Laiv Harmony DAC
Way back in the 20th century, Sony and Philips set the CD standard at the maximum for current technology. A bit depth of 16 bits was the limit of what could be achieved for recording, storing and decoding as an affordable commercial product and was deemed adequate for good sound quality, providing 20kHz of bandwidth. A sample rate of 44.1kHz was chosen to align with video recorders, then used to store digital audio data. Similarly, 48kHz 16 bits was set for professional recording.
Early DACs used resistors to switch the output voltage between to 16 very accurately weighted levels, creating 65,536 steps, in order to decode digital CD data. If only one step were not set accurately, what should be a linear transfer function would not be smooth, giving rise to non linearity and distortion. The best way to implement these voltages accurately was a ladder-like series of resistors having only two values, known as R-2R ladder. Even so, it was difficult to achieve accuracy in the smallest two bits, so some ladder DACs were only built to 14 bit, (but actually resolved to 16 bits with 8 times with over sampling).
All DACs inevitably introduce discrete steps in the waveform at the sampling frequency, these also manifest as spurious ultrasonic images which are normally completely removed by a smoothing, or reconstruction, filter. To do its job, the filter for a 44.1kHz sample rate, (if not oversampled) must be extremely steep from 20kHz to 24.1kHz, the lowest frequency where ultrasonic images appear, potentially introducing phase, amplitude and timing errors (pre and post ringing), which all have their influences on sound quality. However, not having a filter would allow ultrasonic images to pass unhindered, so conventionally, various kinds of filtering are included to eradicate both the steps and their ultrasonic partners.
In the forty-odd intervening years, digital engineers have turned to alternatives which permit gentler filtering, such as oversampling (which shifts the ultrasonic images to higher frequencies), bit stream, one-bit DSD, delta-sigma processing, digital filtering and noise shaping. Since then we have of course also seen recording and streaming at much higher native rates, with greater bit depths in pursuit of wider dynamic range, lower distortion and noise and higher sound quality.
Despite the technical issues, resistive ladder (R-2R), DACs have retained a small but firm following, but why use them in 2025 if it’s hard to do, and why provide a non-oversampling option? Is it just nostalgia – rejecting the technological advances of the past 40 -odd years and turning the clock back to 1985!
Well, not really; the answer is that advocates of R-2R DACs report a more direct connection to the music, and a reputedly more analogue-like sound. However, the main limitation is still the accuracy and stability of the voltages provided by the resistive ladder.
Laiv Harmony
Enter the Laiv (pronounced Live) Harmony R-2R DAC, designed by a team in Singapore headed by Weng Fai Hoh, and manufactured in China. Don’t look for multiple reconstruction (smoothing) filters – there aren’t any, just the option to run in oversampling (OS) and non-oversampling (NOS) modes. In NOS mode, the DAC runs at 44.1 kHz, 48kHz, 96kHz or 192kHz, as per the incoming native rate. In OS mode the Harmony converts all incoming rates, by 16 times oversampling to either 705.6kHz or 768kHz. DSD is only available via the USB connector and these one-bit signals are converted by an Intel FPGA to multi-bit at 768kHz (16x 48kHz), for decoding by the R-2R DAC, as this is Laive’s preference.
According to Weng, the Harmony, in OS and NOS modes resolves around 15-bit resolution at full scale, with an actual dynamic range of about 110dB (approximately 18 bits), based on AP (Audio Precision) testing. He also confirmed that in NOS mode, there really is no oversampling filter in place, meaning it does not remove any ultrasonic, or imaged products, leaving the quantising steps in the waveform..
When I asked why he chose an R-2R DAC, Weng replied “R-2R DACs are indeed more challenging to design compared to delta-sigma models, as they require high-precision resistors with extremely tight tolerances. However, the key advantage lies in their ability to deliver a more natural, organic, and analogue-like sound—something we personally love and wanted to share with fellow audiophiles. It’s all about preserving the purity of the music and bringing out the most engaging and lifelike listening experience.”
Unboxing
Right from opening the box, this little beauty impresses. The packaging has great production values, and the DAC is solidly constructed, using a CNC-machined case, with a gold-finished control knob on the right, an adjacent enter’ button, a large clear, well-lit blue display, which uniquely shows the track number and playing time (which I love), and an on/standby button at the top-left corner.
Finish and build are superb. I would put them at the standard of DACs of at least twice the price. It sits on three cone-shaped feet. To avoid scratching the supporting surface, three recessed pads are provided, together with an alignment card for positioning them. Inputs at the back include coax, Toslink optical, I2S via HDMI connector, and USB, plus RCA phono and balanced XLR outputs and mains input via a standard IEC connector. The Harmony has an equally robust, weighty remote, machined from solid aluminium (including the battery cover), and adorned with 12 gold-finished buttons which select inputs, muting, signal inversion, and over and non-over-sampling modes, plus display illumination and unit on/off. Like the DAC, it’s a thing of beauty and a joy to use.
Inside the Harmony DAC
There are a number of modules inside the Harmony, including an SPDIF receiver for the coax input, and a galvanically isolated USB input module. There’s also an I2S receiver, for use with transports having this output option. Input for this is via a standard HDMI connector. The user can select either the DAC’s clock, or the I2S clock, and to overcome the lack of a standard for this connectivity, the Harmony offers eight automatically selectable pin-connection options, and will only select this source if it is compatible. Unfortunately, I did not have an I2S transport with which to test this input.
The R-2R network comprises discrete resistors (not an integrated circuit), which are selected to 0.05% tolerance, to ensure good linearity, housed in two slender, vertical modules. Audio outputs are provided via both balanced XLR and unbalanced RCA phono sockets on the rear. In pursuit of their purist philosophy, these are connected directly to the R2-R network. There is no output filter and no active output buffer to colour the sound. With output impedances of 1200 Ohms for the XLR and 600 Ohms for unbalanced, this leaves the Harmony sensitive to amplifier input impedance. Care in matching is therefore advised. For instance, passive potentiometers, or preamps with low input impedances, could introduce non linearity due to loading the R-2R.
On the rear panel, in addition to the audio connections, there’s a link to connect or separate the case from the signal earth to improve grounding which can have a significant effect on sound quality.
The sound
Laiv recommend 100 hours of use for running in, however, impetuous as usual, I plugged it in, fired it up, set to NOS, and right from the start it made a big impression. I had connected my trusty re-clocked Teac VRDS 10 SE transport to the Harmony via digital coax, and popped a CD of Dance of the Tumblers from Rimsky Korsakov’s The Snow Maiden into the tray (Reference Recordings, and of course 44.1k/16 bit), and pressed play. I really was not prepared for the explosive dynamics and solid powerful bass which thundered out of my Kudos Cardea 20 Anniversary speakers, via my Musical Fidelity A3 CR power amplifier, and Burston Conductor 160D headphone amp/preamp. It was quite astonishing and viscerally exciting.
To be fair, this is a bit of a show-off piece, but I have rarely heard it sound quite like this – as though the Harmony somehow connected me very directly to the live performance. I then switched to OS on the same section, to find the rendition noticeably less propulsive and dynamic, with slightly softer bas and treble sweeter. It was still superb, but less exciting.
Next in the CD draw was Blue Rondo a la Turk and Take 5 from Dave Brubeck’s Take 5 album (legacy, not re-mastered). Again there was terrific presence and directness with no oversampling, like hearing through the medium of storage to the atmosphere of the original recording session. Switching to OS created a sense of refinement, with smooth textures to snares, cymbals and the like.
Switching back to NOS brought back that sense of directness. It sounded more alive and the performers were somehow more holographically present, but this mode revealed a thinner tonality and was a little rough around the edges. The Harmony’s remote control made it easy to switch seamlessly and rapidly between the two modes, and so the differences were made very clear. I loved the raw excitement and directness of NOS, but oversampling sounded more refined, as though the rough edges had been eliminated to reveal a more sophisticated presentation, particularly of the higher frequencies. So, neither mode was perfect but both offered something special.
Next, I played Cottontail from a JVC CD of Duke Ellington’s Duke’s Big 4. This is no ordinary CD, being produced using JVC’s XRCD, a proprietary process involving over-sampling at 128 times to 20 bit resolution for recording, and downsampling, without using noise shaping, to 44.1kHz, ensuring, they claim, true 16 bits of resolution to preserve low-level information for transfer to silver disc. Again, this sounded more alive in NOS than OS modes. Oversampling in the Harmony gave it a subtler and more languid presentation; for instance, cymbals had a warmer, fuller, more rounded quality.
Switching back to NOS, cymbals were crisper and lighter. I could hear more of the ring and swish from the hits and strokes. So, in this instance, I preferred to leave the Harmony set to NOS. Although the treble was not as smooth, I didn’t feel the need to tame it with OS, as it would have lost the lively, direct presentation.
This was not always the case: with some CDs, the treble became rather too rough and untidy. With these, the smoothing from oversampling was necessary for me to sit back and simply enjoy the music. There were a number of CDs like this, for instance, West End Girls from the Pet Shop Boys Please album was simply too thin and harsh in NOS, and so I switched to OS to tame the dominant upper harmonics. Similarly, with Brian Ferry’s Slave to Love, for example.
With JVC’s Duke, it was the other way around. Oversampling diminished the sense of being there, so it was too much of a compromise. After playing many CDs I came to the conclusion that my preferences were quite variable. I suspected this was at least partly due to differences in the care paid during recording and mastering, the contrast between music featuring acoustic instruments recorded naturally, as opposed to electronic instruments in a studio. For instance, Reference Recordings, are known for the attention they pay to recording and mastering, and the Dance of the Tumblers was definitely preferable in NOS; but it was not that simple.
I loved the smoother deeper presentation of oversampling, but also the lively and dynamic non-oversampling, tempered by its more ragged treble, which was dependent on which CD I played. Unfortunately it seems one can’t have both at the same time. It’s a little like having comfort and sport mode when motoring, and selecting whichever is appropriate. When playing CDs, I would be perfectly happy to cruise along in OS and switch to NOS when the recording or mood warranted it.
It seems NOS is a double edged sword. With synths and studio produced recordings, at 44.1kHz sampling rate, there is likely to be more treble energy, producing stronger unwanted images just above 20kHz with NOS. On the other hand, with acoustic instruments there may be less high-frequency energy and therefore this is less of an issue. Either way, fewer CDs were listenable with NOS.
Unlike many modern DACs there is no choice of reconstruction filters to fine tune the sound, and It was not even possible to compare the sound of NOS with and without filtering, which would have been interesting, and possibly useful when playing some CDs. It’s a pity Laiv decided not to include at least an on/off filter switch, so listeners could decide for themselves, but they clearly believe that having no filter is the best option, presumably because this would have added extra circuitry, instead of pursuing their purist approach.
I also needed to try some higher resolution material, so I pressed into service the Primare SC15 Mk2 streamer, which I used to stream Qobuz, feeding its digital coax output into the Harmony. During the opening Questo Par Rosso, from the remastered 1972 Decca recording of Puccini’s La Boheme (Berlin Philharmonic, conductor Herbert Von Karajan, with Pavarotti, Freni and Ghiaurov – 24bit/96kHz), my initial preference was for the oversampling mode, with its lusher, rendition of the orchestra and how the voices were set back a little further on the stage, (whereas oversampling was more forward and the stage more brightly lit). So, initially the oversampling won, for me. However, as the opera continued, my point of view shifted, and notably the at the opening of the second act, the chorus and orchestra sounded so fresh, vibrant and alive, I felt this was markedly superior, as though a haze had been lifted with NOS providing a clearer ‘view’ of the stage.
Playing Beethoven’s Symphony No 9 (Mito Symphony Orchestra conducted by Seiji Ozawa on Decca), the 24 bit 96kHz recording, in OS mode produced a beautifully lucid and fluid sound from the orchestra. The choir lacked any sense of glare, blare or unnatural sheen, and there was a lovely sense of depth. Switching from OS to NOS brought about a more direct sound – and while tonality was slightly thinner and brighter, the strings seemed fresher and more expressive in the largo, but quite honestly it sounded excellent either way.
I tried more tracks at high resolutions, also with excellent results, and one dramatically illustrating wide sonic variability was Paul Tortellier playing Bach’s Suite No 1 in G Major BVW 1007, on BNF records (Bach: Six Suites for Cello, at 24bit/ 96kHz sample rate). Playing this file the sound was so fresh, his cello so alive and the acoustic impressively realistic in NOS mode that the switch to OS was noticeably flatter and less convincing. It was a stark comparison.
NOS replay was clearly the most energetic, exciting and engaging, as if one could hear all the strings vibrating to create complex overtones, along with the body of the instrument directly energising the atmosphere with pure raw energy, and creating multiple reflections in the reverberant acoustics. Such high sampling frequencies displace aliases much further from the audible frequency band, making a reconstruction filter less essential, or maybe even deleterious. In any case, at these much higher rates, I could easily leave the Harmony in NOS mode and switch to OS only occasionally.
Summing up
The Laiv Harmony DAC impressed me in NOS mode for its lively and dynamic character, particularly higher resolution material, seemingly connecting more directly to the music. Whereas, with CD material, I often preferred reproduction when oversampling, which created a smoother, warmer and deeper rendition of the event. Which sounded best, depended very much on the recording quality, music type and sample rate.
I would be more likely to leave the Harmony set to NOS when playing high resolution material. The recordings to benefit most from NOS, were of acoustic instruments in a natural acoustic, when it seemed to dig right down into the quietest clues to deliver micro and macro dynamics, and in doing so, bring the listener closer to the event. In which case, the Laiv Harmony sounded remarkably alive.
However, results may vary according to partnering equipment and files played: users may have different experiences, and not everyone may warm to its raw, detailed energy. So, the Harmony may divide opinions. Accordingly, auditioning with the same amplification as in one’s own system, or intended system, would be advisable before purchase.
Test & measurement
I thought there might be sound technical reasons for these subjective traits. I would not normally run measurements on a DAC, but the Harmony is so unusual, that I felt compelled to. They are by no means comprehensive.
FFT analysis of CD replay, with Clio 13 showed the sampling frequency at 44.1kHz, plus images around it, extending from 24.1 kHz to 64.1kHz. This is not a fault, but classic behaviour for a non-oversampling, unfiltered DAC (due to having no re-construction filter). The ultrasonic tones should not be audible in themselves, but if intermodulation were to occur in the amplifier or speakers, this could create audible in-band distortion.
Harmonic distortion in the Harmony was a low (0.005% THD measured for -1dB at 1kHz with a test CD), but when the second harmonic distortion frequency exceeded 20kHz, this harmonic tone intermodulated with ultrasonic information creating difference tones below 20kHz. This only occurred for inputs between 10kHz and 20kHz. Below 10kHz, the second harmonic fell below 20kHz and consequently could not intermodulate in this way. So this is purely a high-frequency effect. Fortunately, upper harmonics of acoustic musical instruments are much lower in level than fundamentals, (though this may not apply so much to electric instruments, synthesisers, etc.).
This could explain the variable results with the Harmony and why some CDs, (with more high frequency energy) sounded more pleasing with over-sampling, because this would have pushed all the ultrasonic information to much higher frequencies, and therefore there were virtually no intermodulation products falling into the audio band. Likewise, with higher sampling rate material: FFT testing at 192kHz sampling rate (bandwidth 92 kHz) also showed very low levels of ultrasonic information below 92kHz and virtually no intermodulation products falling in band in either OS or NOS modes. These results actually tie in very well with my subjective findings (as oversampling tended to sound cleaner and smoother at high frequencies when playing CD based files or CDs).
Sine waves with 44.1kHz sample rate and NOS clearly showed the quantized sampling steps at 44.1kHz, which would normally be removed by a reconstruction filter. With oversampling, the sine wave looked extremely smooth, even at 20kHz, with just a trace of high frequency noise super-imposed, showing oversampling doing exactly what it’s supposed to do.
The frequency response and group delay at 192kHz sample rate in OS and NOS were virtually flat up to 89kHz (with very minor deviations up to 92kHz). This meant that the signal was merely delayed, by slightly different times for NOS and OS, so there was virtually no phase shift of high frequencies relative to lower frequencies. According to Weng the delay is caused by a FIFO buffer used to improve jitter. Tone bursts and square waves looked great too.
The Harmony resolved a -100dB dithered sine wave 16 bit CD test tone, at 1 kHz, with its spectral line showing clearly above the noise floor in the FFT (despite the sine wave’s shape being dominated by dither noise). This indicated resolution at the very limit of the 16-bit test CD (at least 16 bits). All in all an interesting and informative set of results.