This is a follow-up post on the topic of HD Audio, the first being posted on June 3, 2014.
HD Audio, for the purposes of these posts, is defined as higher than CD bit depth and bit rate, and no compression is assumed. So anything higher than 16/44 (16 bits, 44.1KHz sampling rate) qualifies, and it’s a “higher the better” game. The typical HD Audio file you can download is 24/96, but there are some as high as 24/192. The file formats are .wav, .aif, ALAC/Apple Lossless, and .flac.
How do you get to play those high res audio files now that you have one? Recall that the whole point of HD Audio is unencumbered frequency response up into the ultrasonic range, low noise floor, and lower distortion.
Fundamentally the goal, then, is to get a high-rate audio file, say 24/96, to play in your system without degradation, without re-sampling that data to any other rate or bit depth, and get the resulting audio all the way to your ears without any additional limits to high frequency response, noise floor, or distortion.
Taking parts of the signal chain one at a time, speakers are mostly analog, passive devices, so there’s not much of a noise issue here with the speaker itself. Not too many speakers have much response up into the 48KHz range, or even the high 20KHz range, and some that do aren’t even spec’ed into that range, so it’s a bit of a question, and quite hard to verify. Sure, you can drive a 40KHz test tone into your speakers at low levels and try to measure it, but even my Earthworks M30 measurement mic is only guaranteed flat to 30KHz. I’m sure there’s some output at 40KHz, but not sure about how much roll off there is, so if I were to measure something at 40KHz, or 48KHz, I have little idea of what level (or loss) is the speaker or microphone. Pretty tricky stuff up there in the ultrasonics, and that energy is very directional, what with a wavelength at 48KHz being .278″, which makes anything measuring .07″ a significant acoustic element. The noise floor for speakers is really determined by the driving electronics and the acoustic noise in the room, and while no power amp has a true 24 bit dynamic range, most falling 30dB to 40dB short of it, it’s not really the limiting factor here. There is no room anywhere that is quiet enough for a real 24bit dynamic range (144dB), which would literally extend from the threshold of pain to the threshold of hearing. A listening room will have a noise floor around NC20, if it’s quiet. NC figures include weighted curves that compensate for progressively reduced low frequency hearing ability, and thus an NC20 room can have real noise above 50dB SPL at 63Hz. NC figures make room noise look good. Fortunately, music doesn’t actually require that kind of dynamic range, and any music recording will fit (mostly) within the bounds of a decently quiet listening room with a system that has the ability to reproduce somewhere around 105dB SPL. And that’s a lot of them. But none come even close to 24 bits of dynamic range capability. In reality, rooms dynamic range is more like 14 bits, if you consider a maximum SPL of 105dB and a noise floor at 20dB SPL.
One more speaker aspect to be aware of. They generate the highest amount of distortion of any element in the system. It’s unusual to find speakers that perform with under 1% THD at all levels and frequencies, which sort of disqualifies them as a pure-pass element for 24 bit recordings, or frankly 16 bit recordings. But they are what they are, and we have to have them.
The amps driving the speakers can be separates or an AVR, both can work, and the demands are no greater for high-res audio files than anything else, especially considering the room and speakers. A typical amp has an optimistic 18 bit dynamic range and having an amp that passes 48KHz well isn’t a big deal unless it’s some form of digital amplifier.
So far we’re limited to 18/96 by the amp, 14/96 by the room and then there’s the speaker with an unknown top end, and fairly high distortion.
Now it gets interesting. If you have a home theater system, you’re fortunate enough to have 5.1 channels. Most HD audio is two channel stereo, but there is are a growing number of surround versions. We’ll come back to this, but for now, let’s say you have a two-channel stereo file. That’s fine, we’ll just play it in two-channel stereo.
Next challenge is to get those HD audio files turned back into analog audio, and to do that you’ll need a DAC (Digital to Analog Converter). A stereo DAC capable of playing those file can be had for anywhere from $149 to $3500 and higher. They usually take the form of a device that you connect to a computer and play out through USB. They’re fine but if you’re already the owner of a recent home theater AVR you might already have what you need. Many products, for example AVRs from Denon, make their on-board DACs available to play audio files either streamed to them via a network or DLNA server, or played directly from a USB memory stick. Some units even can play 24/192 files.
There are lots of ways to get HD audio out of your computer and over your network to your AVR including various player software, NAS devices, etc. We’ll leave the somewhat weird area of which playback method sounds the best to those that obsess about that sort of thing for now.
If your AVR has room EQ or auto calibration, we have a little conundrum here. We have an HD file, at 24/96, but many AVRs run room correction/auto-cal at 24/48. There’s a reason for this, actually several. First, it takes quite a few processor clock cycles to generate the calibration filters for all 6 or 8 channels, and if we double the sampling frequency, we double the processor overhead required to do that. So to keep at least some DSP available for other things, like surround decoding, etc., AVRs usually process for auto-cal at 24/48. There’s also a bigger question. If we did let an auto cal system work at 96KHz, what would it do above 20KHz? There’s no directly audible information there, so what would the filters do? My guess is that we’d just want them to pass ultrasonics, flat, but there’s an opportunity here to equalize the ultrasonically anemic speaker. The problem is, equalize to what target curve? Lots of questions, no answers, so the folks at Audyssey, as one example, just don’t do it.
There are, however, room EQ systems that do operate upward of 48KHz. The miniDSP line is first to mind, with processors that have an internal 24/96 or better structure. The devices themselves are economical, but using them is not. If you have an AVR, there’s no way to insert them into your signal chain, so you’ll be adding external power amps if your AVR has pre-amp outputs. And, setup is a very manual process for which you’ll need computer software and a measurement mic, and a boatload of patience. Not to say the results aren’t good, it’s just a rougher road to get there. Most won’t adopt external DSP just to get room correction up to 24/96.
That means you have a tough choice to make: Either you hear your file without modification all the way to your speakers and give up room calibration, or you keep your room calibration (which is definitely audible in frequencies bands that include the original music) and accept the down-sampled version of your track. This is just the kind of choice that will drive some people nuts. They really want to hear all 24 bits at 96KHz, so they choose killing their AVR’s auto-cal with a “Pure-Direct” mode. That’s fine, if your speakers and room are pretty good to begin with. The way to check is to compare, as best you can, what your system sounds like with and without calibration. If there’s a radical difference, and auto-cal improves things a lot, your choice has just been made: leave the auto-cal in, and take the down-sampling that goes along with it, you’ll have the better sound quality, and a better experience. If, however, auto-cal makes less difference, or very little, then you have a very good room and excellent speakers indeed, and could effectively live without calibration. You can go Pure-Direct and play away all the way to 24/96 (ignoring the above mentioned speaker, room, and amplifier limits).
Just so we understand, if you did decide to get one of those glossy $3500 DACs and plug it into your AVR, what you’ll have done is been able to play HD audio files at full resolution, convert to analog, shove it all into the AVR which re-digitizes the signal at 24/48, and goes to work processing it. I question the efficacy of such a rig. More logically, you’d want that DAC plugged into a similarly glossy and expensive preamp and power amp combination. Unfortunately, this grouping will probably limit you to 2-channel stereo. It’s unfortunate because the difference between 2-channel stereo and 5.1 channel surround is unmistakable, whereas the improvement in sound quality from HD Audio isn’t nearly so obvious to every listener.
Perhaps the best way to get unadulterated 24/96 to your eardrums is via a dedicated DAC and headphones. You still have the problem of knowing the total response of the system, but your chances of getting all that low noise, low distortion, ultrasonic energy to your ears without changing it much are higher when you don’t have speakers, rooms and room correction to deal with. Not to say that headphone EQ is invalid, it’s actually fantastic, but you end up with the same problems. The best of Headphone EQ is not likely to pass ultrasonics either. So take your pick. Great headphones with low distortion and ultrasonic frequency response are available, some are not terribly expensive. And the even mildly exotic USB DACs have capable headphone amps built in.
We are left to ponder this question: With the limited dynamic range, speaker distortion and limited high frequency response of my system, will I hear better sound with HD Audio files? That is the big question. Some will answer absolutely in the affirmative, some will be more undecided. You won’t know until you try some. If you’ve read my previous post on HD Audio, you’re a big step closer to getting some real HD Audio material to try rather than some up-sampled or re-digitized analog material encapsulated into an HD Audio file. Those that original true HD Audio material are driven by a philosophy that pushes them toward high quality recordings, done carefully and with excellent equipment. Those recordings would be good regardless, but it seems that the motivation of producing some real native HD Audio material does push the recording engineer to do his best, and there’s no debating the audible benefit of that. Perhaps that’s the first step to appreciating high resolution audio.
This just in: If you read my Denon Teaser post, you know there are new AVRs about to be released, and we have high hopes for some very cool new features. Among those rumored are 32 bit, 192KHz DACs and internal processing. There simply are no details to be had, no idea if its true…or why anyone would want 32 bit DACs (that’s a theoretical 192dB dynamic range…which if you could reproduce it, would a dynamic range from the threshold of hearing to the clipping point of air, where the rarifaction peak becomes a total vacuum, and hearing is permanently destroyed in a millisecond). There may be a point to 32 bits for internal DSP functions…but who knows. As a Denon dealer, I’m watching this closely and will confirm or debunk the rumor as soon as I have hard data.