

What works for me using my gear will not necessarily be true for that person using their gear. There is no easy way. One must experiment and listen.

Good point.
But applying processing to a 16-bit file will lengthen the word and lose the low order bits, taking the focus and fine detail with them. This is true regardless of the device being used. It is the nature of digital audio.

Advising fellow audio hobbyists in this regard isn't the same as making a sweeping generalization about all devices sharing a single characteristic (e.g. digital EQ or a given sample rate, etc.). A single characteristic is but one of many variables that is not true across all instances.

With something like processing a 16-bit file, we're talking about a phenomenon that will occur in all instances, regardless of the hardware/software used. You can use soundBlade with its wonderful internal engine which creates some of the best sounding files I've ever heard or you can use a free program downloaded from the Internet. Processing a 16-bit file in either one will cause the same damage.

Perhaps we are using different words to say the same thing.

Best regards,
Barry
www.soundkeeperrecordings.com
www.barrydiamentaudio.com

 

BTW, I had never done the dithering test that Barry posted before. It just so happens I ended up choosing the Waves ultra algorithm, which is the second one (by a long shot) from the first one, which is the MegaBit Max Ultra (my second choice).

I would have loved to see/hear iZotope's MBIT+ Ultra one there.

 

Hi Metoo,

Your wish was granted. MetaBit Max Ultra is MBIT+ Ultra.
It used to be called MegaBit Max and is now called MBIT+.

Best regards,
Barry
www.soundkeeperrecordings.com
www.barrydiamentaudio.com

 

Sorry, Charlie. You really gain nothing by upsampling to a divisable amount, then downsampling again, no matter what algorhythm you use. Face it, Sound Forge's SRC isn't very good, and I have used it.

 

Nice to know this. Thanks Barry.

What I noticed between the two I saw as best was that although the MetaBit Ultra one did have the music apparently fading later, I did hear more dithering noise on it. Yet, on the one I chose the dithering noise seemed less noticeable. In any case, it was within my two final choices.

Have you tried/tested the Waves Ultra algorithm? If so, what differences did you notice with the MBIT+ one?

 

Hi Metoo,

If you recall, last year I had put temporary pages up on the BDA web site, one comparing a bunch of SRC algorithms and the other comparing about 14 dither/noise shaping algorithms. Waves Ultra was one of them as I have over a dozen different dither/noise shaping algorithms (and several SRC algorithms).

The one thing the 24-96 web page does not have -at least I haven't found it- is the original, undithered file. (I provided an undithered file on the page I put up, along with the original, unconverted sample rate file on the SRC page.) Without an original to compare to, I find the test becomes which algorithm(s) one likes better but with no way of assessing which is most true to the original, unprocessed file.

In my own tests, I found the Waves algorithm softened timbres, lost some low level information and clouded the soundstage, defocusing it. I find this is true to varying degrees with a lot of dither/noise shaping algorithms. I thought Pow-R 3 provided a result closer to the original than Waves did. And I found MBIT+ to provide one closer still - to this date, the closest I've heard.

All my tests are done by comparison with the original, unprocessed file. My selection is based on which sounds the most like the original, rather than on preference without the original to refer to. For example, if the original is slightly bright, I want the processed file to be slightly bright. An algorithm that removes the brightness might sound "better" but it is not as transparent. (I'd rather choose the EQ myself, instead of having the algorithm do it for me. ;-})

Best regards,
Barry
www.soundkeeperrecordings.com
www.barrydiamentaudio.com

 

Without a doubt being able to listen to the original file is very important because it sets the reference. I agree that in this comparison one goes for that one which one likes more from them all. In my case I did not listen for soundstage, nor did I know - of course - what the original sounded like. I centered mostly on the dithering noise and, secondly, on how much low level detail I heard.

To me, the Waves dithering noise sounded less obstrusive, while there seemed to be slightly more level detail in the MBIT+ one (on the final fade out) although it was really a mixture of sound and noise by that time. Had I listened also for soundstage, which having the original file as reference might have suggested, I might have finally decided on the MBIT+. In fact, it was my choice until the last moment. As I said, the more intruding dithering noise made me finally choose the Waves one.

 

Great read!

 

Thought I'd hijack Barry's excellent thread for a bit by reporting my findings re: digital EQ's for the PC platform.

I really needed a good all-round digital eq so I decided to do some extensive listening tests comparing lots of plugins.

I listened for transient 'smear', transparency and predictability.

The following plugins I found were excellent, non-intrusive and usable:

All in my experience, of course.

-Flux Epure II. This is my current favorite. It's transparent and sounds exceptionally smooth. No smear to speak of and it does exactly what I want.
It comes at a price though: $479. Mac & PC, by the way.

-Sonoris LPEQ. It's linear phase, which means it filters without any phase shifts, a great concept that can only be done in the digital realm. To me, this one's the king of transparency. I guess the designer clearly knows his math and I wouldn't hesitate to use it in mastering in place of my Weiss LPEQ.
Also both PC & Mac, $225

-iZotope Ozone 4. Yes, it's iZotope again. (and not iZotopes) So it's an all-round mastering plugin, but I found if you set its EQ section to 'digital' instead of 'analog' it's not half bad. Really good actually, to these ears. And of course, you get the MBit+ dither thrown in.
Again PC & Mac, $250

There are some others I like, notably the DDMF LP10, which is another linear phase EQ that's almost on par with the Sonoris one. What's great about this one is it's donationware, so you can get it cheap!

At work I use the Algorithmix Red regularly, which some believe is the best EQ plugin in the world. It's fantastic, but it's PC only and prohibitively expensive ($1200) though you can join in a group buy (as we did) and

 

Really? Isn't this the same exact technique used by the best scorers on the SRC site?

 

I have to take a look again. But basically what I have heard from other engineers is that they have been sort of bitten by "linear phase" uneven resampling. That they used them and then later could hear "ringing". It's not the easist artifact to hear. I believe I have heard it on acoustic guitars. Like if you mic up an acoustic guitar (helps if the guitar is fresh strung with strings hanging off the end of the tuning keys so it hits weird ultrasonic pitches and attacks) you can hear the effect on high end attack.

this is from SoX's manual which tells you a lot more about it than iZotope's

"All resamplers use filters that can sometimes create ‘echo’ (a.k.a. ‘ringing’) artefacts with transient
signals such as those that occur with ‘finger snaps’ or other highly percussive sounds. Such
artefacts are much more noticable to the human ear if they occur before the transient (‘pre-echo’)
than if they occur after it (‘post-echo’). Note that frequency of any such artefacts is related to the
smaller of the original and new sampling rates but that if this is at least 44.1kHz, then the artefacts
will lie outside the range of human hearing.
A phase response setting may be used to control the distribution of any transient echo between
‘pre’ and ‘post’: with minimum phase, there is no pre-echo but the longest post-echo; with linear
phase, pre and post echo are in equal amounts (in signal terms, but not audibility terms); the intermediate
phase setting attempts to find the best compromise by selecting a small length (and level)
of pre-echo and a medium lengthed post-echo.
Minimum, intermediate, or linear phase response is selected using the −M, −I, or −L option; a
custom phase response can be created with the −p option. Note that phase responses between ‘linear’
and ‘maximum’ (greater than 50) are rarely useful."

 

Key, thanks for the information. Although it does not address what I was talking about it has led me to look into something else that has cleared some things on the graphs.