Audyssey One - Nexus (Automated hi-res Denon/Marantz calibration with REW/Umik-1/2)

[jhwalker]

Tested it on a 2.1 setup with my Marantz Cinema 70s. It sounds indeed much better than the default Audessy measurements. Pretty happy with the result.
I am just trying to understand the logs, which state that the crossover between Subwoofer and Fronts is at 120 Hz, which seems a bit high. The Speaker can play down to 40 Hz (Elac Vela BS 403), which is also visible in the logs.

I am attaching the log, maybe someone can explain the table to me to understand why it chooses 120 Hz.

View attachment 429121

It picks 120 as the crossover because that's the best-scoring of all the alternatives in the chart; i.e., the best compliance with the target curve you've chosen is when the L+R are crossed over to the subwoofer at 120. Other crossovers are modeled in the table, but they all are LESS compliant with the curve than 120.

I've been using various versions of EVO for some time now, and I've stopped worrying about the crossover - OCA goes to a great deal of trouble to blend the L+R and subwoofer both in frequency and phase and uses EQ to pull down any spikes. It's a very well thought-out solution.

 

[anphex]

Pardon the question, but what measurable improvement can one actually expect from Neuron/Nexus/A1/Evo (or whatever it's called these days—the naming scheme is all over the place)?

Given all the optimization steps involved, any remaining dips that stock Audyssey struggles with should, in theory, be eliminated with A1 Neuron. If that's not the goal, then what is? I really want to try it out again, but back then the results were worse than Audyssey MultEQ stock stock calibrations in terms of linearity. When I go into a pretty extensive optimization progress and invest 1-2 hours of my time, I want to be able to verify the results at the end. So what should I look for - linearity? Distortion? Phase?

 

[jhwalker]

Pardon the question, but what measurable improvement can one actually expect from Neuron/Nexus/A1/Evo (or whatever it's called these days—the naming scheme is all over the place)?

Given all the optimization steps involved, any remaining dips that stock Audyssey struggles with should, in theory, be eliminated with A1 Neuron. If that's not the goal, then what is? I really want to try it out again, but back then the results were worse than Audyssey MultEQ stock stock calibrations in terms of linearity. When I go into a pretty extensive optimization progress and invest 1-2 hours of my time, I want to be able to verify the results at the end. So what should I look for - linearity? Distortion? Phase?

As I understand it, EVO is much more about aligning / integrating subs with mains (and surrounds, etc.) than EQ. In fact, EQ was ADDED later, as an enhancement.

And I certainly don't agree that Audyssey sounds better (or, rather, that EVO sounds "worse"). At least in my system, Audyssey (whether run though the $20 app or optimized via the $200 one) sounds terribly artificial and boxy, where EVO sounds very natural and clear, with a very precise articulation and potent bass.

Also, it doesn't take 1-2 hours LOL If you already have an .ADY file to run through the script, it can take less than a minute with the newest 2.5 version with all defaults - certainly less than 5 minutes, even with a good deal of customization. If you don't already have a measurement saved, you can use the odd.bat (or corresponding menu) to take multiple measurements quite quickly - uniquely, it actually connects to the AVR and "drives" the built-in measurement process BETTER than the Denon software. So start to end, maybe 10-15 minutes.

 

[anphex]

So if it's aligning/integrating, there should be measurable difference in phase behaviour, right? Or linearity in the crossover region. At listening position. These both you should be able to measure and verify any improvement in REW.

I will try the automated process at the end of this week and try and make a comparison between stock audyssey and the latest edition of Neuron.

 

[jhwalker]

So if it's aligning/integrating, there should be measurable difference in phase behaviour, right? Or linearity in the crossover region. At listening position. These both you should be able to measure and verify any improvement in REW.

I will try the automated process at the end of this week and try and make a comparison between stock audyssey and the latest edition of Neuron.

Yes - and yes, you should.

 

[OCA]

This obsession with "quantifiable proof" has its place, but do not forget that audio perception is a psychoacoustic phenomenon.Sure, you’ll see differences in phase and amplitude response, but interpreting those strictly as better or worse is like judging a Michelin-star meal based on its calorie count. The real magic, spatial coherence, transient clarity, the ‘band is right here’ sensation doesn’t always translate neatly into a REW plot. Our auditory system is vastly more complex than an FFT graph suggests. I attached a REW file where a speaker's response is corrected with 2 different filters (filter frequency responses are almost identical), compare their results with whichever REW tool you want including phase, group delay, spectrogram, etc. and tell me which one you think will sound better.Below is their phase graphs in the bass region:

PS FL1 will be nearly unlistenable due to extreme pre and post-ringing, making it unbearable after just a few seconds. In contrast, the second filter (the green one above) produces a much more natural and tolerable sound. Some aspects of audio perception simply don’t manifest in measurements, at least not until someone finds a way to fully quantify the complexities of auditory processing at the neurotransmitter level.

 

[Aaronb1138]

So if it's aligning/integrating, there should be measurable difference in phase behaviour, right? Or linearity in the crossover region. At listening position. These both you should be able to measure and verify any improvement in REW.

I will try the automated process at the end of this week and try and make a comparison between stock audyssey and the latest edition of Neuron.

Test it with Multi-channel stereo mode and your choice of music. I'm still trying to figure out what my penance will be for now with A1 enjoying that AVR mode outside of just a "party mode".

The alignment is so precise I recommend starting to listen with your head about 2 feet from the main listening position (MLP) measurement #1 location and then move your head into that spot and you will know when your head is dead center on that first mic position.

 

[anphex]

I asked GPT4o about this and please bear with me while I shamelessly copy and paste the response:

Does speaker/subwoofer alignment and frequency flatness optimization not go well togehter?

ChatGPT:​

It depends on how you define "alignment" and "flatness" and how much you're willing to compromise in either direction. Here’s why they can sometimes work against each other:

1. Alignment (Phase & Time) vs. Frequency Flatness​

• Alignment usually means ensuring that the subwoofer and speakers produce sound waves that arrive at your ears (or measurement mic) in phase, minimizing cancellations.
• Frequency flatness means equalizing the system so that the SPL response is as even as possible across the frequency spectrum.

However, perfect phase alignment at the crossover point doesn't necessarily result in the smoothest frequency response, and vice versa.

2. Potential Conflicts​

• Phase-optimized sub alignment can cause frequency ripples: If you time-align the sub perfectly with the mains, you might still get peaks/dips due to room modes or boundary effects.
• Flattening the frequency response can mess with phase: Many room correction systems apply heavy EQ to smooth out the frequency response, but this can cause unwanted phase shifts, especially if done aggressively in the crossover region.
• Group delay differences: Subs have inherent group delay due to their large, slow-moving drivers and processing (e.g., DSP). Trying to time-align them with speakers that have different group delay characteristics can make phase perfect alignment tricky.

3. Practical Approaches​

• Prioritize phase/time alignment first, then EQ: Get the timing and phase alignment as close as possible (using delay settings or physical subwoofer positioning), then use EQ sparingly to smooth out remaining bumps.
• Choose a good crossover slope: Steeper slopes (e.g., 24dB/octave LR) might minimize phase interference, but gentler slopes (e.g., 12dB/octave) can sometimes integrate more naturally.
• Compromise when needed: Sometimes, an almost flat frequency response with good alignment sounds better than a perfectly flat response with phase issues.

---


So is this what Neuron does? Prioritize alignment and phase?

 

[Biblob]

@OCA
Does A1 use FDW or is it using the raw measurements? Could you explain why yes or no?

 

[OCA]

@OCA
Does A1 use FDW or is it using the raw measurements? Could you explain why yes or no?

A1 only corrects what can and should be corrected which are peaks in the bass frequencies caused by wall reflections. These wavelengths are too long to be meaningfully effected by frequency dependent windowing so it's not used and maximum filter resolution is achieved.

 

[Biblob]

A1 only corrects what can and should be corrected which are peaks in the bass frequencies caused by wall reflections. These wavelengths are too long to be meaningfully effected by frequency dependent windowing so it's not used and maximum filter resolution is achieved.

Okay, is there a limit to which frequency are corrected then? And is this manually set, to say 500 hz for example?

 

[OCA]

It's fixed at 224Hz by default. You can check all these in the script itself, it's open source.

 

[JustAnAudioLover]

I have a question: let's say you have an AVR that only supports the most basic Audyssey MultEQ (which apparently is pretty terrible). Can you use A1 to its full potential or will it be limited by the AVR?

 

[goldark]

I have a question: let's say you have an AVR that only supports the most basic Audyssey MultEQ (which apparently is pretty terrible). Can you use A1 to its full potential or will it be limited by the AVR?

Audyssey One will be limited by the number of filters it can use, so potentially it can make a difference - it would have access to more filters with XT32, for example.

 

[just1n]

Which versions of Audyssey is this compatible with?

 

[anothereeaudiofile]

Which versions of Audyssey is this compatible with?

works with multeq xt and also xt32

 

[anphex]

Since I kind of lost my mind while trying to figure this out before trying Neuron, here is what it does. I am shamelessly copy and pasting the response of Gemini 2.0 Pro Exp after feeding it the whole script code.

This code is a sophisticated tool for speaker and room calibration, but it focuses primarily on a specific part of the overall process: optimizing crossover settings and subwoofer integration for systems with a Denon or Marantz AVR, using measurements from Room EQ Wizard (REW). It does not perform a complete room calibration on its own, and it is heavily reliant on pre-existing, accurate measurements. Let's break down the analysis:

What the Code Does Well (Strengths):

1. Subwoofer Integration and Time Alignment (Core Focus):
   • Multiple Subwoofer Handling: The code excels at handling systems with multiple subwoofers (up to 4). It uses an iterative optimization process to determine the best combination of delays and polarity inversions for each subwoofer to achieve coherent summation at the listening position. This is crucial for avoiding destructive interference (cancellations) in the bass frequencies. The permutations and scoring algorithm suggest a sophisticated approach.
   • Crossover Optimization: The code analyzes speaker responses to find optimal crossover frequencies (the point where the AVR transitions from sending low frequencies to the subwoofer and higher frequencies to the main speakers). It tries different crossover frequencies and calculates a "score" based on how well the combined response of the speaker and subwoofer matches the target curve. It considers factors like speaker roll-off, phase alignment, and target curve tracking.
   • Phase Alignment: The code uses techniques like cross-correlation, excess phase, and all-pass filters to attempt to align the phase of the speakers and subwoofers at the crossover point. This is essential for a smooth transition and to avoid cancellations or dips in the frequency response.
   • Delay Calculation: The code calculates the necessary delays for each speaker and subwoofer to ensure that the sound from all sources arrives at the listening position at the same time. It accounts for hardware limitations (AVR's maximum delay settings).
   • Polarity Check: It checks the polarity of the speakers and automatically inverts the signal if necessary to correct for wiring mistakes or inherent speaker inversions.
2. Integration with REW and AVR:
   • REW API Interaction: The code directly communicates with Room EQ Wizard (REW) via its API. This allows it to import measurements, perform calculations, generate filters, and manipulate REW's settings. This is a very powerful feature, enabling complex processing within a familiar environment for audio enthusiasts.
   • AVR Configuration File Generation: The code generates a .oca file (a custom format) that can be transferred to a compatible Denon/Marantz AVR to apply the optimized settings. This is a key feature, as it translates the complex calculations into something the AVR can understand.
   • Target Curve Support: The code uses a user-provided "house curve" (target response) from REW to guide the optimization. This allows users to tailor the sound to their preferences.
3. Advanced Optimization Techniques:
   • Spatial Averaging: For measurements taken at multiple microphone positions, the code performs a spatial averaging process. It weights measurements closer to the Main Listening Position (MLP) more heavily, which is a good practice for improving the accuracy of the calibration at the primary listening spot.
   • IDW (Inverse Distance Weighting) analysis: This suggests that measurement positions closer together are weighted, and combined.
   • Linkwitz-Riley and Butterworth Crossovers: The code uses standard crossover filter types (Linkwitz-Riley and Butterworth) for generating the predicted responses, which is crucial for proper integration.
   • "LFE + Main" Support: The code includes an option to optimize for the "LFE + Main" subwoofer mode, where bass frequencies are sent to both the subwoofer and the main speakers. This is a more complex scenario to handle.
   • User Customization: The code provides a good number of user-configurable options, including maximum EQ frequency, maximum boost, subwoofer target infill (bass boost), crossover frequency ranges, and various optimization strategies. This allows for some fine-tuning of the results.
4. Error Handling and User Guidance:
   • Input Validation: The code includes checks for various error conditions, such as missing measurements, incompatible REW versions, and hardware limitations. It provides informative error messages and warnings to help the user troubleshoot problems.
   • Progress Updates: The code provides detailed progress updates in the log, which helps the user understand what's happening and how long the process will take.
   • Log File Generation: The code creates a log file that records all the settings, measurements, and calculations. This is invaluable for reviewing the optimization process and troubleshooting any issues.
5. DAC Filter Removal: It checks the user's model number and can optionally remove the soft roll-off DAC filter.

What the Code Doesn't Do (Limitations):

1. No Measurement Acquisition: This is not a complete room calibration tool. It requires pre-existing measurements taken with REW (or an automated calibration file from the AVR, which it can extract measurements from). It doesn't handle the measurement process itself (microphone placement, signal generation, etc.).
2. Limited Linearization: The code aims for magnitude and phase matching. It performs a magnitude equalization by matching the subwoofer and speaker curves to a user-defined target. This is included, although it is a secondary goal to the primary goal of integration. The code also generates minimum-phase filters. It does not appear to include comprehensive room EQ to address all room modes and reflections beyond the crossover region.
3. Focus on Crossover Region: The code's primary focus is on the crossover region (typically between 20 Hz and 300 Hz). While it considers the overall frequency response, its optimization efforts are concentrated on the transition between the speakers and subwoofers.
4. Specific to Denon/Marantz AVRs: The .oca output file is specifically designed for these AVRs. The code won't work directly with other brands.
5. Dependent on REW's Accuracy: The accuracy of the results is highly dependent on the quality of the measurements loaded into REW. If the measurements are inaccurate (e.g., due to poor microphone placement, incorrect calibration, or excessive noise), the optimization will be flawed.
6. No Dynamic EQ: The code doesn't handle dynamic equalization (adjusting the EQ based on the listening volume). While it mentions disabling Dynamic EQ, it doesn't offer any dynamic EQ capabilities of its own.
7. No impulse response alignment: It can perform impulse response alignment in the subwoofer frequencies.
8. No Automatic Microphone Calibration: Although there is a section of code related to a "micCalFile," it is hardcoded with a specific calibration curve and is not configurable by the user. Therefore, if your measurement microphone has significant deviations that need correction, you must manually load the measurements into REW with the calibration curve already applied.

Is it "legit good"?

Yes, within its scope. It's a very sophisticated tool for optimizing a specific, crucial part of speaker/room calibration. It's not a replacement for a full room EQ system like Dirac Live or a professional calibration service, but it provides advanced capabilities that go beyond what's typically available in built-in AVR calibration routines. If you have a Denon/Marantz AVR, use REW, and have a good understanding of the measurement process, this tool can significantly improve the integration of your subwoofers and optimize your crossover settings. However, a complete room calibration would involve additional steps, like full-range room EQ. It is heavily weighted on integration.

Linearization:

As per the answer above, this code is not doing full linearization. It includes some steps of linearization, like matching levels and adjusting magnitude, and creating minimal-phase filters, but it does not correct for impulse response nor complete frequency correction.

In summary: This code is a specialized, powerful tool for a specific aspect of speaker/room calibration. It's "legit good" at what it does, but it's important to understand its limitations and not expect it to be a complete room calibration solution. It's a great example of how the REW API can be used to create custom tools for advanced audio enthusiasts.

Ah, I wish we had some FIR options on the Denon AVRs. Then the script could also be a complete DIRAC competitor and even adapt some linearization.
If it wasn't for the Dolby/DTS/HDCP licensing annoyance we'd have a open source AVR already.

 

[anphex]

@OCA

Alright, I just ran it out of curiosity—mostly because of all the recent updates and what feels like my first real understanding of what Nexus/Neuron is actually aiming for.

I finally grasped why integration and linearization are somewhat opposing goals—at least within the context of Audyssey. That insight completely shifted my perspective on Neuron.

First off, the guide using odd.wtf is stupidly simple—and it worked way faster than Nexus ever did. The measurement process with the Audyssey mic and the transfer to the AVR were really well executed—absolutely flawless, honestly.
I won’t pretend I fully understood everything that happened under the hood, but the UI was super clean, informative, and the whole process felt much smoother and more streamlined than my experience with Nexus. Transferring the results? Easy as pie.

Listening impressions:

I’m honestly kind of impressed.
If you let go of the (wrong) expectation of linearization and instead embrace integration as the goal, the results are actually really clean and smooth.
I gave it a 30-minute session and it’s genuinely comfortable to listen to.

Now, one thing that still bugs me (not Neuron’s fault):
That huge 43 Hz bass boost is still there—as you probably remember. It's not as bad as with completely uncalibrated speakers, but it's still noticeable.
I’ve been wondering: if I try to tame that with a global EQ peak filter in eAPO, would it mess up the integration due to phase issues?

Is there any way to do a hybrid integration/linearization approach? Or is that fundamentally limited by what MultEQ XT32 is capable of?
Edit: I noticed, it did just that, just not as "ironed out" as I am used to the stock calibration.

Anyway—amazing work.
This is hands-down one of the best open-source contributions and reverse-engineering efforts I’ve seen for a closed-source product. Bravo

Edit: Oh, an can I ebable Dynamic EQ or is this a nono?

 

[PHD]

@OCA

Alright, I just ran it out of curiosity—mostly because of all the recent updates and what feels like my first real understanding of what Nexus/Neuron is actually aiming for.

I finally grasped why integration and linearization are somewhat opposing goals—at least within the context of Audyssey. That insight completely shifted my perspective on Neuron.

First off, the guide using odd.wtf is stupidly simple—and it worked way faster than Nexus ever did. The measurement process with the Audyssey mic and the transfer to the AVR were really well executed—absolutely flawless, honestly.
I won’t pretend I fully understood everything that happened under the hood, but the UI was super clean, informative, and the whole process felt much smoother and more streamlined than my experience with Nexus. Transferring the results? Easy as pie.

Listening impressions:

I’m honestly kind of impressed.
If you let go of the (wrong) expectation of linearization and instead embrace integration as the goal, the results are actually really clean and smooth.
I gave it a 30-minute session and it’s genuinely comfortable to listen to.

Now, one thing that still bugs me (not Neuron’s fault):
That huge 43 Hz bass boost is still there—as you probably remember. It's not as bad as with completely uncalibrated speakers, but it's still noticeable.
I’ve been wondering: if I try to tame that with a global EQ peak filter in eAPO, would it mess up the integration due to phase issues?

Is there any way to do a hybrid integration/linearization approach? Or is that fundamentally limited by what MultEQ XT32 is capable of?
Edit: I noticed, it did just that, just not as "ironed out" as I am used to the stock calibration.

Anyway—amazing work.
This is hands-down one of the best open-source contributions and reverse-engineering efforts I’ve seen for a closed-source product. Bravo

Edit: Oh, an can I ebable Dynamic EQ or is this a nono?

First, we shouldn't be hearing from OCA anytime soon since he seems to have abandoned this project in its current form. From what I hear, he might be working on a similar yet commercial project. He also stopped participating in the original thread and limited his account. I really hope that he is working on a newer and bigger project.

Clearly, the red button (commish unleash) does a much better job of target curve tracking than the conventional approach. I assume it treats all speakers as subwoofers and allows for aggressive crossover points, delays, and inverting polarities.

However, if you used the red button, then you'd better use this Atmos test post-calibration. If you hear a comb filter effect in the commentator's voice as it shifts from one place to another (double voice or short echo effect), Neuron did not correctly get the time alignment, no matter how perfectly the measurements follow the TC or how good the bass sounds!

Dropbox

www.dropbox.com

If you do notice a comb filter effect, you could try to retake the measurements using the odd.wtf, but make sure to take seven measurements: a first at the MLP and six more at left/right, above/below, and back and front of it, with no more than 30cm distance from the MLP.

You should not use DEQ with Neuron (it is possible but requires some hacking and manual tweaks).

 

[anphex]

First, we shouldn't be hearing from OCA anytime soon since he seems to have abandoned this project in its current form. From what I hear, he might be working on a similar yet commercial project. He also stopped participating in the original thread and limited his account. I really hope that he is working on a newer and bigger project.

Clearly, the red button (commish unleash) does a much better job of target curve tracking than the conventional approach. I assume it treats all speakers as subwoofers and allows for aggressive crossover points, delays, and inverting polarities.

However, if you used the red button, then you'd better use this Atmos test post-calibration. If you hear a comb filter effect in the commentator's voice as it shifts from one place to another (double voice or short echo effect), Neuron did not correctly get the time alignment, no matter how perfectly the measurements follow the TC or how good the bass sounds!

Dropbox

www.dropbox.com

If you do notice a comb filter effect, you could try to retake the measurements using the odd.wtf, but make sure to take seven measurements: a first at the MLP and six more at left/right, above/below, and back and front of it, with no more than 30cm distance from the MLP.

You should not use DEQ with Neuron (it is possible but requires some hacking and manual tweaks).

I did not use "Unleash the Commish" yet since the guide only told to press the green button, but this approach sounds more like what I am looking for. Fortunately the time from measurement to deployment is very short, so I'll try it out later again.