Expert review and test of the Schiit Bifrost Multibit DAC by Australian Hi-Fi Magazine. Free pdf download included.
Remember when the first two CD players appeared on the market? The Sony offering was a 16-bit unit, albeit with a DAC that switched between each channel, misaligning them by a small amount. (*See footnote.) Philips, though, felt that 16-bit DACs weren’t sufficiently accurate at that point, so it launched with a 14-bit DAC. The inherent 12dB higher noise floor resulting from that 2-bit shortage still put it below the noise floor of just about any phono pre-amplifier available at the time.
The problem was linearity. Was it possible to create a circuit that could accurately reconstruct analogue signals from digital numbers which spanned four orders of magnitude? That issue soon became moot by means of a nifty bit of side-stepping. As silicon got better (hey, compare the capabilities of a modern computer with an early 1980s model: it’s at least three orders of magnitude faster and more powerful). Instead, oversampling and delta-sigma conversion became the norm. In effect these turn PCM to something like Direct Stream Digital, allowing for a very simple final DAC stage. (Yes, there are additional complications.)
This also dealt with the complexities of creating an extremely sharp anti-aliasing filter, and the drawbacks of such filters.
I bring all this up because the Schiit Bifrost DAC reviewed here—a high-value, high-performance DAC designed for PCM signals up to 24 bits and 192kHz sampling—is the Multibit version.
The standard Schiit Bifrost DAC ($649) uses the Asahi Kasei AK4490, a 256× oversampling delta-sigma DAC chip. The review Schiit Bifrost Multibit model is $979 and uses a very different DAC technique. Since the Bifrost is upgradeable, you can convert the former to the latter, but the board will cost you $549. This is not a user upgrade. Your Bifrost will need to go back to the distributor for installation, and the necessary firmware upgrade for the motherboard.
The only visual differences between the Bifrost and the Bifrost Multibit are stickers, one on the box and one on the back panel of the unit.
I will return to the Multibit thing shortly, but first a little about the unit. It’s a mid-sized (229×58×172mm), three input model that weighs 2.3 kilos.
There’s a USB Type-B socket which you use with a computer (or, if you’re that way inclined, a smart phone) to turn digital audio into analogue, plus an optical input and a coaxial digital audio input. All three support the resolution mentioned—192kHz, 24 bits. There is no support for Direct Stream Digital (neither, apparently, is there in the basic Bifrost, even though its DAC chip supports up to quad speed DSD).
Output is via standard RCA sockets.
The top, bottom and front of the case are formed from one folded sheet of 3mm aluminium.
Three small indicator lights on the front indicate which input has been selected, while a button cycles through the inputs. There are no indicators for showing sampling frequency, so unless you’re using audio software that makes a positive statement about what is being delivered to the DAC, you’ll be proceeding somewhat on trust that you have your computer configured correctly and it is not resampling the audio. A small hard-wired switch on the back turns the power on and off. The FAQ in the slim manual says that you can leave the Bifrost switched on ‘all the time’.
That manual (and the Schiit website) are rather cute. On the back cover next to the company name is a tag line: ‘It Happens’. It notes of the seeming ventilation holes on the top that they ‘Don’t Go Through’, saying: ‘Hey, you try getting FCC-friendly radiated noise on a DAC that has vents in the cover. Go ahead, try. But hey, they look cool, right?’ I confess to having a soft spot for a company that doesn’t take itself too seriously.
What is serious is the technology inside. Instead of an off-the-shelf audio DAC chip, Schiit has gone for an industrial chip, not designed at all for audio applications, and has built its own circuitry around it. Schiit is very up front about this. It describes the system as using ‘Schiit’s proprietary closed-form, time- and frequency-domain optimised DSP-based digital filter… coupled to a precision Analog Devices AD5547CRUZ digital-to-analog converter—a D/A never before used in any other audio product.’
Is this some form of heresy? Shouldn’t a high-quality audio device use components designed for audio products? The data sheet for the AD5547 (Rev.D) suggests as applications test equipment, instrumentation, digitally controlled calibration and digital waveform generation.
It seems to me Schiit isn’t treating digital audio signals as being imbued with some magical musical essence that will be lost without uncertain, golden-eared hand-tuning. No, it’s treating them as things to be converted as accurately as possible into the original analogue signals from which they were formed. It is bringing engineering to bear on that task.
The AD5547 is an R-2R ladder-based DAC chip. That is, good old fashioned technology, but capable of extreme precision thanks to the three-and-a-half decades of development since those early CD players. Since the chip is a 16-bit model, and the DAC is rated at supporting 24-bit signals, I’m guessing that some oversampling still comes into play. The chip actually has a bandwidth of 6.8MHz. But that’s pushing the bounds of my comprehension. What magic this ‘proprietary closed-form, time- and frequency-domain optimised DSP-based digital filter’ pulls off, we’ll see in the test results.
Schiit rates the frequency response of the Bifrost Multibit at 20Hz to 20kHz ±0.1dB, and 2Hz to 150kHz ±0.5dB. The output is rated at 2 volts RMS. THD is rated at less than 0.005% across the full audio bandwidth, while IMD is rated at less than 0.008% and the signal-to-noise ratio is rated at greater than 109dB (referred to 2 volts RMS).
Finally, I will note that Schiit has made its DACs modular and upgradeable. So should there be significant improvements available in the future in either the digital or analogue sections, the internal components can be changed. This, again, would be a job for the distributor. Schiit would like its customers to note however, that although the Bifrost has a 5-year warranty, this covers only labour and parts in the unlikely event that a repair is required: it does not entitle you to upgrades.
In Use and Listening Sessions
As seems to be now standard, no drivers were included in the box. The manual—presumably not too recent a publication—says to go to a specific link on the manufacturer’s website to install them.
But you may not need drivers anyway. Macs have supported USB Audio Class 2 for ages, and so (belatedly) has Windows, since early 2017. The driver page says: ‘Most newer Windows versions already have drivers for our stuff.’ I found that Windows support unsatisfactory when first introduced, but whether it’s this particular DAC, or the several Windows 10 upgrades since then, there were no problems this time. I plugged the Schitt into my desktop computer and instantly the usual USB ‘device connected’ sound emerged from the (already powered up) speakers. I checked the Windows Manage Audio Devices dialog box, and there it was, set as the default device.
So I fired up Foobar 2000, went to its output settings and found the Schiit listed there with three options. ‘DS: Speakers (Schiit USB Audio Gen 2)’, ‘WASAPI (event): Speakers (Schiit USB Audio Gen 2)’ and ‘WASAPI (push): Speakers (Schiit USB Audio Gen 2)’. The first of those simply runs the sound through the standard Windows audio processes, and will resample it if necessary to match the output sample rate you’ve set in the Windows dialog. For hi-fi listening, this is best avoided, but is useful to have if your machine is a general-purpose computer and you need to hear notifications and such.
But for the best performance, you’ll want one of the WASAPI settings. WASAPI stands for Windows Audio Session Application Programming Interface. The standard Windows audio handling has to mix your music with beeps and buzzes from your system, audio from YouTube videos and all that stuff.
So it all has to conform to the same bit depth and sampling frequency. Which means that your audio is often resampled by Windows into a different PCM sampling frequency and bit depth.
What you actually want is bit perfect transmission of the digital audio to the DAC.
The WASAPI settings bypass most of that, and allow the music player to seize control of the audio processes entirel—making them ‘exclusive’. You won’t hear those other sounds. Your 176.4kHz high res audio won’t be resampled to something else. Apparently the handling of WASAPI Event is even cleaner than WASAPI Push, but there’s likely nothing much in it. Just choose one or the other.
There was no ASIO (Audio Stream Input/Output) option, and no need for one.
ASIO is often used to deliver DSD, but since this DAC does not support DSD, it is not required.
Interestingly, switching between sample rates induced a small click from within the unit, as though a relay were activated, suggesting some physical difference in signal routing within the unit according to sample rate. Windows reported that the DAC supports sample rates of 44.1kHz, 48kHz, 88.2kHz, 96kHz, 176.4kHz and 192kHz with resolutions of 16, 24 and 32-bits.
What’s that about 32-bits? Schiit only claims 24-bits. Well, I played music and sounds from Windows with the 32-bit output setting in place, and it worked, so I can only assume that the DAC simply ignores the eight least significant bits.
Schitt’s crack about ventilation holes (or the lack of them) piqued my curiosity so I kept tabs on the case temperature and found it ran only modestly warm and since the AD5547 chip is rated as being suitable for operation at temperatures between –40°C and +125°C, I don’t think there are going to be problems on this front.
I am not going to wax lyrical about the sound quality produced by this DAC, not because it doesn’t deserve it, but because the engineering approach of the company is one of high utility, rather than poeticism, which is the way I like it. The fact is, all music the DAC was capable of supporting sounded as good as it has from any equipment I’ve ever used, and some of that has involved DACs priced at half-a-dozen times the price of this model.
There were simply no defects. There were no weaknesses. My listening involved music with all those sampling frequencies, often switching from one to the other between tracks, and there was no switching noise through the speakers, only that little relay click in the body of the DAC.
Any high-end system that relied on this DAC would be well served.
Let’s face it, when it comes to noise, computers are horrible, horrible things. That’s one of the reasons why one prefers an external DAC: to remove as far as possible the analogue circuitry from the internal electromagnetic stew that fills a computer enclosure.
One of the things an external DAC has to do is keep all that out. So I was a little surprised that my first test produced a signal-to-noise ratio of a modest 99.7dB A-weighted. Examining the noise spectrum, it was clear that there was excessive random noise in the frequencies below 1kHz.
That was with the DAC plugged into my desktop Windows computer via USB. So I switched over to a Mac Mini, and that yielded a measure of –104.8dBA, with a drop in the noise floor of slightly more than 20dB in that bass and lower midrange band. That said, there were still noise spikes at 50Hz and multiples thereof, some of them 20dB or 30dB above the noise floor. A third measurement, this time using the Mac Mini again, but this time feeding the signal via optical cable to the DAC, gave much the same results as with the USB connection.
So, finally, I retained the optical connection and disconnected the USB cable entirely, and again scored a similar headline result (–104.9dBA), but the spectrum showed significantly reduced (for the most part) mains noise, down 20dB for the 50Hz fundamental.
All of which goes to suggest that the USB connection is not fully effective at isolating the system from whatever electrical noise the USB input is being infected with by the connected computer. That’s something I’d like to see corrected.
But let me add, I was using 24-bit signals for those tests, and all those noise levels are below the noise levels you’d typically get from 16-bit audio.
And the noise intrinsic to 16-bit audio is for all practical purposes inaudible. Distortion levels are low. I measured around 0.005% THD and 0.008% IMD, which is bang-on spec. For 96kHz sampled signals, the high-end frequency roll-off is gentle, starting around 10kHz with the output down 0.3dB at 20kHz, 1.3dB at 30kHz and 3.3dB at 40kHz. As far as I could see, the exact same filter is used for 192kHz signals, with output down by 6.8dB at 50kHz and 11dB at 60kHz.
With 16-bit, 44.1kHz signals the noise floor was surprisingly low: –99.7dBA. That was measured with a USB connection from the Mac Mini. The coincidence with the 24-bit Windows signal is indeed just coincidence. That noise floor was lower than normal for 16-bits, although there were some peaks due to the presence of 50Hz and harmonics. The frequency respose was –1dB point was at a very high 21.6kHz.
The Schiit Bifrost Multibit DAC is a fine DAC for those who aren’t possessed of a collection of Direct Stream Digital tracks. It provides high performance. It’s excellent value for money. And I have a real sense that while this DAC is perfectly competent with high-resolution audio, it has been optimised for 44.1kHz, 16-bit sound. Since almost all our music collections are dominated by that standard, that’s very welcome. # Stephen Dawson
*Footnote: I recall the dark mutterings regarding the damage this must do to the sound. In fact, it was a group delay for one channel, easily corrected by putting the delayed speaker 6.8 millimetres closer to the sitting position… if you could even tell which one it was.