IR Properties

Discussion in 'Digital & Modeling Gear' started by Jay Mitchell, May 17, 2019.

  1. biagio1957

    biagio1957 Silver Supporting Member Supporting Member

    Messages:
    78
    Joined:
    Jan 14, 2009
    Location:
    NorCal
    This thread beats the heck out of that long UFO one in The Pub!

    Seriously - whether you like these IRs or not, there is some amazing data and information being passed here. Wow.

    Many thanks to all.
     
  2. vtgearhead

    vtgearhead Silver Supporting Member

    Messages:
    1,009
    Joined:
    Aug 24, 2016
    Location:
    Burlington, VT
    I converted Jay's JMDemoC and JM412 IRs to 44.1 kHz. sample rate and used Kemper CabMaker to convert them to cab presets. Haven't spent much time with the 412, but the 'C' IR sounds really good with Fender amp profiles. Seem to recall some early discussion that suggested this didn't work well, but it seems fine to me and sounds very consistent with what I've been getting on the Helix.
     
    MaxTwang likes this.
  3. burningyen

    burningyen Member

    Messages:
    14,113
    Joined:
    Jan 4, 2002
    Location:
    CT
    I had good luck with antcarrier’s FFIRs in my old Kemper.
     
  4. gigsup

    gigsup Supporting Member

    Messages:
    1,508
    Joined:
    Aug 28, 2016
    I bet they still sound good in new Kempers.
     
    timowens316, MaxTwang and burningyen like this.
  5. kennedydream

    kennedydream Member

    Messages:
    24
    Joined:
    Aug 5, 2010
    Location:
    Carlisle, Cumbria, England
    I've emailed Jay for the IR's as I would really like to audition them. Not sure if I'm too late requesting them.

    From what I've heard from the comparison clips posted in the thread example B sounds like Celestion's own IR using the Neuman Mic's (TLM 107's).

    I have a few IR packs from Celestion and the Room Mic IR's always sounded 'weird' to me. I'm guessing they are best used blended with a NF IR.

    Anyway I'll wait patiently to see if anything appears in my In Box!!
     
  6. Lele

    Lele Member

    Messages:
    1,202
    Joined:
    Jun 26, 2010
    Location:
    Italy
    I think that those "room" IRs have nothing to do with Jay's.
    Because the commercial "room" IRs - as far as I know - contain tons of early reflections, on the contrary - as it was explained in the first 40 posts of this long thread - the target of Jay's IRs is to reproduce the pure sound of the speaker/cab without any influence coming from the environment, that means without any early reflection. This is because when you play with a speaker, the room where you play in will add its own mark (and reflections) therefore it's better if you start with an IR without any other extraneous (truncated) reverb coming from an unknown space.
    And if you play with headphones, it will be easier if you add a reverb space with a separate effect block.
     
  7. kennedydream

    kennedydream Member

    Messages:
    24
    Joined:
    Aug 5, 2010
    Location:
    Carlisle, Cumbria, England
    I realised that b is nothing to do with Jays, in fact I’m sure Jay confirmed that b was a commercial IR. It sounded familiar to me from owning a number of the Celestion IR packs. It definitely sounds like their room IR.
     
  8. MikeRI

    MikeRI Member

    Messages:
    191
    Joined:
    Dec 19, 2016
    That's out of my league of expertise... Any chance of getting a shared copy, so I/we can try it out? Hoping so... Thanks.
     
  9. Jay Mitchell

    Jay Mitchell Member

    Messages:
    5,533
    Joined:
    Oct 13, 2007
    It appears that someone has learned quite a bit from the information I've posted here, although he doesn't acknowledge where he learned it.

    IR Length
    and
    https://www.thegearpage.net/board/index.php?posts/29444860/

    "I've been experimenting with IR length lately and keep finding that I like a shorter length. So I gave some thought to it and I think the reason is that a shorter IR trims off the early reflections."

    IOW, exactly what I've been teaching here: identify the arrival of the first reflection and window the IR so as to remove everything from that point onward. Caveat: in order for that to work well, you've got to have an IR that is free of reflections for a sufficiently long time to give a convincing impresssion of the speaker. Cliff has apparently decided that 10ms - which the original AFX FW was capable of - is long enough. I have always advocated 20ms, which is borne out both by my own subjective experience and by research on human hearing perception. However, as I have also been saying all along, that 20ms must be free of reflections.

    You're welcome, Cliff.

    Oh, and Clark: are you paying attention?
     
    Last edited: Nov 9, 2019 at 10:42 AM
  10. Vitopower

    Vitopower Member

    Messages:
    2
    Joined:
    Sep 29, 2015
    Thank you @Jaymitchell for the opportunity to learn about this technology and to audition these impulses.

    I'm including the following info so that it might help to give the most value to my assessment of Jay's work by lending a different perspective than most of the previous reviews.

    I'm working in a tuned +/-5db studio with broadband acoustic corner, side, rear and cloud absorption, rear fractal diffusion, full range flat response monitoring and monitor decoupling, using Apogee Symphony conversion and clocking. The room sounds balanced and has a very strong phantom center image, but most importantly, music sounds like what I love about music.

    I have been using big name-brand IR for a few years. I always find myself fighting to remove the usual suspects:

    boom / low end artifacts
    stiff or choked midrange and midrange resonance
    icepicky high end


    which means using multiple stages of processing- HPF/LPF, eq/multiband compression to make something resembling music come out of the speakers. Even then, I often have to use timed delays to mask the unnaturalness of the sound.

    In my test I was using a pre-recorded solo

    strat neck single coil and bridge split coil humbucker> custom built fuzz / treble booster that does a sag/compression thing > ef86 AC15 > Suhr reactive load > MixIR2

    ...so I was really putting this IR through the ringer.

    Using "C" IR:

    Holy, wtf...Right off the bat, the sound was completely useable with no processing. It is relaxed, natural, uncolored and artifact free. Most importantly, the midrange is right. I've never heard that before in an IR. Within a few eq moves I had a sound that was worlds beyond the quality of what my previous setup was doing. I blended in about 70% of the "A" IR to put the solo more upfront. This IR also shared the characteristics of "C" and blended effortlessly.

    In mixing, I only had to use one gentle eq with very wide and subtle tone shaping and an HPF to let the solo come up and cut a little. I also used less than a db of multiband compression here or there to compensate for artifacts from the pedal I was using.

    Conclusion

    These results do not represent a slight improvement in IR technology... they represent a sea change. I have no need for the brand name IR I was using. They are obsolete technology in comparison.

    I doubt that anyone in any environment would not be able to hear an obvious change when comparing to other commercial IRs. To me, witnessing this in a balanced, treated mix environment on flat response monitoring is the benchmark.

    So often with music and technology, what works on paper fails to translate into something musical. From what he has been willing to share, Jay shows an advanced understanding of the technical aspects involved. After using these IR firsthand, what makes this so exciting and valuable to me (as a musician who mixes his own music) is that I trust his ear and ability to correctly assess the musicality of his results. This is a rare, subjective thing that comes down to much more than technical knowledge.
     
    Last edited: Nov 9, 2019 at 8:38 PM
    Lele, djd100, Dave Merrill and 4 others like this.
  11. StudioDevil

    StudioDevil Member

    Messages:
    143
    Joined:
    Feb 18, 2015
    When I created the convolution IR processor for Studio Devil Amp Modeler Pro over 10 years ago, I recognized this range of opinion in which impulse length sounded best, so I added a user adjustment for window size. From this I found 5ms and 10ms to sound quite good. However, this was with most IRS which contain some amount of reflections (usually a floor reflection and a secondary ceiling typically). So in these cases the 256 and 512 sample window settings were likely just eliminating those and making them sound “better”. But now that I have been experimenting with my own reflection free recorded impulses, with first reflections outside of 25-30ms, I’m finding that the 2048 settings sound as good if not better than the 1024 and certainly the 512 and 256...which confirms that the likely offender is the notches produced by the combing of the reflections. Sometimes it sounds good though...so can’t say it’s “best” in any general sense. But I like them.
     
    Chocol8 likes this.
  12. Jay Mitchell

    Jay Mitchell Member

    Messages:
    5,533
    Joined:
    Oct 13, 2007
    Ca. 2007, nobody other than myself had ever done a comparison of guitar cab IRs in which the length of the IR was the only variable that was allowed to change. That would require a reflection free IR long enough to window to every length that would be evaluated, and - this is really, really important - at that time, no guitar cab IRs other than mine were free of reflections even to 10ms. When you audition an IR that contains reflections, changing its length changes the number of such reflections in the IR. This effect is not subtle, but it has nothing to do with the minimum length required to capture audible detail in guitar cab sounds.

    That means you can't acquire a 2048-point IR that is free of reflections. That's not a problem, BTW. If you can get clean 20ms IRs, you're doing very well.

    If there are reflections at ca. 25-30ms in the IR, they will be present in the 2048-point (42ms at 48kHz SR) version and absent in the 1024 +(21ms) version. In that case, you can't attribute a sonic preference for one length or the other, nor even the ability to tell the difference between the two, to the difference in length.

    The entire basis for the procedures and conclusions I've arrived at is that the cab block in a modeler is there for the purpose of recreating the sound of the cab, in a manner that is perfectly analagous to the function of the amp block. In that context, it is important to understand that the cabs being simulated are guitar cabs and to consider the properties of such cabs in making decisions about e.g., IR length.

    FYI, the difference between a 512-point and a 1024-point reflection free IR of a guitar cab, while reliably detectable, is quite subtle. The difference between a 1024-point and a 2048-point IR (also reflection free at both lengths) is almost certainly not audibly detectable. It is well beyond the point of diminishing returns.
     
    Last edited: Nov 9, 2019 at 10:45 PM
  13. Chocol8

    Chocol8 Member

    Messages:
    205
    Joined:
    Mar 2, 2019
    Is it possible that the reflections beyond 25 ms are less offensive? At those distances, the reflections are likely at a much lower level than an earlier reflection and (room dependent I assume) should alter the IR much more subtly.

    I get the diminishing returns, but 42 ms vs 21 doesn’t require much more work other than a larger space correct? Somewhere around a 24 foot ceiling with 48 feet between walls or columns should do it or be very close which a lot of gyms, hangars, and industrial spaces should be able to accommodate.
     
  14. StudioDevil

    StudioDevil Member

    Messages:
    143
    Joined:
    Feb 18, 2015
    I found the same ca. 2009 with impulses I had made for me. They were recorded outdoors in a field with a cabinet with its back to the grass and an overhead mic was placed above it. This seemed the most reasonable way to evaluate the impulse length choice at the time anyway. But I also did the same with synthetic cabinet models (IIRs) and had them sampled and windowed into different lengths. My opinion was similar...that I couldn’t tell differences above 1024 and that the spectrum got smoother with 512 and below...sometimes with beneficial tone, which was subjective. Anyway, seems 1024 is a sweet spot for me, but I sometimes “like” the sound of 256 and 512, but often need to compensate for a bass loss with EQ. Fortunately, 2048 is not really a problem. The convolver in my AMP plugin from 2009 actually handled 1M samples quite well even with hardware at that time. Lots of tricks in convolvers to get that to work. We have a few impulses we did in a gym with reflections around 45 or 50ms and they agree with your statements (and mine above as well). 1024 is good. 2048 “might” be overkill but it’s on the edge so I say do it if you can but don’t lose sleep if it’s 1024.
     
  15. gigsup

    gigsup Supporting Member

    Messages:
    1,508
    Joined:
    Aug 28, 2016
    Larger spaces like those are typically noisy places.
    Traffic, planes, AC, motorized industrial hum, fluorescent lights, rain, insects, mammals, all of that adds up to being more... of what you don't want.

    Performance spaces are better candidates. Noise issues have (hopefully) been taken into consideration, and the noise floor is pretty good to begin with.
     
  16. gigsup

    gigsup Supporting Member

    Messages:
    1,508
    Joined:
    Aug 28, 2016
    Why not, right? It's not just about the data for todays modelers, it can also serve as data for future developments no one is considering today. So if you can obtain longer far field IRs, do the best you can do.
     
    StudioDevil likes this.
  17. Jay Mitchell

    Jay Mitchell Member

    Messages:
    5,533
    Joined:
    Oct 13, 2007
    If by "less offensive," you mean "less noticeable," no.
    And, much more importantly, more easily recognized as separate events.
    You really don't have your head around the challenge or the potential payoff here. Maybe you should actually make the effort required to acquire 42ms clean IRs and compare them windowed to shorter lengths. I've gotten that far. The notion of diminishing returns is actually too optimistic. In reality, the returns are vanishing.
    No. Not even close. The actual solid geometry is presented, at least partially, earlier in this thread.
     
    Last edited: Nov 9, 2019 at 9:29 PM
  18. Jay Mitchell

    Jay Mitchell Member

    Messages:
    5,533
    Joined:
    Oct 13, 2007
    There are many, many cans of worms in that kind of testing. To begin with, there's the reflection from the ground, which will arrive within ca. 3ms after the first arrival. So, unless you actually recess the cab into the ground and fill the hole around it so there are no gaps, you'll never get a clean IR that way. This is the kind of mistake that novices in the field make all the time.
    With the mic placed 2 meters from the cab and the cab on the floor, that would require a minimum height to any overhead reflective surface - not just the ceiling, but bar joists, light fixtures, and the like - of 32 feet. Lots of gyms come in shy of that.
    The point is moot. At present, only a handful of folks are getting even 1024-point clean far field IRs. Finding a big enough space is not enough by itself.
     
    yeky83, MIJLOVER and StudioDevil like this.
  19. StudioDevil

    StudioDevil Member

    Messages:
    143
    Joined:
    Feb 18, 2015
    exactly!

    I had a convolver with 64k points so I could capture the room ambience as well. With some clever tricks you can experiment windowing these to get control of close and far sounds. People did this with reverbs a lot when I was just getting into this years ago...for them it was early reflection vs room tail. Here it could be an order smaller, but same idea...cab and room.

    but here’s the thing about the resolution...as you window the impulse...the shorter the window, the blurrier and curvier the spectrum becomes. You lose sharp edges, peaks, and notches in the spectrum as you window the impulse shorter. Sometimes this sounds good. But, as you go from 2048 to 1024, the loss in sharpness is hard to sense even though it’s there.

    nowadays, 4096 convolvers in DSP is kinda trivial but the longer you make the IRs, the more of the room you will get. And since this thread seems to be about “room less” IR capture, it’s really pointless to make them bigger than what you can record without reflections.

    but, that does not mean that if you COULD measure a 4096 point reflection free IR that you would benefit from that in any way that you wouldn’t from a 2048. The frequency resolution gets so fine at this point that it becomes difficult to hear the difference. When you blend IRs together however, that’s another thing...

    I personally like the sound of cabs in rooms because my application for IRs is in recordings and I like my tracks to sound like a cab in a real space. For FRFR reproduction, I imagine reflection free (or what I liked to call roomless IRs in the past) would be best, allowing the reproduction equipment to fill the performance space without any artifacts of the original space/micing.

    maybe someone should try a ground plane measurement outdoors and make some 4096 reflection free impulses and we can compare them.
     
  20. Jay Mitchell

    Jay Mitchell Member

    Messages:
    5,533
    Joined:
    Oct 13, 2007
    That's the classical uncertainty principle at work. That, plus the ability of human hearing to distinguish narrowband response variations, is the basis for the 20ms minimum length. Shorter than that, and you lose audible response detail. Longer than that, and the additional detail is below the ability of human hearing to recognize.

    It's also worth noting here that most modeler users - including Amplifire owners - won't be able to compare 1024- and 2048-point IRs, because the modelers won't convolve IRs longer than 1024 points.
     
    gigsup, djd100 and StudioDevil like this.

Share This Page

  1. This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
    By continuing to use this site, you are consenting to our use of cookies.
    Dismiss Notice