Laiv Harmony R2R DAC Review & Measurements

 

Review

For the full review and my thoughts on the sound watch the video below!

Watch the video

Design and I/O

The Harmony’s most striking aspect is no doubt the physical build. Sporting a chassis made from a single piece of milled aluminium in either matte black or silver, and gold accents for the controls. This combined with its eye-catching shape makes for what in my view is quite a gorgeous looking device!

The display is sharp and quite a bit larger than what’s found on most products, which makes it easy to read at a distance if using this in a speaker setup.

Whilst there aren’t too many settings or things to change, and no built-in volume control, it did make navigating the menu and configuring the device very easy.

The rear of the Harmony is fairly typical for a DAC, offering balanced and unbalanced outputs, and a selection of USB, SPDIF (in optical or coax form) and I2S inputs. What is a little unusual however is the particularly high output impedance of 1200 Ohms on XLR and 600 Ohms on RCA. If we take a peek inside, we can see why; the DAC isn’t actually using a dedicated output stage or buffer! The outputs are connected directly to the ladder itself.

This is an interesting design choice, and one that some could argue is beneficial in terms of keeping out any components from the signal path that don’t strictly need to be there, going for a more ‘purist’ approach that perhaps fits well with the ethos of using a native-PCM conversion R2R ladder in the first place, but it does come with the tradeoff that you’ll need to check the input impedance of the amplifier you intend to pair with the Harmony and ensure it’s sufficiently high.

Most amplifiers have extremely high input impedances, 25,000-100,000 Ohms is fairly typical and will work just fine. However it’s worth mentioning specifically for headphone users that quite a few popular amplifiers that are designed more toward maximising SINAD values often sacrifice input impedance to achieve this. A topping A90 for example has only 2000 Ohm input impedance, which may mean you encounter some unwanted effects if using it with the Harmony.

The internals of the Harmony are just as gorgeous as the exterior, with a hefty transformer surrounded by MU metal shielding, separation between individual modules, and nicely finished accents and branding scattered throughout too. But besides just looking great, there are some features found inside which didn’t get mentioned in the product materials, yet will be a benefit to many users, the most notable of which being the addition of galvanic isolation on the USB input.

This ensures that no noise can pass through from the USB source to the DAC itself, allowing you to connect this to any PC or source without worrying about it being noisy or possibly degrading the performance of the DAC.

Being an R2R DAC, this is not a product that will be targeting THD+N numbers as low as possible, after all R2R DACs are incredibly hard to make accurate to extremely low levels and this is essentially the entire reason delta-sigma DACs exist in the first place.

But how does the Harmony perform?

Measurements

Measurements Include:

  • THD+N (SINAD)
  • THD+N vs Output Level
  • THD+N vs Frequency
  • IMD vs Output Level
  • Low Level Signal Output
  • Reconstruction Filter/Upsampling
  • Jitter
  • Linearity
  • Output Impedance
  • Volume Matching
  • Noise
  • Intermodulation Distortion (IMD)
  • Crosstalk
  • Multitone

Additional measurements and test information available in the full report

Test Setup

  • Audio Precision APx555 B-Series analyzer
  • Measurement setup and device under test are running on an AudioQuest Niagara 3000 power conditioner.
  • Harmony was connected via USB unless specified otherwise.
  • Harmony was warmed up for 6 hours prior to testing
  • OS Mode used unless otherwise noted.
  • USB input, headphone output used unless otherwise specified
  • Tests shown are with a 200kΩ input impedance configured on the analyzer.
  • Exact analyzer/filter configurations for each measurement are detailed in the full reports
  • CH1 (Blue) = Left, CH2 (Red) = Right

Full Measurement Reports

Harmony (OS Mode)

Harmony (NOS Mode)

Results

Dynamic Range (AES17): 110dB

SNR: 120dB

IMD SMPTE: -78dB

Noise Level RMS: 86uVrms (20hz-1Mhz), 3.9uVrms (20hz-20khz)

DC Offset: 294uV active, 212uV idle

Susceptible to intersample overs: Immune to smaller +1dB intersample overs, +3dB overs will clip.

Output Impedance: 600Ω RCA, 1200Ω XLR,

THD+N / SINAD

Disclaimer: SINAD should not be used as a figure to judge the overall performance of a product, it is a figure that combines several aspects of performance into one number whilst not including many others. It should be used as one factor among many when evaluating the overall objective performance of a device, and the type/structure of distortion can often be more important than the level of distortion/noise itself.

1khz 0dBfs Sine - XLR Output:

About 15 bit resolution at full scale, although as shown by the small signal tests further down, actual dynamic range is about 110dB/18 bits.

1khz -3dBfs Sine - XLR Output:

Performance when not at full scale is similar in overall level, though now the harmonic distortion structure is more balanced. The 3rd order content is only higher when the DAC is outputting full scale content.

THD+N vs Output Level

THD+N vs Output level

THD+N vs Output level is fairly consistent. A challenge with many R2R designs is keeping things proportional as output level changes, but the Harmony does this quite well overall. Still a bit of a change around -40dB, but quite significantly less than what is found on some other R2R products such as the Rockna Wavedream for example.

THD+N vs Frequency

(96khz bandwidth used on the analyzer. Don’t compare this directly to standalone audible band THD+N measurements as the measurement setup is not the same.)

20khz Bandwidth

96khz Bandwidth

NOTE: Ignore the random spike on the 96khz graph, this was due to a momentary USB buffer issue.

THD+N vs frequency is overall quite flat, in fact with the 20khz band-limit we see a drop toward higher frequencies, whereas with the 96khz bandwidth it’s flat out to 20khz. This is mostly due to THD+N being primarily limited by harmonic distortion, not noise, and therefore since the harmonics contributing to the THD level extend out quite far, more of these harmonics falling outside the bandwidth of the measurement when set lower will cause the dip we see with the 20khz limited measurement.

Frequency Response

Frequency response is as would be expected flat besides the differences in filter rolloff. NOS mode causes a roughly -3dB attenuation by 20khz. Whereas OS mode is flat up to just over 19khz and then begins a steep rolloff. Use of external oversampling can allow the DAC to extend flat past 20khz.

IMD vs Output Level

IMD does change a fair bit depending on output level, as is typical with most R2R DACs.

Low Level Signal Output (96khz bandwidth)

This test plays a very low level signal through the device to check for any unusual behaviour.

-90.31dBfs 24 bit dithered sine - NOS mode

-90.31dBfs 16 bit undithered sine - NOS mode

-90.31dBfs 16 bit dithered sine - NOS mode

For extremely small signals, we can observe a slight difference in behaviour on the positive side of the signal vs the negative. We are right on the limit of the DAC’s dynamic range here, so at this point we are starting to more clearly see the limitations of the ladder’s accuracy. Though being an R2R DAC, it does not apply any dithering or noise shaping of its own unless you are running it in OS mode. This means that we can quite clearly see the difference at 16 bit between the dithered and undithered signal. The ability to run NOS and lack of any applied dithering when doing so does also mean you may be able to get measurably better performance and linearity by using external oversampling & noise shaping tools like Signalyst HQPlayer or PGGB in ways that aren’t always possible on delta-sigma converters.

Reconstruction Filter (Nyquist/Oversampling Filter)

OS mode

NOS mode

There is only one filter option in OS mode, which does roll off ever so slightly before 20khz, but with decent attenuation at the Nyquist frequency, and fully attenuating shortly after.

NOS mode does not filter out any ultrasonic/imaged products as there is no oversampling filter in place.

Jitter

Jitter performance is moderate. It’s definitely not up to the same level of performance as quite a few competitors, but it’s also not high enough to be concerned about.

Linearity

This test plays a 1khz tone through the device, and steps down the level in small increments. Linearity measures how accurately the actual output of the device reflects the intended output. For example, if the level was stepped down by 20dB, but the output of the device was 19.7dB quieter, there is a 0.3dB nonlinearity, which would show as a 0.3dB rise on the graph.

Bandpass filtered, this is measuring only the accuracy of the 1khz tone itself

No bandpass filter, this factors in noise into the result

Noise

Noise levels are overall very low. Of particular note is that despite the large transformer being used in the linear power supply, mains leakage is well below -145dB, so it would seem the shielding is working very well.

The spike visible towards the right of the graph swaps between 705.6khz and 768khz depending on the base rate of the content you are feeding the device. This is because the DAC itself even in ‘NOS’ mode is running at 16x rates. It could therefore be argued that this isn’t ‘really’ NOS, as it is technically oversampling, but just doing zero-order-hold interpolation which provides an output functionally identical to NOS. But the important factor is the lack of dithering in NOS which allows you to fully benefit from high performance external noise shapers. High performance external oversampling will still have the same effect on any DAC regardless of whether the DAC itself is oversampling or not.

Crosstalk

Crosstalk performance is quite good, remaining below -96dB for the full bandwidth, and going down towards -140dB the lower in frequency you go.

Multitone

The multitone test shows us what the THD+N vs frequency and 1khz FFT already showed us, which is that distortion vs frequency is overall very flat, and the ‘floor’ here is not due to actual noise, but rather the harmonic distortion products themselves.

Conclusion

The Laiv Harmony is not the last word in objective performance, and for that matter an R2R DAC is not what you should be looking for if ultra-low distortion is what you’re seeking. However, as detailed further in the subjective review video, the ways in which the Harmony does add some distortion to your music come across in what are in my view, quite enjoyable ways. It IS a slightly coloured DAC, but coloured in ways that fit quite well with both many of the stereotypes about R2R DACs themselves, and what many people subjectively describe as being desirable aspects.

The one area where it would have been nice to see an outright improvement would be jitter performance, but besides this, I can’t find much to point to as a ‘fault’ with the Harmony. Instead, these measurements are more showing us that it is indeed a slightly less transparent device, but in ways that subjectively may be quite enjoyable, such as a higher, but well balanced harmonic distortion profile.


This is a companion discussion topic for the original entry at https://headphones.com/blogs/reviews/laiv-harmony-r2r-dac-review-measurements
1 Like

So what happens to the audio when you use a DAC with a high output impedance with an amp that has a low input impedance?

It’ll change the FR curve.
The way to think about it is that the output impedance of the DAC forms a divider with the input impedance of the amplifier,

If the output impedance is small relative to the input impedance you get close to a perfect voltage source, I.e. a pretty much the measured straight line response.

IF the output impedance were constant with frequency (which it won’t be) you’d just get a lower level signal to amplify, otherwise some portion of the output impedance vs frequency response curve is basically superimposed on the frequency response. So not a flat line.

You’d need to see the output impedance vs Frequency graph to know what that means, but for most output circuits you’ll most likely you’ll get a boost in bass frequencies and some loss at the upper end.

But I’d be concerned in this case that because the output is coming straight off the ladder you might also get significant change in output impedance with volume, and that would result in none linearity into low input impedance amps.

It’s generally why you want a high input impedance in an amplifier, but for most headphone amps that’s a direct result of the Pot that’s used and the higher the impedance the Pot the harder it is to manage RF noise, so the high SiNAD amps just assume low output impedance for the DAC.

It’s not the only DAC with this “issue”, some very highend DAC’s have similar or even higher output impedance, it just means you have to be careful with the pairing. In the speaker world where preamps are the norm, it’s almost never an issue,

1 Like

Assuming it’s flat vs frequency (which it usually will be), the high output impedance forms an RC (resistor-capacitor) low-pass filter with the capacitance of the cable, which causes high frequency rolloff.

This is one of those weird edge case scenarios where the cables you use CAN matter, although it’s mostly a question of length (since cable capacitance is directly proportional to length, though some cables will be lower capacitance per foot than others). So using short cables will be ideal.

USUALLY, this won’t be significant enough to start having a meaningful effect in the audible band, but it can. A more drastic example is using a passive preamplifier, with those whilst they’re great and in many ways one of the most transparent options for volume control, you need to keep the cables as short as possible to minimise the effect of this RC low-pass filtering.

The distortion profile needs to be more of the even order harmonics as in SET amps;such harmonics are what give the sound presence, one of best sounding amps that exploits this is the Felicks Envy.its a very interesting profile all the same,almost flat linear,never seen that before,usually an exponential like decay with the 2nd or 3rd as most dominant.Thanks for the measurements,explains a lot