SpecWeb 2.3 Beta 1

QuadraphonicQuad

Help Support QuadraphonicQuad:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
Well there is also play-remote, in that case. If you have a smartphone, tablet, laptop, etc. it lets you control SpecWeb via a browser while you sit on the couch in your living room.

Of course, you still need a way to get the audio from your computer to your 5.1 system :0(

Can you say more about "I see no difference"? Are you looking for some visual feedback vs. hearing a difference?
 
Have a look at the attached waveforms. Their name is self-explanatory. They look identical to me. That's why I think I'm doing something wrong.

Screen Shot 2021-04-12 at 7.27.20 PM.png
Screen Shot 2021-04-12 at 7.28.03 PM.png
 
I've just used -c0 -f5 with SpecWeb 2.2 and there is a marked difference. Is there a bug in the beta of 2.3?
Screen Shot 2021-04-12 at 8.29.30 PM.png
 
  • Like
Reactions: GOS
Once again, I'm stumped, this time with version 2.2. -0-100 -c0 -f18 appears to be the same as -0-100 -c0 -f25. For the time being, I have to conclude the difference in waveforms of -c0 -f18 in version 2.2 relative to version 2.3 is not the result of better compliance of parameters by version 2.2, but derived from the overall processing of 2.2 itself. It appears the -f parameters I enter are not heeded.
 
Please LISTEN to your tests, then let me know. The volume of the waveforms probably won't change that much visually.
 
Wow! So, I started with a track from Springsteen's Born to Run. Using 2.2 almost everything ended up in the center and the rest of the channels were very artifacty (I know that's not a word). With 2.3b1 and the default settings the non center speakers sounded cleaner and had more in them with fewer artifacts. A great improvement. I then moved over to Kathleen Edwards new album which had a swirling kind of sound to it (sorry, I don't speak the sound geek language). There was a modest improvement there. The artifacts were decreased a bit. Not as much as I had hoped after listening to Springsteen.

Can you give me an idea of what to expect if I increase or decrease the -y and what a good increment would be. I don't have a lot of time, but I want to help. Any direction by you would be appreciated.
 
Please LISTEN to your tests, then let me know. The volume of the waveforms probably won't change that much visually.
Well, I've done as you requested, but I'm still confused. I must, first of all, admit that -c0 does, indeed, contribute to bring a noticeable surround effect to stereo recordings with heavy emphasis on the centre channel. As for -fx, I've tried, as you know, 5, 18 and 25 degrees. Perhaps I'm going deaf, but I was unable to perceive any differences. I also asked for one of my sons' assistance and he didn't perceive any differences either. My wife, on the other hand, claims 25 degrees is very similar to no "-f" parameter, whereas 18 and 5 degrees sound more quadraphonic. After her assessment, I listened to all three versions again, but, alas, I didn't perceive the purported differences. If there is a difference, for me it's a mystery that the waveforms of different recordings should be, as far as I can see, the same. Even if the overall loudness of the fronts and the rears remains constant, no matter what you do, wouldn't an instrument shift from the front to the rear because of the surround treatment imply some change to the waveform? If so, how come the waveforms are the same?
 
@Infomas; at the bottom of the SpecWeb output (per track) there is:

transient_threshold: 0.0400 percent transients processed: 7.09% # of transients processed: 5785074 total bins processed: 81560786 time slice of a bin: 0.005805 seconds min_mag: 0.0000 max_mag: 0.1219

In this case showing the current default -y value of 0.04, and the percentage of the song that got processed as "transient". That, and your ears, are the feedback for the -y parameter.

There doesn't seem to be a linear relationship between the -y value and percentage and of course it various with musical content, but I have tried values as low as 0.001 or so (note the extra zero, compart to 0.04). Perhaps I should have given it a dB scale. Anyway I'm looking for "It took a -y setting of 'x' to remove all the artifacts, and over all the songs I've processed it seems like 'z' would be a better default than 0.04.

But if the percentage of the song processed as transient becomes greater than say 50% (maybe 20% if we are being conservative?) you end up with essentially "Slice", vs. "arctan" (-M0 vs. -M1), the idea being that "arctan" is the preferred/default method in general. So I guess I'm looking for what the default -y value should be that also keeps the percentage from not being too high.

Hope that helps and glad you are hearing some improvement with this Beta.
 
@EMR; A couple things. I hope I didn't give the impression that -C0 was a preferred setting (4.0 vs. 5.0), I just meant that reducing the width of the center would move more of the sound away from center. I recommend adjusting it via live listening, but I get that that is a challenge in your situation. Listening to JUST the center (center channel solo'd) lets you adjust the center width so vocals there are "full" sounding, yet sounds from what would otherwise be in the fronts are not intruding. Then similar for the front width controls. If you were able to alternately solo the fronts and rears, you could make the front width adjustment in a similar fashion.

Center widths of maybe 15 or so wouldn't surprise me, but less than that would (unless you really wanted "quad"/no center channel).

I haven't tried it but maybe there is a way to get a mixed down to stereo output from your PC with SpecWeb Play (or maybe that's a feature I should look at adding). Nothing solo'd would be 5.1 mixed down to stereo, but you would still be able to solo or mute individual channels (that feed the mixdown).

I guess I don't know if you have a way (short of getting close to individual speakers) of listening to isolated channels in your 5.1 system.

Re loudness and waveforms, there are post separation steps (with the default settings anyway) that are going to affect loudness and therefore the way the waveforms look. There is a "remastering" (essentially a mastering limiter) applied, and then a normalization. These things can be turned off, however, for purposes such as seeing a change in the waveforms.

-v0 (lowercase vee, followed by a zero) would turn off the mastering limiter. -z0 (lowercase zee followed by a zero) turns off the normalization.

One more point on waveforms vs. sound. If you have access to or have seen results from Penteo upmixer, all the waveforms always look "the same". Looking at the waveforms you would think it was just the left and right channels repeated in LS and RS, and yet the sound is different (I can't fully explain that myself
o_O).

Anyway I always try to refer people to the "DKA" approach, although it is somewhat out of date as to today's defaults. It's in the guide, under How To Win With ArcTan (a mini guide from a contributor). I guess I will try to further updated it for the full 2.3 or 2.4 release, but I'll paste the existing langague in here (as I can't seem to get anyone to read the guide ;) ) Note that the actual article in the guide also has graphics and other additional info.

This is a guide we’ve shared in the past with some SBU members who were having a difficult time navigating through the different controls in ArcTan. Working this way has allowed me to work quicker, and has removed some of the guesswork with ArcTan. As with everything else, this is not a foolproof guide to ArcTan, but I do find that it gives me a running start with a majority of albums across genres.
If starting with your standard rock album, mixed without any hard left or right vocal panning, with not a lot of crazy vocal reverb, I have my ArcTan settings at the following (Note, these are not all any longer the default values for SpecWeb):
Image Width: 290
Center Width: 75
Front Width: 90
Mode: Across
Adjacent Speaker: .03 with wrap OFF
Here is where the tweaking starts. I live monitor my rears closely at these settings, but with the rear Slice blend set to 0.0 (which is “off”). If I can eliminate most of the vocal in the rears on a track without doing anything else, or at least enough that they're well-masked unless your ear is right at the speaker, I'm done. If not, I'll try slightly raising my center and front width, but not too much. Too much makes the soundfield too front-dominant. I'll also try "sum" mode rather than "across." If I'm still not there, this is where the ArcTan with Slice blended rears come in.
I then switch to ArcTan/Slice and have my Slice humidity’s at .9 and .95, and start slowly moving the rear blend control away from 0.0 until it sounds good to me. Some tracks take a lot. Some tracks only take a little.
There are some tracks where all this is hopeless because of a more reverb-y vocal, or a track where instrumentation is spare or the vocal is just mixed way up front. This is where I give up on isolating the vocal and see whether "diagonal" mode in ArcTan does a good job at giving me some nice instrumentation back there, but not worry so much about complete vocal isolation. In all probability, the listener is going to experience the vocal up front anyway. I then continue the same process with the Slice rear blend until I like what I hear.
That's pretty much how most of my conversions go. Absolutely use ZAG.

I note he's kind of working from the rears towards the fronts, whereas I have mentioned started with the center and working back.

Again those are no longer all the defaults (with peoples preferences and new methods included) but there is certainly no doubt that DKA's work stands out as amazing upmixing (we used to prefer the word "conversion" to upmixing because of all the negative connotations around poorly done upmixes).

If you want the DKA starting settings they would be:

-i290 for image width to 290 (but that is still the default)​
-c75 for center width​
-f90 for front width​
-D0 Back to Constant power panning vs. today's Zone placement​
-m1 for Across, -m2 for Diagonal vs. today's Zone2​
-M1 for Arctan, -M0 for Slice, -M2 for Arctan with Slice blended rears​
-a0.03 for adjacent speaker to 0.03 (but that is still the default)​
-l and -r are the Slice Blends​
-o -t and -w are the Slice Stage One Humidity, Stage Two Humidity, and Wrap Rears, respectively.​
Hope this helps. Good discussion anyway. Helps me understand what needs to be made more clear and features that might be helpful, etc.
 
So, I'm a bit confused. The percentage doesn't seem to be changing much. As an example, I upmixed Trip to the Fair by Renaissance. Using the default settings, I get a transient percent of 3.86%. Then I do a -y0.001 and I get a percentage of 3.83, and there is very little if any perceivable change to the sound. Does that sound right? Am I doing something wrong?
 
No, like I said, I should probably make that a dB scale vs. linear. But it also depends on the track. If there isn't a lot of "transient" sound then if won't find much more as you lower the -y value. Hmm. but re reading your post carefully I see the percentage actually decreased.

Are you getting swishy artifacts?

Let me try to find a source for that track and check it out.
 
No, like I said, I should probably make that a dB scale vs. linear. But it also depends on the track. If there isn't a lot of "transient" sound then if won't find much more as you lower the -y value. Hmm. but re reading your post carefully I see the percentage actually decreased.

Are you getting swishy artifacts?

Let me try to find a source for that track and check it out.
I'm not 100% sure what exactly a "swishy artifact" is, but I suppose that would be how I'd describe the difference between the upmix and the original; especially in the opening instrumental passage.
 
Well I can go look for short wav examples but some people also describe it as when a snare drum sounds "backwards". Instead of a fast attack and slow release a snare (or other "noisy transient" like maybe some high-hats sounds) get a slow attack and a slow release. So they sort of swoop or swish instead of bang.

That is the specific artifact that this transient processing is supposed to address.

There is another sort of phasiness sounding artifact that happens, when things get spread out across multiple channels. Those can be addressed by adjusting the width controls (as described upwards in this thread) or just my using Slice (-M0) instead of the default ArcTan (-M1) OR DKA's favorite -M2, ArcTan with Slice blended rears.

Note that Slice also doesn't have the first kind of artifacts, and that processing just "transients" with Slice but the rest of things with ArcTan, is what the whole -y thing is about.

Having said all that, I'll listen to Trip to the Fair again, maybe this evening Pacific time. FYI my source is 96 24 from an SACD. So far I have just listened with the default settings, other than -y, but it I was doing an upmix I would be adjusting the widths.
 
Well I can go look for short wav examples but some people also describe it as when a snare drum sounds "backwards". Instead of a fast attack and slow release a snare (or other "noisy transient" like maybe some high-hats sounds) get a slow attack and a slow release. So they sort of swoop or swish instead of bang.

That is the specific artifact that this transient processing is supposed to address.

There is another sort of phasiness sounding artifact that happens, when things get spread out across multiple channels. Those can be addressed by adjusting the width controls (as described upwards in this thread) or just my using Slice (-M0) instead of the default ArcTan (-M1) OR DKA's favorite -M2, ArcTan with Slice blended rears.

Note that Slice also doesn't have the first kind of artifacts, and that processing just "transients" with Slice but the rest of things with ArcTan, is what the whole -y thing is about.

Having said all that, I'll listen to Trip to the Fair again, maybe this evening Pacific time. FYI my source is 96 24 from an SACD. So far I have just listened with the default settings, other than -y, but it I was doing an upmix I would be adjusting the widths.
I think I get it. With TTTF I think it is more of a phasing thing. I'll give that a try later. Thanks for your patience.
 
For some odd reason, SpecWeb has stopped working on my Windows 11 virtual machine. Reinstalling it doesn't seem to fix the problem. How can I diagnose this?
 
New Windows Terminal

I've recently updated my Windows 11 virtual machine to Windows 11 23H2. I've noticed something odd. The default CMD window is now called something like Windows Terminal. It comes with certain improvements and several oddities. One of them significantly affects SpecWeb. Up to now, if I wanted to generate a 5.1 version of a stereo song, the conversion process would show the progress of the conversion on top of a static presentation of labels, channels, Peak and RMS values, et cetera, on a plain and reasonably static CMD screen. Now SpecWeb will cause the Windows Terminal to display a continuously moving or scrolling display for every snippet of conversion it carries out. So, if the Output time goes from 0:53 to 0:54, a new set of 10 lines will be generated with ever changing numbers on a further freshly generated set of 10 lines and so forth. Is there a known way to make that display more static, other than forcing Windows to always resort to its legacy CMD prompt?
 
Back
Top