TrackSpacer - First plugin by Wavesfactory
-
- KVRist
- 138 posts since 26 Feb, 2016
There is still something not quite right with the new version. There is a small bump around 30hz, which will flatten out if low cut is set to around 24db. The noise floor is significantly louder, more so lower down the frequency spectrum.
See images.
See images.
You do not have the required permissions to view the files attached to this post.
- KVRian
- 679 posts since 29 Nov, 2002 from Finland
I found that dividing the default v2 attack & release values by 10 (5 & 30 ms -> 0.5 & 3 ms) produces _almost_ the same results in 2.5.2. They still don't null, but are at least close. But when you get away from those values, it's not that simple any more. This is not necessarily a problem, it's like two different plugins now, as you said. But I understood that the idea was to match the attack & release values between the versions, so that identical values produce identical results, and that's definitely not happening now.c_voltage wrote:Ehh, no, after some tests i can suppose that clarity need be around 0.30ms-0.70ms too, since i found this quick range can be useful as well. (again, i compare with 2.05)
In general, guys, i don't know. Maybe I do not understand something.
-
- KVRAF
- 8511 posts since 5 Aug, 2009
hmmmm, a plugin like this NEEDS to be neutral, just ducking/sidechaining and not alter the sound!
DAW FL Studio Audio Interface Focusrite Scarlett 1st Gen 2i2 CPU Intel i7-7700K 4.20 GHz, RAM 32 GB Dual-Channel DDR4 @2400MHz Corsair Vengeance. MB Asus Prime Z270-K, GPU Gainward 1070 GTX GS 8GB NT Be Quiet DP 550W OS Win10 64Bit
- KVRAF
- 2261 posts since 16 May, 2004 from Soviet Union
Hm, ineresting finding.Captain wrote:dividing the default v2 attack & release values by 10
Well, in principle i agree with you, had the same thoughts after some time.Caine123 wrote:hmmmm, a plugin like this NEEDS to be neutral, just ducking/sidechaining and not alter the sound!
-
- KVRAF
- 8511 posts since 5 Aug, 2009
i have to admit that i didnt use the old version much cause i got the plugin in around december last year and didnt mix much, but this is crucial stuff. i dont use a saturator here but a sidechainer and i dont understand how such things can be brought to plubic in this state? companies should test stuff like this, i dont talk about bugs.c_voltage wrote:Hm, ineresting finding.Captain wrote:dividing the default v2 attack & release values by 10
Well, in principle i agree with you, had the same thoughts after some time.Caine123 wrote:hmmmm, a plugin like this NEEDS to be neutral, just ducking/sidechaining and not alter the sound!
DAW FL Studio Audio Interface Focusrite Scarlett 1st Gen 2i2 CPU Intel i7-7700K 4.20 GHz, RAM 32 GB Dual-Channel DDR4 @2400MHz Corsair Vengeance. MB Asus Prime Z270-K, GPU Gainward 1070 GTX GS 8GB NT Be Quiet DP 550W OS Win10 64Bit
- KVRAF
- 2261 posts since 16 May, 2004 from Soviet Union
Yes, this is a strange circumstance, not often happens.Caine123 wrote:i have to admit that i didnt use the old version much cause i got the plugin in around december last year and didnt mix much, but this is crucial stuff. i dont use a saturator here but a sidechainer and i dont understand how such things can be brought to plubic in this state? companies should test stuff like this, i dont talk about bugs.c_voltage wrote:Hm, ineresting finding.Captain wrote:dividing the default v2 attack & release values by 10
Well, in principle i agree with you, had the same thoughts after some time.Caine123 wrote:hmmmm, a plugin like this NEEDS to be neutral, just ducking/sidechaining and not alter the sound!
But again, when we talk about what does not apply to bugs, we find ourselves in an uneasy position, trying to influence the author's point of view, about how it should work, though he already spent his time to this, and so on. In other words, I'm already inconvenient to ask for something to change again if the author himself does not see an important reason for this.
In general, at least it's good that old version can be opened in host together with new. Many vendors do not support this possibility, during minor numeration.
-
- KVRAF
- 8511 posts since 5 Aug, 2009
sure but why does the author change the neutrality of the old to the 2.5 version?
DAW FL Studio Audio Interface Focusrite Scarlett 1st Gen 2i2 CPU Intel i7-7700K 4.20 GHz, RAM 32 GB Dual-Channel DDR4 @2400MHz Corsair Vengeance. MB Asus Prime Z270-K, GPU Gainward 1070 GTX GS 8GB NT Be Quiet DP 550W OS Win10 64Bit
-
- KVRian
- 1153 posts since 11 Aug, 2004 from Breuillet, France
Well after some testing, I figured out that the software that you are all using to understand what is going on is buggy
I did a test with Fabfilter Pro-Q, that I trust a lot to say the least. I put a bell filter around 30 Hz with 2.5 dB gain in the plug-in, opened in DDMF PluginDoctor. Then, I put the FFT size of the display at the maximum value, and I play with sampling rates either at 44.1K, 48K, or 96K. And I discovered that the frequency response being displayed is sometimes very inaccurate when I move the frequency from 30 to 10 Hz and so on. Sometimes it's right, and then suddenly it "jumps" and you see something which looks like a shelf filter with a lot of gain at DC, and then again it's right. It happens specifically with low cutoff frequencies and high sampling rates. That's a problem well known with frequency response display algorithms and high sampling rates, and of course we saw it as well in the frequency response display from Wavesfactory Spectre during development, but we solved that issue very quickly. It's something linked with filter stability in short.
I saw something like that when I was testing Trackspacer 2.5. What I can say now is that I can't trust PluginDoctor, and that I'm going to see if I see the same thing you all mentioned with right ways to measure the frequency response. At least, nothing seems to be wrong at 44.1K with Trackspacer 2.5 in PluginDoctor, which suggests that everything is supposed to be fine at higher sample rates as well. The fact that just moving a little the lowcut frequency make that bell diseappear as well even with high sample rates strongly suggest the same thing as well (there is no magic involved in the processing, it's a dynamic EQ with a very straight way to set the frequencies !)
For the noise floor that's a different thing, that's because we changed the audio engine in Trackspacer 2.5. That change was necessary to make it easier to maintain, to make it faster, and I had to figure out sometimes a few things all by myself in the original behaviour to make that happen. Or it might just be because original Trackspacer bypass totally the processing when no signal in the sidechain is detected whereas the new one is doing something. If that's the case, then even the original Trackspacer would have that noise floor when doing something on the signal (but you would need a PluginDoctor using the sidechain input to see this)
I did a test with Fabfilter Pro-Q, that I trust a lot to say the least. I put a bell filter around 30 Hz with 2.5 dB gain in the plug-in, opened in DDMF PluginDoctor. Then, I put the FFT size of the display at the maximum value, and I play with sampling rates either at 44.1K, 48K, or 96K. And I discovered that the frequency response being displayed is sometimes very inaccurate when I move the frequency from 30 to 10 Hz and so on. Sometimes it's right, and then suddenly it "jumps" and you see something which looks like a shelf filter with a lot of gain at DC, and then again it's right. It happens specifically with low cutoff frequencies and high sampling rates. That's a problem well known with frequency response display algorithms and high sampling rates, and of course we saw it as well in the frequency response display from Wavesfactory Spectre during development, but we solved that issue very quickly. It's something linked with filter stability in short.
I saw something like that when I was testing Trackspacer 2.5. What I can say now is that I can't trust PluginDoctor, and that I'm going to see if I see the same thing you all mentioned with right ways to measure the frequency response. At least, nothing seems to be wrong at 44.1K with Trackspacer 2.5 in PluginDoctor, which suggests that everything is supposed to be fine at higher sample rates as well. The fact that just moving a little the lowcut frequency make that bell diseappear as well even with high sample rates strongly suggest the same thing as well (there is no magic involved in the processing, it's a dynamic EQ with a very straight way to set the frequencies !)
For the noise floor that's a different thing, that's because we changed the audio engine in Trackspacer 2.5. That change was necessary to make it easier to maintain, to make it faster, and I had to figure out sometimes a few things all by myself in the original behaviour to make that happen. Or it might just be because original Trackspacer bypass totally the processing when no signal in the sidechain is detected whereas the new one is doing something. If that's the case, then even the original Trackspacer would have that noise floor when doing something on the signal (but you would need a PluginDoctor using the sidechain input to see this)
- KVRAF
- 2261 posts since 16 May, 2004 from Soviet Union
Ivan, i forgot to ask, another thing what disturbed me thru compare - i attach screen from video in which we see that frequency curve are very different during the same freq impact (the kick drum). That pushed me to an think - Is there any change in the internal EQ algo in the new version? Like reducing the number of bands, or averaging\decrease the sensitivity of the source capturing.
Or is it just a changes in visualization engine?
Or is it just a changes in visualization engine?
You do not have the required permissions to view the files attached to this post.
Last edited by c_voltage on Wed May 30, 2018 6:52 pm, edited 1 time in total.
-
- KVRist
- 138 posts since 26 Feb, 2016
Hi Ivan,Ivan_C wrote:Well after some testing, I figured out that the software that you are all using to understand what is going on is buggy
I did a test with Fabfilter Pro-Q, that I trust a lot to say the least. I put a bell filter around 30 Hz with 2.5 dB gain in the plug-in, opened in DDMF PluginDoctor. Then, I put the FFT size of the display at the maximum value, and I play with sampling rates either at 44.1K, 48K, or 96K. And I discovered that the frequency response being displayed is sometimes very inaccurate when I move the frequency from 30 to 10 Hz and so on. Sometimes it's right, and then suddenly it "jumps" and you see something which looks like a shelf filter with a lot of gain at DC, and then again it's right. It happens specifically with low cutoff frequencies and high sampling rates. That's a problem well known with frequency response display algorithms and high sampling rates, and of course we saw it as well in the frequency response display from Wavesfactory Spectre during development, but we solved that issue very quickly. It's something linked with filter stability in short.
I saw something like that when I was testing Trackspacer 2.5. What I can say now is that I can't trust PluginDoctor, and that I'm going to see if I see the same thing you all mentioned with right ways to measure the frequency response. At least, nothing seems to be wrong at 44.1K with Trackspacer 2.5 in PluginDoctor, which suggests that everything is supposed to be fine at higher sample rates as well. The fact that just moving a little the lowcut frequency make that bell diseappear as well even with high sample rates strongly suggest the same thing as well (there is no magic involved in the processing, it's a dynamic EQ with a very straight way to set the frequencies !)
For the noise floor that's a different thing, that's because we changed the audio engine in Trackspacer 2.5. That change was necessary to make it easier to maintain, to make it faster, and I had to figure out sometimes a few things all by myself in the original behaviour to make that happen. Or it might just be because original Trackspacer bypass totally the processing when no signal in the sidechain is detected whereas the new one is doing something. If that's the case, then even the original Trackspacer would have that noise floor when doing something on the signal (but you would need a PluginDoctor using the sidechain input to see this)
Thanks for your response. I have also seen the behaviour that you mention in Plugin Doctor, but it seems to be a bug, which corrects itself after a restart. I also measured several other plugins during the same session to see if it was PD or TrackSpacer and saw no anomalies in other plugins. I am running at 48kHz.
I'm not trying to be a pain. I love TrackSpacer, it is one of my top plugins. Although it would be great to have less CPU hit, I want the new version to work as well as the old one; to be something that I can rely on.
Paul
-
- KVRian
- 1153 posts since 11 Aug, 2004 from Breuillet, France
The internal EQ algorithm does the exact same thing than before, and the number of internal bands is exactly the same as well. However, the spectrum analyzer algorithm is a new one, so the difference you see there is probably related with the refresh speed and dynamics.Ivan, i forgot to ask, another thing what disturbed me thru compare - i attach screen from video in which we see that frequency curve are very different during the same freq impact (the kick drum). That pushed me to an think - Is there any change in the internal EQ algo in the new version? Like reducing the number of bands, or averaging\decrease the sensitivity of the source capturing.
Or is it just a changes in visualization engine?
Hello !Hi Ivan,
Thanks for your response. I have also seen the behaviour that you mention in Plugin Doctor, but it seems to be a bug, which corrects itself after a restart. I also measured several other plugins during the same session to see if it was PD or TrackSpacer and saw no anomalies in other plugins. I am running at 48kHz.
I'm not trying to be a pain. I love TrackSpacer, it is one of my top plugins. Although it would be great to have less CPU hit, I want the new version to work as well as the old one; to be something that I can rely on.
Paul
It's not a bug in the development sense of the word, which can appear and then diseappear totally after a restart (until the next time). It's basically a DSP mistake, so under some given conditions you'll see it happening 100% of the time. To see it, you need to use an EQ at 48 kHz or more for the sample rate, and play with the controls under 30 Hz. The frequency response will display often some random "anomalies" which will diseappear when you move a little your controls.
Again I'll investigate properly if there is nothing wrong in the low frequencies with Trackspacer (with my own measurement methods obviously) but I'm quite confident there since moving a little the low cut control is already removing the issue in PluginDoctor, and that there is no way such a little move could remove something like a bump on its own.
-
- KVRist
- 138 posts since 26 Feb, 2016
Hi Ivan,Ivan_C wrote: Hello !
It's not a bug in the development sense of the word, which can appear and then diseappear totally after a restart (until the next time). It's basically a DSP mistake, so under some given conditions you'll see it happening 100% of the time. To see it, you need to use an EQ at 48 kHz or more for the sample rate, and play with the controls under 30 Hz. The frequency response will display often some random "anomalies" which will diseappear when you move a little your controls.
Again I'll investigate properly if there is nothing wrong in the low frequencies with Trackspacer (with my own measurement methods obviously) but I'm quite confident there since moving a little the low cut control is already removing the issue in PluginDoctor, and that there is no way such a little move could remove something like a bump on its own.
I have tested several EQs with PluginDoctor @48kHz sample rate & tried various control manipulations under 30Hz, but I am only seeing the bump with TrackSpacer.
Thanks
Paul
- KVRAF
- 2261 posts since 16 May, 2004 from Soviet Union
Ok, thanks.Ivan_C wrote: The internal EQ algorithm does the exact same thing than before, and the number of internal bands is exactly the same as well. However, the spectrum analyzer algorithm is a new one, so the difference you see there is probably related with the refresh speed and dynamics.