Category Archives: Audio

Bits and Dynamic Range

When digital audio came out I wondered how the number of bits per sample correlated to the amplitude of waves. I imagined that the total expressible range was independent of the size of the smallest discernible gradation. Since this appeared to be a trade-off, I wondered how anyone decided what was a good balance.

Later I realized this is a false distinction. First: the number of bits per sample determines the size of the smallest gradation. Second: total expressible range is not a “thing” in the digital domain. Third: if the total range is a pie of arbitrary size, dynamic range is the number of slices. The smaller the slices, the bigger the dynamic range.

Regarding the first: to be more precise, bits per sample determines the size of the smallest amplitude gradation, as a fraction of full scale. Put differently: what % of full scale is the smallest amplitude gradation. But full scale is the amplitude of the analog wave, which is determined after D/A conversion, so it’s simply not part of the digital specification.

Amplitude swings back and forth. Half the bits are used for negative, the other for positive, values. Thus 16 bit audio gives 2^16 = 65,536 amplitudes, which is 32,768 for positive and negative each (actually one of the 65,536 values is zero, which leaves an odd number of values to cover the + and – amplitude swings, making them asymmetric by 1 value, which is a negligible difference). Measuring symmetrically from zero, we have 32,768 amplitudes in either direction. So the finest amplitude gradation is 1/32,768 of full scale in either direction, or 1/65,536 of peak-to-peak. 16-bit slices the amplitude pie into 65,536 equal pieces.

Here’s another way to think about this: the first bit gives you 2 values and each additional bit doubles the number of values. Amplitude is measured as voltage, and doubling the voltage is 6 dB. So each bit gives 6 dB of range, and 16 bits gives 96 dB of range. But this emphasizes the total range of amplitude, which can be misleading because what we’re really talking about is the size of the finest gradation.

So let’s follow this line of reasoning but think of it as halving, rather than doubling. We start with some arbitrary amplitude range (defined in the analog domain after the D/A conversion). It can be anything; you can suppose it’s 1 Volt but it doesn’t matter. The first digital bit halves it into 2 bins, and each additional bit doubles the number of bins, slicing each bin to half its size. Each of these halving operations shrinks the size of the bins by 6 dB. So 16 bits gives us a bin size 96 dB smaller than full scale. Put differently, twiddling the least significant bit creates noise 96 dB quieter than full scale.

To check our math, let’s work it backward. For any 2 voltages V1 and V2, the definition of voltage dB is:

20 * log(V1/V2) = dB

So 96 dB means for some ratio R,

20 * log R = 96

where R is the ratio of full scale to the smallest bin. This implies that

R = 10 ^ (96/20) = 63,096

That’s almost the 65,536 we expected. The reason it’s slightly off, is that doubling the voltage is not exactly 6 db. That’s just a convenient approximation. To be more precise:

20 * log 2 = 6.0206

So doubling (or halving) the voltage changes the level by 6.0206 dB. If we use this more precise figure, then 16 bits gives us 96.3296 dB of dynamic range. If we compute:

20 * log R = 96.3296

We get

R = 10 ^ (96.3296 / 20) = 65,536

When the math works, it’s always a nice sanity check.

Summary

The term dynamic range implies how “big” the signal can be. But it is both more precise and more intuitive to imagine the concept of dynamic range as the opposite: the size of the smallest amplitude gradation or “bin”, relative to full scale. Put differently: dynamic range is defined as the ratio of full scale, to the smallest amplitude bin.

With 16 bits, that smallest bin is 1/65,536 of full scale, which is 96 dB quieter. With 16-bit amplitudes, if you randomly wiggle the least significant bit, you create noise that is 96 dB below full scale.

With 24 bits, that smallest bin is 1/16,777,216 of full scale, which is 144 dB quieter. With 24-bit amplitudes, if you randomly wiggle the least significant bit, you create noise that is 144 dB below full scale.

Typically, the least significant bit is randomized with dither, so we get half a bit less dynamic range, so for 16-bit we get 93 dB and 24-bit we get 141 dB.

Practical Dynamic Range

Virtually nothing we record, from music to explosions, requires more than 93 dB of dynamic range, so why does anyone use 24-bit recording? With more bits, you slice the amplitude pie into a larger number of smaller pieces, which gives more fine-grained amplitude resolution–and, consequently, a larger range of amplitudes to play with. This can be useful during live recording, when you aren’t sure exactly how high peak levels will be. More bits gives you the freedom to set levels conservatively low, so peaks won’t overload, but without losing resolution.

However, once that recording is completed, you know what the peak level recorded was. You can up-shift the amplitude of the entire recording to set the peak level to 0 dB (or something close like -0.1 dB). So long as the recording had less than 93 dB of dynamic range, this transforms the recording to 16-bit without any loss of information (such as dynamic range compression).

In the extremely rare case that the recording had more than 93 dB of dynamic range, you can keep it in 24-bit, or you can apply a slight amount of dynamic range compression while in the 24-bit domain, to shrink it to 93 dB before transforming it. There are sound engineering reasons to use compression in this situation, even for purist audiophiles!

To put this into perspective: 93 dB of dynamic range is beyond what most people can pragmatically enjoy. Consider: a really quiet listening room has an ambient noise level around 30 dB SPL. If you listened to a recording with 93 dB of dynamic range, and you wanted to hear the quietest parts, the loud parts would would peak at 93 + 30 = 123 dB SPL. That is so loud as to be painful; the maximum safe exposure is only a couple of seconds. And whether your speakers or amplifier can do this at all, let alone without distortion, is a whole ‘nuther question. You’d have to apply some amount of dynamic range compression simply to make such a recording listenable.

High Bit Rate Audio

When CDs first came out in the 1980s they sounded lifeless. I still have several in my collection from those years and they still sound bad. In some ways they were better than LPs: no background rumble or hiss, much cleaner and tighter bass, uncolored midrange, and consistent sound quality unlike LPs that sound best in the outer groove with sound quality gradually deteriorating as the record plays and the needle moves toward the inner groove. At the end of the record, just when the orchestra is reaching is crescendo finale, you gear audible distortion or dynamic range compression because the inner groove can’t handle the dynamic range. CD avoided these issues, yet by “lifeless” I mean the high frequencies and transient response on CD were quite poor, much worse than LP.

Over the 1990s, CDs improved until around the year 2000 I found the best CDs had surpassed LPs. CD high frequency and transient response had vastly improved, plus CDs retained the other advantages they had all along. By this point, the best CDs of live acoustic music sounded more natural and real, where the best LPs sounded like an artistically euphonic sonic portrayal.

Looking back, the reason for this transformation of CD audio quality seems to be the use of poorly implemented anti-aliasing filters in the early days. Over the 1990s, we owe the improvement in CD quality mainly to digital oversampling and more transparent anti-aliasing filters, and partially to better implementations of dither and noise shaping.

At the same time, around the turn of the century, high bit rate formats came out: SACD and DVD-Audio. Various engineering and acoustic reasons were given for these high bit rates, most of which were based on well-intended yet fallacious understanding of digital audio, some on blatant pseudo-science.

The best explanation I’ve seen comes from a video by Monty Montgomery, and on his website, where he debunks the most common misunderstandings about digital audio. However, in his zeal to shed the light of math and engineering on this subject, he overstates the case in a few areas. Here I describe those areas. However, while I dispute these points, generally I do agree with him. He’s essentially got it right and is worth reading.

Audible Spectrum

Monty says, Thus, 20Hz – 20kHz is a generous range. It thoroughly covers the audible spectrum, an assertion backed by nearly a century of experimental data. This is mostly, yet not quite true. The range of human hearing is closer to 18 Hz to 18 kHz. It’s common for people to hear below 20 Hz, but almost nobody above the age of 15 can hear 20 kHz. For example, at age 50 as I write this, my personal hearing range is from around 16 Hz to 15 kHz.

Ironically, this actually strengthens Monty’s case. Digital audio has no problem going lower than 20 Hz, and we only need to go up to around 18 kHz to be transparent.

The Human Ear: Time vs. Frequency Domain

The ear is a strange device. Highly sensitive, yet also non-linear, and it can also be inconsistent and unreliable. Our keen perception of transient response is more sensitive than one would expect, given the upper threshold of frequency tones we can hear.

For example, consider castanets. They have lots of high frequency energy, to 20 kHz and above. If you listen to real castanets–not an audio recording, but an actual person snapping them in front of you–the “snap” or “click” has an incredibly crisp, yet light and clean sound. Most recordings of them sound artificial with smeared transients, because these recordings don’t capture those high frequencies well. They’re lost somewhere in the microphone, the position of the mic to the musician, or the audio processing.

I have an excellent CD recording of castanets (it’s a flute quintet, but several tracks feature castanet accompaniment) that has energy up to 20 kHz. It’s one of the best, most realistic castanet recordings I have heard: clean, crisp yet light. Almost perfect sounding. As a test, I’ve applied EQ to this recording to attenuate frequencies above 15 kHz. I can differentiate this from the original in an A/B/X test. In the filtered version, the castanets don’t sound as crisp or clean. It’s hard to describe, but they sound slightly “smeared” for lack of a better word. The effect is subtle, but consistently noticeable when you know what to listen for, and listen carefully.

Yet as mentioned above, I can’t hear frequencies above 15 kHz, so I can’t hear the frequencies I attenuated. How is that possible? I believe it’s because the ear can detect transient response timing that requires higher frequencies to resolve, than it can hear as tones. Put differently: take a musical signal of castanets (or anything else with very high frequencies) and apply a Fourier Transform to convert to the frequency domain. The highest frequencies you cannot hear as pure tones. But if you filter them out, it distorts the original waveform in the time domain, rounding off sharp transients and causing pre-echo. The ear can detect these artifacts.

The moral of this story: well-engineered digital audio does perfectly capture any analog signal that has been bandwidth-limited to the Nyquist frequency. But, some caveats apply:

  1. Bandwidth-limiting the signal can create audible distortion. Anti-alias filtering with a steep slope creates audible time domain distortion in the pass-band.
  2. Higher sampling rates (alternately, oversampling) give a wider transition band, making a gradual filter slope, reducing this pass-band distortion.
  3. The frequencies needed for transient response to sound transparent, may be higher than the frequencies that people can hear as pure tones.

Of course, these points are not unique to digital audio. To get transparent transient response, every step in the recording chain must preserve high frequencies. You must use microphones with extended high frequency response, position them close enough to the musicians to capture the frequencies, etc.

Lossy Compression

Monty says: a properly encoded Ogg file (or MP3, or AAC file) will be indistinguishable from the original at a moderate bitrate. Whether that is true depends on one’s definition of “moderate”. Trained listeners of high quality recordings on high quality equipment may need bit rates higher than Monty suggests.

A/B/X testing the highest quality recordings in my collection, I can reliably distinguish MP3 up to about 200 kbps rates, using LAME 3.99.5, which is one of the best encoders. Most MP3s are done at 128 to 160 kbps thus could be differentiated from the original.

There is some truth to the “moderate bitrates are sufficient” viewpoint. Most MP3s are of rock, pop or electronic music, inferior quality master recordings that are compressed, clipped, and heavily EQed. The low 128 to 160 kbps rates may be transparent for this content. But that’s not relevant to us; here we’re talking about high end.

In short, if you are a trained critical listener of high quality recordings on high quality equipment, you can hear the difference of MP3 and other lossy compression, unless you use the best modern encoders at higher than average bit rates (say 256 kbps minimum to be safe).

I’ve also got a few thoughts on dynamic range and 16 vs 24-bit. That’s a whole ‘nuther discussion.

Conclusion

What Monty says about digital audio is true, generally speaking. He’s done a great job of debunking common myths. High bit rate recordings are over-hyped and can actually be counterproductive. However, there are some caveats to keep in mind:

  1. High bit rate recordings often do sound better, because when they are being made, extra care and attention is used throughout the entire recording process.
    • But if you took that recording and down-sampled it to CD quality using properly implemented methods, it is likely to be indistinguishable from the original.
  2. High bit rate recordings may be sold as “studio masters”, not having dynamic range compression, equalization or other processing often applied to CDs.
    • This is related to (1), and the same comment applies.
  3. High bit rates can offer subtle improvements to transient response.
    • But not all do, because the limiting factor for transient response may be the microphones or other parts of the recording process.
  4. High bit rates can sound worse, because they may capture ultrasonic frequencies that increase intermodulation distortion.
  5. The differences that high bit rates make (improvement or detriment) are subtle and most people don’t have good enough equipment or recordings to hear the differences.

Parting Words

Engineers may want to record at higher sampling rates with more bit depth to give headroom for setting levels and other processing. But their final result can virtually always be transformed to 44-16 without any audible compromises (distortion, compression, or loss of information). Yet in some areas, 44-16 while sufficient, is barely sufficient, which means it requires careful well engineered re-sampling, anti-aliasing filters, noise-shaped dither, etc.

High bit rate recordings, when done carefully, can offer slightly better transient response for certain types of music. But to the extent they actually do achieve this by accurately capturing higher frequencies that improve transient response (which is rare), this HF content is a double-edged sword that brings the risk of higher IMD distortion. Of course, high quality well-engineered audio gear (DAC, amp, speakers, etc.) mitigates this risk.

Some practical guidelines:

  • If the original recording was made in the 1980s or earlier, there is no point to high bit rates. Ultra high frequencies are already non-existent or rolled off, transient response is already imperfect, dynamic range is already limited. Here, the 44-16 standard is higher fidelity than the original.
  • If it’s rock, pop, electronic, there’s probably no point to high bit rates. It’s already heavily processed and there is no absolute reference for what this kind of music is supposed to sound like. Classic rock/pop albums get re-released every few years with different re-masterings that all sound different. One version may have better bass or smoother mids, but that is not a 44-16 limitation. Which release is “best” is not a limitation of digital bit rate, but only a matter of opinion.
  • If it is acoustic music recorded in natural spaces, a high bit rate recording may be useful. Even if the bit rate alone doesn’t help things, the entire recording is probably (though not always) made with more careful attention to detail and high engineering standards.

Overall, I don’t worry about it. The quality of a music recording depends far more on the mics used, their placement, the room it was recorded in, mixing and mastering, than it does on the bit rate. And 44-16 is either completely transparent, or so close to transparent that even on the highest quality equipment with the most discerning listener, limitations in other areas of the recording process make the differences moot.

Back to the HD-580 – For a While

My Audeze LCD-2 fell off my desk at work and got pranged so they’re going back to Audeze for repair and, incidentally, upgrade to the 2016 drivers. My home pair hasĀ  these drivers and they are a subtle improvement over the 2014.

In the meantime, I’m listening to my trusty old HD-580s. Original 18 year old drivers, though I’ve replaced the headband and ear pads, and the cable, a few times over the years. They’re clean and play, fit and look like new.

First impression: these HD-580s are nice headphones! Smooth mids, nice timbres, well balanced. They really were the very first audiophile headphone, SOTA for 1999, a whole different league apart from Grados and the like. But compared to the Audeze:

  • The low bass is rolled off
  • The bass is not as tight
  • The mids are a tad boxy, not as open sounding
  • The high treble is rolled off

Overall, they sound a tad muffled and slow compared to the LCD-2. Conversely, the LCD-2 has:

  • Wider bandwidth: deeper bass, higher treble
  • Better detail & articulation throughout the range
  • More natural, realistic voicing

A gentle parametric EQ helps widen the HD-580’s apparent bandwidth:

  • +3 @ 25 Hz, Q=0.67
  • +3 @ 14 kHz, Q=1.5

I’m enjoying this trip down memory lane. I listened to these same HD-580s during most of the 10,000 hours I put into Octane Software back in the day. They sound nice, but I will be very happy to get my Audeze back.

Audio History

I loved music and was fascinated with audio electronics since I was a little kid. Later I became interested in the physics of sound.

I bought my first audio component in the 1980s in college, a Harman Kardon integrated amplifier. It was simple and cheap, had no tuner, only 40 WPC output, but it did have a phono amp (MM only) and decent gain stage. To find good speakers, my friend Shawn and I visited the local audio store and listened to several different speakers (Klipsch, Polk, and a few others) with a variety of music. We both liked the Polk 10Bs best. They had the smoothest least colored sound for my limited budget. My musical taste at the time was about half classical, half rock.

Back in those days digital audio and headphones were not an audiophile option. Good headphones simply didn’t exist and digital audio was so new, consumer CD players were expensive and tended to have poor reproduction of high frequencies and transient response. Because of this, there were no good cheap paths to high quality sound, like we have today.

I didn’t have a turntable, they were too expensive. But I did get a good CD player, an Onkyo DX-530 which was one of the first CD players to use oversampling, which improved the high frequency and transient response by enabling more gradual slope Nyquist filters.

This little system lasted me through college with many hours of satisfying listening. Then, my junior year in college, the local audio store went out of business and I got their used demo pair of Polk SDA-2 speakers. This was a big upgrade from the 10Bs, and the price was so good it was almost an even trade when I sold the 10Bs.

After graduating from college I was ready for a decent turntable. I visited the local audio store and auditioned a couple of different turntables & cartridges for several hours, picking a Thorens TD-318 MK II with an Ortofon MC-3 high output MC. That was in 1991. That HK integrated amp only had a low-gain MM phono amp, and my budget didn’t allow for a low ouput MC. The high output MC was a little on the bright side, but it had the smoothest, least colored sound compared to the MMs.

This little system lasted me for several years, until around 1995 I got a new job and promotion and my budget was ready for an upgrade. I auditioned a couple of different power amps and pre amps at the local audio store and ended up taking home an Adcom 5800 power amp with a Rotel RC-990BX pre amp, which had a dual-stage phono amp, so I could now try low output MC phono cartridges. And I had enough power to fully drive those Polk SDA-2 speakers.

At this time, digital was improving but to my ears, good vinyl still had more natural sounding high frequencies and transient response. But only good vinyl – like heavy 280-220 gram pressings, half-speed masters, etc. I started collecting MoFi half-speed masters, Cheskys, Audioquest, Telefunken, Wilson Audio, Classic, Water Lily, and other audiophile vinyl. I didn’t have the budget for much, so I carefully selected and treasured each new addition to the collection.

In the late 90s I replaced my Onkyo DX-530 with a Rega Planet CD player. I read so many good things about it, I thought it must be great. I never really got into this CD player, I think the old Onkyo was actually better. The Rega had a distinct sound that grabbed one’s attention at first. But upon further listening it was to my ears, congested and the high frequencies were all wrong. I ended up selling the Rega about a year later. It was so popular, it was easy to sell. I replaced it with a Rotel RCD-1070. Nothing special, but a solid well engineered good sounding player.

Fast forward a few years to 2000, when I sold my first startup (Octane software) and was ready for another audio upgrade. I already had reference quality amplification so this time it was the speakers. I visited the local audio store with my best albums and spent all day listening to every fine speaker system they had. I also did a bunch of research in audiophile channels. I ended up picking Magnepan 3.6/R speakers, as they had the most natural, linear, uncolored midrange and treble of any speaker I listened to. The Adcom 5800 had plenty of power with enough refined clarity to make these excellent speakers really sing.

About a year later I designed and built my own ladder stepped attenuator to replace the preamp. This added a level of clarity and transparency to the system — no active preamp is cleaner than a single metal film resistor in the signal path! And I learned a little about analog audio circuits, grounding and soldering. Now I didn’t have a phono amp anymore. I did a bunch of research and picked up a DACT CT100, which is an excellent reference quality flexible phono amp, but just a circuit card. I designed and built a power supply for it (dual 12V batteries), with a small chassis, cabling & grounding & connectors. I was delighted with the sound, a noticeable upgrade from the Rotel pre amp’s phono amp, which was quite good to begin with.

This new level of transparency revealed the limitations of the Rotel CD player so I looked for alternatives, knowing that DACs were constantly improving. I ended up with another Onkyo, a DX-7555. It had a more refined sound with more natural midrange voicing.

After we moved from Orcas Island to Seattle my listening room changed. I used test tones, microphones and measurements to tune my new audio room. I built floor-to-ceiling height 22″ diameter tube traps for the rear corners, RPG acoustic foam 4 layers thick strategically located on the wall behind the listener, careful room and speaker arrangement, and ended up with a great sounding room that was within 4 dB of flat from 40 Hz to 20 kHz. It wasn’t perfect though. There was a small rise in the mids around 1 kHz, likely inherent to the Mag 3.6 speakers, and the lowest bass octave was from 6 to 12 dB down. Notwithstanding these limitations, it was a great sounding room.

I kept this system for about 10 years, from 2005 to around 2015. Then I replaced the ladder stepped attenuator with an Oppo HA-1 DAC, using the digital outputs from my source components. And I got a Behringer DEQ 2496 and used its pure digital parametric EQ to tame the 1 kHz bump and lift the bottom bass octave. This put the in-room system response within 3 dB of flat from 30 Hz to 20 kHz, which is comparable to a good recording studio. The sound is fantastically natural: detailed yet smooth and not bright, bass is deep, yet controlled and fast, natural voicing through the mids with seamless transition to high frequencies.

Finally, in Jan 2018 I sold my turntable, vinyl, and related analog equipment. I just wasn’t using it anymore, since I had all those recordings on digital, and the sound quality of digital had improved so much, while great LPs do sound great, I no longer felt that they sounded any better than great digital.

Mike’s Best Vinyl LP Records

UPDATE: Mar 2018: These are all sold!

As I’m liquidating my vinyl and playback equipment, I’ve sorted through all my LPs and found about 100 of them to be half-speed masters, heavy vinyl, 45 RPM single sided, Japanese Press, Mobile Fidelity, Chesky, Wilson Audio, Telefunken limited edition pressings, or other such. Many are out of print, all are in mint condition – no scratches, cleaned with the Nitty Gritty 2.5FI, played only on properly aligned high end equipment.

I’ve got a few hundred more LPs not shown in this list, many of which are nice, but they’re standard quality. I’ll probably sell them in bulk for $1 each somewhere.

Here’s the list of my best LPs. Items already sold are highlighted in RED: lpListHighQuality-1712

Vinyl LP Cleaning Solution Recipe

I covered this topic about 10 years ago, offering a recipe for fluid to clean vinyl LPs. I still use that recipe in my Nitty Gritty; here’s a summary and a few more tips.

It has 3 ingredients, one of which is optional:

  • Distilled Water
  • Isopropyl Alcohol
  • Wetting Agent (optional)

Most wetting agents are soaps which contain fragrances and other non-essential ingredients that you don’t want polluting your record cleaning fluid. I’ve stopped using the wetting agent and it still works just fine. If you use a wetting agent, all it takes is a couple of drops for a small batch.

Alcohol is a solvent that may degrade the seals of record cleaning machines. To avoid damaging the machine, keep the alcohol below 20%. That seems to be a conservatively safe level, and it doesn’t take much alcohol to do the job so adding more won’t necessarily get records any cleaner.

Two kinds of isopropyl alcohol are commonly available: 70% and 91%.

  • Recommended: Conservative formula (< 20% alcohol)
    • With 70%: 1 part alcohol to 3 parts water = 17.5% alcohol
    • With 91%: 1 part alcohol to 4 parts water = 18.2% alcohol
  • Aggressive formula (< 25% alcohol)
    • With 70%: 1 part alcohol to 2 parts water = 23.3% alcohol
    • With 91%: 1 part alcohol to 3 parts water = 22.8% alcohol

As for cost (as of Jan 2018):

You can buy 91% isopropyl for about $3.50 per quart, and distilled water for about $1 per gallon. That makes 1.25 gallons of fluid for about $5. Nitty Gritty charges about $80 for 1 gallon of their solution, which is for all practical purposes the same thing.

JDS Element vs Meier Corda Jazz

This is a detailed comparison of the Corda Jazz with the JDS Element. I own one of each and listen to them almost every day, along with an Oppo HA-1. I’ve reviewed each of them separately.

TL;DR Summary: If all you need is a pure analog headphone amp, get the Corda Jazz. It has all the clean neutrality of the JDS Element, but richer, sweeter, more refined. If you want the flexibility of having a DAC and analog RCA inputs and outputs too (even if you won’t always use them), get the JDS Element.

Similarities

Cost: Both cost the same (about $350).

Provenance: Both are built by very small independent companies. Both are designed and built with a no-bullshit engineering philosophy.

Sound Quality: Both have excellent reference-quality measurements and great subjective sound quality. Differentiating them in a properly done level matched DBT is possible, but requires careful listening.

Gain: Both have adjustable gain separate from the volume knob–a switch for high vs. low gain. This enables them to drive anything from efficient IEMs that only need millivolts and milliwatts, to big power hungry planar magnetics.

Power: Both have > 1 watt max continuous power output, enough to drive almost any headphone on the planet, except for electrostats which need a dedicated voltage step-up transformer.

Reliability: I’ve used both near daily for more than a year with no problems.

I believe any pragmatic audiophile (myself included) would be happy with either one, so long as he valued sound quality and neutrality over fancy knobs, glowing displays and the exclusivity of limited production boutique equipment. Actually, each of these does provide some of the latter exclusivity despite their low price, being less common than mass-produced gear from major manufacturers. When people see one on your desk they ask, “What the heck is that?”

Differences

DAC: Advantage: JDS Element
The Element has a DAC; the Jazz doesn’t. The Element’s DAC is clean, but USB-only and does not run in async; it relies on the source (your computer) to clock the data. JDS claims async mode doesn’t provide any audible benefit, and their measurements support that claim (though that doesn’t necessarily make it true). I do note, when using theĀ  Element’s DAC from my computer, occasional (once every few minutes) “tics” or brief drop-outs that are not in the source material and occur seemingly randomly. These don’t happen when bypassing the DAC and using the analog input. This behavior is consistent with the notion that the clocks (computer source vs. Element DAC) are slightly different and it occasionally re-syncs. This may happen less frequently or never on other computers.

Flexibility: Advantage: JDS Element
In short: in addition to being a headphone amp, the Element also has a DAC and can serve as a preamp. The Jazz is only a headphone amp; it has no DAC and cannot serve as a preamp.

If you need a DAC that can drive line-level analog output (for example to a different device), and also a headphone amp, the JDS Element does the job. You can use the Element with any computer or device having a USB connection; you don’t need a fancy sound card.

Both the Element and the Jazz have unbalanced analog RCA input jacks and can be used as a simple analog headphone amp. In the Element, this bypasses the DAC.

The Element also has analog RCA line-level output jacks, which the Jazz lacks. This makes Element quite flexible as a line-level DAC, an analog headphone amp, or a preamp. When turned off, the DAC is still on and it routes the USB input to the analog RCA outputs. So you can use the Element as a DAC with line output, and as a headphone amp, leaving both plugged in at the same time. However, it will only drive one or the other, depending on whether it’s turned on. Put differently, think of the Element as “always on” for DAC, line input and output, and its power switch controls the headphone amp.

The Jazz is nothing more than a pure analog headphone amp. It has no analog RCA outputs to drive another analog line level device. It can’t be a DAC, nor can it be a preamp. That’s because of the Jazz’s active balanced ground drive.

So what is active balanced ground drive? More on that below, but in short, it improves the S/N ratio. The drawback is that the – output carries part of the signal, so an unbalanced analog input shunt this signal to ground. This makes the Jazz attempt to drive a 0-ohm load, which can blow the fuse or damage the amp. Use the Jazz only to drive headphones — not other audio components!

Volume: Advantage: Corda Jazz
The Element has an analog potentiometer volume control. It’s a very good one: smooth, wide range, well balanced, but still a pot. The Jazz uses a stepped attenuator triggered by an analog pot; there is no pot in the signal path, only metal film resistors. It has about 30 steps, each about 1.5 dB apart. One can argue whether a stepped attenuator makes any audible improvement, but there’s no question it’s a superior design: cleaner signal with perfect channel balance at all volumes, and unheard of at this price.

Imagery: Advantage: Corda Jazz
The Jazz has a mode to artificially create a more natural stereo image from normal (non-binaural) stereo recordings. It’s a switch that blends channels with phase delay depending on the difference in L / R channels. I’ve used these before and they’re usually gimmicky. Meier’s is not a gimmick. It’s the only one I’ve heard that improves the image while getting out of the way of the music being otherwise sonically neutral or nearly transparent. I said “nearly” transparent. It does make the tone a tad less rich, a small emphasis in the upper mids to lower treble. I usually leave it off, except on recordings with extreme L-R separation, where for example a singer or instrument is entirely in one channel or the other. These are hard to listen to on headphones, and this switch fixes that.

Signal Isolation: Advantage Corda Jazz
Both amps have unbalanced analog output to the headphones. But the Jazz adds a twist: active balanced ground driving. Signal ground to the headphone is not the 0 V frame ground that it would be with standard unbalanced. Signal ground contains some of the L and R signal combined, such that the net signal at each speaker of the headphone (difference between + and -) sums to pure L or pure R. Because the ground contains some L and R signal, the net field around the cable is near zero (not exactly zero, as it is with balanced). This isolates the signal better, immunizing it to hum or other electrical interference, improving the S/N ratio. Some might say this also makes the load easier for the power supply to drive, but the power supply is already over-engineered with its 10 W toroidal transformer.

Build Quality: Advantage Corda Jazz
Both have great build quality, but the Jazz is a small step higher both inside and out: the case, switches, knob, power supply and other internal components. The Element is by no means cheaply made, it’s a pleasure to view and handle. But the Jazz is a step up.

Sound Quality – Subjective Listening
Both sound great: clean, neutral, detailed and fast without brightness, deep bass without being bass-heavy. Both are dead silent even at high gain full volume – no hum or other background noise. Both have excellent measurements comparable to professional reference gear. Not having a DAC, the quality of the Jazz depends on the source. I compared the Jazz & Element using an Asus Xonar DX sound card to drive the Jazz, and driving the Element with a USB bit stream. I used Audeze LCD-2 headphones (with the 2016 drivers) and extremely high quality recordings of a variety of music, mostly acoustic.

In this configuration, I preferred Jazz in overall sound quality, which countered my expectations since good solid state amps are so hard to differentiate in blind listening tests. The Jazz has the same level of clarity and detail as the Element, yet at the same time sounds slightly more rich in the bass and sweet in the mids and treble. Call it more musical, yet without any loss of neutrality or clarity. I emphasized the word slightly because the difference is subtle. Upon first impression they sound identical, though I feel well trained experienced listeners using excellent recordings would detect the difference consistently with careful listening in an ideal quiet environment. Indeed, the differences were big enough to overcome my expectation bias that there would be no differences!

That said, if I needed the flexibility that the Element provides–listening to music from a laptop where I must stream bits over USB because I can’t install a high quality sound card, or I needed to use it as a preamp in addition to a great headphone amp, or I was using great but less than reference quality headphones like Sennheiser HD-600 instead of the Audeze–I would grab the Element without hestitation. To put some meat to that statement, I purchased my Element and it’s not for sale.

By comparison, the Oppo HA-1 is the best of both worlds and more. Its analog amp equals or exceeds the Jazz, which is a high bar. Its Sabre ES9018 DAC is fantastic and has coax, toslink and USB inputs. It is fully balanced with both line level and headphone outputs, yet also has a single-ended outputs. It also has great flexibility with numerous inputs and outputs. The only feature it lacks is the Meier’s headphone image circuit, but I only miss it on those rare recordings with artificially extreme L-R separation. But the HA-1 is big and bulky, weighs about 15 lbs, costs 3x the price of the Jazz or Element and is no longer made.

Audio EQ Settings

Since I went to the dark side, and started using gentle parametric EQ to correct the FR of headphones & speakers to neutral, I want to collect the EQ settings here. My general philosophy is to make subtle corrections. FR response deviations are often related to other forms of distortion like phase or ringing. Amplitude corrections big enough to restore completely flat response can exacerbate those other factors. Thus I stick to subtle, gentle corrections that improve neutrality yet preserve the original character of the sound, short of restoring perfectly flat response.

When boosting levels, remember to apply a reduction in overall gain to avoid clipping. Take the biggest amplitude boost at any frequency and cut by that amount. This is in the voltage (not power) domain, so dB = 20 * log(V1/V2).

For example, if you apply +2.5 dB @ 4.5 kHz, then you must reduce overall gain by 2.5 dB, which is a gain ratio of 1 / (10^(2.5/20)) = 74.98%. Or just use 74%. When it comes to gain ratio, to avoid clipping, lower is safer. So always round down (truncate the decimal), not up.

Sennheiser HD-580

Characteristic FR: flat down to 50Hz, then roll off bass at about 6 dB per octave. Gentle dip of about 5 dB between 3500 and 9000 Hz. Narrower dip of about 6 dB between 10 and 17 kHz.

Parametric EQ Correction:

  • +4 dB @ 25 Hz, Q=0.67 (2 octaves wide — 1 each side)
  • +4 dB @ 14 kHz, Q=1.3 (slightly under 1 octave wide)
  • Gain: -4 dB = 0.63 = 63%

Subjective Difference:

This simple EQ transforms the venerable HD-580 making it more open, natural and neutral. Deep bass extension without impacting mid-bass and linearity. Slightly crisper transients and “air”. Even though the midrange is untouched, it sounds a touch more open, less boxy.

Audeze LCD-2

Characteristic FR: flat from zero to about 2 kHz. Gentle dip of about 7 dB between 2k and 9k. Flat from 9k on up.

Parametric EQ Correction:

  • +2.5 dB @ 4500 Hz, Q=0.67 (2 octaves wide — 1 on each side)
  • Gain: -2.5 dB = 0.74 = 74%

Subjective Difference:

Normally the LCD-2 sounds perfectly natural, yet a touch soft like listening from the 5th instead of 1st row, and correspondingly soft on detail. This EQ brings the LCD-2 back to the 2nd row and restores some lost detail, but without affecting its near-perfect voicing. It transforms the LCD-2 into the perfect headphone!

The 2016 and later LCD-2 are a touch brighter and need less EQ. Say 2 dB for these, and 3 dB for the 2014 model. CSD plots show the LCD series headphone tends to ring or resonate around 4-5 kHz, which is the trough of their natural response curve, so you don’t want to boost this freq too much.

Magnepan 3.6/R

Characteristic FR: depends on the room. My listening room has floor-to-ceiling tube traps 2′ diameter in corners behind the listener and 4″ thick RPG acoustic foam on the wall behind the listener. This EQ is busier than the headphones, which is unavoidable with in-room speakers, though I still managed to keep the rates gentle.

Parametric EQ Correction:

  • +3 @ 32, Q=0.67 (2 octaves wide — 1 on each side)
  • -2 @ 90, Q=1.41 (1 octave wide)
  • +3 @ 240, Q=0.4 (1.5 octaves wide)
  • -2.5 @ 1000, Q=0.67 (2 octaves wide)
  • +3 @ 3000, Q=1.41 (1 octave wide)
  • Gain: -3 dB = 0.7 = 70%

Of these, the bold-faced ones correct anomalies inherent to this speaker. The rest are room corrections. That is, the Magnepan 3.6/R is a near-perfect speaker when set up properly in a good room, but it rolls off the low bass and has a gentle lift around 1 kHz. The other 3 settings are corrections to my room.

Ubuntu VLC DAC Audio

I recently got a JDS Labs Element DAC + headphone amp. I drive it from my Ubuntu desktop using VLC as the audio player. It’s plug and play – no drivers needed. However, best results come after applying a few tips:

VLC Audio Device: The DAC has 17 output devices that appear in VLC. Which one to use? Use Pulse Audio if you want to hear a mix of all audio on the computer. Pulse Audio mixes all sources and resamples them if necessary to a common rate. Use JDS Labs Element DAC, USB audio direct hardware device without any conversions if you want to hear the audio track in its native sampling rate & bit depth, and nothing else. I prefer this for best sound quality.

VLC Output Module: use Pulseaudio audio output if you want to hear a mix of all audio on the computer. Use ALSA audio output if you want to bypass Pulseaudio to hear the audio track and nothing else.

VLC occasionally stopped playing and popped up an error saying “Device or resource busy”. If you’re using ALSA, only one app at a time can use the device. For example, if the browser tries to play a video it can steal the device from VLC. Also, VLC seems to have a bug in which it occasionally steals the device from itself when switching tracks. Adding a udev rule made this happen far less often. Add a file called 41-jdslabs-dac.rules to directory /etc/udev/rules.d. Make the contents like this:

# JDS Labs Element DAC
SUBSYSTEM=="usb", ATTR{idVendor}=="262a", MODE="0666", GROUP="plugdev"

This makes the JDS Labs DAC accessible to any Linux user.

Audio Glitches: Occasionally, once every hour or so, the audio will stop for a moment, then resume. I believe this is because the JDS Element uses USB adaptive mode, not async. This makes it compatible with more computers. Some people claim that adaptive move has more jitter and lower sound quality, but measurements belie this claim.

More audio glitches: Occasionally I would hear tics in the music, as if the computer CPU were too busy to deliver audio. Re-nicing the VLC process to -15 fixes this.

Review: Audeze LCD-2 (2016)

What? Another LCD-2 Review? Why? Here’s the background.

I got a second pair to use at work, again from the Headphone folks in Montana. I’ve owned this headphone since 2014 and already reviewed it twice: once when I first got them, again later when I EQed them. Audeze never rests and is constantly improving their products. But they don’t change the model numbers. The LCD-2 has gone through several variants with names the community invented because Audeze didn’t see fit to name them:

  • LCD-2.1: the original version – creamy sound, smooth linear mids with rolled off treble
  • LCD-2.2: same linear mids, improved treble response, yet still on the warm side of neutral
  • LCD-2F 2014: introduction of Fazor, improved detail and transient response, but some people report the treble sounds wonky
  • LCD-2F 2016: lighter re-tuned drivers, further improved transient response

The 2016 LCD-2 is similar to the 2014 overall, with excellent reference quality sound. Since I linked the prior reviews above, here I’ll describe only the differences. Compared to the 2014, the 2016 LCD-2 has:

  • Bass: cleaner, tighter, faster but neither attenuated nor amplified. This is hard to imagine because the 2014 bass was excellent to begin with. Somehow they improved it.
  • High Treble (9+ kHz): cleaner, faster and slightly amplified. A good recording of castanets shows the 2014 was already excellent, but the 2016 is even better. Treble is shelved up a touch and brighter compared to the 2014, but the 2016 is not bright sounding.
  • Mids: different – described below

Comparing the midrange is more complex and takes more than a few words. The 2014 midrange is incredibly smooth and natural and has a slight presence emphasis compared to the 2016. This presence is subtle and to put it in perspective, the Sennheiser HD-600 (a great headphone in its own right) has far more presence sounding boxy or nasal in comparison. I like the 2014 mid presence on small ensemble acoustic music; it brings out the natural timbres of acoustic instruments and voices. But with large ensemble works and big complex music, this presence becomes a slight glare that veils the music. The 2016 lacks this presence, yet it also lacks the glare that comes with it on big complex music. The 2016 still voices acoustic instruments in a natural, realistic way – it’s not midrange suck-out like some headphones have.

So when it comes to the midrange, both do extremely well, yet I prefer the 2014 for small ensemble acoustic music and the 2016 for bigger more complex music.

Overall, the 2016 is better than the 2014 in many ways, but not in every way. The 2016 is more open, faster and more resolving – all good. Yet the 2014 has a special intimacy and realism to the midrange voicing of small ensemble acoustic music.

Note: I contacted Audeze and for $400 they will upgrade any model of LCD-2 to the latest version, which includes new ear pads of your choice and return shipping. I’m leaning toward upgrading my 2014… but haven’t yet decided. If the 2016 was better in every way, I would. But the decision isn’t that easy.

Alternatives

In my experience, planar magnetics wipe the floor with conventional drivers in terms of overall sound quality – both in headphones and in speakers – so I limited my search to them.

I auditioned the HiFi Man HE-500 a few years ago. It was a great headphone but had a weird midrange response that didn’t voice acoustic instruments properly. I looked at other HiFi Man models but none of them have the truly linear frequency response I’m looking for.

I gave the Focal Elear serious thought. Sure, it’s a conventional driver. But it had such rave reviews I considered it. Yet it also had some decidedly non-rave reviews and the specifications showed non-linear frequency response, transient ringing and higher distortion. No thanks.

Finally I decided to keep it simple. I like my LCD-2 so much, why not first try the latest version? If I didn’t like it I could always return it and move on to something else. I found a pair from Headroom, the headphone folks in Montana, that was an open box, so I got a lower price, yet new with full warranty. I’ve been a customer of theirs since 1999 because they are knowledgeable, honest and have a generous 30-day try-return policy.