The No.1 Website for Pro Audio
denise introduces Dragon Fire: A new way to color and tame your track dynamics
Old 2 weeks ago | Show parent
  #61
Lives for gear
 
bgrotto's Avatar
Quote:
Originally Posted by cuzzinlouie ➡️
Should I not expect a flat frequency response as long as all bands are at 0 though?
Depends on the input signal and the detector. Looks like your input signal is hotter in the hf, hence, more hf compression. But if that visual is just due to the weighting of your analyzer, one might suspect that the detector is simply more sensitive to higher frequencies.

Were you feeding it pink noise? White noise?
Old 2 weeks ago | Show parent
  #62
Gear Nut
 
🎧 5 years
white
Old 2 weeks ago | Show parent
  #63
Lives for gear
 
bgrotto's Avatar
Quote:
Originally Posted by cuzzinlouie ➡️
white
Pink noise = equal power in all freq and thus a better test. Try again w pink noise
Old 2 weeks ago | Show parent
  #64
Gear Nut
 
🎧 5 years
Sorry, I beg to differ:

https://en.wikipedia.org/wiki/Pink_noise

Pink noise or ​1⁄f noise is a signal or process with a frequency spectrum such that the power spectral density (power per frequency interval) is inversely proportional to the frequency of the signal
Old 2 weeks ago | Show parent
  #65
Lives for gear
 
bgrotto's Avatar
Quote:
Originally Posted by cuzzinlouie ➡️
Here's what I am getting at 48KHz, but I am getting similar results at 44.1KHz.

-In the PD images I am actually getting even higher aliasing levels, but these can be argued over when looking at a static image, so I left them out on purpose.
-In the RX tests (20 - 23000Hz sine sweep) I am effectively measuring the aliasing levels only, circled in red at the bottom of the image. RX lets me listen to just the lasso-selected aliased partials.

Please let me know if you still think that I am misreading these tests. If proven wrong, I am willing to make any adjustments necessary.
Hello again! Thanks for turning me onto the analysis functionality of RX8. I hadn't used that before, and it's very cool.

Anywho, I stumbled upon a possible (probable, even) flaw in your test, which is that when using that lasso tool to select a specific range, and especially if you're looping that range (which seems to be the case, looking at your screen shots), you must select a zero crossing point for the loop start and end locations, otherwise, you get a small 'click' at the end of playback, which will show a considerably higher signal level.

To illustrate that point, you can simply sweep a sine wave with no processing, then lasso the silent area below the generated signal, and you'll find the meters indicate a level of as high as -60db for a sine wave at -18.

Anywho, I dove deep into this, cuz it's a pretty fascinating look at thing. Here's what I found (with the big fat caveat that this is literally my first time using RX for this, so it's likely I'm overlooking one or many things):

- Using your initial settings, I found I was mistaken about the ~20k aliasing tone being -120db, as it appeared in PD. It's tough to say for sure using the RX spectrogram (at least, with the shallow understanding i have of it thus far), but the ~20k aliasing tone seems to be between -75 and -80. From around 15k downward, there are no aliasing tones higher than about -95db.

- With the same settings and the drive at 100%, it looks like the loudest aliasing tones are between -45 and -50db, way up in the 15k - 20k range. Below that, the aliasing seems to live around -70.

- Using your third settings, with attack at 0, drive off, and RMS at 0%, i see very little aliasing. Seems to be hovering around the -90db mark.

- Those same settings with RMS at 1% show effectively no aliasing at all.

Thanks again for hipping me to this usage of RX

ETA - just noticing that had RX set to 44.1, not 48k. I suppose that means the above figures would be a bit lower yet as compared to your findings.
Old 2 weeks ago | Show parent
  #66
Gear Nut
 
🎧 5 years
Quote:
Originally Posted by bgrotto ➡️
Thanks for turning me onto the analysis functionality of RX8. I hadn't used that before, and it's very cool.

I am glad you find this useful too. Staring at Plugin Doctor for extended periods of time ...does things to me

I was aware of the initial level burst, that's why I took the screenshots when the playhead was towards the end of the selected range, giving the meter a little time to settle. That said, this is far from reliable, so I decided today to generate that sine sweep over 20 seconds instead of just 2, and now I can confidently say that my results were pretty accurate. Actually, the aliasing levels on the meter are even worse, according to the new tests.

I am not going to clutter your thread with even more images though. Suffice to say that, as far as compressors goes, this one generates unacceptable levels of aliasing. Unacceptable for me, that is.

The way I usually test something like this:
1. Quickly load the vst in PD/harmonicAnalysis for an initial check, and if I see aliasing getting above -90db or -80db, I don't even need to look further, because I know that this is audible.
2. If the PD test is inconclusive or hard to decipher, I move to RX and do the lasso trick and listen. I then decide if I can live with what I am hearing or if the vst has to go.
I usually don't even look at the meter, I've only pointed it out as some sort of visual proof.

Still, if the sine sweep goes over 20 or even 10 seconds, I can say with 100% confidence that what the meter is showing during the second half of the range is 100% accurate, as that initial burst is long gone and the meter had more than enough time to settle.

Another tool we can use for this is SpectraLayers. It works almost identically to RX in this regard, but there is a fade-in which gets rid of the non-zero crossing issue. Plus, it also supports VST3, which RX does not, as far as I know.
Another advantage is you can set the min-max visible range. It has some disadvantages too though.
Old 2 weeks ago | Show parent
  #67
Gear Maniac
 
Quote:
Originally Posted by cuzzinlouie ➡️
I was aware of the initial level burst, that's why I took the screenshots when the playhead was towards the end of the selected range, giving the meter a little time to settle. That said, this is far from reliable, so I decided today to generate that sine sweep over 20 seconds instead of just 2, and now I can confidently say that my results were pretty accurate. Actually, the aliasing levels on the meter are even worse, according to the new tests.
When doing this kind of test in RX it's useful to just get a static frequency perhaps with a small wobble and add some fadeins and fadeouts. I also add a bit of silence at the ends, because sometimes plugin behave weird in RX and truncate audio. In the first picture I have -18dB 4K tone with 10%@5Hz modulation.

Setting DragonFire to defaults with 100% drive and -42dB threshold so that I get at least a bit of reduction creates a barely noticeable harmonics and lots of aliasing. You can compare the color/level of the wavy shape around ~12K with the rest of the second picture. If I cut out everything above the first generated tone I am still left with peaks around -45dB as seen in the third picture. It's the same in single-band mode with threshold set to achieve similar amount of gain reduction.

On the other hand, if I apply -20dB of gain before DragonFire and then +20dB after it with the threshold set to -62dB for similar GR, then I get much clearer harmonics and much less noticeable aliasing as shown on the fourth picture. If we cut out the 4K tone and above then the peaks are around -60db, so still a lot, but that is with drive on 100% and its much better than before.
Attached Thumbnails
denise introduces Dragon Fire: A new way to color and tame your track dynamics-denise-dragonfire-rx-01.png   denise introduces Dragon Fire: A new way to color and tame your track dynamics-denise-dragonfire-rx-02.png   denise introduces Dragon Fire: A new way to color and tame your track dynamics-denise-dragonfire-rx-03.png   denise introduces Dragon Fire: A new way to color and tame your track dynamics-denise-dragonfire-rx-04.png   denise introduces Dragon Fire: A new way to color and tame your track dynamics-denise-dragonfire-rx-05.png  

Old 2 weeks ago | Show parent
  #68
Gear Nut
 
🎧 5 years
Thanks! That's a very interesting way of looking at this, as it represents something closer to a real life situation (as opposed to a sine sweep). I'll definitely play with it too, but first I need to figure out how to interpret what I am seeing. Also, good point about adding silence at the beginning and end.

Regarding this particular VST though, I don't expect different results. The fact that it distorts less with lowered input gain is to be expected, and the same result can be achieved by simply lowering "drive". I could be wrong, but that's what I would expect.

If it were just one particular setting/knob/value that's causing trouble, one can learn to avoid it, but with this vst its more than just one combination of settings that can push aliasing to really high levels.
Old 2 weeks ago | Show parent
  #69
Lives for gear
 
bleepbleep's Avatar
 
🎧 10 years
Quote:
Originally Posted by cuzzinlouie ➡️
Also, I am little surprised that nobody mentioned the HF loss (maybe I missed it).
it could be the 'weighting/roll off' style of the spectrum.
some spectrums use the default 'flat' setting, others use different 'weighting' which ends up producing these contours.

it's just a guess though, it wouldn't be good if it really was killing all that high end.
Old 2 weeks ago | Show parent
  #70
Gear Addict
 
Denise.Audio's Avatar
 
Andrew Scheps Mixing technique in the DF

Hi all,

we wanted to share with you how you can execute the the world famous Andrew Scheps vocal mixing technique using the Dragon Fire. Allow out engineer, Enrico Tiberi to show you how to do just that.



All the best.
Old 2 weeks ago | Show parent
  #71
Gear Maniac
 
Quote:
Originally Posted by cuzzinlouie ➡️
Thanks! That's a very interesting way of looking at this, as it represents something closer to a real life situation.
Please don't joke in this way, somebody might misunderstand.
Quote:
Originally Posted by cuzzinlouie ➡️
Also, good point about adding silence at the beginning and end.
Silence sometimes helps. Fadeins and fadeouts always help, it's a form of a window.
Quote:
Originally Posted by cuzzinlouie ➡️
Regarding this particular VST though, I don't expect different results. The fact that it distorts less with lowered input gain is to be expected, and the same result can be achieved by simply lowering "drive". I could be wrong, but that's what I would expect.

If it were just one particular setting/knob/value that's causing trouble, one can learn to avoid it, but with this vst its more than just one combination of settings that can push aliasing to really high levels.
I agree that this is what one would normally expect. However, I think it's a bit different with DragonFire. It seems that the amount of saturation is controlled by the drive parameter and the gain reduction the plugin applies. Please observe that in the fourth picture in my previous post the third harmonic at ~12K seems stronger than in the second picture. It's hard to tell for sure because of all the noise, but the one from the 4th it not weaker, despite the plugin being driven with a 20dB quieter signal.

My guess is that there is a pre-compression saturation stage with its internal "amount" macro parameter controlled by something like drive * GR, and that stage aliases badly. With lower inputs and appropriately adjusted threshold you still get similar amount of saturation, but with less aliasing. I don't know, but to me it looks more like a bug rather than a feature.
Old 2 weeks ago | Show parent
  #72
Gear Maniac
 
@ Denise.Audio Mind chiming in on what we are seeing/hearing with the examples we posted and perhaps if there are any plans to improve it?
Old 2 weeks ago | Show parent
  #73
Gear Nut
 
🎧 5 years
Quote:
Originally Posted by bleepbleep ➡️
it could be the 'weighting/roll off' style of the spectrum.
some spectrums use the default 'flat' setting, others use different 'weighting' which ends up producing these contours.

it's just a guess though, it wouldn't be good if it really was killing all that high end.
The analyzer weighting, or tilt as it's called in ProQ3, should have nothing to do with it. I am simply applying a little bit of compression to white noise, with all bands at zero.
White noise = random signal having equal intensity at different frequencies.

Anyway, just to remove any second-guessing, here's the same test with tilt set to 0. You can also observe the HF drop even in dragon's analyzer. The lower the threshold, the higher the HF drop.

Also, I could not find a way to balance the freq response from within the dragon , other than setting the LPF pretty low, around 4k. Still not perfectly flat, but close.
Attached Thumbnails
denise introduces Dragon Fire: A new way to color and tame your track dynamics-denisehfflattilt.jpg  
Old 2 weeks ago | Show parent
  #74
Gear Addict
 
Denise.Audio's Avatar
 
Quote:
Originally Posted by random musican ➡️
@ Denise.Audio Mind chiming in on what we are seeing/hearing with the examples we posted and perhaps if there are any plans to improve it?
We'll mull over this and get back to the forum when we can.
Old 2 weeks ago | Show parent
  #75
Gear Maniac
 
Quote:
Originally Posted by cuzzinlouie ➡️
White noise = random signal having equal intensity at different frequencies.
White noise does have equal intensity, yet our hearing understands frequencies, with some simplifications, logarithmically. You want same intensity in each octave, each doubling the frequencies. From musical perspective white noise is skewed toward high frequencies and as bgrotto suggested previously, pink noise is better suited for such musical testing.
Old 2 weeks ago | Show parent
  #76
Gear Nut
 
🎧 5 years
The Fletcher Munson curve would be even closer to how we are hearing, but this has in my opinion little to do with the test at hand, which is measuring the HF drop, in which case white noise seems to me most appropriate.

If we're indeed focused on how we're hearing these changes, and keeping in mind that the Fletcher Munson curve changes with loudness, then we should expect this VST to adjust the threshold in certain frequencies based on the level we're listening at, which of course would be unreasonable, let alone undesired.

What I am basically saying is that I would not want a compressor to make these decisions for me, even if the tweaks were made according to how our hearing works, which is not the case here either.

Last edited by cuzzinlouie; 2 weeks ago at 04:24 PM..
Old 2 weeks ago | Show parent
  #77
Lives for gear
 
bgrotto's Avatar
Sidechains are generally not linear, and the way each compressor “hears” and reacts to incoming signal is a massive part of each unit’s tone.
Old 1 week ago | Show parent
  #78
Gear Maniac
 
majoraxis's Avatar
 
🎧 10 years
Excellent Plugin

I am really loving Dragon Fire!

The processing is high quality enough to do mastering with it. I was able successfully replace a multiple band compressor and eq plugin and I did not look back.

When purchasing it I went for 3 plugin in total to get the 30% discount when purchasing 3 or more. I choose Dragon Fire, Bad Tape and Sweeper. You can turn of the "bad" parameters on bad tape and it sounds great. It is worth trying if you are in the market for a tape plugin.

Is it possible to have an auto release option, a look ahead option, oversampling options to choose from, a channel linking/unlinking option and an off button for the clipper?

Thanks!

Mark
📝 Reply
Topic:
Post Reply

Welcome to the Gearspace Pro Audio Community!

Registration benefits include:
  • The ability to reply to and create new discussions
  • Access to members-only giveaways & competitions
  • Interact with VIP industry experts in our guest Q&As
  • Access to members-only sub forum discussions
  • Access to members-only Chat Room
  • Get INSTANT ACCESS to the world's best private pro audio Classifieds for only USD $20/year
  • Promote your eBay auctions and Reverb.com listings for free
  • Remove this message!
You need an account to post a reply. Create a username and password below and an account will be created and your post entered.


 
 
Slide to join now Processing…

Forum Jump
Forum Jump