Directstream Mk 2 observations

Very valid comments!

The MK2 has not had as smooth a run as it deserved. Part of that was a late-caught design issue where a feature that was intended to lower the noise floor (adjustable relay-based attenuation of the audio transformer input signals) was actually creating a whole lot of other noise due to reflections of the multi-MHz signals. Removing those components altogether means the unit now runs at maximum analog level all the time so the noise floor is capable of creating hiss in more sensitive systems just like the MK1. An analog pre-amp functioning as an attenuator is probably a necessity in the majority of systems. I ended up with a Benchmark LA4 for my MK1 and it rarely goes above -20dB.

I’ve yet to have the pleasure of hearing a MK2. Just about everybody seems to agree that it sounds better than its predecessor when both are in stock form – and that’s what I’d expect since I can’t name a single area other than the attenuation where the MK2 doesn’t have a clear advantage on paper. A reasonably modded MK1 however is generally said to hold its own against the MK2 and the modding potential of the newer unit is significantly more limited.

The big hope is for future evolution of the MK2 firmware to help it pull ahead by taking better advantage of its dual FPGAs. My big question is whether that potential includes a lowered noise floor in the audio band.


Just so I understand…

The MKII has a “high”, audible analogue noise floor (heard as hiss) that cannot be appropriately attenuated?

In discussions prior to the DAC’s release, I have a recollection of Ted saying the absolute level of the MK2 noise floor without attenuation was lower than that of the MK1 with its -20dB attenuator engaged. I can’t easily provide evidence of that statement nor measurements to confirm it.

However, if it’s approximately correct then what we have is a machine that in medium sensitivity systems and quiet environments will produce some white noise through the speakers if connected directly to the power amplifier. The higher the sensitivity and quieter the background (and better the listener’s ears) the more audible that noise will be. It’s similar in principle to plugging a tape deck directly into a power amplifier – except the digital volume control in the DAC can only lower the music volume, it can’t reduce the hiss volume.

That applies to both MK1 and MK2 by the way. The solution, if needed, is to add a pre-amp in both cases so that the music and the hiss can be reduced by analog means.


Aside: it might be worth noting that a “pre-amplifier” originally took the tiny signal produced by a turntable and made it bigger, so that it was going to produce something loud enough to listen to after the power amplifier multiplied it by its fixed amount and sent it on to the speakers. These days we have higher output levels from our line sources and much more gain in our amplifiers so we generally use our “pre-amplifiers” to make the signal smaller and avoid it blasting the house down.


In JA’s original review he noted audible noise with Mk2 while using LS50s which are not terribly sensitive speakers. He said the noise was NOT present on the Mk1 and did not change with volume level or ground setting. He specifically recommends NOT using the digital volume control with high sensitivity speakers due to the noise.

So, no, the noise at the DAC output cannot be attenuated with the built in digital volume control, because digital volume control only reduces signal, not noise.

As @dvorak explains very well, a good preamp with analog volume control can be used to lower both signal and noise. This means that at lower volume levels, the noise will be reduced.

Another way to look at it, an analog volume control retains SNR as you reduce level, a digital volume control (even lossless!) will reduce signal without reducing noise, therefore reducing SNR at lower volume levels. In practice this means with a digital volume control the level of hiss is constant regardless of volume position. This is why digital volume control is fine with a low noise DAC, if the level of noise is low in an absolute sense it will never be audible. Unfortunately, the Mk2 is not a low noise DAC.


I think this is a bit of a tempest in a teapot then because the noise CAN be attenuated with the use of a preamp. Therefore, digital level adjustment is just an extra feature that some (many?) customers will find useful for getting the best synergy out of their system components (DAC, pre. amp, and speakers. :man_shrugging:

Like most things in high-end audio, YMMV.

OTOH, if the reported noise is significant and audibly degrades the sound at normal listening levels with or without a preamp, then it seems like more design work is in order.

I am going to talk about how my system handles both vinyl and the Mk2 DAC. I hope this is helpful.

My MC phono cartridge produces a signal that about 0.2 mV ( 1 mV is one thousandth of a volt ). I use a step up transformer ( SUT ) to increase the voltage by 10x to 2 mV. I use a phono preamp ( the is often called a phono stage if it is part of the preamp ) to increase the voltage from the SUT by 500x to 1 volt. ( The phono preamp aso applies the RIAA equalization to the signal. ) This goes to one of the inputs on my preamp. ( I will return to this later. )

I set the volume control of my DS DAC Mk2 at 100. Because of how the digital volume control works the 100 setting gives me the best S/N out of the Mk2. I also turn on the galvanic isolation of the Mk2’s analog output. This greatly reduces the noise on the analog output signal. This output signaal is more than 1 volt.

My preamp is all analog ( no digital ). It has an analog volume control that goes from 0 to 99 in integer steps. Each step raises the preamp’s output by about 0.7 dB. I typically listen to music at about 75 dB SPL ( sound pressure level ). To achieve this SPL I typically play SACDs at a preamp volume setting of 25 to 35 and for vinyl a preamp volume setting of 40 to 50. My preamp has unity gain at a volume setting of 66. Thus, as mentioned above, my preamp is attenuating the signals I get from both Mk2 and my TT system. I could use the digital volume control on the Mk2 and reduce the signal from it so that I can play SACDs at the same volume setting as the vinyl. This would be choosing convenience over better S/N. I choose the better S/N.

Finally, in the period from 2021 to now, my system has been unchanged except for trading -in the DMP and Mk1 DAC for the PST and Mk2 DAC. The last to arrive was the Mk2. I can say with complete certainty that in my system the MK2 DAC with original firmware sounded better than the Mk1 DAC. And, with the Massive firmware the Mk2 DAC sounds so good that when I play vinyl a record from MFSL and then play the MFSL SACD of the same music ( I always match volume level when I do this with an SPL meter ) which is made from the same master tape as the vinyl was made, it is very hard to say which sounds better.



For me the question comes down to is it competing with the best in the market or with the Mk I? Clearly it is a step up from the Mk I. From my perspective the early released Mk II DACs had a handful of bugs, and the unit was released before all the issues had been addressed. Similar to the PST transport.

1 Like

Apropos of maybe nothing other than my own curiosity, are you using a Conrad Johnson preamp? Your description certainly sounds like it.

There are some consistencies in noise and other problems with the MK2 and PST. I haven’t had a problem with either units since day one. Only the MK2 did not play DSD256 which was fixed right away with a firmware update. I hear zero noise 1 foot from my speakers at normal listening level. I do use a preamp mainly because I start to max out the attenuators in the MK2 with some recordings. Now I have more efficient speakers I probably don’t need a preamp, but it sounds so much better with my preamp in the circuit so it stays. My main concern is SQ, and it’s there were the MK2 really shines. With really good recordings, the sound is so immersive, I coundn’t ask for anything more. As a comparison, even a Esoteric N-01XD could not keep up in resolution. I would not worry about noise or measurments, and just enjoy the sound this marvelous machine can produce.


Yes, but this not the place to discuss preamps.

Understood. As I said, just curious.

I forgot about that, here is the post → DirectStream DAC MKII Released - #430 by tedsmith. Unfortunately, this claim seems to be completely false.

The Mk2 has more noise than the Mk1 relative to full-scale output (or in terms of SNR), I’d estimate around 6 dB worse on the -90 dBFS Stereophile measurements below.

To make matters worse, as the Mk2 also has higher full-scale output voltage (2.85 V for Mk1 vs 3.72 V on Mk2, balanced outputs), the noise in absolute sense will also be about 2 dB higher than that. So overall you are looking at about +8 dB noise floor compared to the Mk1. And you have no analog attenuator on the Mk2!

To try and quantify this a bit, in the original Mk1 review, JA estimated resolution at about 5 dB better than 16 bit, so let’s say SNR of around 101 dB. Further firmware improvements seemed to have gained about 3-5 dB of additional SNR, so let’s say Mk1 has full-scale SNR of around 105 dB. So, audio band residual noise is approximately 2.85 x 10^(-105/20) = 16.0 uV. If you apply the 20 dB analog attenuator that reduces the noise floor to 16.0 uV x 10^(-20/20) = 1.6 uV, which is really quite excellent, although you are limited to 0.285 V output.

Giving the Mk2 the same treatment, let’s say SNR is 6 dB worse than the Mk1, so 99 dB. This aligns with JA’s claim of roughly 16 bit resolution. Audio band residual noise is therefore 3.72 x 10^(-99/20) = 41.7 uV.

In summary:

Mk2 noise floor: 41.7 uV
Mk1 noise floor w/o attenuator: 16.0 uV
Mk1 noise w/ attenuator: 1.6 uV


I should stay out of this but I see a couple of issues … First, the graph above for the Mk I does not indicate if the 20 dB attenuator is engaged which is what dvorak stipulated. Also, even though the Mk II is shown as higher noise than Mk I, that noise is still below -125 dB. Can you hear that? I can’t. Second, do you own a Mk II? All I saw was the Okto dac8 pro which has been out for several years and was below $2k. I’ve never heard it, nor even read about it (there is a Stereophile review from 2020), but since it’s not all the rage in the audiophile world I suspect it measures great and sounds, well I’ll just say less great.

My point here seems to be you are trying to convince us of the folly of our ways in buying the Mk II I think you should give up. I love mine and your attempts to tell me how poor it is our wasted. Sorry for being blunt …


Ditto to what you said here. Thank you so much.

I suggest that anyone interested in this noise issue start in the Mk2 Released thread at post 421 and read down through post 432. It is a very complex issue. I do wish that Ted Smith would join us here in this thread and give us his thoughts on the noise issue and JA’s Follow Up in Stereophile.

1 Like

I agree and my analysis considers this. I first calculate Mk1 noise without the attenuator and then assume the noise is exactly attenuated by 20 dB, in reality the attenuator may add some noise and the noise reduction may be less.

It is very clear that the Mk2 has a higher noise floor than the Mk1 without the attenuator engaged (~8 dB more) and with the attenuator engaged it has significantly higher noise (~28 dB more).

The noise is not below -125 dB, this should be clear from my analysis and tells me you do not understand the concept of FFT gain. See here for more info on the concept, but basically the longer the FFT, the lower the apparent noise floor will be on the FFT, but the actual noise level stays the same.

As an example, here is a -60 dBFS tone from a DAC captured with different length FFTs (16K vs 128K vs 1M). Based on appearance you might say the 1M FFT has ~15 dB lower noise than the 16K FFT, but in actuality they all have the same 20-22K noise level of about 7 uV or about 62 dB below the -60 dBFS tone for a full scale dynamic range of 62 + 60 = 122 dB. It is just a different visual representation.

You can compare relative noise level if the FFT length is the same, but you can’t read the noise level directly off the FFT.

I do not. However, I have done extensive listening tests with a variety of DACs and can say that with the speakers and amplifiers I use, 41 uV residual noise from a DAC is extremely audible. Remember that any noise at the DAC output will be multiplied by amplifier gain, therefore the noise at the speaker terminals is much higher than 41 uV.

I am not sure how the Okto is any way relevant to the discussion at hand, and I am not going to try to convince you of its sound quality over the internet, actual listening would be required to do that. However, while a lot of measurement stuff is not audible, noise certainly is and I can say without question that the Okto has much lower noise (3.5 uV) than the Mk2 and has no audible hiss.

I am not trying to convince you of re-thinking your purchase, I own and enjoy all sorts of gear that doesn’t measure particularly well.

However, it is clear that PS audio has made repeated false claims about the noise performance of the Mk2. I think this is important information to be aware of and I question anyone using the Mk2 without an analog volume control.


I appreciate that openness. I am a MK2 early adopter with a beta build. I am very happy using the MK2 myself and I do prefer its sound to my ESS-based Okto and exaSound DACs. With Massive, I also finally preferred the MK2 to a Bricasti M1 and sold the M1. However, I would be much more comfortable selling my ESS-based DACs to other people as I know that they will not have noise issues. The Bricasti also had no issues when outputting directly to power amps. It is really normal today to be able to use a DAC without a preamp and not to have the hissing noise that I experienced when not using a preamp or not using -20dB attenuators between the MK2 and a power amp. And not all of us are US-based with the option to trade units back in to PSA.

1 Like

Carousel, Those of us that have both analog ( usually at TT ) and digital sources almost always use a preamp to switch between sources and adjust volume levels between sources.

With respect to the attenuator on the Mk I, I recall discussions here years ago on noise and use of the attenuator to match systems with high output, but not specifically that it reduced the noise floor 20 dB. What I do know (I owned the Mk I for nine years) is that with it engaged it closed the sound down, i.e., lost openness. Was that due to noise or something else? I don’t know, I just didn’t need the attenuator. Apparently unlike you, I never had an issue with the noise floor of the Mk I or Mk II being audible at the listening seat.

With respect to the FFT, no I do not really understand them and should not have written that bit. That was a quick analysis that I regretted after posting. Noticed too late the 1k pulse was peaked at -90, I should have made easy observation. I have a mechanical engineering background, electronics are something I never got a full grasp of.

I mentioned the Okto only because that was the only non PS Audio DAC that got more than a passing mention by you in this entire thread. It was DAC 2 in an August 2023 reply to JK Richards. You also recommended it in August 2023:

As for PS Audio making false claims, I’ll leave that to the lawyers. Your posts seem sincere, but I can’t help but read into them (perhaps incorrectly). In that regard I checked the ASR website and there is an mdsimon2 (first name Michael, like you!) there who has 20 pages of posts. Some of which were in a Okto Owners thread. While I have not done much browsing of the ASR site, there is a lot of animosity here regarding the site. A few folks from ASR have come here to raise some hackles, successfully … until they were shooed away. While I have no direct experience, others from this forum did post on ASR and were blocked for not toeing the “Measurements rule” edict. I don’t want to ban you, you are entitled to free speech and all that it entails, but I like Paul McGowan and PS Audio and get annoyed when folks come here and knock them down. Yes it’s a personal problem …