KVR Audio

Satch1 · Post by **Satch1** » Fri Oct 24, 2014 2:08 pm

Edit: Original post removed. I feel I was a bit unfair and disrespectful. Sorry Steve!

Ingonator · Post by **Ingonator** » Sat Oct 25, 2014 10:19 am

jc_vt wrote:There is a long list of videos about Serum by Dorian Day on YouTube, and one I just looked at does some great comparisons among diff synths for purity of signal

http://youtu.be/MtvpAxTe75E

I have little problem with CPU load myself, but that for some may be the price of inherent quality noted here

I have tried to make some tets like in the video myself using the Spectrum analyzer of Live 9 like in the video.
I was a bit confused by the noisy baseline which i found in almost any synth i checked including e.g. Diva at highest quality settings which is strange.

Then i used the free Voxengo SPAN Signal Analyzer and with e.g. Diva the noise in the baseline of the Signal was gone (actualy it loks very clean now...). Looks like this is more a problem of the signal analyzer plugin than with the synths.

I could get a slight amount of noise in SPAN when i set the refresh time for the signal dispaly very short (but by far not as much noise as in Live 9).
In that case you could also a "jumping" signal like in some cases at the video. A slightly higher refresh time seems to be better to not getting unwanted noise or a "jumping" signal. The time used in te video in SPAn is close to the shortest one and the default is much higher. in Live 9 even with the highest refresh time the signal (or at least the baseline) is still very noisy.

Here is a direct comparison of the same Diva patch (basic Saw with Minimoog modules at Divine quality) with the Live 9 spectrum analyzer and the free Voxengo SPAN using the maximum refresh time at the Live 9 analyzer (200 ms) in both cases and using higher octaves (C6):
https://dl.dropboxusercontent.com/u/532 ... PAN_01.png

In Voxengo SPAN with a higher refresh time the noise even disappears.
With higher refresh time value you will only see "real" peaks added by e.g. aliasing but not not noise based on the settings in the Spectrum analyzer.

I also found that checking for aliasing makes much more sense at higher octaves (e.g. C5 or C6 instaed of C3) as even with synths that are know to produce aliasing (like e.g. PPG Wave 3.V where this actually is a "feature"...) it could happen that this is not noticed in midlle octaves.

This is how aliasing actually looks like using PPG Wave 3.V in "2.2" TruePPG mode (refresh time at 200 ms like in the example above):
https://dl.dropboxusercontent.com/u/532 ... rum_01.png

in lower octaves this is much less obvious.

UPDATE/CORRECTION:

It looks like i finally found the parameter which makes the spectrum analyzer in Live 9 and Voxengo SPAN look similar. It is the "block size" parameter. Switching this to the same value as in the Live 9 analyzer at the video gives the same result for me in SPAN now. The default value in SPAN is 2048 while in the video 8192 was used.
Switching this to a lower value in Live 9 also reduces the noise comparable to the comparison i did above (where i used the default block size of 2048 in SPAN).

While this could a failure from my side the question still is what is a proper value.

Usually a setting where almost ANY synth including high quality synths like Diva look "noisy" at the baseline could not be the right one...

So even if the anayzer settings in the video were right (or not?) there are still two points there that should be corrected: Using real synths instead of samples and doing the aliasing test at high (and maybe also low octaves) instead of uisng a middle octave.

As mentioned in my test above also the refresh time of the signal still has an influence on the result while it is less than the block size value.

When the block size value used in the video is a usual value the "noisy baseline problem" should be included in the interpration of the results as almost any synth will look "noisy" then at the baseline of the analyzer.
IMO also a higher value (maybe something between like 4096) should still be useful to detect additional peaks from actual aliasing.
The guy who made the video did not seem to be aware of this as he interpreted the result for e.g. Diva by having some strange noise there. With higher block size values also a "jumping signal" seems to happen more often whcuh the guy did not seem to know too (in SPAN this is mentioned at a tool tip when you adjust the block size value).

Satch1 · Post by **Satch1** » Sat Oct 25, 2014 10:43 am

Ingonator wrote:
I also found that checking for aliasing makes much more sense at higher octaves (e.g. C5 or C6 instaed of C3) as even with synths that are know to produce aliasing (like e.g. PPG Wave 3.V where this actually is a "feature"...) it could happen that this is not noticed in midlle octaves.

Yeah I used to think aliasing only really happened in the higher octaves, but I've heard some synths that sound like they're aliasing in the lower octaves. I don't know whether this is technically aliasing or not, but it sure sounds like it.

tooneba · Post by **tooneba** » Sat Oct 25, 2014 12:17 pm

he wasn't directly using synth when monitoring analyzer? I can see he is using warp enabled sample in that video. Have someone pointed out it? I think it should cause aliasing from daw not only from synth.

Ingonator · Post by **Ingonator** » Sat Oct 25, 2014 12:49 pm

edited

Sampleconstruct · Post by **Sampleconstruct** » Sat Oct 25, 2014 4:10 pm

One last demo combining three fresh Serums:

http://soundcloud.com/sampleconstruct/l ... demo-sonic

PavoniMaster · Post by **PavoniMaster** » Sat Oct 25, 2014 7:33 pm

CURVE settings in the MOD matrix are brilliant and being able to see it's effect on the wavetable is even better!

Sampleconstruct · Post by **Sampleconstruct** » Sun Oct 26, 2014 4:56 am

Epilogue:

http://soundcloud.com/sampleconstruct/s ... demo-sonic

PietW. · Post by **PietW.** » Sun Oct 26, 2014 10:48 am

Ingonator wrote:
jc_vt wrote:There is a long list of videos about Serum by Dorian Day on YouTube, and one I just looked at does some great comparisons among diff synths for purity of signal

http://youtu.be/MtvpAxTe75E

I have little problem with CPU load myself, but that for some may be the price of inherent quality noted here
I have tried to make some tets like in the video myself using the Spectrum analyzer of Live 9 like in the video.

Where is the purpose in these tests. I want to hear music, not see.

Ingonator · Post by **Ingonator** » Sun Oct 26, 2014 11:32 am

I found something new and updated my post with the screenshots above wit ha comment.

bob.baffy · Post by **bob.baffy** » Sun Oct 26, 2014 2:52 pm

+1 for a "draft" mode -- that would be so useful for me while composing and tweaking/stacking sounds.

PHY6 · Post by **PHY6** » Sun Oct 26, 2014 3:21 pm

Anyone tried to load galbanum waveforms into serum ?
I own the Alchemy waveform pack...when I load them it doesn't do the waveforms justice as they should...wrong format I guess !?!

replicator · Post by **replicator** » Sun Oct 26, 2014 4:13 pm

Hi,

I am currently demoing Serum and like the sound quality and the ease of use. I discovered pretty high cpu usage when using higher release times in unison mode and when modulating wave table positions (1x oversampling). I am using a quad core. I think I can live with that, but one thing I feel Serum is lacking is more presets that can show more of the potential of the synth. I will definately put it on my buy list and are looking forward to test it further in demo mode.

Best regards,
Replicator

rustman · Post by **rustman** » Sun Oct 26, 2014 6:39 pm

I was using Cthulhu to drive 8 instances of Serum yesterday. VST performance realtime was varying between around 50% and 70% roughly. No heavily stacked/high polyphany stuff, but still, I was happy to see it run.

hwangman · Post by **hwangman** » Mon Oct 27, 2014 3:52 pm

I tried the demo on my iMac over the weekend (dual-core, 4GB of RAM), and my CPU usage was frequently around 40%. This was with nothing but a single instance of Serum in a blank/new project in Ableton. I enjoyed the sounds but I'll have to pass for now. I'd want to layer the sounds but it seems pretty clear that my rig can't handle it. Maybe someday.

Official Serum thread!