Ayre Acoustics QX-5 Twenty D/A processor

Throughout the summer and fall of 2016, I worked on a project with Stereophile contributor Sasha Matson, recording, editing, mixing, and mastering—for release on 180gm LP, CD, and high-resolution download—an album of works that Sasha had composed for various chamber ensembles: Tight Lines. As you can read in the article we published about this project, for the vinyl release we decided not to master the discs directly from the hi-rez files, but to create an intermediate analog tape master. Feeling that audiophiles would want an LP that at some stage was "analog," we therefore needed to choose a D/A processor to drive the Studer open-reel tape recorder we were going to use. As my colleague Michael Lavorgna had very positively reviewed Ayre Acoustics' QX-5 Twenty digital hub ($8950) for our sister site AudioStream.com in November 2016, and Sasha's own system is based on Ayre components, we borrowed Michael's review sample. I played the hi-rez master files on my MacBook Pro using Pure Music 3.0, and we used an Ayre KX-R Twenty preamplifier to increase the QX-5's balanced output level to match the Studer's input sensitivity.

Following the analog transfer session, the QX-5 Twenty was returned to Ayre, but I was so impressed with its sound quality that I asked to borrow a sample at a later time for a full review in Stereophile. And so, in April, the very same unit that Michael Lavorgna had reviewed, and that we'd used for Tight Lines, took up residence in my system.

What It Is
In three videos, Michael Lavorgna talked about the QX-5 Twenty with Ayre's principal development engineer, Ariel Brown, and software engineer Brendan Boyle; and separately, in two one-on-one conversations, with Charley Hansen, Ayre's founder and head of innovation. I won't repeat much of what was said in those discussions, but in brief, Ayre refers to the QX-5 Twenty, released in July 2016, as a digital hub because, while it is at heart a D/A processor, its full complement of digital inputs and its 100-step digital-domain volume control allow it to perform as the prime component in a digital music system. (In this respect, the QX-5 is a radical departure from the QB-9 processor, which established Ayre as a major player in the computer-audio world but was USB only and lacked a volume control.) In addition, as the QX-5 is both a certified "Roon Ready" component (footnote 1) and adheres to the DLNA/UPnP protocols, it can access both locally stored music files and streaming music services such as Tidal and Spotify.


From the outside, the QX-5 Twenty is styled to match the other models in Ayre's 5 Series, with a central alphanumeric display flanked by two buttons, and a circular control ring on the right. At the base of the display panel is an array of headphone jacks: ¼" and 3.5mm stereo phone jacks for conventional unbalanced use, and two 3.5mm jacks for balanced operation, using the protocol first seen in the PonoPlayer and featured in Ayre's Codex D/A headphone amplifier. The rear panel has balanced and unbalanced analog outputs and as many digital inputs as one could wish for: two AES/EBU on XLR jacks, three galvanically isolated S/PDIF on BNCs, three TosLink optical, an asynchronous USB Type B for connection to a host computer, two USB Type A ports to connect the supplied WiFi antenna or a USB flash drive (footnote 2), an RJ45 Ethernet jack for connecting the QX-5 to a local area network (LAN), and two AyreLink ports on RJ-11 jacks.

Inside the case—which is tastefully illuminated in red from multiple LEDs presumably used as voltage references—are four transformers, each with multiple windings and fed from the AC line via Ayre Conditioner RFI power-line filters. The analog circuitry is fed by a linear power supply. There are four main circuit boards, and two more behind the front panel. The largest board on the QX-5's base carries the power supplies, the transformer-coupled serial digital input circuitry, and the master clock, FPGA, and D/A modules. To its left is a small board carrying the Type B USB port and the USB receiver circuitry, this based on an XMOS chip running Gordon Rankin's Streamlength code. Above the main board's center is an Ayre-designed Ethernet receiver board, this carrying a smaller board piggyback sourced from South Korea's ConversDigital. This small board carries the USB Type A ports. Finally, a large board above the main board's right-hand side carries the zero-feedback, fully balanced analog output circuitry, this based on Ayre's Diamond topology and using discrete transistors rather than op-amp chips.

The QX-5's 100-step volume control has a step size of 1.0dB. According to Hansen, this control can maintain the full resolution of any source file all the way down to –60dB.

The heart of the QX-5 Twenty is ESS's new ES9038PRO DAC chip, though Ayre doesn't use this chip's built-in reconstruction filters. Instead, the two filters, labeled Measure and Music, are both minimum-phase types running on an FPGA chip with, unusually, the 16x oversampling of the audio data performed in a single pass. In an interview published in 2012 on AudioStream.com, Charley Hansen said that "virtually all oversampling filters are a concatenation of several 2x filters in series. For example 99.9% of the time an 8x filter is made from three 2x filters in a row, simply because it is the cheapest way to do it. . . . [However], there are always rounding errors at each oversampling operation. If you perform it all at once, the rounding errors are minimized. But if you perform 16x oversampling as a series of four 2x stages, then the rounding errors are compounded four times. . . . Performing all of the calculations in a single pass (rather than a concatenation of 2x stages) requires more computational 'horsepower'."

Ayre implements the filters in a Xilinx FPGA, which provides the necessary horsepower. In an e-mail, Hansen told me that, during the development of Ayre's digital products, "We have custom filters whereby we can load whatever coefficients we want into [the FPGA] . . . an external switch allows us to load different filter coefficients and hear the audible changes they make. This is a great test, as it allows for everything else to be held completely constant. I spent four solid months doing nothing but listen[ing] to the effects of various filters—corner frequency, stop-band response, minimum-phase vs linear-phase, apodizing, sharpness at the 'knee,' windowing functions, interpolation rates (eg, 2x vs 4x vs 8x vs 16x), dithering functions . . . every single factor I could think of." There is no MQA decoding, however, as Ayre is still evaluating the codec.


When digital data are converted into an analog signal, of prime importance is the quality of the clock that controls the conversion of each digital word. The QX-5 uses two "doubly rotated" quartz-crystal oscillators, one for 44.1kHz-family data, the other for the 48kHz family, these jointly developed by Ayre and the Russian company Morion. These oscillators are said to have extraordinarily low phase noise.

A feature of the QX-5 that awakened memories from a quarter-century ago is that it can decode HDCD recordings. Developed by Pacific Microsonics and later sold to Microsoft, HDCD was a proprietary way of embedding data in a Red Book CD's LSB that could, in theory, extend the CD's 16 bits of dynamic range to 17.6 bits. If a true HDCD-encoded datastream is detected by the QX-5, a green LED on the right of the display window illuminates and the letters hd are displayed in front of the sample-rate identification of 44.

Setup and Use
Installing the QX-5 Twenty in my system was as simple as plugging it into the AC, connecting it with an Ethernet cable to my router, an AES/EBU cable to my Ayre C-5xe disc transport, a USB cable to the Aurender N10 server, and balanced interconnects to my power amplifiers, and turning it on. The Fing network utility on my iPad mini identified the QX-5's UPnP name as "Ayre QX-5," its UPnP service as "AVTransport," and its MAC vendor as "ConversDigital." When the QX-5 Twenty is turned on and is connected to a network, it checks with Ayre's FTP server to see if a firmware update is available. (Firmware updates can also be installed via a USB drive.) And yes—when I turned the QX-5 on for the first time, it did indeed find and install a firmware update.

Before I listened to music, I checked the Setup Menu. This is accessed when the QX-5 is in standby (right-hand button glowing green) by holding down the left-hand button for three seconds. Then, using the right and left arrows on the rotary control, you can scroll to choose among options, and select and exit submenus with the right and left buttons, respectively. I started off in Preamp mode (volume control active), with the Music reconstruction filter applied to the USB, AES1, and Network inputs, and USB data sourced from the Aurender server.

A caution: The QX-5 Twenty runs hot—my review sample's top panel measured 106.8°F (41.6°C). Make sure it's well ventilated.

My first impression was of detail. And more detail. The QX-5 Twenty dug deep into the LSBs of the data it decoded.

My first task was to examine the difference between the Music and Measure filters. Using "Diamonds on the Soles of Her Shoes," from Paul Simon's Negotiations and Love Songs: 1971–1986 (ALAC file ripped from CD, Warner Bros. 2-25789), I repeatedly swapped between the two filters. I consistently preferred the quality of the bass-guitar line with Measure: the instrument sounded "rounder," fuller in tone. But the imaging was more palpable with Music, the acoustic objects more easily differentiated. The ostinato guitar riff—di do-whi-do di—that appears in every second measure of "Diamonds" stood clear of the other instruments and voices with Music, with a slightly better delineation of the pitches.

Footnote 1: "Roon Ready," indicating the highest level of Roon integration available to hardware manufacturers, means that the Roon Advanced Audio Transport (RAAT) is customized for and embedded in their devices. For more details of how RAAT differs from other network streaming protocols, and the design philosophy behind it, click here.

Footnote 2: Music on the USB flash drive can be streamed to the QX-5 Twenty using the mconnect control app.

Ayre Acoustics, Inc.
2300-B Central Avenue
Boulder, CO 80301
(303) 442-7300

mrkaic's picture

How can this product be superb? Figure 9 shows that the THD is around 0.1 percent. This is unacceptably high, almost tube preamp territory. The THD number should be at least 10-100 times smaller.

David Harper's picture

correct me if I'm wrong guys,but if a recording is,at any stage, digital,then it is not analog. In order to be analog a recording must be 100% analog, from beginning to end. If the master is a digital file,then cutting a vinyl disc from it is a digital recording on vinyl. Defeats the whole purpose, I would think. Not that pure analog would necessarily sound better although in theory it could. Maybe there's something elusive that's captured by an analog recording that's lost as soon as it's converted, at any point, to digital.Something that we don't know how to specify or measure because we don't know what it is.

tonykaz's picture

Geez, I just had a little trip down Memory Lane with those Thiel CS1s.

Which -- prompted me to have a look at Mr. Atkinson's history with Amplifiers.

Hmm, what happened to that Krell Amp you brought back from England?

I discovered Mono Amps to be the Ideal as seems the case with ya'll. I'm now going further down that path by thinking that Active loudspeakers are the natural ideal.

Do I see plenty of Ayre gear in your System Photos ?

I suppose we all prefer what works well and kinda stay with it, is Ayre a worthy Brand to build with and stay put? I'm getting that idea from Stereophile writers.

I hope you review those new Schiit Mono Amps & matching Pre-amp.

It's fascinating to read that the MBL Amp is Class D

Tony safe in Michigan

John Atkinson's picture
tonykaz wrote:
Hmm, what happened to that Krell Amp you brought back from England?

I still have it and occasionally use it. Recently I took it to John DeVore's place to use for a screening of the movie "Forbidden Planet," which was where Dan D'Agostino got the name "Krell" from.

John Atkinson
Editor, Stereophile

David Harper's picture

Isn't that movie great? Maybe someone (Ridley Scott?) will make a new remake. Think about that!!!!

pma's picture

John Atkinson wrote


Overall, Ayre's QX-5 Twenty digital hub offers superb measured performance.

and I have to say that it is really difficult for me to agree with his statement. I understand that there is to be a kind of 'political correctness' in reviewers writing, however the measurements of non-linear distortions and filter mirrors and aliases say the opposite.

Johnnyjajohnny's picture

I have to agree. This DAC measures shockingly poor, and then at this extraordinary price! "Extortion" is really the right word here.
Someone once said that two segments of hi-fi products are extremely poorly made: The exceptionally cheap products, and the exceptionally expensive products. Ayre seems to be an example of this. But this is probably one of the reasons why digital haters prefer Ayre to so many other products, because they are as imperfect as turntables and cartridges - they muddy up the sound, distorts and twists it so it finally sounds as poor as analogue. And I'm saying this as someone who collected vinyl records for 15 years.
Compare the Ayre DAC to the €250 Topping D50s (yes, only €250), which measures infinitely much better. The highest harmonic in that DAC is at -120, compared to -65 dB in the Ayre. And the filters in the Ayre are terrible as well!
I suspect that the roll-off with 44.1 kHz data has conciously been made that way to "show" listeners that the format is poor and that hi-res sounds better - which it doesn't with a competent DAC (there will be no audible difference).
And there are many, many other cheaper products that measure almost as well as the Topping. The Odac, the SMSL SU-8 or the Cambridge Audio DacMagic are just three examples. Granted, not all of these were available at the time when the Ayre came out, but there were PLENTY of other DACs that were much better made. And it's honestly sad that this product is being celebrated so much - even by John Atkinson, who sees all these artefacts.

dumbo's picture


In the measurements section of this review you mention that

"Apple's AudioMIDI utility revealed that, via USB, the Ayre accepted 24-bit integer data. The optical inputs locked to datastreams with sample rates up to 96kHz, the AES/EBU and S/PDIF inputs to streams of up to 192kHz."

My question is if this seems strange or not given the fact that the DAC Chip is a 32bit device? Is there something in the connectivity chain that would prevent your PC from seeing it as a 32bit device or is there something that Ayre did during the design process that would prevent an End User from "seeing" the fact that the DAC Chip was a 32bit device?

I'm thinking along the line of wanting to use JRiver's 64bit software VOL control. If the Ayre could accept a 32bit feed from some external source that had VOL control capabilities then one could put the Ayre in DAC Mode and bypass the sub-optimal internal VOL control while at the same time maintaining 24bits worth of resolution for a long as possible (ie..maybe -60dbfs) in a DAC Direct to AMP configuration.

On the other hand, if for some reason the QX5 does only accept 24bit Data then I'm left scratching my head as to why they bothered with a 32bit DAC in the first place.


John Atkinson's picture
dumbo wrote:
if for some reason the QX5 does only accept 24bit Data then I'm left scratching my head as to why they bothered with a 32bit DAC in the first place.

I'll put to one side whether or not the QX-5 accepts 24- or 32-bit data via its USB port. (I don't have a way of examining the behavior of its Ethernet connection.) But the reason for having a DAC chip that will decode 32-bit data is that as soon as you operate on the data before they are presented to the chip, you will increase the bit depth. For example, if your volume control works with 24-bit coefficients, it will output 48-bit data with 24-bit audio. It is optimal to reduce the bit depth as little as possible, even when redithering as you do so. So in theory, reducing 48 bits to 32 bits is better than reducing it to 24 bits.

John Atkinson
Editor, Stereophile

dumbo's picture

Thanks for your reply JA, much appreciated.