I have a Roland quad capture audio interface, Realtek integrated sound in my PC and Nvidia sound in my graphics card.
 
It is a new I9 PC. 
I am trying to set it up to make (Cakewalk/Adobe/raytracing) instructional and gaming videos. 
 
I plan to use OBS studio to record.
 
I want the switch from when I make a gaming video to when I make a Cakewalk instructional video as least complicated as possible. As simple as changing a profile in OBS and not having to reroute sound devices, hardware plugs and drivers.
 
Example: When I am in Cakewalk I would like to use the same mic (plugged into the Roland quad capture with phantom power) that I am using to narrate the video as I use to sing and lay down vocal tracks in the video.
 
I would think it should be as simple as creating an audio track in Cakewalk and having it input the same mic.
 
No matter what I do I can't seem to get the quad capture to play its audio without it also including the mic input in the signal.
 
I am so confused now between "the main", input 1 + 2 and input 3 + 4.
 
I have a USB analog to digital audio converter box also that has left and right RCA inputs to a coaxial RCA output, this would output the Realtek or Nvidia sound card into its own input on the Roland quad capture. 
 
Though, this seems like the wrong way to do things and I am afraid of a feedback loop and blowing up my whole computer! lol 
 
There must be a way to isolate within the Roland quad capture the mic input from the Cakewalk master output so that OBS can separately see them.
 
I am confused between the "direct" output switch in the quad capture driver software and the playback/input knob.
 
Any help in clarifying this would be greatly appreciated.
 
I don't expect to get all of this answered very quickly. It seems like the kind of thing that may take a month to figure out.  :)
 
OBS studio is fairly easy to set up it is getting Roland quad capture to isolate the master output of cakewalk and my games from the mic input on the quad capture that is giving me the most frustration.
 
If I switch to a USB mic then why use the quad capture at all?
 
I like the quad capture and I am familiar with how it works and I like having a real physical knob to attenuate the volume.
 
I just can't figure out how to set the quad capture to isolate the mic and master outputs into separate tracks.
 
Any help on this would be greatly appreciated.
 
Someone in a Youtube video suggested using VoiceMeeter Bananna, there must be a way to do this without that program. 
 
Thanks in advance.

I do not think I understand what you mean by all that right... but:
 
Each audio device has hardware (!) "Output(s)" and sometimes "Inputs". At the beginning, they are separated, till something route one to another. There are several ways to route hardwire(!):
1) with real cables you can connect some Output to Input, on the same device or between devices. Note that you can not connect Input to Output with a cable. F.e. you can connect Realtek output to Roland input, but you can not connect Roland input (mic) to whatever using this method.
2) with hardware mixers (I remind, I write about hardware only so far). Roland (and Realtek) have build-in mixers. On Roland that is "Direct monitor". But do not be confused by "Direct monitor" "Mono" button. It defines weather input should be mixed as mono (so to the center) or "as is" (1->left, 2->right). The amount of direct monitoring is controlled by the "Mix" knob. "Playback" (leftmost) position means no direct monitor, "Input" (rightmost) position means no playback (so direct monitor only).
 
Now to the software. Here it is a bit more confusing. When you set the "Input" in software, it does not directly mean some Inputs hardware. In general, in software(!):
1) software mixer can route Input to Output directly. In Windows standard audio settings (for inputs) or custom GUIs of interfaces.
2) the "Input" for software can be hardware Input or software Output (!). OBS can record Mic input of Roland, but it can also record (supposed to be sent to) output for Roland/Realtek/NVidia. Note that in the second case, it records only signal generated inside PC for the output. F.e. if you Direct monitor Mic on Roland but record the output for Roland in OBS, you will hear the mic in speakers (connected to Roland) but it will not be recorded by OBS (since it will be "mixed" with PC output on hardware level). If you not direct monitor Mic on Roland but record Roland input in OBS, you normally will not hear mic in speakers (but see (3)) but it will be recorded.
3) software can "direct monitor" inputs to outputs. F.e. in Sonar that is "Echo" button of the track. The effect is like in (1).
 
But... In general, sharing one hardware device between several programs in parallel is a bad idea. Imagine one computer keyboard is used by 2 persons. Sometimes that is not possible at all ("exclusive" mode WASAPI, ASIO for most interfaces), can produce some troubles (audio glitches, no sound for one of the applications) or work on acceptable level.
I do not have Roland Quad Capture, but having Roland e-drums and VS-20, I guess any Roland fall into "produce some troubles" category (all Roland interfaces in practice use almost the same Edirol technology).
M-Audio I have fall into "can work on acceptable level" category. As long as bit depth is equal.
RME... can do everything. If you are serious about your setup, think about Babyface Pro. It will take you may be 2 month to understand, will cost quite a lot including cables (note you will need special XLR->TS/Chich if your speakers do no have XLR inputs). But it will work with "one click switch" once setup correctly, without noise and reliable.
 
With what you have, it is also possible to route things. But not perfect.
For example:
* set Sonar to use windows driver (not ASIO, not WASAPI), do not forget to increase buffers. Monitoring throw Sonar will not be possible, so you will need to use "Mix" knob on Roland.
* set Windows to output to Roland.
* set OBS scenes to record what you need, Roland's input alone or Rolands output (so everything mixed)
 
Or...
* set Sonar to use Roland in ASIO
* set Windows to Realtek
* connect SPDIF output from Roland to Realtek (if you can) or normal/headphones output of Roland to Realtek.
* turn "Mix" knob all the way left and orginize required routing of mic into used for chained connection Roland output inside Sonar or use "Mix" knob to control mic volume to OBS.
 
There are several other variants... Once you get the idea, I guess you can find what you need 
 

azslow3
 
Or...
* set Sonar to use Roland in ASIO
* set Windows to Realtek

I am not familiar with OBS studio, so will only talk to the hardware/Windows settings at this point.
 
The above "may" be your answer, but I am also not clear what the OP is really doing. Windows will run multiple interfaces, but will not let a single program run multiple ones simultaneously, so the above suggestion is preferred, and how I run my system. Are you recording with ShadowPlay, or is OBS studio a variant of that on the audio side?
 
First, the Nvidia HD drivers are only going to do anything if you run HDMI output to a monitor that can play sound (not preferred in most cases), so they only give another resource for Windows to manage and get confused with. You can go into the "Device Manager" and disable those (easiest), but forget offhand if they are required for ShadowPlay.
 
Second, to achieve the above (at least in Win7) you must explicitly tell Windows to use the Realtek driver by default. To do this:

1. Open your sound properties with both Realtek and the Quad on (so both devices are listed).
2. Even if Realtek is "default," set the Quad to default, then shift the default back to the Realtek (highlight them and choose the "set default" near the bottom of that screen).
3. If the Quad is set to default, you only need to highlight the Realtek and set that to default.

With that done, Windows will now default to Realtek for any program which needs audio output and manage conflicts, but makes the Quad reserved for a program you want to use ASIO with (i.e., SONAR). Essentially, a recording program that has ASIO selected inside the program itself is the only thing that will see the Quad.
 
From there, open SONAR and configure it to use the Quad with your mic connected to it, and you should be able to record that mic in SONAR while recording in ShadowPlay, but will need to use a video editor to mate the two audio track up after the fact.
 
You can achieve recording both in SONAR with a physical loopback but this requires discipline and proper setup so that you do not blow an input on the Quad doing so. I will not post details on that at this point, since I am not clear what you are trying to achieve and am not sure how familiar you are with setting things up (no offense to you, just protecting your hardware).
 
Could you please let us know the specifics of what you are trying to do?

Thank you both for your awesome posts! They will take me a month to study them.
 
This is, I think, by far the most complex discussion I have ever had online.
 
I plan to really learn a  lot in this process. This takes more than just learning how to set it and leave it alone but a deep understanding so I can make changes on the fly.
 
It is sort of simple what I want to do but then again it is extremely complex.
 
I want to be able to set windows to record a "live" XLR microphone track (mono), a "real-time" stereo mix track and a video track simultaneously from Cakewalk.
 
I would like the live mic selectable and useable in Cakewalk and the realtime mix track supplied by cakewalk while the video track is supplied by a USB webcam.
 
I want to record all three and save them as a multi audio/video mp4 file.
 
I could just try and mix the mic and audio mix together and have one stereo file but that is not how I like to do things. Especially when I do not plan to stream this but I plan to edit it later because I want it all in 4k. 4K is too big to stream. So with post editing I could drag any out of sync tracks either way to align them to the video and each other in the video editor.
 
I do not want to use a USB mic it must be an XLR mic recording.
 
Does anyone have a complete schematic for the Roland quad capture. :) let's get on the same page here.
Having a schematic for the Realtek sound card would be helpful too. 
 
THANK YOU!!!!  
 
Last night I uses "asio 4 all" windows drivers instead of my roland drivers and suddenly a lot of things had ASIO listings in Cakewalk,  but still nothing seemed to talk to one another and with various configurations my quad capture kept giving me the message that my "device was in use".
 
At another point I got my mic isolated but the cakewalk mix had my mic in it also...
 
I have two little devices also i got on amazon. one is an analogue to digital converter and the other is a digital to analogue converter. The are both a little smaller than a pack of cigarettes (usb powered) and both have RCA ins and outs. So I could make use of the  3/4 inputs and outputs of the Roland quad capture. The stereo RCA outs could go into the realtek line in. Or I could use a mixer and feed the Phantom powered mic into the 3/4 ins if need be. I was up till 4 AM and still no luck. I don't have a phantom powered mixer (yet) also. It would only need one channel.
 
Yet I am still not sure if the quad capture would pass these inputs out separately or mix them together before exiting the audio interface. 
 
I suppose with a phantom powered mic I could send that into the line in of the realtek but then I would no longer have it for use in cakewalk unless I could route it in somehow like a midi synth then I could record the mic before it enters the quadcapture.
 
Just confused and scattered thoughts at this point.
 
Maybe I will buy a "very cheap" preamp and phantom powered
 mixer. 

I just dug this out of the closet and YAY it is phantom powered!
Behringer Xenyx 802 Premium 8-Input 2-Bus Mixer with Xenyx Mic Preamps and British EQshttps://www.amazon.com/Behringer-802-Premium-8-Input-Preamps/dp/B000J5XS3C
 
I wonder if I can send a mono signal from this mixer into the mic input jack of the Realtek?
 
Would it work and would it be wired right to accept a mono 1/8" jack into the PC?
 
I could set the Realtek mic input in cakewalk to not enter the main mix but just record and monitor.
 
Would there be a lag in Cakewalk and would my vocal be offset?
 
An I9 should be able to handle it right?
 

RexRed
I could set the Realtek mic input in cakewalk to not enter the main mix but just record and monitor.
Would there be a lag in Cakewalk and would my vocal be offset?

Sorry, still unclear for me
 
Cakewalk track -> OBS
Mic -> Cakewalk -> ?
Mic -> OBS
At the same time?
 
https://www.amazon.com/Behringer-302USB-BEHRINGER-XENYX/dp/B005EHILV4/ref=sr_1_4?s=musical-instruments&ie=UTF8&qid=1517664241&sr=1-4&keywords=behringer+mixer+usb
Can give you:
Mic -> USB -> OBS (without realtek) and in parallel
Mic -> (Analog) -> Roland -> Cakewalk
 
But...
------------------------------------------------------
But I have just tried with my Babaface Pro:
Mic -> Cakewalk (separate track, recording mic) -> Monitor (separate output) -> Headphones
Mic -> Headphone (directly, without software)
Mic -> OSB
Cakewalk (backing track, playing ) -> Separate output -> OSB (so, no mic from Cakewalk)
Firefox -> OSB
Firefox -> Headphones
Firefox -> Cakewalk (separate track, recording, no monitoring, so not Firefox from Cakewalk to OSB)
ALL that AT THE SAME TIME!
I was switching on/off particular routes to check that it works as expected.
I guess that can cover everything you can imagine to do
 
The difference: you roland is 4x4 and is not multi-app capable (till degrading latency and quality). BFPro is 12x12x(12x12) (last 12x12 are software inputs and outputs, they are decoupled from first 12x12 hardware connectors), with build in digital mixer. So you can create 12 (6 stereo) separate mixes from 24 input sources (12 hardware inputs + 12 software OUTPUTS). And you can use any set from that mixes as the INPUT for software in addition to OUTPUT to hardware.
With one mic, monitors and headphones as the only "hardware" that sounds like an overkill. But to organize the routing I have mentioned, I had to use more than 50% of that "matrix".
 
I must admin that before I put BFPro on my table (a) I was sure I do not need it (b) it is too expensive (c) I already have 8 (!) interfaces in the room (Realtek, SB, M-Audio Audiophile, M-Audio 410, Behringer USB Mixer, TC Vocal processor, Roland RD-11, Cakewalk VS-20).
But since I have it on my table, I am convinced (I) I use more its functionality then I have thought I will (II) for stability, latency and functionality, it is well worse the money.
For your case:
* you can route the way you like, you can use it in parallel by several applications. So whatever you have in mind, you can be convinced that your system can manage that.
* you can sing/speak with monitoring throw the DAW and till you are crazy about latency, you will not hear ~4.5ms delay
* you will not loose the sound in the middle of something (with all my other interfaces, that has happened more then once)
 

Here is a routing diagram I created in Adobe Illustrator. This is how the universe was made (haha, just kidding). Let me know what you think of this way of hooking things up. It took me several entire days to figure it all out. 
 

There is at least one problem:
Since Cakewalk <-> ASIO Roland is "closed" system, you can not get sound out of it other then with directly connected to Roland speakers. I mean you can not capture it into OBS nor route throw realtek to speakers.
So you will need to connect Roland output to Realtek input, better digital, but analog can work as well.
 
For me, there will be yet another problem. While many people claim Realtek has seriously improved audio quality, I can not use SB nor old Realtek with my studio headphones, there is too much noise (from low hum up to high pitched inducted from HDDs, mouse and other components). It is in fact better with my 2016 notebook, but still not usable for me.
 
For DIY fun, the schema can work. But you have i9, 3 monitors and want 4k. Yet you build sound schema on under $100 audio with $1 (realtek chip) primary component. Roland interface is more or less fine, but it was thought for other jobs.
 

The Realtek is set to record in 24 bit and unless you are Neil Young and able to hear flea farts you will not be able to hear the noise floor in any recording. Even a 5 dollar sound card recording in 24 bit records far above the inherent noise floor. If you are getting noise in your recordings it is your CPU fan and I suggest getting some long, second gen USB cables and moving your PC further away and some sound proofing foam inside your PC to deflect fan noise and low noise fans. The PC fan noise and keyboard clicks would far surpass any noise made by the computer's electronics.
 
Also, I send my line out of the PC into a: 
PYLE-PRO PHE400 Hum/Noise Eliminator 2-Channel Box with TRS Phone and XLR Inputs/Outputshttps://www.amazon.com/gp/product/B00BARTW42/ref=oh_aui_detailpage_o09_s01?ie=UTF8&psc=1
 
This eliminates a huge hum factor.
 
If you are hearing any hum in your output or recordings after all of this you have a faulty device along the chain or you are setting your record volume way too low.
 
All electronics put out hum and it is that very hum that is being added back into recordings to give them more analogue warmth...
 
Also one could add a second Roland audio interface if need be but I am still learning the system and learning the ins and outs. Also, having the PC keyboard set master volume and mute is a nice feature.
 
If over time I notice any degradation in the final sound product I will go for a second external interface replacing the realtek internal interface altogether.
 
Still recuperating from the I9 price tag though. :)

RexRed
The Realtek is set to record in 24 bit and unless you are Neil Young and able to hear flea farts you will not be able to hear the noise floor in any recording. Even a 5 dollar sound card recording in 24 bit records far above the inherent noise floor.

I am sorry to disappoint you... but there is no single audio interface in the world with can record in 24bit...
Before you start to think that I am crazy, please read on. You will understand.
 
There is an audio input part (analog), than ADC (analog to digital converter), then you get some bits. Once you are at bit (digital) level, you can transfer everything precise. Modern ADCs are more or less ok, already not 24bit precise, but ok. But analog part is what can be from worse, throw good up to very good. To get some ideas how good it can be (that is not the whole story), look at SNR characteristics of the interface. It will immediately limit which bits can be real and which can not be. Calculation is simple: 16bit -96dB, 24bit -144dB.
Now have a look at specification (keeping in mind they want best numbers, sometimes putting the device is unrealistic, special tuned environment). For Realtek: http://www.realtek.com.tw/products/productsView.aspx?Langid=1&PNid=14&PFid=24&Level=4&Conn=3
(select the chip you like).
For older one, you will find something like -85dB. Under 16bit! Better -104dB. So claimed 17bit, in perfect condition.
Top audio interfaces can do -118dB. ~20bit
 
I have older Realtek. Not only it is under 16bit, that is not so bad. But analog circuits on it was bad isolated. The noise I was talking about is not mechanical, it is inducted electrically from other circuits. Including digital.
 
I do not try to convince you waste money. I just try to explain what you have, what you can have and where is the difference. If not welcome, I have other things to do.