Linn DS streamers: Majik DS(/1), Klimax DS/2 and Klimax DSM/2 measurements. (Examples of "high end" audio.)

The Linn Klimax DS/2 here at home for auditioning back in 2024.

Readers might recall about a year ago, we ran a blind listening test using hi-res captures of the analog output from different DACs to see if there was a clearly audible difference between a very cheap DAC (the Apple USB-C headphone dongle) versus much more expensive "audiophile high-end" streamer DACs which were the Linn Majik DS(/1) and Klimax DSM/2 Katalyst. If you don't know the results of that online study, here they are, and the subjective comments around what was heard.

While I showed 0dBFS 1kHz THD+N results for the Majik DS and Klimax DSM/2 back then, there was quite a bit more data collected as part of the preparation for that Internet Blind Listening Test that I did not publish. Even though my friend linnrd (who has contributed to these pages), as suggested by his screen name, happened to be quite the Linn fan, I appreciate that he is not blinded to the complexities of human preferences. He's well aware that both subjective impressions and objective performance play their part and he was totally open to me running measurements on his Linn digital gear to examine for technical accuracy.

For this post, let's dive deeper into the technical performance of some Linn streaming DAC gear. Here are the devices I'll be showing results of...

Image of Linn DS models tested - front and back.

In the image above, I've included general stats on the 3 Linn DS devices. Of note, I've also included the MSRP which is important when we consider the concept of "value".

As someone not immersed in Linn culture, I noticed that they like to spell names in their unique way with "k" instead of "c". They are based in Scotland which doesn't explain this idiosyncrasy 😉. I assume the "k" spelling is just to exude a mystikal touch, not because of any actual linguistic rationale (as per Wiki).

We basically see 3 "generations" (if I may call them that) of ethernet-enabled DACs ranging from a model that was released back in 2008 - the very first Majik DS/1 upgraded with a Dynamik Power Supply (spread-spectrum, filtered switching device, €480 MSRP) based on the Wolfson WM8740 chip DAC. In the middle is the Klimax DS/2 with Exakt update that allows digital connections to Linn's active speakers released in spring 2015 (almost exactly 10 years ago). I'm not sure which DAC this uses but I've heard that it's probably dual Wolfson WM8741's. Finally, the most recent one of those above is the Klimax DSM/2 with "Katalyst" architecture (AKM AK4497EQ) released in the fall of 2016, produced until March 2021 at which point it was replaced by the Linn Klimax DSM/3 with its huge jump in price! When new, the DSM/2 had an MSRP of around $20k, the DSM/3 boosted that number to an even more impressive US$39k MSRP with it Organik DAC architecture!

It is only the latest of these, the Klimax DSM/2 (Digital Streamer with Multimedia inputs) that provides for other digital connections like S/PDIF and HDMI. I assume it's capable of PCM playback from the HDMI input and can act as a repeater/switcher for the HDMI 1.4 (upgradable to HDMI 2.0) out. I see that there are HDMI upgrades including a Linn Surround Module which can decode bitstream Dolby Digital and DTS variants, not the newer Atmos or DTS X though.

These are primarily PCM streamers with the ability to handle up to 192kHz sample rate. FLAC is recommended by Linn. Some of these streamers like the Katalyst and later units can handle DSD also (see here).

Whereas the lower end Majik feels in the hand like a typical consumer sheet metal box weighing in at less than 10lbs, the Klimax boxes are pretty nice looking, understated, classy, chunky pieces of aluminum, each weighing over 20lbs. The Klimax DS(M)/2's have a small utilitarian LCD screen on the front, notice the small font used to indicate volume.

Volume at 50.

The measurements were done with my E1DA Cosmos "stack" consisting of the Cosmos Scaler, and ADC. All 3 streamers were sent test signals over Roon, bit-perfect (unity gain, no DSP) over a 25' length of ethernet cable. Since these tests were done in preparation for the 2024 listening test, these results were with early 2024 firmware (Davaar 4.104.581); later on you'll see that I remeasured the Klimax DS/2 at my home more recently with the latest 2025 update.

As a start, let's jump into an exploration of the 1kHz 0dB THD+N/SINAD for the 3 DACs:

As you can see, we're looking at good levels of performance here. RCA output as per the device specs are at ~2Vrms, and XLR ~4Vrms. Right-left channel balance is very good with <0.2dB difference. For all 3 units, I found the left channel to have slightly lower noise than the right. Isn't it interesting that over the generations, as the price went up, the company would of course claim the sound got better, and subjective reviewers also claim "better" sound, the measurements also improved? Who says objective performance doesn't correlate to subjective claims? 😉

The first generation Linn Majik DS+Dynamik PS only has single-ended output and scores a THD+N of around -100dB. This looks to be a little better than ASR's measurement of the Majik DS/1 Streamer+Amplifier with about SINAD 96dB. Presumably without all the extra inputs and amplifier section in that variant, the noise and distortions are lower in this ethernet streamer DAC.

As for the Klimax'es, the DS/2+Exakt is capable of around -106dB THD+N XLR out and the DSM/2 with Katalyst measures down to -110dB using XLR.

Good numbers from DACs capable of better-than-CD resolution.

Hi-res DACs these days should have no issue with frequency response. Let's take a look at that as well as the filter quality using the "Digital Filter Composite" graph (extension of Reis test) that I've shown over the years consisting of 19+20kHz intermodulation tones, white noise with 0dBFS peaks that have intersample true peaks up to +3dB, white noise with -4dBFS peaks (no intersample overs), and digital silence to look at noise level. I'll measure out to 96kHz to make sure no abnormally high ultrasonic stuff:

While I only plotted the impulse response for the Linn Klimax DS/2 Exakt, all 3 were linear phase filters implementing steep cutoffs. None of the three showed any evidence of intersample overload of the White Noise with 0dBFS Peaks indicating that they can handle intersample peaks at least to +3dB. That's great.

My friend was curious how Linn handles intersample overs. Let's have a peek at the Klimax DS/2 with a test tone that pushes up to +8dB True Peak to see where it clips:

That's quite an impressive amount of intersample overhead! The blue channel is a 0dBFS (0dB True Peak) wave whereas the green is our test signal with potential peak at +8dB if the DAC can accommodate. The Klimax DS/2 can handle up to +5.3dB. No wonder we don't see any issue on the Digital Filter Composite graphs above. This is really quite excellent, probably the highest intersample overhead device I've seen thus far at recommended peak volume level.

For a comparison, here's what I found for the recently discussed Topping DX9 15th Anniversary DAC with its internal AK4499EQ DAC using F-1 "Sharp Roll Off" linear filter setting:

Notice that the Topping DX9 DAC "only" has around +2dB intersample overhead which is typical of AKM DACs. (Let's discuss more in the Summary portion below.)

How about jitter performance with ethernet input?

Nice! Across the board no issues with jitter at all. With an asynchronous interface like ethernet, we would not expect any problems (just like with a good implementation of modern asynchronous USB, jitter should be insignificant), even 10 years ago with the Majik DS/1.

For the DSM/2 Katalyst, since it has TosLink input, I had a look at the J-Test with that and note that it's really quite excellent also. They likely implemented an internal buffer and reclocker to achieve this level of S/PDIF temporal stability. 

With that TosLink input on the DSM/2 Katalyst, let's explore a bit further using loopback and a stepped-sine measurements of THD+N vs. Level:

Looks nice. Well controlled low noise and distortion as one would expect with the AKM AK4497EQ DAC. And with that data we can plot out the linearity:

Excellent performance to below -120dBFS.

Audiophiles can be obsessive when it comes to the idea that certain things will add "noise" to the audio output. This is why devices often can turn on/off display screens for example. Since the DSM/2 Katalyst has HDMI in/out switching, there is a software switch to turn off the HDMI function. Linnrd and I were wondering if this changed noise performance at least out to 48kHz:

Since balanced XLR provides excellent noise rejection, we looked specifically at the single-ended RCA output. As you can see, turning the HDMI function on/off made no difference as far as we could tell with an i7 laptop computer plugged into the HDMI in. Who knows, maybe other devices could cause more noise. The absence of 60Hz hum on the RCA out is also excellent.

Let's try out a more complex signal then - here's the 1/10-Octave MultiTone 32:

Hmmmm, that's strange, what's happening with the Klimax devices?

Notice the high amounts of distortion with only 80-85dB of distortion-free range. We made sure there was no DSP turned on (eg. Linn's Space Optimisation) or volume attenuation below the recommended unity gain "80" level. The older, supposedly lower end, Majik DS/1 was not affected and shows a much more typical clean hi-res MultiTone FFT.

Staying with the theme of more complex signal testing, on most of my DAC and amplifier measurements, I've been using the Triple-Tone TD+N as a measure of resolution in anticipation of picking up issues which might not be seen on the simple 0dBFS 1kHz tone:

The Linn Klimax DS(M)/2 devices seem to be good examples of when the 1kHz tone doesn't pick up anomalies! Like the 1/10-Octave MultiTone above, we see that the Klimax'es do a poor job with complex signals scoring a rather weak TD+N of -94 to -97dB even with XLR out. Notice how even the much less expensive first generation Majik DS was capable of lower TD+N on the RCA out compared to the Klimax DSM/2 XLR!

I don't think I've seen this reported elsewhere. Unfortunately, measurements of the Klimax DS such as here and here didn't quantify performance using more complex signals like the triple tone or show the multi-tone FFT.

Summary:

1. We already know from the blind listening test last year that hi-res capture of music from these streamers generally sounded the same, even compared to an inexpensive Apple USB-C dongle. This is consistent with recent discussions around audibility of modern hi-res DACs. As serious audiophiles, I think it's important to be honest with ourselves and the limits of human hearing, recognizing that it's OK to unburden ourselves from the cultural "Golden Ear" myths that have been perpetuated in this hobby over decades.

2. With a simple 1kHz THD+N, we see a nice improvement in the scores from the least expensive Majik to the most expensive Klimax DSM/2 Katalyst. The Katalyst device is based on the excellent AKM AK4497EQ sigma-delta chip - this is the smaller 2-channel sibling of the AK4499EQ used in the recently measured Topping DX9 but it looks like Linn uses their own digital filter rather than one of the built-in AKM settings.

While the Linn Majik and Klimax THD+N/SINAD numbers are absolutely fine, they're not particularly impressive these days. We have many sub-$1000 DACs capable of -120dB or lower THD+N in the 2020's.

3. Kudos to Linn for that jitter-free S/PDIF TosLink J-Test result on the DSM/2 Katalyst. Also the excellent linearity and clean THD(+N) vs. Level stepped sine measurement. Their steep linear-phase filter also provides good suppression of ultrasonic imaging distortions.

Also, it's impressive to see the Klimax DS/2 having more than +5dB intersample overhead (alas, not tested on Majik or DSM/2, but we know they're above +3dB). However, realize that if you're listening to music with frequent, high level intersample peaks when this extended overhead might be of benefit, you're likely listening to highly dynamically compressed music (think DR4 or worse recordings) that's clipping all over the place - these are not "audiophile-quality" recordings! As I discussed when I explored "intersample overs" back in 2018, it'd be nice to have something like +3dB overhead but if we're actually listening to high quality recordings, this is "mostly academic". In other words, there's no need to be worried, and why I don't routinely measure this.

But suppose you still insist on obsessing about this 😁, then just use some volume normalization like ReplayGain with reasonable target volume like -18dB LUFS (EBU R128) which is what I normally do. Alternatively, turn down the DAC level by something like -3dB, or if you're using Roon, turn on the "Headroom Management" DSP feature which defaults to 3dB of basically guaranteed overhead.

[On a side technical note, notice that Linn's "unity gain" is set at 80 which if we interpret that as 80% of maximum DAC voltage, is +2dB on top of the filter overhead. This could be why we see that extended ability to manage overload peaks. If we capped the Topping DX9 DAC by -2dB output level, we'd see a total 4dB overhead which is more than enough.] 

4. There's something weird happening with those Klimax DS(M)/2 streamers when it comes to complex signal reproduction. We see this both in the 1/10-Octave MultiTone 32 and the Triple-Tone TD+N tests. It looks like there's an exacerbation of low-level intermodulation distortion. I wonder if this has to do with the precision of Linn's digital filtering/DSP.

As it has been a year since I captured the results above and there have been new firmwares released, let me repeat a few of those tests on the Linn Klimax DS/2 + EXAKT, before and after updating with the latest "Davaar" firmware available for this hardware (version 4.110.556 - Feb 4, 2025 release).

Again, notice the little "80" bottom left of the screen denoting volume level. Linn's 0dBFS/unity gain level.

As you can see, with a simple 1kHz 0dBFS test signal, there has been no significant difference between the older (Nov 2023) and newer (Feb 2025) firmwares.

And with a more complex Triple-Tone test (1:1:1 at 48/960/5472Hz):

Again, no difference with new firmware, it's still not normal with those low-level distortions. Likewise, here is the 1/10-Decade MultiTone 32 with the new firmware which has the same "dirty" signature regardless of old/new firmware:

For context, here is what the 1/10-Decade MultiTone 32 signal should look like when it's reproduced accurately, as per the Topping DX9 DAC (BTW, Triple-Tone TD+N for the Topping DX9 is -117dB):

The Linn Klimax DS/2 and DSM/2 are interesting examples I think of how we should make sure to check more complex signals than just 1kHz THD+N/SINAD.

Who knows, since the 1kHz tone is commonly used, manufacturers might purposely optimize their devices to look good on this simple test (like how GPU drivers might optimize for benchmarks like 3DMark). Music is complex so it makes sense to run more complex tests, emphasizing the results of those over a simple 1kHz tone.

I know that my friend linnrd contacted Linn about the anomalous findings last year (since Spring 2024) but only received a response that the issue was being forwarded to their engineers with no further update since. Considering my positive experiences with contacting companies like Oppo, Sabaj, and Topping over the years when presented with concerns typically resulting in an updated firmware fix, it's a bit disappointing that a company like Linn selling what I think most would consider to be expensive "high end" devices aren't more accessible to tell us whether they looked into this or can address it. Given that neither I nor my friend would claim that the Klimax DS/2 or DSM/2 sounded "bad" as is given the low-level nature of the distortions, even an acknowledgement from the company that they're aware and the effect is perhaps an intentional result of their digital processing would have at least been something! Oh well.

I must say that despite the disparaging remarks on forums and in magazines against "Chi-Fi" companies, realize that such companies often will focus more on objective performance such that when approached with issues, in my experience, they actually have the skills to look at the reported concerns and fix them with quick turnaround time if possible. To me, that's a much better technical customer support experience than what we saw with Linn's customer engagement presumably trying to get in touch with their engineering team in Scotland!

Needless to say, while there are many good points about the Linn Klimax DS/2 and DSM/2 streamers like low jitter and impressive intersample overhead, I would not be completely happy with some of the results at Linn's asking price.

5. Of interest, there was a correlation with increase in analogue output level as the price increased. While looking at the RCA output levels, this is what we saw for 0dBFS 1kHz sine output voltage (volume 80):

Majik DS/1              1.96Vrms     0dB
Klimax DS/2 + Exakt     2.01Vrms  +0.2dB
Klimax DSM/2 Katalyst   2.10Vrms  +0.6dB

Whether this is just coincidental or intentional, there's about a 0.2-0.4dB bump in output level with each higher level, more expensive, device tested. So if one were to do comparative listening, be mindful that some of the subtle differences could just be volume increases acting as a slight bias towards a "better sounding" higher end product. I'd be curious if the latest DSM/3 (with much higher price tag) pushed that output level even a little more. 🧐

For those interested in having a listen to the hi-res AMPT standard FLAC recording:

Linn Klimax DS/2 Exakt DAC AMPT (173MB)
24/96 capture through RME ADI-2 Pro ADC

While these are Linn's older streamer/DACs tested in this post, they're still relatively recent with the Klimax DSM/2 in production up to March 2021. The newest models now use Linn's custom "Organik" DAC architecture which looks like a discrete sigma-delta design + FPGA upsampling. I've only seen subjective reviews like this, this and this. The latest Klimax DSM/3 has a nice large front display, cool top control knob, but has an asking price of US$39k. I dunno guys, at that kind of price, I surely hope it handles complex signals significantly better than the Klimax DS/2 and DSM/2!

[It looks like Linn can upgrade the Klimax DSM/2 to the Organik DAC, priced at US$7k as per this review. Robert Harley of The Absolute Sound claims:

"It was clear that the Organik DAC is a major advance over the Katalyst, and a worthwhile upgrade for existing DSM owners. There were no sonic tradeoffs with the Organik; it was better all around—and in ways that are vitally important musically."

Why am I not surprised that Harley would say this? Is it even possible for a purely subjective audiophile magazine reviewer to claim that there has not been yet another level of improvement with each upgrade from an esteemed company when they've been provided access to the latest and greatest technology to grace their magazine content!?]

Here's Linn's Organik DAC video about "our most natural sound". Maybe it is, maybe not. Regardless of the veracity of this claim, I do like the computer graphics in the advertisement:

There's one last thing that I might look at with the Linn Klimax DS/2 while I still have it borrowed here at home. It's their Space Optimization DSP to see what they're doing if I can get that going - ran into some issues with the website last year when I last tried. 

Whether it's blind listening evaluations or more technical examinations as presented here, we are reminded that expensive "high end" audio, especially DACs, really are not capable of magikal sounds. There's really nothing like the oft-dramatic, idealistically amazing loss of "veiled" sound that the poetic subjective reviewers try to impress upon us when they speak of increasingly expensive luxury devices.

As usual, I remind everyone that there's nothing wrong with being impressed by build quality, appearance, brand name, and perhaps features that might be available when buying the more expensive product guided by subjective preference. Just remember that an expensive Rolex might be more desirable but doesn't tell time any better than a utilitarian Timex, and while the music we listen to are like the myriad wines we can taste and enjoy, the wine glass itself (sound system) only contributes a certain amount to that experience once we're sure the hardware is clean/transparent.

Epic "How It's Made: The Klimax DSM" video from chunk of metal!

Thanks again linnrd for the opportunity to examine these devices!

--------------------

Alright Northern Hemisphere audiophiles, I hope you're enjoying your summer with some cool tunes and good times with loved ones.

Here's a cute parting song from a local Vancouver band (originally from Kelowna, British Columbia). Yukon Blonde with "Adore You" from their upcoming album Friendship & Rock 'N' Roll due in September:

For organ lovers, check out Anna Lapwood's recent album Firedove (2025). The streaming multichannel/Atmos version's full, dense, room-filling sound in the louder portions also can convey subtle nuances when needed in the softer parts leveraging the expanded number of channels. An enjoyable album for classical / crossover / soundtrack listeners.

Here she is on "Time" originally from the Inception soundtrack: