Atmos on the Flip Side | Nuprime H16-AoIP Setup

[BestGreed]

For a specific need requiring free signal re-routing, I identified and purchased the Nuprime-X H16-Aip, which arrived on October 31st. I believe I am the first user in my region.
The H16-Aip addresses AoIP system decoding needs at approximately one-third of the budget, offering significant value even compared to modifying integrated AVRs (e.g., adding a Dante expansion via I2S for AVR5).
My setup needs to support PC gaming at a desktop, then rotate 180 degrees for movie viewing. This approach maximizes space utilization in a small room.
I have devised a solution that, while somewhat unconventional, is readily achievable and cost-effective.
The core architecture is: PC-AVD-DSP-Interface-Speaker

Specific configuration:
Main PC: 13700k + 4070
AVD: Nuprime H16-Aip
x86 DSP: 11800H + 2060S with IP-KVM
Interface: Shure XMWANI8 (x2)
LCR: Yamaha HS8(x3)
Surround: Yamaha HS4(2 pairs)
Top Surround: Yamaha HS3(2 pairs)
Subwoofer: Yamaha DSR118w
Switch: FortiGate FG-60E
Mic: B&K type-4958A, along with a Presonus IO24, was sent to a laboratory for full system calibration.
Software: DVS license, VB Matrix, EQ Apo, SoundID license
Display Monitor: HKC OG27QH
Projector: Dangbei S7 Ultra Max

I also plan to try QuickEQ, and may borrow a desktop-suitable Dirac license from a friend.
I will update this thread on the process. Given my limited environment/equipment/technical proficiency, and not being highly proficient in some electro-acoustic related issues, debugging these items may be a challenge for me.
I'm sorry I can't provide more pictures of this product, mainly because the silver version I received is unlikely to be released as the official version.

 

[dziemian]

Please share your experiences. I am on the same way

 

[BestGreed]

Busy with work lately, but managed to make some progress that I’d like to share.

To introduce my room and layout: the space is roughly 6m (L) × 3.8m (W) × 2.55m (H), and I’m only using about half of it (I know this is not ideal). With a rough model created using DARDT and SolidWorks, the layout is approximately shown below.

 

[BestGreed]

I’ve now connected all the hardware, soldered the XLR cables, made the network cables, and mounted everything into the rack.

I did a simple connection test between the H16 and the interface, and everything worked correctly—at least there was sound.

Hardware topology

 

[BestGreed]

For the first round of tuning, I only ran a basic REW sweep.

As expected, due to room effects, the rear top channels and the center channel show some peaks.

I originally planned to use QuickEQ for calibration, but found that the software cannot connect to ASIO, which means it doesn’t support calibration above 8 channels.

So I tried SoundID Reference calibration instead, but the sonar-based mic localization performed very poorly in my system. It barely detected mic positions, and further measurements were impossible.

 

[BestGreed]

To get the system running as soon as possible and verify feasibility, I opted for Dirac + BC. Everything went smoothly. I ran the Dirac plugin inside a DAW, and the whole process was very stable.

During the process, I also ran into a small issue: since Dirac does not support B&K-format microphone calibration files, I had to convert the calibration file manually by following an online guide.

The Dirac measurement results were not very different from REW. Following a friend’s suggestion, I designed the filters accordingly.

For now, the system works reasonably well, although Dirac didn’t perform as well as I expected. To my ears, there is some unnaturalness in the high frequencies, and it seems the front top channels may have degraded in quality. I don’t intend to investigate this further at the moment. I will try EQ + rephase later when time allows. Since obtaining ART on PC is difficult, I probably won’t be purchasing Dirac.

The chart shown here is the REW verification measurement taken after applying the Dirac filters.

 

[BestGreed]

I can confirm that I’ve achieved my initial goal: I feed the H16-Aip’s 7.1.4 signal as the bed layer into the Dolby Atmos Renderer bus inside the DAW, and successfully rotate the sound field by 180 degrees. This allows me to maintain good spatial audio localization even when using the display on the opposite side of the room.

I can share the software topology later. It is more complex, and for effective re-routing I’m currently using a 48×48 matrix. If I eventually move to a convolution engine to run FIR + IIR processing for results closer to Dirac + BC, I may need additional channels.

 

[opcod]

Hi

This look quite an interesting setup. So with the calibration, are you gonna run as :
You take measurement in sonarwork(same as REW)
-then pas the output to Equalizer APO
-that will then provide the 7output channel
And then assign those channel somehow to the hdmi port. Then over the 'aoip controller' assign to those Shure decoder unit.

Or is this Nuprime give you 7input and you will pass over a matrix, maybe VB banana, then out of Equalizer Apo to all the in of the nuprime..? if that make sense..

Also, while it look interesting for a basic setup, just having any AVR receiver , they do in majority 9.1 and over. Do you feel limited with that unit, as you just mention you feel missing channel ?
They show 16out port on the connector, but not provide 16out, maybe the version nuprime h17 or so will do full 16ch...

Thanks for details

 

[BestGreed]

Honestly, it looks like you have no clue how desktop auto-calibration tools like SoundID Reference work, especially in multi-channel or immersive setups.

For systems exceeding 8 channels, desktop calibration tools generally cannot operate through WDM, and the only practical option is to run the correction as a VST.

These tools also do not export filters for use in external convolution engines, which makes VST insertion the only viable workflow.

During measurement, the calibration software must access the full speaker matrix via ASIO, otherwise only 8 channels are available.

This is why WDM-based routing is not applicable for immersive audio.

Even with maximum effort, WDM still provides only 8 channels, which limits the system to something like 5.1.2, while my setup requires at least a 7.1.4 layout.

Using the NuPrime hardware allows me to obtain a full 7.1.4 output over HDMI (rendered from MAT 2.0), which is the key to the entire solution.

The device simply delivers the digital audio over AoIP/AES3; nothing else special is involved.

In my configuration, running standalone EQ-APO is not feasible.

To achieve a 180-degree sound-field rotation, I must feed the original 7.1.4 bed from the H16 directly into the Atmos Renderer and re-render it there.

Only at the end of the Atmos bus chain can a VST calibration module be inserted (EQ-APO’s companion plugin, Dirac, SoundID, or anything else).

Finally, the processed audio is routed through a matrix to DVS, re-enters the Dante network, and is then delivered to the Shure audio interface.

Hi

This look quite an interesting setup. So with the calibration, are you gonna run as :
You take measurement in sonarwork(same as REW)
-then pas the output to Equalizer APO
-that will then provide the 7output channel
And then assign those channel somehow to the hdmi port. Then over the 'aoip controller' assign to those Shure decoder unit.

Or is this Nuprime give you 7input and you will pass over a matrix, maybe VB banana, then out of Equalizer Apo to all the in of the nuprime..? if that make sense..

Also, while it look interesting for a basic setup, just having any AVR receiver , they do in majority 9.1 and over. Do you feel limited with that unit, as you just mention you feel missing channel ?
They show 16out port on the connector, but not provide 16out, maybe the version nuprime h17 or so will do full 16ch...

Thanks for details

 

[BestGreed]

My Dirac license happened to expire right on Black Friday. To be honest, the prospect of manually aligning the phase and timing for every single speaker and the subwoofer was giving me a massive headache.

Fortunately, while researching EQ APO, I stumbled upon CamillaDSP, another excellent convolution engine. Crucially, it features ASIO integration, meaning I can chain it directly after the DAW.

I also discovered an interesting behavior in QuickEQ: for systems exceeding 8 channels, it employs a channel-multiplexing strategy. For example, in a 7.1.4 layout, after measuring the base 7.1 layer, it reuses output channels 1-4 to play the test tones for the height channels.

Armed with this knowledge, I devised a 'free,' semi-automated calibration workaround. The trick is precise timing: the moment QuickEQ finishes the first 4 channels and starts sweeping channels 5-8, I immediately mute outputs 1-4 and re-route them to the height speakers. This way, by the time the software starts measuring the height layer (multiplexed onto outputs 1-4), the signal path is already correctly switched to the ceiling.

Thanks to the powerful routing capabilities of the VB Matrix, even though QuickEQ is limited to WDM, managing the signal flow feels effortless—like the audio is right at my fingertips.

However, I do have some concerns. With the current trend, the audio routing inside this x86 DSP host is starting to look like a tangled ball of yarn. Each audio stream passes through nearly 10 endpoints back and forth. I’m not sure if this will introduce negative side effects. Once I have everything dialed in, I plan to measure the full-system latency and the specific DSP processing latency to be sure.

Here is the predicted signal path for just the Center Channel in this 'A-chain':

Main PC (Source) -> AVP (HDMI via MAT 2.0) -> x86 DSP (Dante Input) -> VB Matrix (ASIO Routing) -> Studio One (DAW Input) -> Dolby Atmos Renderer (7.1.4 Bed Rendering) -> VB Matrix (Loopback) -> CamillaDSP (Convolution/EQ via ASIO) -> VB Matrix (Final Output Routing) -> x86 DSP(Dante Output via DVS) -> Audio Interface

 

[VoidX]

I also discovered an interesting behavior in QuickEQ: for systems exceeding 8 channels, it employs a channel-multiplexing strategy. For example, in a 7.1.4 layout, after measuring the base 7.1 layer, it reuses output channels 1-4 to play the test tones for the height channels.

You can export the measurement signal as a 12 channel WAV, and start playing it back after you start a measurement. That's the old version, the new can simply remeasure channels, so you have a lot of time swapping.