DMGAudio TrackComp 2

[3ee]

Sorry, but I can't help but notice that my comment (about using it as intended, not switching models with audio running through it, still causing volume spikes) was just ignored.

Oh, sorry, should have mentioned. I've improved this for the 2A model, which was by far the worst offender.
Much much harder to do for the others. Any other major offenders?

Dave.

The 76D seems to have a lot lower volume than the other processors on the same threshold, which is noticeable with and without autogain.

Worst case scenario: without autogain, switching from "all-in" to DMG 4:1 or lower is a serious boost in volume. With autogain on, it is boosted even more for a moment, because autogain takes a second to catch up to the model switch.

What Q-Bias value do you have? I noticed that the volume differs significantly based on the Q-Bias value..

[omiroad]

It was set to the default for what I was describing.

[Krzysztof Oktalski]

Just to confirm, you're saying you've got volume discrepancies on the 76D when QBias is set to 0? I'll investigate this later.

[e@rs]

Is it me or the 76 mode uses more than 3x the CPU needed for the other models (in 1.01)?
I think it wasn't that hungry in the first version...

[Krzysztof Oktalski]

Is it me or the 76 mode uses more than 3x the CPU needed for the other models (in 1.01)?
I think it wasn't that hungry in the first version...

That's correct, it's 3x more complex than any of the other models. It wasn't so hungry in the first release because of a bug which got fixed in 1.01 (and simultaneously doubled the CPU usage).

[DaveGamble]

Hi all,

1.02 out now! https://dmgaudio.com/trackcomp

Changes:
- Fix GR metering in mono instances
- Fix "manual" sidechain listen mode
- Fix very obscure crash loading in Logic X
- Make slider behaviour smoother
- Remove spurious DC bias in 76D mode (resolves issues with excessive distortion)
- Report correct version number in Cubase VST2
- Fix ProTools crash when selecting Ext sidechain but none is routed
- Jumpstart resource loading to prevent weirdness in ProTools
- Alt-mousewheel on snap-to sliders will cycle through values.
- Extend E-Channel release time, and mark hardware ranges.
- Flatten low-end response for 76D (assume higher input impedance on interface)
- Jumpstart 2A GR to minimise pops when switching models
- Manual edits
- Improve 2A attack behaviour
- Support VST3 GR meter in Studio One

Enjoy!

Dave.

[cturner]

Delighted to see 1.02 arrive, but it was Carnap that whispered not to dally any longer with the acquisition. Best wishes for the holidays!

[omiroad]

Just to confirm, you're saying you've got volume discrepancies on the 76D when QBias is set to 0? I'll investigate this later.

(All of the following is tested on v1.02)

Yes, the 76D with Q-Bias 0.00% at 4:1 -14.0 dB threshold is quieter than DMG at 4:1 -14.0dB.

At 4:1 -14.0dB threshold, the same input results in:

[Model : Peak Max, RMS Max ]
DMG: -7.8, -7.7
E-Channel: -5.9, -7.7
G-Bus: -7.2, -8.4
76D: -13.0, -15.6
2A (Comp): -8.6, -10.7

Going from E-Channel to 76D on the same settings boosts the volume significantly. With autogain on this is quite startling.

---

Issue with ratio when switching between models:
1. Set 76D to Ratio 4:1
2. Set G-Bus to 4:1
3. Switch back to 76D
Result: 76D ratio is now set to 8:1.

4. Set 76D back to 4:1
5. Switch back to G-Bus
Result: G-Bus ratio is now set to 2:1.

---

The 2A attack fix in 1.02 helps a lot, which is definitely a step in the right direction. I believe this plug-in will get there, but it's not there yet.

[DaveGamble]

Hi,

Just to confirm, you're saying you've got volume discrepancies on the 76D when QBias is set to 0? I'll investigate this later.

(All of the following is tested on v1.02)

Yes, the 76D with Q-Bias 0.00% at 4:1 -14.0 dB threshold is quieter than DMG at 4:1 -14.0dB.

At 4:1 -14.0dB threshold, the same input results in:

[Model : Peak Max, RMS Max ]
DMG: -7.8, -7.7
E-Channel: -5.9, -7.7
G-Bus: -7.2, -8.4
76D: -13.0, -15.6
2A (Comp): -8.6, -10.7

A wide spread of numbers is to be expected here. If you could get them to match, what would be the point of having different models?

In the case of the 76D being SO far off, I wonder whether you are feeding a wide signal into a stereo instance. If you unlink the sides, do the numbers line up tighter (NB: they do for me)?
The 76D is (something like) peak linked, because that's how the hardware works. The other units are summed, which is generally what people are used to nowadays. That results in the 76D reporting almost twice the GR for wide signals.

The 2A attack fix in 1.02 helps a lot, which is definitely a step in the right direction.

I was unsure whether anyone would be able to detect the change to the 2A attack, which increases precision of the GR signal from 3 decimal places to 20; this corresponds to a maximal change of about 0.01dB in cases of heavy compression. I made the adjustment because I could, and felt like it was better workmanship, not because I believed it might be audible. If you're able to discern a difference, might it be possible for you to share some of your test audio with me? You may have found something with dynamic structure that renders the change more apparent, and that would be very interesting to have!

Dave.

[omiroad]

A wide spread of numbers is to be expected here. If you could get them to match, what would be the point of having different models?

I expected the peaks to be different more than the overall loudness to be different with the 76D. If it's how it's supposed to be, then if gain reduction instantly compensated for model changes then the issue is smaller.

In the case of the 76D being SO far off, I wonder whether you are feeding a wide signal into a stereo instance. If you unlink the sides, do the numbers line up tighter (NB: they do for me)?

Setting stereo linking to 0% does not change anything for me. (it actually makes it slightly quieter)

The 2A attack fix in 1.02 helps a lot, which is definitely a step in the right direction.

I was unsure whether anyone would be able to detect the change to the 2A attack, which increases precision of the GR signal from 3 decimal places to 20; this corresponds to a maximal change of about 0.01dB in cases of heavy compression. I made the adjustment because I could, and felt like it was better workmanship, not because I believed it might be audible. If you're able to discern a difference, might it be possible for you to share some of your test audio with me? You may have found something with dynamic structure that renders the change more apparent, and that would be very interesting to have!

Sorry, I thought the attack fix is what fixed the volume burst when switching to 2A. I can e-mail you the audio I tested with, but as far as I can tell it's nothing special.

[DaveGamble]

I expected the peaks to be different more than the overall loudness to be different with the 76D. If it's how it's supposed to be, then if gain reduction instantly compensated for model changes then the issue is smaller.

Would you be up for dropping me an email with some material that reproduces this?

Sorry, I thought the attack fix is what fixed the volume burst when switching to 2A. I can e-mail you the audio I tested with, but as far as I can tell it's nothing special.

Aha! I misunderstood, sorry! When I read your statement I assumed you were discussing accuracy of the hardware model.
The volume burst is being fixed by the "jumpstarting" process.

Am I right in thinking that your principal concern is discontinuity in loudness when switching between units?

Dave.

[pekbro]

Excellent 3 Band Track Eq. Build this you will... Depends on this, the fate of the universe does...

[omiroad]

Would you be up for dropping me an email with some material that reproduces this?

No problem, I sent it to you.

Am I right in thinking that your principal concern is discontinuity in loudness when switching between units?

That's exactly right, as it can cause the volume to raise unexpectedly when the listening level is adjusted to be comfortable with the current model. Auto-gain worsens this instead of helps because of the ~1 second delay after switching.

Besides the model switching, another issue is that without autogain on, adjusting the S/C HPF can cause the preview to be unexpectedly louder because it's at the original volume. Especially with a decently high threshold setting this is a big jump. It's a shame to turn the preview feature off because of this though.

[bundoo]

how does this do on the distressor front?

[3ee]

While playing with v1.02 yesterday I encountered a strange bug with the DMG model.

I was experimenting with large amount of RMS size 200-300, heavy gain reduction like more than 20dB and auto make-up gain and on tweaking some parmeters (the knee IIRC) had very loud sound burst which triggered Reaper's auto-mute option and (not sure how loud the peak was but I have set Reaper to auto-mute tracks that go above +12 dBfs)

The make-up slider disappeared from the GUI and came back once I reopened the window

added: ... can't seem to reproduce the bug, will keep trying and let you know if something.