What exactly do IR loaders do - or go with no IR at all?

[Igor Samurovic]

I've noticed there's a great difference among IR loaders' frequency responses, such as keFir being much darker than MixIR and LeCab 2 being somewhere in the middle. Does anybody know if those plugins (especially interested in LeCab 2 since I am using it) do anything else than add an EQ curve to the signal? I've come up with an idea of creating a 'perfect tone' for my needs by simply using a ton of EQ instead of impulses, since I know exactly which frequencies I want in and which ones I want out, so is there a serious downside to this?

 

[JonWormwood]

Give it a swirl.

 

[JakeAC5253]

Most IR loaders will add a certain coloration to the IR that most times can't be dialed out. When I make IRs, I find that when using KeFir, the resulting tone sounds much closer to the tone that the IR sample actually came from. Rather than calling KeFir darker, I would think the others are brighter because they add their own coloration. The major drawback of KeFir though is that its GUI is not very convenient to use, but sound for sound it is the most honest.

 

[Igor Samurovic]

https://dl.dropbox.com/u/31334613/Render/sampleriff.mp3 here's a sample, no IR has been loaded here, all you hear is extensive EQing. I like this because it gives me much more power over the tone! The tone is a bit fizzy but that's a preference.

 

[deLuther]

+1 for KeFIR.
There is some test, I did long ago:

Comparison to convolution in Wolfram Mathematica.
Another picture, waveform comparison of LeCab 1 (red), LeCab 2 (yellow) and Mathematica (blue, reference):

 

[Igor Samurovic]

Woah, I've been avoiding keFir for years because of its dark tone, and now it turns out to be the best one? A way to feel stupid.

 

[JakeAC5253]

Woah, I've been avoiding keFir for years because of its dark tone, and now it turns out to be the best one? A way to feel stupid.

Don't feel stupid. Unless you make your own IRs, there is really no way for you to have known. Differences in IR loaders' tone is one of the aspects of using IRs that is not very well known, though I wish it were. People might be much happier with their IR tones if they did!

 

[deLuther]

Exactly! The only way is to compare convolved tone with miced, closest one wins.
If you compare impulse loaders alone, then you should compare against trusted convolution reference.

 

[Igor Samurovic]

Well, it's great that I now know keFir is going to result in much more natural tones than others, thanks guys
However, one of the questions still stands > can the effect of 'having a cabinet' be achieved by EQ alone?

 

[JakeAC5253]

However, one of the questions still stands > can the effect of 'having a cabinet' be achieved by EQ alone?

In theory, an IR is just an EQ with a very high amount of sample points. As evidenced above, you could probably EQ yourself to a tone that is useable in some sense if you like the resulting tone. As far as sitting down and saying "I am going to EQ myself Mesa cab SM57 tone today" I don't think so. A real IR is going to have a much higher level of both natural coloration and accuracy that you can't approximate with EQ.

Thanks deLuther for the test results, that is great!

 

[Igor Samurovic]

Well, in that case, that's great news for me as I do not want to replicate any cab/mic tone combination, I want to tailor something that I will actually like, so that will not be problematic! In my tone, I've put a lowpass at around 5khz that should simulate the cab a bit, and I realized from 5khz up there's now a straight line on my spectrum analyzer! And I really like that sort of air balance.

 

[LokiFoki]

IRs do quite a bit more than just EQing your sound. That's why they can produce reverb and delay sounds.
IRs of guitar cabs are very short and don't introduce much audible reverb but it's still there. Like resonating lows or smeared high frequencies.

I experimented with using EQs instead of a impulse respones, but I couldn't get the high mids/highs right. They were either too grainy and fizzy or too low. EQs can't smooth out frequencies (I'm talking about smearing/oscillation...whatever). That's why simple frequency-corrected line outputs on amps usually sound either harsh or dark.
So you probably don't want to use EQ-corrected guitar sounds on their own. But mixed with ir/mic sound they can bring something new to the table. They sound really dry and precise. Great for industial sound etc...

 

[deLuther]

Even IR made from tone matching standpoint (i.e. as pure very high resolution EQ) can introduce some artificial reverberation, especially when length is high enough.
For example I`m used such approach to imitate sound of real Dual Recto with ENGL E412XL cab with LeCto.
Original recording (in studio): 04032012_CCS_MesaDR_E412XXL_SM57.mp3
LeCto with matching impulse (65k): 04032012_TS999_LeCto_MesaDR_E412XXL_Conv.mp3
And another approach by using direct frequency matching (no artificial reverb): 04032012_TS999_LeCto_MesaDR_E412XXL_STFT.mp3
Matching impulse: MesaDR_E412XXL_65k.wav
Settings used:

DI can be found here: SPM100_vs_AR133.rar
(Also comparison between famous BSS AR-133 DI box and one good russian preamp with Hi-Z).

 

[Robert W]

Here's a three way comparison between KeFir, LeCab, and MixIR2. Admittedly it's not a very scientific comparison, but I do believe it illustrates the inherent tonal differences between the three different loaders.

KeFir
https://dl.dropbox.com/u/17738100/scKEFIR.mp3

LeCab
https://dl.dropbox.com/u/17738100/scLeCab.mp3

MixIR2
https://dl.dropbox.com/u/17738100/scMixIR2.mp3

The same combination of amp settings and cab sims was used for all the clips. The only thing that's different is the use of the impulse loader.

 

[deLuther]

Properly coded IR loaders should not have tonal differences, that`s all.
Main question is: how far is particular IR loader from ideal implementation.

 

[onqel]

I'm no expert on IR loaders, but what I know is that the IR samplelength does matter when you want to preserve the low end response of the loaded impulse. Personally I want to use at least 1024 samples to get more low end details and more of the room-flavour, when you cut down on the samples you also cut away the acoustics of the room.. I might be biased but I tend to think that the longer response smoothes out the sound a little.. cheers!

 

[Igor Samurovic]

Thanks for the responses, this made me realize and think about many things I didn't even think about. I see how this can be used to create peaks in the frequency spectrum, to add voicing to a 'miced cab' impulse via resonant frequencies of the amp/cab/anything in the construction. I've tested what onqel said and it truly seems to be that way, I cannot hear a difference when the track is solo'd but it does seem to have an impact by smoothing the whole mix out, and is especially noticeable when you put the track's volume down! Also, I must say it's great to see the creator of the best amp VST ever around here I use x50 almost exclusively (may turn to 456 and lecto for saturation/alternate lead tracks though).

@Robert: the first two actually sound pretty close in the mix, except for MixIR2 which sounds vastly different.

 

[deLuther]

onqel, you are absolute right, shorter impulse length reduces low end definition, it`s a fact.
This is the same as using FFT analyser with different block sizes, frequency resolution = sample rate/block size.

 

[Kazrog]

Convolution is a mathematical equation, and there is only one right answer per input sample / IR pair. However, many audio plugins for realtime processing will use math rounding for greater CPU efficiency / lower latency, which is why you see a fair amount of deviation with the output results of different plugins - each implementation is unique.