I've had an issue ever since I started using the Nu, that when recording with 384000 sample rate, there's DC offset present in the recording.  I was told before, DC offset is corrected by the Nu at sample rates at 192000 and below, but not in the highest rates.  I always blamed my source but today, I made a recording of the line in with nothing connected and the same DC offset is present.  To be clear, I use the card as an ADC for recording; I'm referring to DC offset being present in the Line In, not the line or headphone out. 
 
What might cause this and how might I remove the offset?

tigerdognew
I've had an issue ever since I started using the Nu, that when recording with 384000 sample rate, there's DC offset present in the recording.  I was told before, DC offset is corrected by the Nu at sample rates at 192000 and below, but not in the highest rates.  I always blamed my source but today, I made a recording of the line in with nothing connected and the same DC offset is present.  To be clear, I use the card as an ADC for recording; I'm referring to DC offset being present in the Line In, not the line or headphone out. 
 
What might cause this and how might I remove the offset?

First I would check is to make sure everything is properly grounded and you don't have a ground loop. Make sure any PSU wires are routed away from any audio signal cables so you're not inducting current or noise.

I talked to our friends at Audio Note, UK, and it reminded me of another user who asked this question before.  The answer is that this is basically due to how the AK5572 works. 
 
Per Audio Note:

"There is always a DC offset internally, plus that from drive electronics, but, at lower sample rates it filters it out internally. At 352.8kHz and 384kHz, the internal filter is disabled to reduce DSP overhead.  I did consider applying an analogue cancellation scheme but decided against it due to additional complexity etc. and concerns about asymmetrically losing resolution.
 
The simple solution is to use a plugin high pass filter in the DAW. I have done that, and [another customer who used the card to record high frequency animal calls] did the same successfully.
 
AKM uses an internal DC offset to reduce idle tones, it's quite normal and ordinary.
 
Here is an excerpt from an ADI article:
 
'...(The) frequency of the idle tones is a function of the dc input. This opens up the possibility of introducing enough dc offset on the A/D input to move the idle tones out of the bandwidth of interest to where they will be filtered out by the decimation filter.'
 
Which is what is being done here, except it's only visible when the conversion frequency gets so high the internal filter is disabled.
 
There are two ways to get rid of it, in DAW software. Firstly cancellation, secondly filtering."
 
Hope that helps.

tigerdognew
Lee, many thanks.  I think I had asked this question before, but understood it to be a function of my external equipment that the card wouldn't filter.  When I saw it present without input, I questioned my sanity.  The explanation above makes things more clear to me.  If I read it correctly, the internal DC offset is intentionally designed into the AKM ADC to eliminate unwanted behavior.  at 352.8 and 384KHz sampling rate, some external component in the digital signal chain must be inserted between the card and the recording software.  I'm not quite sure how to do this with Linux ALSA and Audacity.  I can post-process with filters in Audacity, but the DC offset is present in the initial record, and limits to about 36dB of dynamic range, not ideal either.   Given my personal setup, it sounds like I'll need to record at 192KHz.

Have you tried this?  https://manual.audacityteam.org/man/dc_offset.html

The card doesn't have the analog bandwidth to capture anything even close to 300 kHz, so why are you using that high of a sample rate? It's really pointless and silly. Take the analog bandwidth times two, and that is the maximum sample rate needed to guarantee that all data is captured. 192 kHz is already way more than necessary for the analog bandwidth of the card. Much less, consider your audience is likely a human who can only hear up to 20 kHz.

tigerdognew
Lee, thanks for going the extra mile to look for solutions!  The Audacity "normalize" effect that's referenced is a non-real-time, post-processing effect.  It does what it says on the label, but only on audio that's already been recorded. If I could figure out how to upload a screenshot, I could show the Nu's DC offset effectively limits dynamic range to only 33 dB.   Even my turntable setup provides dynamic range greater than 33 dB, so the offset makes accurate captures impossible - peaks simply clip.  Since I work in Xubuntu Linux, the Windows native offset canceller cannot be used. 
 
Here's a link to the screenshot, hosted in Google Photos:
 
 

Is this screenshot with nothing connected as you stated? (Possible grounding issue?) I'm honestly curious how this is possible given that most cheap-dac's ($20 range) can get to -65 dB or so before hitting a noise floor. Most of the external units I've worked with in recording are capable of hitting -92 dB without any side-effects whatsoever but that's round about the limit of 16 bit audio depth anyway and anything lower for 24 bit work is gravy. Most of the recordings we ever made (music industry) were in 24/96 but then later run through an analog summing device during the mastering process and captured at the desired native target. So we ended up with several sub-masters at 16/44, 16/48, 24/48, 24/96, and etc if applicable.
 
I also wouldn't be too surprised if the 3.5mm jack isn't causing noise issues just due to it's size and relative location vs a bunch of other EMI emitting devices. The design and location of these jacks may be the cause of your issue in this case but I couldn't say definitively of course. That said, I wouldn't trust this card with high-end recording personally but are you able to get better results by switching to 24/48 or 24/96 and letting the DSP do the rest of the work? Some of the best interfaces I've ever had the pleasure of using only ever got to 24/96 and I've absolutely had interfaces that were only capable of 16/48 and would have taken a quality 16/48 over a lesser 24/96. It's all in the design of the supporting hardware. Even then, I couldn't even think of a use for a sample rate higher than 192k outside of an electronics lab (oscilloscopes come to mind). I'm not doubting your abilities, and much respect if you can hear the differences above 48k, but for better or worse I certainly can not personally tell. Bit depth is a different story though but that's mostly around audio engineering and what you can do with a signal after it's captured. In the case of guitar pedals and such with wicked compression and the like.. .you'll want a clean, low floor to clean up from or you get a lot of strange (but sometimes desirable depending) effects. Even the flaws can be used to great effect. However, from the aspect of audio preservation, I would want to stay as far away from EMI as possible and use balanced cables where possible. It's a bummer that the I/O on the card wasn't all RCA jacks at the very least but as a consumer device I understand the choices they made. Even still.. quite interested in picking one up myself despite the potential drawbacks here.
 
For what it's worth though, I would be somewhat interested in seeing loopback test results from this card. Yes, I know those tests are subjective but it's a basic starting point for understanding some of the characteristics of how the internals perform. Search for an article from hifizine about the focusrite 2i2 circa 2012, it'll give you a glance at the process and results from other consumer devices. The 2i2 is a popular low-cost interface a lot of musicians use on the go.
 
Are you able to confirm that the DC offset is not present when in 24/48 or 24/96? That would kinda bum me out for use with a PC mic, but I would think the dac can filter that internally without involvement of external software so long as the DSP has enough time to think about it vs clocking through high sample rate data.
 
I'm honestly interested in picking up one of these EVGA Nu cards up for myself but mostly to use as sound output for when I don't want to power up my rack or just to leave headphones connected to. Just reading through and trying to see how much of an issue coil whine (emi issues etc) are on this card but that's for a different thread.
 
 

-