Using convolver universally in HTPC

[antani]

I think it's possible to limit the correction to a specific frequency range, but I never did it.

If the measurement is really fine, I suggest you to listen for a while with this set up. The first time I heard a phase and amplitude balanced reproduction I was upset, because it sounded so strange compared to what I was used to. Now I can't listen without equalization.

If you really want to modify the frequency response I suggest to work on the target curve in DRC. In my environment I was used to reduce 3-4 dB in the higher frequency range.

[mchabana]

Hi,

sounds like a good idea, i will try and let you know.

Rgds,

Al

[tarsonis]

antani wrote:

You need to set up Directwire in this way:

It will route the audio streams from a WMD program like Powerdvd towards ASIO.
Then you need a VST host such as . You should create an Asio project.
The project should look like this:

The 5.1 channels will be processed in this way, from left to right:
- a convolution engine, Voxengo Pristine Space but you can use Convolver as well
- a parametric equalizer (SpinEQ)
- a time delay, to adjust speakers distance (Voxengo Audio Delay)
- a bass manager (Kelly Industries) to boost LFE channel by 10 Db and to route bass frequencies towards the subwoofer.

I doubt you can get a more sophisticated sound elaboration .

This is another example of final result, amplitude graph is slightly decreasing because I set the target curve in this way:

Hey guys. Sorry for rolling up this thread. I have three important questions.

1. I have to have a Stereo Impulse response for DRC, that is used to calculat the filters.
I open it with the convolver Plug or Voxengo Pristine Space. Is it right?
2. What for is the EQ and the time delay plugin? I tought the convolver plugin makes the equilisation and time correction?!
3. I got a null by 227hz to 245hz (-20db only left speaker) and a big null by 314hz to 405hz (~ -12db both speaker but not similar). My Room is treated to the maximum and I couldn't solve it. Would DRC correct this problem?


Hope for help

[antani]

tarsonis wrote:

.1. I have to have a Stereo Impulse response for DRC, that is used to calculat the filters.
I open it with the convolver Plug or Voxengo Pristine Space. Is it right?...

Right, the filter created with DRC needs a convolver.

tarsonis wrote:

...2. What for is the EQ and the time delay plugin? I tought the convolver plugin makes the equilisation and time correction?!...

The parametric equalizer isn't needed, I used mainly for testing. The time delays however are necessary, DRC don't compensate for the different distance of the loudspeakers from the listening position. Remember to place the mic in the exact center between left and right speakers, even few centimeters can create problems.

tarsonis wrote:

...3. I got a null by 227hz to 245hz (-20db only left speaker) and a big null by 314hz to 405hz (~ -12db both speaker but not similar). My Room is treated to the maximum and I couldn't solve it. Would DRC correct this problem?...

Difficult to say, but I am pretty confident you will get a consistent benefit.

[tarsonis]

Well, thanks for that information. I was confused about the EQ. So I already have optimised my speaker posiotn so I won't need the time delay. I will test it only with the convoler vst plugin.
For me the best correction would if DRC only corrects the mids < 2Khz. Is there anyway to setup the impulse that the convolver only work on that range.
Or should I only measure that range and export that result?

[antani]

DRC is designed to work full range. Limiting the correction to a specific frequency range is difficult, maybe impossible.

By the way I don't really see the reason for that, DRC is working pretty well even with high frequencies. If the need is just to correct for example a subwoofer, I think it's much simpler to create a minimum phase filter with REW. Actually this is what I did for the LFE channel of my home theater setup, while I used DRC for the other 5 channels.

[tarsonis]

Hey I`m really happy . I got my own DRC Filter. After several try, also with audiolense, nothing worked. The informations on the net are very confusing to me. And made me sick. The Voxengo Plugin is not freeware and stops playing all the time.
The ConvolverVST on soundforge doesn' start in my applications (live / wavelab). "Couldn't load the module" Is there any other VST Plugin out there?

[jlohl]

Is there any other VST Plugin out there?

you may try this http://www.ohl.to/about-audio/audio-...s/crossvolver/ and this http://www.ohl.to/about-audio/audio-softwares/align/

[antani]

tarsonis wrote:

... Is there any other VST Plugin out there?

I guess you can find many, such as this one. With Mac I'm using LAConvolver. However Pristine Space is a pretty stable one (but not free).

[antani]

jlohl wrote:

you may try this ...

Crossvolver and Align are interesting applications, but they are not VST convolution engines.

[tarsonis]

Hey guys. Thanks for the advice. Yeah your right that aren't VST, only the keFIR is a VST. It looks pretty nice. Maybe it works with my sequenzer. But is this a good idea to mix the dry level with the correction, like they say?
And a really important question. Is it useful to use a eq after the convolverm because some mid frequncies are still to loud? Or is it a psychoacoustical thing that there isn't any correction made?!
As I only used a eq to solve this, it sounds really strange.

[antani]

tarsonis wrote:

...Is it useful to use a eq after the convolverm because some mid frequncies are still to loud?....

You can of course use an eq to shape the sound the way you like. But, if you created a correct filter (maybe you can verify it with some sweeps in REW), I'd try to listen for a while without any further modification.

[jlohl]

Crossvolver and Align are interesting applications, but they are not VST convolution engines.

in fact both programs can be used standalone or as VST (just load the .dll file into another VST host). They are not convolution engines but they integrate "convolver".

[tarsonis]

I used the normal.drc. I didn't change anything, but the filename. I verified the correction with virtual cable -> host -> roomEQ and a sweep. I could hear it too. About 5db by 211hz and 300hz.
The right LS

But what a EQ ? Linear Phase EQ? I got the fear that it makes it worse than it its...

And did anyone try to make more than one measurement in a sweetspot area and averange them with roomeq. And than make a correction filter?

[antani]

It's not bad, but I think you can get a better result. In my environment I have a flat frequency response +- 1 dB, with a smoothing of 1/3 oct. Try the other .drc presets. But in particular pay attention not to clip the measuring sweep (I'm using 8 sweeps of 1 M).

DRC can create minimum phase, linear phase and mixed phase FIR filters. All the .drc presets are mixed phase. Mixed phase means the filter is correcting the minimum phase component and a part of the excess phase component. It's a sort of compromise between minimum and linear phase, with a short preringing.

If you try to average the sweeps, you will only get a bad filter. DRC is not Audyssey (luckily ).

[tarsonis]

Well, I used 1/12 smoothing. I think I can get this better too. Maybe I can come close to your holy flat frequency response +- 1 dB. I want this too. ^^
Do you use an extra eq after the convolver?

So I understood, you use your own manipulated preset.drc? How long did you need to tweak it, till you get to your result?

In my situation I use my monitors primary for musik making/listing and of course, a ultra flat response is really important to me. So I should target the linear phase option not to lose the possibilty to hear phase cancelecing in the mix. Do you have a tip what I should try to manipulat in these *.drc files?

Edit:// I tried the ARC system. I didn't work for me. Always a unbalanced stereo correction...first DRC correction works really nice, up to the mids. I really like DRC. Hope to get the rest tuned up.

[antani]

Tweaking DRC is far from simple. In Italy, with the support of Denis Sbragion, we are now building a graphical interface to DRC. It will show only the fundamental parameters. But it's not completed yet, it'll take another month or so, before beta testing.

One good starting point, at least for me, is optimized.drc. Be sure to deactivate the psychoacustic target (PTType=N) if you really want a flat response.

If you are keen about phase distortion, maybe you should go linear phase, setting PSFilterType=L. I sometimes use this kind of filter. The reverse of the coin is that you might introduce audible pre-ringing. By the way I was never capable of identifying even severe phase distortion on a blind ABX test. But I guess with mixing you might have problems. You will get a latency of 0,7 seconds, it's up to you to see if you can handle it.

I'm not using any extra eq. I did in the past but eventually it was useless.

[tarsonis]

Hey, thanks for the advice. Sounds really good. DRC should get a GUI.

Well my host got latency compensation. I hope the latency is the smalles problem. But, btw.
How do Audyssey IK ARC make it that there isn't any latency and high cpu usage?! Don't they use a similar convolver technolgie?!

//Edit: optimized.drc. I haven't that file. Which version do you use. I got somethin like extreme, minimal, erb, etc..

[antani]

Your vst host can compensate automatically only the latency reported by the vst plug-in, normally associated with the calculation latency.

A linear phase filter has the characteristic to have a symmetric impulse response. That means it has the same length before and after the impulse. You can easily check this opening the filter with Audacity.

This kind of latency is normally not automatically compensated.

For music listening isn't so important, because it'll only create some delay for example between play and music. I am using linear phase crossovers with my loudspeakers, and the latency doesn't bother me. But it can create problems for audio/video and for mixing.

Audyssey uses minimum phase filters (i.e. the impulse is at the beginning of the filter), moreover with a small number of taps. That's why it doesn't introduce latency. The drawback is that this kind of filter can't correct the excess phase component of the system, introducing some phase distortion.

You are right the latest version of DRC doesn't include optimized.drc. You can get it from release 3.0.0.

[tarsonis]

Okay. Good information. I tried to play around with the working first correction and I didn't really feel a constant latency. A bit of variation. Is there a way to calculate this latency, because my Host got a separate driver latency compensation?

So my second correction is a mess. This time a took 8 x 1M sweeps for every speaker. I searched the Internet and used the optimized.drc with PTType=N & PSFilterType=L options.
Really strange result. It feels/sounds like the sound comes from the left and the middle, but not from the right. Also it looks different to the first one.







The measured FFT results are looking great for both speaker. +-2db. Too good to be truth.
Better than the first correction I measured, but it sounds ugly.





What could went wrong? If I could I would test it again, but I have to test it tomorrow. Hope this is a stupid mistake made by me, otherwise

[antani]

The 2 filters are very different in term of gain, and this is a little strange. But they are also producing a good result, accordingly to the frequency response.

Obviously they will produce a different level of output. Just try to level the 2 channels using the channel wet control of Pristine Space (or the volume control of the vst host). Use a phonometer or the mic, with pink noise. You also should be very careful not to clip the audio card.

Every FIR filter has its own gain. You can also try to use the A gain (automatic gain)option of Pristine Space (but I don't trust it).

The compensation of the gain is always required, to avoid clipping and equalize the channels.

[antani]

tarsonis wrote:

...Is there a way to calculate this latency, because my Host got a separate driver latency compensation?

The latency reported to the vst host by Pristine Space is written at the bottom right of the main page. In your case is 4096 samples. On top of it you have the delay of the linear phase filter, in your case 65536/2 = 32768 samples, 0,743 seconds at 44.1 KHz.

[tarsonis]

uiii, now I can really hear it. YOu are right. The first Filter didin't involve such latency. Makes me not happy. .

I made a third correction and I still have this weird right/left relation. RoomEQ tells me "the impulse peak is not where it should be...". PANIK the first time I saw the red genelec led appear. The sweeps was around 18000 -20000hz, when it happens. I made all measurements by 20hz to 20Khz for the impulse. Is it wrong?! Should I go to 21Khz?!

Or maybe it is to flat, so the ls get to much gain on the highs?! I allways read that you have to have a roll off on the highs, when you sit more than a 3.2 foot away. My sweetspot is 4,2 foot away from the speakers.

I used the normal.drc for the first time with this option "Target respose stage/PSPointsFile" set to this file bk-2-44.1.txt. In the optimized there is this target file pa-44.1.txt used. I think that wasn't good?!

The corrected measurments look similar to my last post.
But they can't tell the truth, because when I listen to my music referenzes and some mono voices I have to gain the right channel, although the corrected measurement tells me something else. But it is still unbalanced. I have to make a shelving filter on the right ls about +2db, than it is close to normal.

The first Filter was looking so
The log sweep was 512k long, maybe that is the reason for that small latency in this filter?!
I try around a bit to find the error.

[antani]

Going 10 Hz -21 KHz might be a good idea. You can create your own target curve just editing the target .txt file, it's very simple. I used to decrease the highs by 3-4 dB.

The latency was due to the linear phase. The longer the sweep the better the resolution.

[tarsonis]

Ok. Thank you for your patience and explanation. So it not important what a latency I setup in my host and voxengo space. The filter latency is added after the dirver/vst latency and is not known by the host.
So my next idea is to tweak my (nearly) optimized.drc to get a filter that is close to my current result but doesn't introduce such a big latency, to be useful to make music. And to use the "HD" filter for listing.
So I tought it less filter taps would be a good solution. But the first normal.drc file I used got the same value for RTFilterLen, changes are in the LowerWindow and FinalWindow option.
It is really complicated, when you got fear to blow your speaker.
Some values in the manuel aren't told.

Okay. Sometimes it is easier than you tought. I made my own target file!


EDIT:// Hey when I setup the offset of voxengo to around 0,737sec, there is no latency. I pull it near the highest spike.
So it looks like the impulse in my post #73 or in your example in #32 and not like #70. This gap (voxengos visualisation) of ~740ms, before the first spike appears, is important?
In doesn't sound different or my ears are tired. It sounds like the sound starts 0,740ms earlier, not more. o_O

EDIT:// So here is what I did and the result is amazing. I took a impulse from roomEQ. This time without normalization and did the drc correction process as usual. After I put both impulses together. I cut the area where is only silence.

And look. The impulse is closer to the start.

Now I have no really noticable latency like in my first filter. Could be better, but it is a huge change of about 700ms.