Schiit Gungnir Multibit impressions

Is there some sort of disadvantage to setting it to 32/192 regardless?

 

I "think" that the Gungnir Multibit will always use the filter unless the input is at 352.8K or 384K. Half that for Bifrost Multibit.

 

Other than adding extra zeros to the ends of long strings of numbers, no. Unless you have a computer that makes lots of mistakes and accidentally throws the occational one in the information.

 

I'm not the most knowledgeable on the subject of DACs around here so some parts of my answer might be wrong:
1. The 32 and 24-bit capable DACs that are advertised almost everywhere are actually only 5 to 8-bit Delta Sigma DACs using modulation. The 32 and 24-bit specifications that they flaunt around are actually the capability of processing 32 and 24-bit digital signals without dithering. the 19 and 21 bits that Schiit specify are actual, measureable DAC performance (a 16-bit signal theoretically has around 96db or dynamic range,etc...) E.g: A TV can process wide-gamut colors but cannot display them due to the display's performance. Currently, physics limit even the best measuring equipment from going past 22 bits of accuracy so Yggdrasil is probably as good as it can get without using theoretical and immeasureable specs.

2. MegaComboBurrito Filter is a closed-form filter that upsamples/oversamples your incoming 44.1 Khz/48 Khz signal into the native sample rate of the DAC which is then outputted as an analog signal(Bifrost Multibit:176.4/192 Khz, Gungnir Multibit/Yggdrasil:352.8/384 Khz). There exist Non-oversampling DACs but most people's impressions of them are that they sound rolled-off to hell (overgeneralization). All you need to know is that this filter has been in development for a long time and was optimized for time and frequency accuracy and has been regarded as one of if not the best upsampling filters. Also, this filter can only upsample by multiples of 2 (important but perhaps too technical for this explanation).
Moving on from that, setting prefences to higher than the sample rate of your played file will enable Window's Directsound upsampling, which is widely regarded as trash.
For example, setting it to 32/192( 32bit is fine and won't affect your signal in any negative way) will upsample your 44.1 Khz into 192 Khz using an asynchronous filter (can upsample into a rate that is not a multiple), which will then be doubled into 384 by the Gungnir Multibit, creating an imperfect signal due to the low quality filter used in Windows( Macs are better at this and Linux has configurable filters). The only way to make sure your music will only be upsampled using the Schiit filter is to use WASAPI exclusive/ASIO in order to feed to digital signal directly to the DAC without the influence of Windows. So, as a guide:
-Set your prefences to 32/44.1 Khz if you listen mostly to 44.1 Khz music (99.99999% of digital music) in order to prevent Windows samplling your music if you're using Spotify or Tidal. (EDIT: Tidal now has WASAPI support)
-Set your media player to output WASAPI exclusive (event>push)/ASIO in order to bypass Windows altogether and let the DAC do its work (switch between multiples of 44.1 and 48 Khz according to the incoming signal)(recommended).
-Games and apps are hit and miss since no game/app developer supports WASAPI/ASIO in its games and there isn't a consensus on what sample rate games/applications use for audio.

 

Your Gungnir Multibit is doing its job without you having to set anything in your sound settings. That said, you probably want to use the WASAPI mode from your player (foobar, musicbee, and jriver all support this). This has the effect of bypassing the processing that the OS does to be able to mix multiple sound sources, which you don't usually want when listening just to music.

The number of bits (19 on Gungnir Multibit) is somewhat irrelevant since it really just amounts to how far down the noise level gets pushed. ie there is no real benefit for a DAC to process higher bit depths. This probably would benefit from more explanation if you want.

The filter is always in use. Basically, almost all modern DACs oversample the typical bit rate to something several times higher than 44.1 (Gungnir Multibit does 8x), in order to make it easier to filter off the noise without messing up the phase and other things. Oversampling means interpolating samples to "fill in" the samples between the original rate, say 44.1, and the higher rate, say 352.8 or 386 khz. Those inbetween samples have to be computed to match what the waveform should be, and this is what the MegaTaco does so well.

This also probably needs more explaining, if you want more.

 

You don't want to let the OS upsample the sample rate of your music, because it does a terrible job of it. This is the purpose of the fancy filters on DACs, especially the BigEnchiladaFilter - it does the upsampling interpolation extremely well. It's where the magic happens.

Edit: upping the bit depth doesn't hurt though, in general.

 

@Ash1412 and @lm4der, you guys are rockstars. Thank you very much for the explanations - super helpful.

One last question (I think): If I set my settings to 32/44.1, I don't have to worry about WASAPI?

 

This (32/44.1 or or 24/44.1) is generally the optimal setting if you are NOT using WASAPI. In this case you are using the OS mixer, and all you are doing is setting the output bitdepth to essentially match the bitdepth that the mixer will output, so that it doesn't get processed back down to 16 bits. This also assumes your source samples are 44.1, because you are trying to leave that part alone.

In this mode, in theory, the OS mixer shouldn't be messing up the samples too much. However WASAPI bypasses the OS mixer and goes direct to the kernel and out to the DAC. So in theory the WASAPI mode incurs zero processing of the samples, which is technically better. Most people run WASAPI either because they can hear the difference, or just because it is technically more direct and cleaner.

 

Unless you listen to 48 khz music then you probably won't. I set Windows to output 32/48Khz since I do most of my critical music listening using foobar WASAPI so my music won't be affected and I read somewhere that most games, like most movies, output at 48Khz. It all depends on you usage. I do plan to obtain a Mac some time in the future though, just for the sake of better driver compatibility and better OS upsampling.

 

Right, so @Ash1412 is setting his OS mixer settings to 32/48khz, because when he is listening to games or youtube, etc, and allowing the mixer to do its thing, he is optimizing for the idea that a lot of sources are 48khz. CD's are, however 44.1 - but he is using WASAPI for his critical listening, which bypasses that 32/48 setting completely and leaves the samples untouched.

BTW, I'm not sure I can hear the difference between the Windows mixer at 24/44.1 vs. WASAPI, but this kind of thing is pretty subtle.

 

Great to have all this detail. Do any of you know whether the advice above applies to the Mac also? I'm on El Capitan and have my Mac outputting at 16-bit/44.1 to my Gungnir Multibit (via MIDI audio device settings), but I'm not sure if OS X sends the audio unadulterated or if I need to do something special in settings.

 

Yeah, all the above applies for the MAC. The MAC has a better mixer than Windows, at least in terms of the upsampling algorithm. The equivalent to WASAPI is Core Audio, and you want to enable exclusive mode for it.

 

Also by using software like http://odeus-audio.com.au/Odeus/AsioLinkPro you can essentially bypass Windows by making it think there's an ASIO driver installed, the tool will take care of the rest. This essentially makes the entire PC, from applications, games to web browsers all use ASIO instead of DirectSound. You still need to pick 1 setting because this way you can also use multiple applications at the same time with ASIO, while ASIO normally uses exclusive mode, but that's it. No more fuddling with DirectSound and if you use Foobar2000 or JRiver, just pick the original driver so that either uses ASIO or WASAPI to truly output in the native format.

After doing this, I'm never going back. Whether it's my imagination or not, there's no OS sampling or interference and I honestly hear an improvement. This made streaming audio services much, much better sounding.

@Boops
Nope, opposed to Windows, MAC entirely does everything on DAC side, the OS doesn't interfere.

 

Is there a way to enable that mode in iTunes (or some other native setting)? Or do i have to use a different player app to do it?

 

One advice. You should set the bit depth to the highest possible, since converting from 16 to 24/32 bit is lossless and 24/32 bit is better for digital volume control

 

Well, the MAC still uses a mixer to manage multiple streams. You want to set your music player to use Core Audio in exclusive mode.

 

Yeah, good point about the reason for the extra bit-depth - this gives you more headroom so bits don't fall off the bottom when using software volume control.

 

But does it upsample? I was always in the assumption that there was no resampling going on hence why there's no ASIO on it. But the fact that you need to put it on exclusive mode kinda answers that I guess.

 

Yeah, the MAC mixer is still tampering with things to some degree. Not sure if it upsamples by default. For some reason I think the MAC is smarter on this, and automatically uses the source sample rate, rather than always upsampling. But that too may be player dependent.

 

I concur. There's simply no way for sound streams of different factors,e.g a 192Khz music stream and a 44.1 Khz music stream, to play at the same time without being converted into one stream with one sample rate ( the one you set in the settings). This makes me wonder why the heck don't all content creators use either 44.1 or 48 Khz for everything and then let the samplers convert everything to 176.4/192 with much better results than linear resampling?