Audio Research I/50 integrated amplifier The DAC Module

John Atkinson reviewed the DC1 DAC module in February 2024 (Vol.47 No.2):

I concluded my September 2023 review of this tubed integrated amplifier by writing that the Audio Research I/50 "is the perfect amplifier for use in a relatively small room with a high-quality source and speakers, especially if the latter have a fairly high impedance. The I/50 has a touch of that 'tube magic' but without going whole hog, as so many of the current crop of tubed amplifiers do." The review sample just had balanced and single-ended line inputs and costs $5500, but toward the end of the review period I was sent the optional DC1 DAC module, which adds $1000. As well as coaxial and optical S/PDIF inputs, this module has a USB port and a Bluetooth antenna to allow streaming from computers and smartphones. The module appears to use Asahi Kasei Microdevices' 32-bit, two-channel AKM 4490 DAC chip.

Audio Research's Dave Gordon emailed me a 22-page pdf that offered comprehensive instructions on how to install the DC 1 DAC module. When I started reading the pdf, I was alarmed to read "Installation of the DAC Module should only be attempted by electronic technicians or other qualified service personnel. Audio Research strongly recommends owners DO NOT attempt to install the DAC Module." However, as I have designed and built a number of electronic products over the years, I decided that I did qualify as "service personnel" and set about the installation. I wore a grounding strap on my wrist to make sure I didn't inadvertently zap the module with static electricity.

The module comprises a double-sided printed circuit board stuffed with both surface-mount and through-hole components. One end has a plate with the input ports and antenna connector; the other has two multipin connectors. I placed the I/50 upside down on my test bench—it was stable sitting on its transformer cover—and removed the bottom panel with its fan, carefully unplugging the cable that connects the fan to the amplifier's main circuit board.

Before I started the installation, I noted that the instructions warn that the I/50's high-voltage power supply "may contain hazardous DC charge for several hours after the product has been turned off!" I connected a Fluke DC voltmeter to the test points described in the pdf and found that even though the amplifier had not been used for a couple of weeks, 39.5V was still present. I used the 1k ohm resistor that was included with the module kit to discharge this residual voltage.

Once the voltage had dropped below 1V, I removed the blanking plate on the I/50's rear panel, replaced a grounding cable with a shorter one to make room for the DAC module, and mounted the two nylon standoffs, checking my work as I did so. I then plugged the two multiconductor cables in to the module's connectors and those on the main board and inserted the module as instructed, taking care to dress the two multiconductor cables so that the connections remained secure.

Page 6 of the pdf says, "if the I/50 is used with high-sensitivity loudspeakers," a 2.2µF capacitor that is included with the module "may need to be installed to reduce noise from the DAC module, especially with the Bluetooth input." As the review sample of the I/50 already had this capacitor installed, I was able to skip this step.

Holding my breath, I connected a TosLink cable from my Audio Precision SYS2722 system to the module's optical input and powered up the I/50. After the warmup countdown, I selected that input and fed it data. I slowly turned the volume up and was gratified to see a signal appear at the speaker terminals. I breathed in, powered down the amplifier, then replaced the bottom cover, making sure that the fan cable was correctly dressed as described in the instructions.

I used my regular KEF LS50 standmounts to audition the Audio Research I/50's DAC module, connecting the speakers to the amplifier's 4 ohm output taps with AudioQuest Robin Hood cable. The source was initially CDs played on an MBL N31 CD player-DAC, using its TosLink output.

Listening to CD data, the tonal balance was a little darker than I was used to with the I/50 handling analog data. However, even without the 3dB boost in the bass I use with Roon and network streamers to extend the small KEFs' low frequencies, the sound of the I/50's digital inputs had sufficient weight on solo piano recordings. Maria João Pires's piano, on her live CD of Brahms's three Op.117 Intermezzi (from The Wigmore Hall Recital, DG 479 0965), sounded warm, with a satisfying left-hand register. I followed Pires with Robert Silverman performing these delicate piano works on Stereophile's Intermezzo CD (STPH003-2), recorded by Kavi Alexander in a Santa Barbara church in 1990. Silverman's Steinway is lighter balanced than Pires's Yamaha but had the same palpable quality as the Wigmore recording with the I/50's DAC module.

The subtle hall acoustic on both these piano recordings was well-resolved, perhaps a little more so than I was used to. I played the CD of the Portland State Chamber Choir's 2020 album Translations (Naxos 8.574124), which I had co-engineered. The dome of ambience surrounding the musicians and singers was clearly presented, but again perhaps slightly more audible than I was expecting with this very familiar recording.

I briefly compared the sound of CD data being decoded by the I/50's DAC module with the MBL N31's analog outputs fed to the Audio Research's balanced inputs, with the levels matched with the 1kHz tone at –20dBFS on my Editor's Choice CD (STPH016-2; no longer available). Not a fair comparison perhaps, as the N31 costs $18,500, but the expensive player's analog output with Molto Molto, a CD of three works composed by Sasha Matson for jazz orchestra (Stereophile STPH023-2), sounded more transparent than the I/50's TosLink input. While the soundstage was less forward with the MBL, the drums at the end of Matson's Symphony No.3 were reproduced with a little more force.

I must admit, however, that the differences between the MBL and the I/50's DAC module were not nearly as great as I had anticipated. It is possible that the sounds of both digital sources decoding CD data were dominated by the Audio Research amplifier's intrinsic character. Even so, props to the I/50's DAC.

For hi-rez playback, I used Adobe Audition to play files at their native resolution and sample rate on my MacBook Pro, sending USB data to the Audio Research DAC module's USB input, buffered with an AudioQuest Jitterbug FMJ. Again, I played Molto Molto, this time the 24/96 files from which I had prepared the CD masters, comparing them with the CD data fed to the I/50's TosLink input. The differences were still smaller than I was expecting, but the lows via hi-rez USB had more weight than they had with optical CD data. Kickdrum had slightly better articulation with 24/96 USB, and the overall presentation had a little more space between the instruments.

I don't have a D/A processor that costs the same $1000 as the Audio Research DAC module, but I do have a $349.95 AudioQuest DragonFly Cobalt. This has a measured resolution very similar to that of the I/50's DAC (see below), though it is limited to sample rates of 96kHz and below. I plugged the DragonFly into the MacBook Pro's other USB port, again via an AudioQuest Jitterbug FMJ, and connected its single-ended analog outputs to one of the I/50's line inputs. (The DragonFly's volume control was set to the maximum to facilitate level matching with the I/50's USB input, and it wouldn't have had any problem driving the I/50's single-ended input impedance of 4.4k ohms.)

Listening again to Matson's Symphony No.3 at 24/96, tonally the two USB sources sounded very similar in the midrange and treble. However, the double bass and kickdrum had more weight with the I/50's digital module, the DragonFly sounding somewhat woolly in comparison. At first, I felt that the soundstage presentations were identical, but after repeated comparisons with a 24/192 needle drop of "Die Tänzerin," from Ulla Meinecke's Wenn Schon Nicht für Immer Dann Wenigstens für Ewig (German RCA 426124), I could see/ hear more clearly into the reverb-drenched picture presented by the Audio Research's USB input, even after I had set Audition to resample the file to 96kHz so it could be played on the DragonFly.

Apple's USB Prober utility identified the I/50 DAC module as "Audio Research HD USB Audio\000\ 000\000\000\000\000\000\000\000\000\00" and confirmed that the USB port operated in the optimal asynchronous isochronous mode. Apple's AudioMIDI utility revealed that the I/50 DAC module accepted 16-, 24-, and 32-bit integer data sampled at all rates from 32kHz to 384kHz. The coaxial S/PDIF input accepted data sampled at rates up to 192kHz, while the TosLink input was limited to 44.1kHz and 48kHz.

The digital inputs preserved absolute polarity (ie, were noninverting) from the I/50's loudspeaker and headphone outputs. With a 1kHz signal at –20dBFS and the I/50's volume control set to the maximum of "45," the level at the 8 ohm output tap into 8 ohms was 3.818V, from the 4 ohm output tap into 8 ohms it was 2.652V, and from the headphone output it was 201.3mV. To avoid clipping the power amplifier stage, I performed all the subsequent measurements from the headphone output. (With this in use, the I/50's speaker outputs are muted, though the 6550 output tubes remain powered.)

Fig.1 Audio Research I/50 DAC module, impulse response (one sample at 0dBFS, 44.1kHz sampling, 4ms time window).

Fig.2 Audio Research I/50 DAC module, wideband spectrum of white noise at –4dBFS (left channel red, right magenta) and 19.1kHz tone at 0dBFS (left blue, right cyan) into 100k ohms with data sampled at 44.1kHz (20dB/vertical div.).

The I/50's DAC module offers a single reconstruction filter. Fig.1 shows the filter's impulse response with 44.1kHz data. There is considerably more Nyquist-frequency ringing after the single sample at 0dBFS than before, something I have seen before with some other D/A processors (footnote 1). The magenta and red traces in fig.2 show the filter's ultrasonic rolloff with data sampled at 44.1kHz. These traces reach full stop-band attenuation just above half the sample rate (this indicated by the vertical green line), with the aliased image at 25kHz of a full-scale tone at 19.1kHz (cyan, blue) suppressed by 48dB. The harmonics associated with the 19.1kHz tone all lie at or below –80dB.

Fig.3 Audio Research I/50 DAC module, frequency response at –12dBFS with data sampled at: 44.1kHz (left channel green, right gray), 96kHz (left cyan, right magenta), and 192kHz (left blue, right red) (1dB/vertical div.).

Fig.4 Audio Research I/50 DAC module, 24-bit data, spectrum, 10Hz–1kHz, with noise and spuriae of dithered, 1kHz tone at 0dBFS (left blue, right red; linear frequency scale) (20dB/vertical div.).

Fig.3 shows the frequency response with data sampled at 44.1, 96, and 192kHz. With the two lower sample rates, there is a gentle rolloff above 10kHz with then a sharp rolloff just below half of each sample rate. The gentle rolloff continues with 192kHz data, the output reaching –3dB at 48kHz and –6dB at 65kHz. There is a bass boost reaching +0.8dB below 50Hz; this was not affected by the level of the volume control. Channel separation was moderate, at >60dB in both directions below 1kHz, falling to 40dB at 20kHz. Though a spurious tone at 60Hz was present in the I/50 DAC module's noisefloor (fig.4), this lay at only –110dB ref. 1kHz at 0dBFS. The level of random noise was somewhat higher than I was expecting and didn't change when I reduced the volume control setting by 10dB.

Fig.5 Audio Research I/50 DAC module, spectrum with noise and spuriae of dithered 1kHz tone at –90dBFS with: 16-bit data (left channel green, right gray), 24-bit data (left blue, right red) (20dB/vertical div.).

Accordingly, an increase in bit depth from 16 to 24, with dithered data representing a 1kHz tone at –90dBFS, dropped the DAC module's noisefloor by around 10dB (fig.5). This implies a resolution close to 18 bits. Odd-order harmonics were present with 24-bit data (blue and red traces), which implies that the least-significant bit was being truncated.

Fig.6 Audio Research I/50 DAC module, 24-bit data, HF intermodulation spectrum, DC-30kHz, 19+20kHz at 0dBFS peak, sampled at 44.1kHz.

Fig.7 Audio Research I/50 DAC module, high-resolution jitter spectrum of analog output signal, 11.025kHz at –6dBFS, sampled at 44.1kHz with LSB toggled at 229Hz: 16-bit TosLink data (left channel blue, right red). Center frequency of trace, 11.025kHz; frequency range, ±3.5kHz.

Intermodulation distortion with an equal mix of 19 and 20kHz tones, each at –6dBFS, was very low, with the difference tone at 1kHz lying just above –100dB (0.001%, fig.6). I examined the I/50 DAC module's rejection of word-clock jitter with high-level 16-bit J-Test data, fed via the USB port and the two S/PDIF inputs. Fig.7 shows the spectrum of the DAC module's output when it was fed TosLink data. (The coaxial and USB inputs behaved identically.) The odd-order harmonics of the undithered low-frequency, LSB-level squarewave lie at the correct levels, though the level of the noisefloor is higher than I usually find. (Because the J-Test signal is undithered, any noise that appears between the odd-order harmonics is due to the product being tested.)

Summing up
The AKM DAC chip can offer excellent measured performance, but in the Audio Research I/50, its ultimate resolution appears to be limited by the amplifier's behavior. However, the DC1 module's sound quality with CD data didn't leave me feeling I was missing anything, and it sounded respectably satisfying with hi-rez files. The DAC module adds value to what is already a relatively high-value tubed integrated amplifier.—John Atkinson

Footnote 1: See, for example, fig.6 here.

AR Tube Audio Corporation (Audio Research)
6655 Wedgwood Rd. North, Suite 115
Maple Grove
MN 55311
(763) 577-9700

stereophilereader's picture

Interesting how your findings re - using the 4 ohm taps echo Martin Colloms' conclusions when he tested Quicksilver tube amps in HiFi News back in the 80's.

Expofan's picture

Nice review.

How would this perfrom with a 16 ohm load? Valencias, for example.

georgehifi's picture

Both these GoldenEar (EPDRs of 2.15 ohms at 59Hz and 1.53 ohms at 135Hz), and Kef(EPDRs of 1.66 ohms between 135Hz and 140Hz) speakers were not the best load for a tube amp such as this to be reviewed with, it may have faired a lot better with speakers with easier loads.
Both the GoldenEar and Kef have EPDR impedance loads down below 2ohms in areas where power is needed measured by JA himself.

Cheers George

Glotz's picture


Jason Victor Serinus's picture

Hi George,

The JVS Music Room Enlargement Fund (MREF) is a mere $25,000 short of its goal. It would be great if you would contribute a large chunk of that, plus extra money for the assistants whom I will need to hire monthly to help me move and properly place two sets of speakers.

Thank you so much,

georgehifi's picture

If your a paid reviewer. Then they are tools of the trade, & "retail" tax deductible with the "right receipt". On top of getting them @ cost, not that anyone does that

Thanks George

texanalog's picture

Does your statement stand in contradistinction to JA's comment that "...neither loudspeaker is a particularly demanding load, with impedances that remain above 6 ohms over much of the audioband,..."?

georgehifi's picture

I understand Stereophile can only use the speakers they have on hand at the time. But maybe it's time to have 2 pairs of resident speakers, easy load, and hard load, to give the readers a bit better indication of how amps really sounds with both loads?

Cheers George

Jason Victor Serinus's picture

I'll be sure to check in with you next time my shoulds list runs low. In the meanwhile, what do you charge for accounting services?


georgehifi's picture

Saying about tradesman and their tools.
Should be, maybe they should just get the right ones.

Cheers George

hollowman's picture

Sorry, this is off-topic, but since this is the most recent entry by JA, I thought it may get prioritized in the "Reply to" email notification system ;)

This query is about JA's Measurements of DACs ... notably:

"Fig.8 DAC brand-model-x spectrum of 50Hz sinewave, DC–1kHz, at 0dBFS into 600 ohms (left channel blue, right red; linear frequency scale)."

I've often noted JA refer to this as a "punishing" test and if a d/a unit performs well, it is "bomb proof".

I assume -- correct me if I'm off -- that this is because DAC's output is not "experiencing" its expected high-Z pre-amp load? Hence, it's output stage is forced to drive more current, concomitantly "overloading" that stage a bit. Yes?

Does the output stage exhibit thermal stress (heat, FLIR glow)? Is this a bit like amplifier clipping?

Thanks in advance for JA/anyone's feedback!

georgehifi's picture

"Does the output stage exhibit thermal stress (heat, FLIR glow)? Is this a bit like amplifier clipping?"

Very much so, there are some amps that have stupid low input impedances 1khom-10kohm as yes this taxes some sources output stages./ Gone are the great days when the industry standard input impedance for amps was 47kohm, and tubes were 100kohm and anything could drive them.

Cheers George

hollowman's picture

Note that my query was specifically about JA’s 600 ohm DAC and/or disc or file-player test.

JA may (???) have introduced this test after RH left. Not sure??? Even JA's early digital reviews did not include the 600-ohm test, but it got added in later.
Pre-amps and headphone amps also include similar distortion tests, but not the "low Z" load variety.

MhtLion's picture

Many tube amplifiers I know drives speakers better than what their spec wattage may suggested. However, this particular amp may be opposite? Not sure how to interpret that this amp struggles a bit at the specified impedance terminals. My question is does I/50 exhibit clipping with the moderate load which is normal for many modern speakers?

liguorid42's picture

"I was alarmed to read "Installation of the DAC Module should only be attempted by electronic technicians or other qualified service personnel."

Since the amp chassis contains high voltages, they obviously needed to cover their posterior.