Here are some interesting test files ...

[Talos2000]

... I got them from BitPerfect. They are computer-generated square waves which produce a beeping 688Hz tone by toggling one bit on and off while leaving the other bits all off. I'm not sure what broad use they might have, but I found it interesting that I could see which individual bits produced an audible sound and which did not. On my system (see signature line below) I could hear Bit 20 quite clearly, but Bit 21 was totally inaudible.

 

Anyway, I thought they might be of interest....

 

(This is the first time I have tried the attachments tool here, so apologies if I screwed something up)

Bit Test Tracks.zip

[ssgp2]

Bit 15 is my threshold, curious about what it means....

[Talos2000]

Bit 15 is my threshold, curious about what it means....

 

I'd say it means your system's noise level is less than 96dB below your DAC's maximum output. That would suggest very poor performance. For example, my $50 PC speakers driven by the motherboard's built-in sound card also has a Bit 15 limit.

 

Your system signature does not suggest a system having such feeble performance, so I would look elsewhere. Is it possible that you have dialled in a large dose of digital volume control? Every 6dB of attenuation will lose you one bit of resolution.

 

When playing the quietest tracks, you have to crank every volume control - especially the digital ones - up to max.

[ssgp2]

Ok, I checked my speakers and the input gain are adjusted 1 notch below the 0 gain(3 notch to -10db)

I'll try adjusting those and see what happens.

[ssgp2]

Cranked up all gains and could barely hear the bit 19.

[Talos2000]

Cranked up all gains and could barely hear the bit 19.

 

OK. That's quite a bit better. Amongst other things, it says that IN PRINCIPLE your system should be able to resolve some differences between (well-recorded) 16-bit and 24-bit music.

 

Beyond that, I'd want to see more data from a wider range of users before drawing any more elaborate conclusions.

 

These files have initiated some ideas in my addled skull, but the thoughts are still grinding around. I'll post them when they distill into something that might be defensible.

[cjf]

Hrrmm interesting, I cant seem to hear any of them. Does this mean I'm dead and no has told me yet? Maybe those long nights at 105db finally caught up to me?

 

Tested them on my secondary system in my office using Media Center 17 playing thru Benchmark DAC1 USB, line out to old school Adcom 535 amp then out to Infinity Outrigger speakers.

 

Maybe I'll try my luck again on the main system this weekend

[crisnee]

Hrrmm interesting, I cant seem to hear any of them. Does this mean I'm dead and no has told me yet? Maybe those long nights at 105db finally caught up to me?

 

Tested them on my secondary system in my office using Media Center 17 playing thru Benchmark DAC1 USB, line out to old school Adcom 535 amp then out to Infinity Outrigger speakers.

 

Maybe I'll try my luck again on the main system this weekend

 

You might be dead, but don't take it too hard, cause you write a mean post for a dead guy, or it could have something to do with settings in your player.

 

-Chris

[crisnee]

If I put my ear near the speaker I could hear bit 22 beeping away; that's with a MF V-Dac and a Hiface.

 

-Chris

[crisnee]

I decided to send my results in to Richard from BitPerfect as he asks for them in the included pdf, and I'm curious if he has any comments about my unexpected results.

 

The pdf states that the vast majority of hi-fi's would not sound much beyond 16 bits and that one would not likely hear 21 bits even in the best systems, never mind 22. So I'm curious as to his comments/explanation, as although I have good equipment it is by no means the elite stuff.

 

-Chris

[jhwalker]

I can hear bit 20 coming from my Lenovo laptop via DragonFly to FoxL speakers or headphones. Bit 21 is inaudible. Will be interested to see same / worse / better from my home system when I return from my travels.

[Talos2000]

Interesting results - especially fron crisnee. Not sure how to interpret that. Hearing Bit 22 would seem to suggest a system with stunning performance that can resolve -132dB. That seems unlikely, but he's hearing it so there must be an explanation somewhere!

[firedog]

My system makes a small amount of background noise - you only hear it if you put your ear close to the speaker. I think almost all the noise is from the power amp.

 

At normal volumes I can hear the 14 bit tone from my listening position. At very loud volumes (what I listen to a symphonic work at when I want it to sound "real") I can hear the 15 bit playback at my listening position.

 

If I turn the volume up about halfway between that and max position (louder than I would ever listen), I can hear 16 and 17 (just barely) at my listening position.

 

If I totally max out the volume, I can hear 18-19 from my listening position. 19 is very faint.

 

At max volume, I can hear 20 if I get close to the speaker, standing a few inches away. I can hear 21 just as a distant very faint barely perceptible beep if I put my ear an inch from the speaker.

 

I'm pretty encouraged by this. My DAC claims SNR of 128db, so apparently that's not a lie.

 

I'm about to get a new amp which should be much quieter than my present one. It will be interesting to see if I can hear more or more easily with the new amp.

[ssgp2]

My result were all at the max volume of the DAC.

SNR from the DAC2 manual >115

Test was conducted at the listening position, 90" from the speakers

Room was measured at 40db with an IPad Apps(decibel ultra).

 

To recap;

 

- first test was conducted with the speaker gain at -3db = bit 15 was heard, (might of heard 16 but not clearly)

 

- second test was conducted with speaker gain at +10db(max) = bit 19 was barely heard

[sjoc2000]

Up to 20 bit can hear dit dit dit..., at 21 can still hear but sound changes to di dah di dah...

 

22 and nothing. Of course the test from 19-21 require that I have my ear approx. 6-8 inches from speaker.

 

This report is guaranteed to be from a non-dead person.

 

Jim

[goldsdad]

... I got them from BitPerfect. They are computer-generated square waves which produce a beeping 688Hz tone by toggling one bit on and off while leaving the other bits all off. I'm not sure what broad use they might have, but I found it interesting that I could see which individual bits produced an audible sound and which did not. On my system (see signature line below) I could hear Bit 20 quite clearly, but Bit 21 was totally inaudible.

 

Anyway, I thought they might be of interest....

 

(This is the first time I have tried the attachments tool here, so apologies if I screwed something up)

 

 

My prediction is that someone at some point is going to have their system set up so that system sounds go through the same output as music, either through ignorance or by accident. When they're testing, for example, bit 20 with the volume cranked to max and their ear about an inch from the speaker, a system sound is going to destroy the speakers. The lost speakers won't matter, though, because the person is going to be deaf or even dead from shock.

[crisnee]

My prediction is that someone at some point is going to have their system set up so that system sounds go through the same output as music, either through ignorance or by accident. When they're testing, for example, bit 20 with the volume cranked to max and their ear about an inch from the speaker, a system sound is going to destroy the speakers. The lost speakers won't matter, though, because the person is going to be deaf or even dead from shock.

 

I bet it's happened already (not with this test) and the victim is now looking for a mono only elitist equivalent of CA, if he/she survived the initial shock. I know I was pretty nervous doing the test; I didn't get close to the speaker until the sound was going, poised with mute button at the ready.

 

-Chris

[RobbieC]

I can get to 18 or 19 testing several headphones on a MacBook Pro using Audirvana+ via DragonFly. Interestingly, some headphones produce a very low level tone with the DragonFly (at a frequency less than the 688 Hz tone used for the test) while others do not. The tone is only audible while (and after) playing a test track and then only with the volume turned all the way up.

 

I can hear what amounts to a very low, low level white noise with my best headphones as the volume is turned all the way up, but not the unwanted tone. I've thought about it and have concluded that the fault lies in the particular affected headphones and is not a problem with the DragonFly I have. Could be inferior shielding in those headphones (which incidentally happen to be the least and second least expensive ones I have).

 

The noise canceling headphones I have had the least performance with bit 18 the last audible while my best pair made the 19th bit audible. Think I'll spring for some Shure 1840's; I think there's more performance to be had from this little DragonFly.

 

Btw, I left repeat selected when I was doing the 16 bit test and turned the volume up pretty loud toward the end of the test. When digital silence was playing I was listening very intently for anything in the silence and it wasn't until the middle of the track that I realized that the next track up would be the very loud -3 dBFS one. So I started to lower the volume...but of course it doesn't lower instantaneously and I didn't make it. Very disconcerting in headphones! I'm lucky I got it down some because it is very loud through the DragonFly's amplifier at full volume.

[BitPerfect Support]

I see these test tracks are spreading

 

I really must emphasize that these were created for the specific task of providing audible test tracks that produce sounds by energizing one specific bit – and that bit only – in a true PCM ladder DAC. At BitPerfect we have a large and growing collection of reference DACs supplied by some generous (and in some cases quite extraordinarily generous) manufacturers. Mostly we use these to ensure that BitPerfect works as intended with the widest variety of DACs, but sometimes we get other requests…

 

The files are NOT designed to be general purpose audibility test tracks for consumer use. As some posters have pointed out, you can cause a lot of damage both to equipment and ears. Please think very carefully about what you are doing, particularly when you crank up the juice to play the very low-level tracks. Our STRONG recommendation is to place these tracks individually into a playlist of one, and play them from there.

 

Just a few comments on these files, and how you might interpret them. But in order to do that I need to clarify a few things about PCM DACs. And in order to that I need you to know about Delta-Sigma modulation. This is very long-winded, so feel free to skip the next three paragraphs to get to the meat.

 

Delta-Sigma Modulation (DSM) is an alternative format to PCM for encoding analog signals in the digital domain. Those of you whom are familiar with SACD will know that this format uses a system called “Direct Stream Digital” or DSD. DSD is a specific implementation of DSM. Whereas PCM seeks to represent the exact value of the analog waveform at a very specific instant in time, DSM seeks to represent the analog waveform by a massively high-speed stream of single bits which are either on or off. The larger the amplitude of the analog waveform being sampled, the higher the percentage of bits that get turned “on”. PCM is much easier for the layman to understand. We understand what the digital data represents, and we understand what the analog filters have to do after the digital data stream is converted to raw analog output. It is quite easy to sound knowledgeable when we talk about it. It is quite easy to argue a case for why, for example, 24/96 might sound better than 16/44. DSM is way different. It is much harder to describe what the data stream represents, where in the data stream the music lives, what the resolution is and how we can specify it, even what each individual bit brings to the party. It is much harder to describe what the role of the analog filter is, and how it accomplishes the magic which it brings to bear. But this isn’t the place to expand on this, and I’m not the person to do it

 

Why does anyone bother with DSM, then? Two reasons. First, is that DSM can – in principle – encode the analog waveform with a much lower signal-to-noise ratio than a PCM encoding with the same data density. Second, is that DSM is far, far easier to implement than PCM, both on the ADC and DAC sides of the coin. The implication of this second point is far-reaching. Today, the huge majority of ADCs and DACs work in the DSM domain, and use digital format conversion to convert between DSM and PCM. So, the ADCs do the actual analog-to-digital conversion in the DSM format and then convert the DSM data stream to PCM. Likewise, the DACs convert the PCM data stream to DSM, and do the actual digital-to-analog conversion in the DSM regime.

 

None of this would be worth going to all this trouble to explain if it was not for one significant issue. You see, DSM and PCM are so alien, one to the other, that although they are both digital representations of music, there is no direct bit-for-bit correspondence between one format and the other. There is no mathematical correspondence. An algorithm does not exist that can convert back and forth between PCM and DSM and always get the exact original back. In other words, the conversion is lossy. This is not surprising, for example there is information in the DSM data stream that cannot be encoded in the PCM data stream, and vice versa (for example, the SNR that can be captured in PCM is independent of frequency, whereas the SNR that can be captured in DSM rises with frequency). Furthermore, since the algorithms that convert between DSM and PCM in both ADC and DAC applications must perform their conversions in real-time, they are bound to employ algorithms known by the term FIR (Finite Impluse Response) which most listeners believe has a negative impact on the sound quality compared to those known by the term IIR (Infinite Impulse Response) which can really only be employed for off-line format conversion.

 

So, back to the test files. Let’s look at the Bit-16 test file which is encoded in 16/44.1 format. What it encodes is, in musical terms, a 688Hz square wave with an amplitude of -96dB and a DC bias a smidgeon short of –V_max. That is because its role in life is to exercise Bit 16 only. If, in musical terms, a signal was wanted that produced a 688Hz square wave at -96dB you would omit the massive DC bias. But that would require a different bit pattern. And I can produce that if there is a demand for it. For most listeners, i.e. those who do not use a true PCM ladder DAC, this signal has to go through a PCM-DSM conversion process before the actual DAC conversion. Each DAC manufacturer has their own proprietary conversion algorithms, and you cannot be certain how those conversion algorithms handle the specific case of a minimum-amplitude signal with a maximum-amplitude DC bias. In other words, would the resultant encoded DSM signal still be encoded at -96dB?

 

Another question to consider is the music playback software. The ubiquitous iTunes, for example has a “Sound Check” feature which attempts to “equalize” the playback volume so that all tracks have the same perceived record level. That was initially implemented as a simple per-track volume adjustment. But the latest versions of iTunes implement a much more elaborate dynamic range reduction system which could have the effect of elevating the -96dB signal in the digital domain by several dBs. So if you use iTunes, you should be sure to turn off “Sound Check”. Other media players may have similar features.

 

I could go on to talk about the effect of the separate SNRs of the DAC, pre-amp, and power amp, but this is already way to much.

 

P.S. Apologies to those of you who have sent us e-mail. Our e-mail service is proving to be unreliable, and we are going to have to do something about it. Unfortunately, it may take some time … .

[JonP]

Listening on my home-built DAW and headphones (soundcard is Xonar ST), I was able to hear bit 23 quite clearly with the volume settings maximised. I could also hear bit 24 but only if I also turned the gain of the equaliser to maximum as well - so not just master volume and mixer volume maximised, but all three. Nevertheless, the sound of bit 24 was indeed coming through on the headphones and I could (just) hear it.

 

This result needs to be qualified by stating the obvious - at those volume settings I would have caused myself significant hearing damage almost instantly if anything had come through the headphones apart from the file being tested, since the gain would have to have been at an insane setting (more than 144dB since I could hear the tones).

 

On my 16 bit loungeroom system (which has a Chordette Gem 16 bit DAC), I was able to hear bit 19 - this of course was my dithered version of the 24 bit, bit 19 file - the dithering used was iZotope. At bit 20, all I could hear was the noise aded by the dithering algorythm - the tones were completely inaudible.

 

I actually think this is a danagerous test in one respect. Unless you are 100% certain nothing is going to come through your system apart from the test tones, both the equipment and one's hearing could be damaged.

 

Nevertheless, I found the test interesting because although I can easily hear the difference between 24 bit files and their 16 bit equivalents, I had no idea how far down those last bits really are. It just goes to show how what goes on far, far below the threshold of audibility has an effect on what we can actually hear. At my normal listening levels which are quite low, I would never have my volume control high enough to hear bit 13 or 14 on this test, and even in that situation it would be on account of a very loud peak produced by a large orchestra.

[kumakuma]

I actually think this is a danagerous test in one respect. Unless you are 100% certain nothing is going to come through your system apart from the test tones, both the equipment and one's hearing could be damaged.

 

I agree. I was going through them using my SBT and I pressed the next track button at the end of the last track which caused my system to go from the last track (16-bit digital silence) to the first track (Bit 1 -6db).

 

Won't be doing that again...

[DigiPete]

Genelec 8260a:

 

 

@ 2.1m/7':

 

14 bit (-84dB) OK

Volume @ -40dB - std. critical listening

 

20 bit (-120dB) OK

Volume @ -0dB - max.

 

 

@ 0.2m/8" to the side of the speaker - I dare not stand closer ;-)

 

22 bit (-132dB) OK

Volume @ -0dB - max.

 

So the rig is highly resolved and powerful, no surprise but interesting!

Also, anything beyond -100dB may be a waste of effort in my every day life, who knows?

Is this for real?

[Jud]

To BitPerfect support: Thank you, that was very informative, at least for someone at my not terribly sophisticated level of understanding.