iFi Audio iPurifier3
Sometimes, it pays to do research…I had a problem that needed fixing. To be frank, it was a rather trivial, First World Problem. Join me in a lengthy tale of woe, you may find it an interesting process…
You see, I wanted to listen to headphones from my living room couch – understandable enough. To complicate matters, I wanted to use a piece of audio jewelry that had been sitting in storage since we moved some three years ago: the Woo Audio WA7 Fireflies (2nd gen), a headphone amplifier (HPA) plus DAC with a minimalist signal path and several convenience features. The fact that its visual appeal trounces most HPA/DACs is a point in its favor. After all, I wanted it on display on a side table in our front room, a space already burdened with several larger pieces of high end hi–fi gear. All well and good, with one caveat: my Mac mini server sits across the room from the side table… What to do?
Just How Long Is It?
The WA7 Fireflies Mk. II – henceforth referred to as the WA7.2 – has standard connectivity spigots; USB and line level on RCAs. The really inexpensive solution would have been to purchase or construct a long headphone cable. Only, there is an inherent problem with that… I’ve built several headphone extensions and, even with relatively fancy wire, the sound quality was not to my liking. You can’t simply ignore all that capacitance and inductance that a long cable brings. I could have purchased a long version of my reference headphone cables, Wireworld’s Nano-Platinum Eclipse, but the cost would have been prohibitive even with a courtesy discount. Again, what to do?
Instead of extending the amplified analog signal, why not extend the USB interconnect? The USB input of the WA 7.2 natively supports sample rates of up to 8x (384 kHz) linear PCM, as well as DSD2 or double speed DSD (DSD128). If I connected it to Ye Olde Mac mini, then I could switch between my reference DAC, the extremely capable exaSound e22 Mk. II and the WA7.2. And after all, it’s digital so “it doesn’t matter.” Hah! I only wish that were the case.
While I am not a strict objectivist as are many in my pro audio day job, when I do hear something of interest, I will satisfy my curiosity until I’m sure what I heard was valid. Once satisfied that I’m not imagining it, whatever “it” is, I’ll go with my subjective response until proven wrong or misguided. For example, cast your mind back to early in 2014, when I wrote a review of VUE Cables’ USB interconnect. As a result of that review, I ended up purchasing their VU-3 and still use it today as my reference. Alas, VUE Cables are no longer in business and besides; a long USB cable is an even more stupid solution that a long headphone extension‡…or is it?
Also in 2014 (must have been something I ate), I wrote another review of an interconnect technology for which I have great respect; optical fibre. In that instance, it was Corning’s first generation, self–powered active glass fibre Thunderbolt interconnect, a 10 meter example. Even today, with a bit of competition in that product category, glass fibre extenders are pricy.⚙︎ As the son of Great Depression survivors, and someone who lived through the dotBomb of 2008, I’m basically a scrooge. So, even if my HPA/DAC had a Thunderbolt interface, that was a non–starter. “But wait!”, as they say in cheesy TV ads…how about an active USB extender? I’ve always liked Monoprice for their get–the-job-done lines of utilitarian products, so I settled on their 33 foot USB-A to USB-B 2.0 active extender. This assembly can be had for the princely sum of $15 including shipping, so I was all in! The active extender is a relatively stiff, actively driven, bus–powered copper cable contrivance, and easily reached from server to table. w00t! Mission accomplished, or so I thought…
Now, the WA7.2 is a capable little box, though certainly not reference grade. Due to its relatively low cost, it errs mostly on the side of omission; rolling off the top octave and softening transient response a bit. That said, the combination of HPA/DAC plus extender sounded even less appetizing than the WA7.2 alone. That’s not surprising as the USB interface is as sonically important as any other subsystem in the signal chain. Now I had my headphone amp next to my comfy seating, but it didn’t have the quality I had hoped for.
Purification Ritual
As already mentioned several paragraphs back, I asked once again, “What to do?” Almost all the pieces were in place save one. The WA7.2 sounded better when plugged directly into a host than through 33 feet of cable, albeit active cable. This could only mean one thing: the eye pattern♆ was degraded. I thought, “There’s a fix for that!” I wish that were the case in real life, but in Audiophile Land at least, isn’t there always? The solution is a USB “cleaner” of some kind, which range in price from DIY cheap and cheerful, to the exceptional but relatively spendy offerings from UpTone Audio. Being the cheap ass that I am, I went for the low end to start by reaching out to Lawrance Lee at iFi USA, who offered to send me an iPurifier3 eval unit. The iPurifier3 is their entry level cleaner. It’s delivered as an in–line active device, with an input at one end, and output at the other, and nicely finished aluminum casework in between. It has two LED idiot lights on top. They indicate the presence of adequate power (blue) supplied to the unit, along with valid signal present at the USB receiver (green).
Although they wouldn’t tell me anything substantive about what it actually does to the incoming data, iFi’s web site mentions that the “…hub repeater chip is only the starting point for us (and) the reclock is only a part of the equation in the iPurifier3. The active noise cancellation and power conditioning clean up the power to the USB (if it draws USB power – OMas). Additionally, there is a filtering of ground-borne noise (but no ‘IsoGround’ to deal with low-frequency ground loops).” Let’s pick apart that statement…
Triune Aspect
There appears to be three aspects to the iPurifier’s function:
- Reclocking
- ANC or Active Noise Cancellation
- Power Conditioning
Let’s look at each in turn…Reclocking is basically the process of receiving an incoming clock signal, in this case a dirty or degraded one, and recreating from that a clock signal that is clean. Clean usually means the voltage level transitions, square waves that represent logical one or zero, are as fast or vertical as possible, free of ringing, and are being “clocked out” at a constant frequency. Wait, isn’t the USB signal derived from a quartz crystal–referenced timebase?
How could the USB clock have gotten messed up somehow? In a word; jitter, specifically timing jitter. Timing jitter is the deviation of a timing reference from its supposed nominal frequency. Say what? Imagine a clock that, instead of ticking once per second, actually ticks at 1.1 ticks per second or eleven ticks every ten seconds. At least that clock would be consistently fast. What’s worse is when a clock frequency varies randomly; sometimes faster and sometimes slower but only occasionally correct. That is, in fact, the way most all quartz clocks actually work. Their variations are small enough that we slow analog beings don’t notice. However, digital circuits are very sensitive to clock variations. When a clock “tick,” a voltage transition, is expected and arrives early or late, the whole system performance suffers.
The most common causes of variations in a timing circuit are mechanical vibration, induced noise, power supply variation and temperature variation. Of those, temperature is the common dominant factor. Quartz crystals, miniature slabs of high purity quartz, will vibrate at a characteristic frequency derived from their physical dimensions, when presented with the right electrical stimulation. As these tiny “tuning forks” heat up, they expand and, as they cool down, they contract. Since the resonant frequency at which they sing is determined by their size, as is the case with most resonant physical objects, temperature variations cause deviations in their resonant frequency.
This is why better quality gear uses a TCXO, or temperature–controlled crystal oscillator, where some effort has gone into correcting for varying behavior with temperature. For really expensive applications, the quartz crystal is “ovenized,” or contained in a heated enclosure to maintain a constant temperature. A MIL–spec, miniature, ovenized crystal oscillator enclosed in a hermetically sealed package to provide highly stable clock output can cost a couple of thousand dollars or more, while a basic oven-controlled crystal oscillator will run you around a hundred or more. Given the price point of the iPurifier3, we can assume that a plain ol’ quotidian time base is employed. When asked where the re–clocker derives its clock reference, I was told that “…there is a nice external clock that drives the reclocker…inside the hub chip.”
Next up is noise, the bane of designers everywhere especially when it modulates your data signal and causes more errors in data recovery. After all, it’s accurate recovery of the audio data at the USB receiver that makes for a successful and good sounding USB interface. Noise comes in many forms and at many frequencies and amplitudes. Suffice it to say that some form of active and passive circuitry is used to, ahem, purify the data lines and clock. This might include capacitors to reduce power supply noise and shunt RF to ground, and maybe inductors to filter out high frequency noise.
The last leg of the iPurifier tripod is power conditioning. The iFi folks wouldn’t tell me anything useful about their power conditioning either, except that it is part of the design. Given that the iPurifier3 has no external power supply, it must derive its power from the usually rather filthy USB power line. That source is noisy and not as tightly regulated as one would want. The result? Yet more performance degradation. So, the iPurifier3 provides some form of active regulation and noise filtering to insure the data lines, a differential pair in geekspeak, are cleaner on output than they were on input.
Oblique Obfuscation
When I asked for generalities about what the above mentioned hub repeater does; why is it needed and what are the benefits, I received the following rather eccentric reply, verbatim, from iFi’s marketing team; “Check out the ipurifier marketing materials, basically it recreates a whole new much nicer USB signal and also improves the USB power supply a lot.” Uh, okay. Brilliant. So, we have to guess that the hub repeater used is a low cost monolithic integrated circuit. Here’s what one vendor, Microchip Technology, says about their obsolete hub repeater chip, the model USB2502: “When connected to a high-speed (USB) host, the…downstream facing ports (the output) can operate at low-speed (1.5Mb/s), full-speed (12Mb/s), or high-speed (480Mb/s)…The USB2502 supports both bus-powered and self-powered configurations. In bus-powered mode, all power is derived from the upstream facing port and no external power supply is required. An external USB power distribution switch device is used to…limit current and sense over-current conditions.” That particular hub product uses an on–chip oscillator employing a “low cost crystal.” In simple terms, a chip much like the USB2502 is the heart of the iPurifier3. It performs re–clocking, and most likely some of the power conditioning in that it’s designed to be immune to typical crappy power found in bus–powered applications.
Performance
I don’t take kindly to “black boxes” that have magic inside. Unfortunately, the folks at iFi made the technical aspects of the iPurifier3 as opaque as possible…not good. To me, high end hi–fi is a meritocracy. Gear should sell itself, and a manufacturer should be forthcoming about the unique differentiators that make their product special. When asked what, concretely, the iPurifier3 does, iFi’s response was unsurprisingly less than informative. They said, “Much better USB signals, use the DAC USB receiver doesn’t have to work so hard, and hence create much less jitter in the system. Also clean up majorly the USB Power…”.
The key tidbit you’ve been waiting for is this: did the iPurifier3 audibly improve the sound quality? The skinny little white dude’s effect is subtle but effective on some pairings. My reference DAC (exaSound e22 Mk. II) doesn’t seem to benefit from it, most likely because its fancy USB implementation makes it pretty hard to improve, but it does improve my most “difficult” DAC setup; that Woo Audio WA7 Fireflies Mk. II via active USB extender. With a variety of cans, principally my Audeze LCD-3 with Wireworld Nano-Platinum Eclipse cabling and Westone W60 in–ears, my impressions were better transient response or less temporal smear; hard to tell subjectively. Those aspects of a high fidelity recording I refer to as sparkle and air were both subtly improved. In keeping with a reduction in jitter, I also noticed tighter and more impactful bass. One last aspect that I became aware of was soundstage. Without the iPurifier3, the soundstage shrunk a tad, becoming slightly miniaturized.
Moving from a simple USB implementation to something more sophisticated, I next tried my midline desktop DAC. With iPurifier3, the Mytek Brooklyn became smoother and pleasantly more burnished. Unlike the Fireflies, I didn’t notice an improvement in low frequency reproduction or transient response. However, the soundstage was a bit more expansive. The really salient upgrade with this combo was a reduction in grain, a general relaxation especially in the midband. Frankly, this change for the better was addictive. Using Amarra Luxe to play a variety of MQA–encoded content streamed from TIDAL, I had to forego my customary EQ “tone” controls† but the forfeit was worth it.
Unless you have a bottom feeder DAC, I doubt the iPurifier3 will make a huge difference in the sound, since the rest of the system probably isn’t highly resolving. Where the iPurifier3 does lend an understated polish is to a more accomplished but still mid–priced rig. It won’t radically alter your listening for the better, but it may just be the droids you’ve been searching for. It certainly added an auspicious touch of refinement to both of my casual converters. Unfortunately, with a $129 street price, it’s more spendy than those that need it most will be able to justify.
As the more well heeled buyer moves up the product line, “What,” you might ask “…are the benefits of the more elaborate digital enhancer family members from iFi?” The company claims “…even better USB signal, even nicer power supply.” To that I’d suggest you save your shekels, then trade in your less than ideal DAC for one with a more elaborate USB design, which will also entrain a whole host of other internal upgrades that you most definitely will hear.
On Review
iFi iPurifier3 USB Audio and Data Signal Filter
US$129 (USB Type A and B available)
iFi audio USA
1085 Blair Ave
Sunnyvale CA 94087
+1 (800) 799-IFIA
Footnotes:
⚙︎ — An example: as this article “went to press,” Wireworld had announced their Stellar, the first optical HDMI cable to receive HDMI® 2.1 ultra high speed certification. Prices range from US$450 for a 5 meter up to $700 for a 30m length. Try throwing HDMI 2.1 100 feet over copper! I don’t think so…
‡ — Why would a long USB cable be an even more stupid solution that a long headphone extension? In short: the cliff effect…Digital signaling has only one of two states, (see ♆ below) so if the signal becomes extremely degraded, the receiver cannot “read” the signal transitions any longer. At that point, the receiver simply gives up and stops outputting valid data. One moment the transmission medium is working, and the next it isn’t. It “falls off a cliff,” so to speak, quite abruptly. By contrast, an analog signal will gradually degrade, with the signal–to–noise ratio becoming smaller and smaller until the signal is perceptually swallowed by noise. Analog degrades “gracefully,” as we say in the biz, whereas digital either works or doesn’t. There is no Mr. In Between. With a headphone extension cable, the sound quality will progressively degrade as the length is increased. With a digital extension, the signal would subtly degrade until suddenly…it simply stops working!
♆ — An eye pattern is the name given to a graph of many consecutive frames of data superimposed on one another. This graph looks somewhat like an eye, with an empty space in the middle where the data is not in transition. Noise and jitter will blur the edges of the “eye,” those regions where data is switching from high to low or low to high. As the degradation increases, the eye “closes.” That is, the clean delineation of the two states becomes less and less distinct. At some point, the eye closes enough that the receiver can no longer read the signal, and the receiver stops transmitting recovered data.
† — In order for an MQA DAC to fully unfold or decode an MQA stream, the data must be bit–perfect. Any DSP applied to the original data stream will alter the data, resulting in a DAC’s inability to decode the MQA metadata. The result is partially decoded MQA audio. Amarra Luxe, as with Audirvana and other MQA–capable player apps, can carry out the first “unfolding” or decoding stage but a DAC must perform the second and final decode step. That second unfold yields sample rates past 2x; 88.2 and 96 kHz. So, 4x and 8x sample rates (176.4, 192, 352.8 and 384) need two decode steps.
As to the tone control I usually have in circuit, it’s my company’s nonpareil linear phase parametric equalizer, thEQred. When I first receive a product for review, I’ll listen in bit–perfect mode without any EQ or other modification. Once I’ve settled in for extended listening though, I like to correct voicing to taste. Especially with headphones, I usually need a low Q, low frequency shelf to lift up the bottom by about 3 dB.
“I don’t take kindly to “black boxes” that have magic inside.”
Amen, brother!