Fosi Audio DS2 Portable DAC & Amp Review

[roladyzator]

That's great, thanks very much, totally makes sense.

I also saw there is a volume setting in the usb audio pro about volume and either hardware, software or none but I'm not sure what it really does.

Hardware is the setting you should use.
This is what allows UAPP (amongst others) to use 1V output out of Apple dongle or the Samsung dongle (including the EU version for the Samsung).

Device with physical volume buttons usually ignore this, and the volume works in the software mode anyway

 

[PJR918BMW]

I have read the review and have indeed gone out and bought one of these.
But I still can't believe what I have bought. If I understand correctly this thing can be attached to dual balanced cabled and fed straight into pretty much any home hi-fi amp (does it have enough gain?) With the measured performance of this bit of kit I can attach it to my phone or any other bitstream/pcm streaming source and knock pretty much any other DAC out of the park?
I know I am old and behind the times but I find that truly staggering.... In hi-fi terms it's like Musk catching falling buildings with little arms sticking out of a tower!

 

[jasonhanjk]

I have read the review and have indeed gone out and bought one of these.
But I still can't believe what I have bought. If I understand correctly this thing can be attached to dual balanced cabled and fed straight into pretty much any home hi-fi amp (does it have enough gain?) With the measured performance of this bit of kit I can attach it to my phone or any other bitstream/pcm streaming source and knock pretty much any other DAC out of the park?
I know I am old and behind the times but I find that truly staggering.... In hi-fi terms it's like Musk catching falling buildings with little arms sticking out of a tower!

It's balance output is 4Vrms. So if you connect to an amp of 10X gain and 8 ohm speaker, the possible max power would be 200W.
Do take note when using android phone, using app like Hiby will bring out the full potential of this dac.

 

[MC_RME]

Here is part 2 of my measurements for unbalanced (TRS out. Balanced aka Pentaconn is here: https://www.audiosciencereview.com/...s2-portable-dac-amp-review.57063/post-2102639). I again wrote down the max values that the DS2 can provide at the edge of clipping - harmonics are about to rise but just do not:

600 Ohm: 2.03 V, +8.4 dBu, 114 dB THD+N, 6.9 mW

300 Ohm: 2.03 V, +8.4 dBu, 114.0 dB THD+N, 13.7 mW

50 Ohm: 1.8 V, +7.3 dBu, 110 dB THD+N, 64 mW

32 Ohm: 1.59 V, +6.2 dBu, 109 dB THD+N, 78 mW

16 Ohm: 1.17 V, +3.6 dBu, 101 dB THD+N, 85 mW

Here is the dashboard view for 600 Ohm:

And here the THD+N sweep. I left the balanced one in the picture, with dotted lines.

SNR: Base noise is a fixed -117 dBu, which is about 1.1 µV, the typical limit for the AP, so not the real noise of the DS2. I redid the base noise measurement with a +34 dB amp to get the AP out of the equation. Now I measured -120 dBu (rms unweighted). Of course I had to redo the balanced one this way too, which turned out very cumbersome due to various ground related side effects. It improved from -115.5 dBu to around -117 dBu. That would match theory, with balanced having 6 dB higher output level but 'only' 3 dB better noise.

The SNR on the unbalanced output is -120 dBu (rms unweighted) plus the respective output level (for example 120+8.4= 128.4 dB, 120+3.6= 123.6 dB. All rms unweighted. dBA = +2.3 dB).

The Pentaconn output delivers ground, so adapters balanced to unbalanced will work (if they are wired correctly internally). Using such an adapter to get the unbalanced signal from the balanced output instead of the unbalanced output does not bring any advantage, as expected and proven by the measurements.

During the SNR measurements I noticed that the DAC switches its output off (Edit: it switches off immediately when playback stops, but the software I used before has an otherwise useful 'keep alive' feature, which caused the 3 seconds reported before) – some kind of energy saving? It’s not a simple mute, the SNR gets much worse in this state because the output stage is open (this is not an issue in real-world usage, with an open output the headphone does not play anything, and the process does not cause a click noise. But measurements are really screwed up). So for the base noise measurement I played a 24 bit file that only includes dither to keep the DS2 output alive (constant dithered playback is default on the AP, so one would not easily notice that behaviour there).

In that 'off' state the output impedance rises to 10 kOhm, no matter balanced or unbalanced. I also measured the output impedance in normal operation: Balanced 0.6 Ohm, unbalanced 0.5 Ohm. One would expect other values here (like 2 x 0.5 Ohm for Bal = 1 Ohm), but not.

Finally I used the DS2 with Windows and now fully understand the frustration of others in this thread - the ramp-up effect, which on my unit even includes a right to left volume shift, is quite annoying, and IMHO a deal-breaker for many. As I already wrote I am lucky as I don't have that effect with my iPhone and Neutron as player, but I think one can't recommend the DS2 until Fosi has fixed that issue.

 

[melowman]

@MC_RME Awesome!! Thank you very much!
Ain’t there an issue with the sweep graph? It does not match the results in your table.

 

[MC_RME]

Good catch! The reference impedance was set to 16 Ohm. Correct pic is now up. Thanks.

 

[vaidd]

Was this measured on high gain or low gain?

From their online instructions, the volume-up and volume-down buttons can be pressed simultaneously to switch between "high level mode" and "low level mode."

Fosi Audio Product User Instruction DS2

Design by GemPages

fosiaudio.com

Spoiler: Screenshot from fosiaudio.com

 

[mc.god]

Was this measured on high gain or low gain?

From their online instructions, the volume-up and volume-down buttons can be pressed simultaneously to switch between "high level mode" and "low level mode."

Fosi Audio Product User Instruction DS2

Design by GemPages

fosiaudio.com

Spoiler: Screenshot from fosiaudio.com

View attachment 404740

High gain for sure, othewise it would't deliver 2V unbalanced - 4V balanced.

 

[melowman]

There’s no high/low mode in DS2 2024; that was present in the previous ESS version only. (Btw this mode was acting on impedance: https://www.reddit.com/r/inearfidelity/s/9mUSIlqXhx)

 

[mc.god]

Here is part 2 of my measurements for unbalanced (TRS out. Balanced aka Pentaconn is here: https://www.audiosciencereview.com/...s2-portable-dac-amp-review.57063/post-2102639). I again wrote down the max values that the DS2 can provide at the edge of clipping - harmonics are about to rise but just do not:

600 Ohm: 2.03 V, +8.4 dBu, 114 dB THD+N, 6.9 mW

300 Ohm: 2.03 V, +8.4 dBu, 114.0 dB THD+N, 13.7 mW

50 Ohm: 1.8 V, +7.3 dBu, 110 dB THD+N, 64 mW

32 Ohm: 1.59 V, +6.2 dBu, 109 dB THD+N, 78 mW

16 Ohm: 1.17 V, +3.6 dBu, 101 dB THD+N, 85 mW

Here is the dashboard view for 600 Ohm:

View attachment 404585

And here the THD+N sweep. I left the balanced one in the picture, with dotted lines.

View attachment 404608

SNR: Base noise is a fixed -117 dBu, which is about 1.1 µV, the typical limit for the AP, so not the real noise of the DS2. I redid the base noise measurement with a +34 dB amp to get the AP out of the equation. Now I measured -120 dBu (rms unweighted). Of course I had to redo the balanced one this way too, which turned out very cumbersome due to various ground related side effects. It improved from -115.5 dBu to around -117 dBu. That would match theory, with balanced having 6 dB higher output level but 'only' 3 dB better noise.

The SNR on the unbalanced output is -120 dBu (rms unweighted) plus the respective output level (for example 120+8.4= 128.4 dB, 120+3.6= 123.6 dB. All rms unweighted. dBA = +2.3 dB).

The Pentaconn output delivers ground, so adapters balanced to unbalanced will work (if they are wired correctly internally). Using such an adapter to get the unbalanced signal from the balanced output instead of the unbalanced output does not bring any advantage, as expected and proven by the measurements.

During the SNR measurements I noticed that the DAC switches its output off after 3 seconds when no playback happens anymore – some kind of energy saving? It’s not a simple mute, the SNR gets much worse in this state because the output stage is open (this is not an issue in real-world usage, with an open output the headphone does not play anything, and the process does not cause a click noise. But measurements are really screwed up). So for the base noise measurement I played a 24 bit file that only includes dither to keep the DS2 output alive (constant dithered playback is default on the AP, so one would not easily notice that behaviour there).

In that 'off' state the output impedance rises to 10 kOhm, no matter balanced or unbalanced. I also measured the output impedance in normal operation: Balanced 0.6 Ohm, unbalanced 0.5 Ohm. One woud expect other values here (like 2 x 0.5 Ohm for Bal = 1 Ohm), but not.

Finally I used the DS2 with Windows and now fully understand the frustration of others in this thread - the ramp-up effect, which on my unit even includes a right to left volume shift, is quite annoying, and IMHO a deal-breaker for many. As I already wrote I am lucky as I don't have that effect with my iPhone and Neutron as player, but I think one can't recommend the DS2 until Fosi has fixed that issue.

Many thanks for your measurements, great job.
The output shutdown after 3 second is the cause of the ramp-up effect when coming back on, my Sonata BHD behaves in the same way, I suppose they are almost identical implementation. I can add that the Sonata, when in "sleep" mode, catches up audible emi interference from my laptop, probably from the wifi or the ssd, producing a low but audible buzz noise that is greatly amplified if i tighten my hand around the iem cable while reducing if i touch the mtal part of the dongle body or the metal chassis of the laptop. Ironically it happens only if the dac is set into low or medium gain, disappearing in high gain (it has 3 level). As soon as something start playing coming alive from "sleep" mode, no more noises at any gain setting, so I can imagine how it can mess up measurements.

p.s.
I just ordered this other dongle, quite different dual CS43131 implementation, we'll see if it shows the same behavior.

 

[MC_RME]

Interesting one. More options than Fosi (I tried that one's key combinations but none of them worked with DS2), so it is a different design. Awaiting your report!

 

[MC_RME]

The output shutdown after 3 second is the cause of the ramp-up effect when coming back on

That might be. I just reproduced this again and did not see the 3 second timeout anymore, must have been the software I used before. Now with AIMP if you quickly stop/start a song (not pause) the ramp-up comes every time - and the output immediately goes into off mode on stop.

 

[Hammeredklavier]

Finally I used the DS2 with Windows and now fully understand the frustration of others in this thread - the ramp-up effect, which on my unit even includes a right to left volume shift, is quite annoying, and IMHO a deal-breaker for many. As I already wrote I am lucky as I don't have that effect with my iPhone and Neutron as player, but I think one can't recommend the DS2 until Fosi has fixed that issue.

Thanks, it's nice to have this confirmed by someone who knows what they're looking at. I don't understand how this can happen. Surely no one would release a USB DAC without making sure it works properly with Windows?! I know it's made 'to a price' but if it doesn't work properly then it isn't really worth anything.

 

[mc.god]

That might be. I just reproduced this again and did not see the 3 second timeout anymore, must have been the software I used before. Now with AIMP if you quickly stop/start a song (not pause) the ramp-up comes every time - and the output immediately goes into off mode on stop.

This is interesting... If it's something application dependent if anything I'd expect the opposite with some applications sending like constant keep alive signals preventing the unit to go into sleep mode. As an example, I'm using my sonata BHD on Linux, and monitoring the pipewire audio chain I have this:

1 - nothing playing

2 - mpv playing a file

if i pause plying, this remains the same for those 3 seconds, during that resuming play is seamless.

After 3 seconds, the output is shutdown while mpv still open:

3 - mpv paused after 3 seconds

if i resume playing now i hear the ramp-up issue.

This happens with almost any standalone audio\video application
With browsers it's a bit different, for example playing youtube in chrome i have this:

4 - chrome playing

but if i pause the reproduction, it keeps the output alive much more than 3 seconds, around 15 seconds within which resuming play is seamless. After 15 seconds in pause then chrome completely disappears from the chain and the output it shutdown immediately, going back to like the screenshot 1 while chrome is stil open in pause.

Then we have some applications like audio editors that behave still different, keeping the output on as soon as they are opened even though nothing playing, as we can see with Ocenaudio:

5 - ocenaudio open not playing anything

So an application that immediately shuts down the output after stopping seems a bit strange, but clearly it depends on the type of signal the apps sends to the os and the dac. Did you use stop or pause in AIMP? Did you set AIMP in directsound, wasapi exclusive or asio? I tested the Sonata on windows just briefly long ago to check the Asio driver but didn't test if the issue is there with asio or wasapi exclusive, I just remember that it has the problem when playing OS system sounds, so as directsound device.

 

[MC_RME]

I used ASIO with AIMP, and disabled the default rampings (fade-in/out) to clearly hear the delayed start from the DS2, not the app.

 

[mc.god]

I used ASIO with AIMP, and disabled the default rampings (fade-in/out) to clearly hear the delayed start from the DS2, not the app.

OK, so it could be something also related to os, maybe the savitech bridge is set to immediately shutdown after receiving a stop signal from the driver, and windows directsound and linux pipewire are programmed to send that stop signal after 3 seconds while ASIO, which has exclusive access, could be programmed to send an immediate stop signal to the dac. But this is only my humble speculation trying to understand the different behavior, I will check again with Sonata BHD with ASIO under Windows when I'll have to boot up my now almost unused desktop pc.

 

[OK1]

Thanks Amir and Fosi Audio. Great product. With downloadable drivers. Measures very well. For the discerning. Cost effective. Impressive. I have been using a TempoTec Sonata BHD (not the Pro version which was favourably reviewed on ASR by Amir). My BHD sounds really great, satisfied customer, with only one strange behaviour - the stereo image shifts a bit to one side, but this improves and is more centered, when I unplug it from USB-C on my laptop and plug it back again (with a wait of a few seconds, between both actions).

Great time to be alive., with superb products at affordable prices.

 

[Emlin]

Thanks Amir and Fosi Audio. Great product. With downloadable drivers. Measures very well. For the discerning. Cost effective. Impressive. I have been using a TempoTec Sonata BHD (not the Pro version which was favourably reviewed on ASR by Amir). My BHD sounds really great, satisfied customer, with only one strange behaviour - the stereo image shits a bit to one side, but this improves and is more centered, when I unplug it from USB-C on my laptop and plug it back again (with a wait of a few seconds, between both actions).

Great time to be alive., with superb products at affordable prices.

I too hate it whenever my stereo image shits, especially when it does it on the carpet!

 

[OK1]

I too hate it whenever my stereo image shits, especially when it does it on the carpet!

Corrected. Thanks. Meant to say shifts!

 

[Chr1]

Very interesting concept however.