SPECWEB (Now 2.2)

[zeerround]

OK thanks for that.

IMO what you are hearing is a difference in loudness, not a difference in bass. With the default settings the surround comes out 2.43dB quieter than the stereo. However the center channel is reaching full peak value so we can't just turn up the gain. What is needed is more compression, via lowering the limiter threshold, but this version of SpecWeb doesn't have that feature as all the limiter settings are automatic.

There is one parameter that can get us much closer, and that is turn off true peak detection in the automatic limiter settings. You'll have determine if the results sound distorted in any way (the max peak value will still be 0dB but the intersample value will come out higher than that of the original stereo in this case [both are > 0dB]).

If you want to try that (can't be done by ini file, my bad) make a copy of your SpecWeb shortcut and right click and select properties and add -k0 as a parameter at the end of the command string.

OR

if the command string has -A you can just inter -k0 when SpecWeb says "Enter Options or hit return to accept all defaults".

In my experiments on your file that gave a surround output that less than half a dB quieter than that Stereo (probably not enough to notice). In AB tests the bass sounds the same to me, at that point, but here's another test I did:

The left hand spectrum is the Stereo (at a random point in the song) and the right hand spectrum is the Surround (mixed back down to stereo) at the exact same moment in the song.

That's as close as you're going to get without doing separate manual "remastering" on the output of SpecWeb.

SpecWeb is designed to get as close as possible to the original loudness, without manual settings, but it is not perfect. The formulas I developed to set the limiter parameters were based on my selection of 31 "regression" tracks from all types and styles of music, including any tracks that gave anyone trouble.

I'll add this one to the list going forward.

The "full" version of Spec (also free, but runs inside Plogue Bidule) http://www.surroundbyus.com/sbu/viewtopic.php?f=8&t=266 let's you set all the remastering parameters you could want, but it is not as push button easy as SpecWeb. Your choice.

Cheers,
Z

 

[zeerround]

Or, if you just want more bass, change the specweb settings to increase the LFE channel (less technically correct than the above, but hey, it's your music so...)

 

[RobJam]

Thank you for your detailed reply zeerround!

 

[zeerround]

Sure. It's good for me to revisit this stuff or it fades from my mind.

Another tip, if you add the -V1 option you'll get more feedback on the dynamics of the stereo vs. surround:

Here we see the (extra) blue line that tells us the original stereo had a true (intersample) peak of 0.02 dB and its loudness was -9.7 (pretty loud, producer mashed it pretty hard) and the Surround had a true peak of 0.09 dB and was less than 1/2 dB quieter.

-V1 is not the default because it takes longer to process taking those measurements.

 

[maddhatter]

Hi, (new member) just joined the QQ forum (although I have been trolling for years), I decided to make an inquiry about SpecWeb as an update maybe in the works for the near future. I really love your app, and have been using it for years yielding great results, so I must 1st thank you very much and also congratulate on this fine GREAT upmix utility. I was wondering if there was a 2hr max limit on source audio files, since I received my 1st and only error a while ago (when upmixing the 1974 CSNY live concert (196:11) min.), I received a “max sample count exceeded” msg near the end of the upmix process (around 122:22 minutes). Unfortunately I didn’t verify it assuming there was an upper limit so I did a simple work-around by splitting my source file into 2 parts then edit rejoining the 5.1 upmixed parts to a single (.w64) result, so it really wasn’t a big issue, just extra work. If however there really is an upper limit to audio source length, is it an arbitrary one? if so, I was wondering if possible you could please extend or remove this limit? Thank you. I'am really looking forward to the next version, can't wait. (Note: I always use the command-line approach over the drag & drop method as I find it faster, bypassing the 2-pass since I check my audio source file beforehand for any brick-walling)

BTW I posted a reply on Reddit a few years ago to an old inquiry regarding upmixing, in the ‘audiophile’ section, mentioning your SpecWeb utility. It didn’t go over very well, received with negativity, even so far as regarding it as spam advertising, even after mentioning it was free (sorry should have said donationware). It occurred to me later after looking at some other postings in this ‘forum’, that they were only interested in bragging about how expensive the equipment they had recently bought , or how many points they would rack up. Anyway I left that group and Reddit immediately afterwards, seems mostly a bunch of narrow minded snobs there anyway, myself I have difficulty listing to stereo anymore, whether it is in high resolution or not.

 

[zeerround]

Yeah "upmixing" is often poopoo'd by groups without actually checking it out. Lots of attitude out there. Checking on the max sample count now...

 

[zeerround]

OK "max sample count exceeded " is from deep in the FFT code I didn't write. It looks like it happens when an unexpected "End Of File" happens, so I'm not sure. I've never seen that error in practice.

I'll do some testing...

 

[zeerround]

What was the file format and sample rate and bit depth of the single 196:11 file? You can test by playing your input file in foobar2000 or whatever, and see if it plays all the way to the end. If it cuts off early, you are past the 4GB limit in a format that doesn't support it.

It it was over 4GB in size, and not in a format that supports audio files larger than 4GB I can see how that error would occur.

Formats that SpecWeb can read, which support files larger than 4GB are:

W64 and RF64

I don't see a crisp answer for flac.

 

[maddhatter]

Hi zeerround, that was a quick response! Well, it was several months ago when it occurred, the audio source was 24-bit, 192kHz, in either a flac or W64 format. (I've never used RF64), hope this helps. Don't trouble yourself about this error too much as there is a simple way around it, just wanted to let you know in case you hadn't already known.

 

[zeerround]

You can test your stereo input file by playing it in foobar2000 or whatever, and see if it plays all the way to the end. If it cuts off early, you are past the 4GB limit in a format that doesn't support it.

 

[maddhatter]

Gottcha, I should have checked if my file was corrupt. (I finally found my backup file in flac format).

ok, so I just fully decoded it successfully with EAC3to and here is the log results. Whether EAC3to decodes flac thru errors I cannot say.

FLAC, 2.0 channels, 3:16:12, 24 bits, 5110kbps, 192kHz
Decoding FLAC...
Writing W64...
Creating file "C:\Crosby, Stills, Nash And Young - CSNY (1974) (24-bit, 192 kHz)\Crosby, Stills, Nash & Young - CSNY.w64"...
The original audio track has a constant bit depth of 24 bits.
eac3to processing took 4 minutes, 55 seconds.
Done.

I also was able to successfully load the 7G flac fully into Soundforge without errors and play it.

Maybe its something with your flac decode (BASSFLAC).

I will try to re-verify this error using flac 1st, and then with .w64. If flac fails and w64 passes then it;s probably BASSFLAC. If it does't fail on either then maybe my backup is not the same as the one I used when got the SpecWeb error, or it was something else in either case it's not a valid error. I need to 1st make some space available on my HD to do this test to verify the SpecWeb error, so it may take awhile before I get back to ya.

 

[zeerround]

I did get a crash trying to duplicate. Since I didn't start in a cmd I didn't see the error and SpecWeb just went away so no I am on run number 2. 6GB .w64 input file at 24-96. I used AudioMuxer to join the 40 flac files into one .w64 file. 196:31 here ;0)

I'll keep you posted.

By the way you need to play the end of the input track to be sure it plays all the way to the end. Many tools will happily write a >4GB file, without reporting errors, that won't playback past 4GB, however, since I did get a crash I'm leaning away from that being the problem here.

Of course, one could run the 40 input files through SpecWeb and join the outputs in AudioMuxer, etc. If you want a single file, and need to avoid this error.

 

[zeerround]

Oh by the way BassFlac is only used for "Play". libsoundfile is the audio file io for non "play" operations but yeah the bug could be in that.

I was able to duplicate your crash with a .w64 input file at 122:22. Interesting that it happens at that time despite us using different sample rates.

 

[maddhatter]

Alright, I've tested with both flac and .w64 and they both failed with the same error at 122:22 and I did screen shots of both which I've attached.

I'm sure the source audio is not corrupt. Anyway not sure if you could call these errors or just a limitation. I know errors since I authored buggy software myself, a little old app call 'tranzcode', maybe you heard of it, abandoned it once a better app became available (Eac3to) which also has much more capabilities.

I might get in trouble hear since this is not relevant to this thread, but yes, I'm an oldtimer who had previously corresponded with a few guys on another forum (SNDINMYHD, and EOH whom both have passed on now, let them rest in peace), and also with a prominent member here on QQ (Bob Ramano). Bob, I have an Acura TL (2010) with ELS for several yrs now, also I've installed a separate 5.1 system that shares the same speakers via relay board, based on part of a Sony Blu-ray player, an amplifier, small TFT screen display, an HDMI Audio De-Embedder, Relay board (with 16 relays on it), and a couple of HDMI cables, and small Remote. I can play 6 channel 24-bit, 192 kHz flac via USB in my car using this system, but it's currently disassembled since I need to upgrade the Blue-ray part so that it has enough current to power a small 3TB Seagate HD so I could store and playback a library of 5.1 music in my car during the warmer months of the year.

 

[zeerround]

Thanks for reporting this, it got added to the todo list.

I doubled the range of a few key variables (“shotgun” approach) but no joy. Will try to seriously debug.

 

[maddhatter]

Hi, I realized afterward that I could run another test to rule out anything with the audio content of the file, so I cut the last 75 mins (1:15:00) of the file and saved it to a new test file (.w64 format) and then ran it through SpecWeb. It just finished running fine without producing any errors, so it's definitely an audio duration (length) issue. Too bad you have to wait awhile till it runs just over 2 hours to do each test for troubleshooting purposes. It took me quite a while to run these tests because of high resolution. Good Luck!

 

[zeerround]

So I found and fixed that, however then I hit another stop at 186:24.

Variables that are of type "long" are running over their range.

https://docs.microsoft.com/en-us/cpp/cpp/data-type-ranges?view=vs-2019
Besides just increasing everything to type long long, I'll need to see what the impact on performance is.

 

[zeerround]

Yea!

I ran this with LFE turned off to save time, but the error was on individual channels so that doesn't affect the test.

input file was 6.6GB 96-24.

output file was 20GB 96-24.

SpecWeb 2.0 (now in alpha testing but I'm not done adding features) has built in up/down sample so this actually processed at 192K sample rate, and that's why I got the errors at the same time stamp you did, even though your input file was 192k.

Anyway the new time limit is somewhat arbitrary but it is 10,000,000,000 samples in a channel so ~14.5 hours @192 KHz ~63 hours @44.1KHz.

Again, thanks for reporting this.

 

[maddhatter]

That was a quick fix, and variable type overflow makes perfect sense here. Thanks allot, I can wait till the next release, and I guess I could live with 14.5 hrs @192k & 63 hs @ 44.1k

 

[Ottoguy]

Is there a way to disable any added reverb/delay. I've noticed some is added.

The thing is I'm using the program for something different; I'm running (mainly 60s tracks) through this in order to be able to downmix them to stereo again. I do this because the old stereo mixes are awful: drums all in the left channel, bass all in the right etc...

What settings do you think would be best for this? I guess with as little reverb, delay and compression as possible, as well as minimized bleed from channel to channel. Thanks in advance, great program!