A Study on Hilt Sound Resonance

[Arkhan]

Thanks for posting the screenshots of the meter output, btw.

It gives a very good indicator of where the audio in most "saber idle hum" is present.

Below 250, not much
Above 4K, not much

Given that the small speaker has little problem producing higher freq sounds, I think we'd be safe assuming that any difficiencies in volume at higher freq are in the sound itself.

For the low range, if we output'd saber sounds into a real speaker with a nice 20-20Khz response range, and ran the test meter, we could see what portion of the limited low end freq response is due to the small speaker, and what portion is due to the actual sound samples themselves.

The samples may well have been run through compressor/limiter prior to committing to asic, to make most use of limit sample resolution and sample rate.

Hard to say what the sample resolution is (6-bit?) but it would seem that its at most, an 8K sample rate based on nyquist limit.

6-bit, 8K, Mono doesn't take up very much space on the asic, for sure.

tightwads could have given us better sound for two pennies more.

[Sunrider]

One thing to keep in mind is the meter is using an (A) weighting filter. This means the frequencies in the middle. (250hz to 2khz) are weighted heavier than frequencies at the high and low end. So the 1db change at 125hz is in reality more than only 1db.

[pointoforigin]

Speaker size and placement are always an issue in the small space we have in a hilt, we all know this. By using a "correct" size chamber we get more of the frequencies that we want to hear, better.

But!

Even CFv5 only outputs sound at 16 bit 22050 hz. For those who understand about sample rates, that makes for... not brilliant audio. Workable, but not brilliant. Not to make Erv' sound bad, cos it's obvious he's not, and he's keeping file sizes down. Just saying that if even the BEST board out there isn't even CD quality then...

For those who don't understand much about digital sample rates the basic rule of thumb is that to reproduce a given frequency, you must have a sample rate double that given frequency. So with the given maximum sample rate of 22050 hz for the CFv5, the maximum useable frequency is 11.025khz. This is not to say that it won't reproduce sounds higher than that, it will, they will however be mis-represented. They won't sound "right". This is because a waveform needs two samples - a compression and a rarefaction - to make the wave be erm... wavy

Bitrate plays a part too - bitrate is basically a measure of how loud one of those samples is. A lower bitrate will mean that lower volume sounds - such as idle hum - are also misrepresented. If we had say a 63dB sound, but the bitrate meant that it measured in lots of 10dB, it'd turn into 60dB. Not a brilliant example, but if you turned that into a 3db sound, it'd get knocked down to 0. 16 bit is CD quality bitrate. If there's boards running 6 bit then... oh God... I may have to burn them, being an audiophile...

There's more to go on about here, but it'd get waaay to technical for the layman lol!

I think the speaker, in our situation, has less to do with the quality of the sound than the digital sample and playback rate of the sound.

Arkan makes a good point about testing in this regard.

That said, getting the speaker chamber right always makes it better

And using a passive crossover? Absolutely brilliant!

I do believe I will steal this idea... *Yoink*

[Novastar]

Nova has mentioned that he reduces the sample level below 80hz. Maybe he could explain more. I'm guessing a two speaker setup could candle this frequency range better than the single speaker.

I generally remove or drop frequencies from my sound samples that are ~60Hz or ~80Hz and below and what not.

Sometimes, this affects the waveform and I can then boost the signals we're actually going to HEAR.

For example, some samples with tons of bass sound great on a computer or whatever--a GREAT speaker setup. Put it into CF and... OF COURSE the bass disappears, but what you can find is that the 60Hz and below DOMINATED the waveform... so I remove it, since it's futile to have it there.

My opinion, though I guess.

[Jordandau]

Removing the that stuff leaves you with more headroom too. Do you compress the signals at all? Also boosting a multiple of a frequency can help make it seem like you still have that info. Example: Since you roll off 60hz, try boosting it's multiples 120, 180, 240 etc.

[erv]

@sunrider : thanks again for running those tests ! excellent work !

@pointoforigin : the "rule of the thumb" you're mentioning about sampling rate is the Nyquist-Shannon sampling theorem (I think it's important enough so that the men are named). Your explanation slightly confuses readers IMO since for the reconstruction phase (digital to analog), everything is dictated by the medium, ie the sound file in the case of CF. And the sample rate of course. So there's not real need of talking of the mis representation of high frequencies since they aren't in the digital representation of sound.
Now for the "my bad" time, I'll confess CF doesn't have any reconstruction filter (cardinal sinusoïd filter) since a low pass filter is doing the job in the compromise of price/board size/sound quality. Who know's I might make a version with a version with a CODEC chip featuring a proper reconstruction filter but I doubt it will add actual quality if not listened to with headphones.
The lack of a reconstruction filter affects a bit the high frequencies though, but there's no spectral aliasing since the media, when resampled to 22k got a anti aliasing filtering applied prior to resampling (at least that what my resampler in sound forge does).

http://en.wikipedia.org/wiki/Nyquist...mpling_theorem

The bitrate explanation above (no offense meant) is wrong. Bit rate is on top of everything a measurement of data rate on a said medium. On a MP3 decoder, it indicates the amount of data that feeds the decoder, and when the same file is streamed it gives you a good idea of what your internet connection must have to receive the streaming properly. So bitrate isn't linked with sound or loudness in general.
In an uncompressed sound file (like WAV), and for CF, it's 22050 samples per second, and each sample is on 16 bits each, so bitrate is 44.1 kbps.
Since you're talking about dB in your bitrate explanation, I think I understand what you're trying to say.
The other aspect (after sampling rate/frequency) is the 16 bit thing. In digital audio we usually call it the dynamic range and it's expressed in dB but it has nothing to see with "loudness". The resolution of the samples defines the quantification of the samples. When looking to the direct output of the DAC (or the waveform in an audio editor), a high resolution looks almost like a continuous waveform, while a 8 bit sample (used buy most of the "other" boards ) look like a stairway, we then say it's a coarse quantification. It's like comparing a 1MP and a 6PM picture taken with a digital camera and zooming 2:1 on both.

Dynamic range in audio mastering / editing / mixing is a good indicator of the potential "expressiveness" of a device or media, ie how detailed the NUANCES can be. Recording a song with ppp to FFF (like a pizz) requires that good dynamic range, otherwise you'll adjust your gain in such a way that you get one part of the scale, and not the other.

Dynamic range can be INDEED expressed in dB. For 16 bit :
Range = 20 log(2^16) = 96 dB

16 represents neither a bit rate nor the "CD" quality. CD quality is name after the combination of a sample rate that allows to represent frequencies high enough for what our hear can recognize (22050 Hz max th.) AND a quantification that ensure the quantification noise is low enough not to be detected (16 bits). There's no rate in there, just to be accurate.

Now, back on topic : the lower boundary of the speaker we provide (here at TCSS, or mines, same thing) is said to be 300 Hz. They go (they say) up to 20Khz (sorry of course, no actual freq response graph lol). But having the output waveform going to half of that (11k), it makes a pretty nice device. Now, that micro tweeter is just perfect to help with the high frequencies. Next step is having the freq range really split up to separate the use of the 2 speakers, with a filter on each (can be passive ones due to the low power involved).

thanks again for this topic sunrider !

[pointoforigin]

I am but a student, and Erv' is the master. lol! Thanks Erv, I'm not always good at putting what I mean into words. I think I may have used the wrong word in my explanation - "bitrate" is indeed how many "bits" per second - I was meaning how many "steps" in the staircase each individual sample has, which IS governed by 16 bit, 8 bit, whatever, in the file type. I also should have specified dB FULL SCALE rather than leaving it at dB, because full scale is the digital measurement, in the digital environment.

Yes, full credit to Mr. Nyquist and Mr Shannon.

If that anti aliasing algorithm in your software is good, then hell, there's another reason why CF is awesome

Pretty sure those speakers will produce lower frequencies, just relative to the rest of the frequencies they produce it probably isn't very loud. Taking Erv's figures of 300hz - 20khz, that's probably at either -3dB or -6dB from normal. A frequency response chart would be great
Which brings us back to Sunrider's original topic - the mounting of the speaker and the chamber that it's in can make up for the low volume that we perceive at those frequencies. Which is awesome.

[erv]

no worries, and thanks for the kind words about CF ! after searching in the arcane of my brain for the proper english word you were referring to the bit depth of the sample, which is better describing the sample properties.
No will to be called Herr Professor though,
but since that thread is about digital audio general knowledge, I thought it was important to detail all that !

[Sunrider]

Thanks guy's

Now that I have a good platform I will do more testing when I get a good sound board. I think I can still improve the response with different size caps but will have to wait to see.

[Matt Thorn]

Sorry to chime in a month late.

Wow. Just, "Wow." Sunrider, you have really been pushing the envelope and raising the bar (sorry 'bout the mixed metaphors) on so many technical aspects of sabersmithing. My thanks. And a question.

A 4.7uf ceramic capacitor is soldered on the positive leg of the bigger speaker then connected to the tweeter. This is needed to only pass higher frequencies to the smaller speaker. Making it only produce higher frequencies. I used a surface mount 0805 size part .

The capacitor is "then connected to the tweeter"...where, exactly? A simple wiring diagram would be much appreciated. When you say "surface mount 0805", do you mean something that looks like this?

Thanks!