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

Skeptico Saloon: An Objectivist Joint

Discussion in 'Sound Science' started by gnarlsagan, Jun 27, 2013.
97 98 99 100 101 102 103 104 105 106
108 109 110 111 112
  1. uchihaitachi

    Er no. Firstly the product description mentions the sonic design as such. Secondly as an owner of the K10 and having talked to the creator of the IEMs, aka Dr Moulton, there is supposed to be more treble presence in the K10. The N6 is more similar to the Heir 8A, which I also own.
  2. KamijoIsMyHero
    Well go ask Moulton if more drivers is a gimmick then. Even Moulton doesn't hang out here and there were multiple threads he could have offered insights on the tech.
  3. dprimary

    LOL hard to fit in IEM. A modern touring sound system dwarfs the Dead's wall of sound driver counts with around 1200 in the main PA and around a half million watts. If anyone figured out the amount of processing power in system the government might ban exporting them to some countries. 
  4. dazzerfong
    So, can we conclude that multiple driver IEM's aren't a gimmick if they are designed properly (ie. minimise crossover). That way, everyone's happy.
  5. Steve Eddy

    Hehehe. I know. Just that in modern systems, the drivers aren't so dazzlingly on display as in this old Wall of Sound system. They're usually hidden away inside folded horns.

    And as for fitting inside IEMs, who cares about IEMs? Give me surface area, baby!

    Peace. :D

  6. Joe Bloggs Contributor

    My thoughts on that (quoted from my blog):

  7. Steve Eddy

    Some of this I can agree with. For example, I never quite understood why measurements should be made on a dummy head with the diaphragm essentially where the eardrum would be. At least not for headphones where people's pinnae and ear canals are all different. Seems to me you'd want the diaphragm just outside the pinnae to capture the sound field before it reached the pinnae. I mean, you don't measure loudspeaker performance using a dummy head with the microphone diaphragm where the eardrum would be. Sure, with a headphone you need to replicate the relatively small cavity that the drivers are operating into, but not everything else.

  8. miceblue
    Ah, I shall revisit this thread again. : )

    What exactly do these numbers represent?

    I usually define absolute dynamic range as the difference between the softest and the loudest points. That is quite literally how dynamic the music is and your amplifier would need to be able to produce sound from the softest part to the loudest part without clipping and with the proper slew rate if the music demands it.

    For calculation purposes though, that's not a practical metric since software usually can't identify/tell you the quietest points within a given music piece unless there's complete silence like in some synthesised music. Using a track's RMS and peak values (thus crest factor) are much more indicative of the dynamic range for this purpose. The RMS value represents the continuous/sustained load on an amplifier, whereas the peak represents a sudden power surge requirement from the amp. This makes complete sense to me for determining a music track's dynamic range, how much change is there in the music from the average level? The crest factor is also commonly used for speaker amps to determine how powerful of an amp you'll need. A VU meter is also indicative of the RMS values and average volume level of a given music piece for loudness monitoring.

    The DR utility does something completely different from anything above and I don't see how it can be indicative of a given music track. The largest 20% of the RMS values doesn't tell you anything useful for either the music track nor the requirements for an amplifier; it isn't indicative of how loud the music track is from the peak values for a dynamic range calculation and it makes the track seem artificially louder than it really is, as evident by the much lower RMS values determined by Audacity, Musicscope, and Foobar's ReplayGain values.

    From the spec sheet posted above, 132300 samples are analysed in a 3-second block for a 44.1 kHz sampled file for each channel. For those 132300 analysed points, one RMS and one peak value are determined. The RMS calculation isn't really an RMS value to begin with since the radicand is multiplied by 2 when RMS calculations don't have that factor, no? So it's erroneous right there and is already inaccurate. If you have a 3-minute song, that means 60 RMS and 60 peak values are generated (one for each 3-second interval) for each channel, which means a blknum of 60. The RMS-sum is just the total RMS value (single number) for the whole channel, which is based on the inaccuratly calculated RMS values. To get the top 20% values, they divided the total RMS's radicand by 0.2 * blknum (12 in this case).

    Say you have these made-up 60 values from the RMS calculation:
    Ten values of -6.7
    Ten values of -7.3
    Ten values of -6.4
    Ten values of -5.8
    Ten values of -6.9
    Five values of -5.3
    Five values of -5.0
    Sum of Squares = 2469.35
    Sum of Squares / 12 = 205.78
    Top 20% Total RMS value = sqrt(Sum of Squares / 12) = 14.35

    With our simple numbers, we can double-check this.
    Reordering the values from smallest to largest:
    Ten values of -7.3
    Ten values of -6.9
    Ten values of -6.7
    Ten values of -6.4
    Ten values of -5.8
    Five values of -5.3
    Five values of -5.0
    Obviously the top 12 values representing the top 20% RMS values are the five -5.0 values, the five -5.3 values, and two of the -5.8 values.

    Sum of Squares for top 20% = 332.73
    Top 20% Total RMS value = sqrt(Sum of Squares for top 20%) = 18.24

    18.24 ≠ 14.35

    So even the "top 20%" RMS values they calculate are inaccurate from the samples they take.

    The spec sheet says:

    What would be an example of when small peaks would greatly alter the calculated RMS value in such a way that the true RMS value isn't good enough or indicative of the whole track? I guess I just don't understand why the DR utility uses the highest 20% of calculated values. Literally no other sources I've seen and read use this convention and instead use the crest factor for dynamic range determination, which makes sense to use unless you can provide a concrete example of when the DR utility would be a better convention to use.

    To me, the crest factor (peak minus RMS value in dB) is a far better indication of the dynamic range than what the DR utility outputs.
    For example, here's a song from a video game soundtrack that I really like:

    Here's its corresponding decibel waveform in Audacity:

    The light-blue area represents the loudness of the song, or the sustained/RMS values. If you just eyeball the approximate average level of this area, the left channel looks to be at around -13 or -14 dBFS and the right at around -14 or -14.5 dBFS. Subjectively, I hear that the right channel seems to have less stuff going on, so it makes sense that the right channel has a slightly quieter average volume level/average RMS value.

    You can use Audacity to objectively determine the average of the RMS values by doing a "Contrast" analysis.

    Indeed the right channel is quieter based on the average RMS value it calculated (-13.7 dBFS vs -14.3 dBFS).

    For peak values, you can use Audacity's "Amplify" (without clipping) effect. This will bring the song to as loud as it can get, 0 dBFS, without clipping. Both the left and right channels can only be increased by 0.1 dB for this particular song, so that means the peak values for each channel are -0.1 dBFS.

    From the average RMS values and the peak values, you can determine the crest factor for each channel.
    Left: -0.1 dBFS peak - -13.7 dBFS RMS = 13.6 dB
    Right: -0.1 dBFS peak - -14.3 dBFS RMS = 14.2 dB

    We can average the two crest factors to get the crest factor for the whole song: 13.9 dB

    So this represents how much variation the song has between its average loudness level and its peak levels.

    What does the DR utility output?
     Left Right
    Peak Value: -0.10 dB --- -0.10 dB 
    Avg RMS: -10.77 dB --- -11.38 dB 
    DR channel: 8.95 dB --- 9.87 dB 
    Official DR Value: DR9
    That doesn't seem right to me....
  9. uchihaitachi

    He is an industry insider and will obviously have limitations on what he can disclose.

    Gimmicks aside, I merely wanted to understand how driver counts impact performance from an objective standpoint which you clearly fail to comprehend.
  10. KamijoIsMyHero
    I only needed a fair amount of multi-driver IEMs to sample and conclude multiple drivers can be used to enhance performance. For someone who has 
    A technical explanation of driver count affecting performance would be useless if you yourself can't form a conclusion whether or not it can impact performance. You have had a lot of evidence, a statistical advantage compared to my experience in IEMs, at hand to answer your own question but it seems you can't comprehend such evidence. It would be pointless for an expert on the subject to explain further details to you.

    As I have said before this is the worst place to ask such question. People that know how to implement multi-BA IEM wont bother with this forum. Go ahead and wait for one though, I merely offered an answer from my experience and you called it unhelpful since it was anecdotal. Then you post this bit
    If Moulton himself didn't give you any objective evidence to support that statement, then you are a hypocrite for telling me anecdotal claims are unhelpful. If you don't trust 10K cable performance claims from manufacturer due to lack objective evidence, then you shouldn't trust Moulton's either.
    Of course, I would trust his claims myself but for different, and less hypocritical, reasons than yours.

    You may but **** that dude.
  11. dazzerfong
    He's asking a genuine question, and all you're doing is saying 'yeah it's better coz I said so.' He's asking if so, why: if you don't know why, don't go all medieval on anyone who offers an alternative opinion. An anecdote isn't much of use, frankly when he's trying to understand why:
    uchihaitachi likes this.
  12. stv014
    The multiplication by 2 is there so that the RMS level calculated for a 0 dBFS sine (rather than square) wave will be 0 dBFS. Without that, it would be -3.01 dBFS.
    In other words, if the entire input is a simple sine wave at a constant level, the calculated dynamic range will be 0 dB. But it will be 0 dB even if only 20% or more of the input is a constant sine wave, and the rest of it is silence. Therefore, it is a pessimistic algorithm, because it will rate the track as having poor dynamic range if any substantial part of it looks like being heavily compressed. It also ignores a high peak in a single block due to using only the second highest peak value. If you take 5 minutes of high quality classical music, and paste a couple minutes of compressed metal or pop at the end of it, then the reported dynamic range will be based mostly on the latter, where the 20% highest RMS values are likely to be found.
    The RMS sum is not based on the dB values, but rather on the power, which is the correct way to combine the RMS level for multiple blocks. That is, the correct overall RMS for the 12 blocks is:
    10 * log10((10^(-5 / 10) * 5 + 10^(-5.3 / 10) * 5 + 10^(-5.8 / 10) * 2) / 12) = -5.25 dB
  13. uchihaitachi

  14. KamijoIsMyHero

    You can't get medieval on an internet forum...it isn't like I am trying to convince him or that my answer is a end all, be all. I just offered a genuine answer as I am skeptical that multi-drivers can't offer a better performance than a single dynamic driver.
  15. uchihaitachi

    Nobody said that multi drivers don't offer a better performance. I queried whether it was a gimmick and if not how it could be superior in an objective sense. From an engineer's perspective. Yet you kept on insisting it was superior by ear which is essentially completely irrelevant to what I was querying about. Then you proceeded to brand me a hypocrite, and insinuated incompetence, as I should know better from my hearing experiences and shouldn't bother asking.
97 98 99 100 101 102 103 104 105 106
108 109 110 111 112

Share This Page