New Neumann MA 1 - Automatic Monitor Alignment

[Deleted member 12642]

I had a brain fart and realized the dsp is in the kh80 and the sub would be unaffected. Just used to doing it from the pc I didn't even think about it.

 

[DJBonoBobo]

You could still use the MA1 on the KH80 for the whole spectrum and integrate the sub afterwards. You would have to do the crossover and room eq for the sub by yourself, of course, but i guess it is much easier if everything above 80hz is calibrated already.

 

[BDE]

Review of the MA1/ Neumann KH 750DSP + KH310
Sound and Recording KH750 MA1

 

[DJBonoBobo]

Review of the MA1/ Neumann KH 750DSP + KH310
Sound and Recording KH750 MA1

Ah, there it is! But it is without the MA 1 actually. Just the KH750 + 310 + iPad-App. For linear phase FIR-filtering of the KH310 you don´t need the MA 1.
What i find interesting is the almost perfectly linear FR of the KH 750 solo up to 700Hz. I wonder for what cases this is useful.

 

[thewas]

Ah, there it is! But it is without the MA 1 actually. Just the KH750 + 310 + iPad-App. For linear phase FIR-filtering of the KH310 you don´t need the MA 1.
What i find interesting is the almost perfectly linear FR of the KH 750 solo up to 700Hz. I wonder for what cases this is useful.

Also I find the highpass filter response interesting

 

[DJBonoBobo]

Also I find the highpass filter response interesting

Why? Is it different than typical highpass filters?

 

[thewas]

Why? Is it different than typical highpass filters?

That its not continuously dropping like the corresponding symmetric lowpassfilter (red curve).

 

[DJBonoBobo]

Ok. Maybe it is not really a highpass but a bell filter? I know nothing about filter design, though.

 

[thewas]

Ok. Maybe it is not really a highpass but a bell filter? I know nothing about filter design, though.

What interests me more is why they chose/designed it this way.

 

[Deleted member 12642]

Got my ma-1 mic in, or is it the p48? it says both so who knows. The cost of this thing is insane for what you get. It comes in a thin cardboard box in some foam with no case, I would've expected some way of safely storing the mic for $250. The mic itself is very small, half the size of the emm6. It's build doesn't instill confidence but it's probably going to be a use it and put it away mic so whatever.

Running test was easy, same as anything else out there. It had a couple confusing parts though.

1. Setting output level is weird, it mutes the speakers and plays what I assume is an internal loopback measurement to determine output gain? I'm assuming it wants the monitor set to the lowest input sens. Mine were on stock settings so 100db, at the level the software picked this would likely have damaged the speakers, at this point I'm very aware of the relationship between my interfaces volume knob and how it relates to the speakers output so I didn't bother running it anywhere the softwares desired output. It's weird because later on it has you set the mic gain and speaker output together in real time and I find that's far more useful.

2. When identifying monitors I have them selected properly but in the actual measurement it shows the opposite speaker being measured? Left makes the left monitor logo blink, right makes the right one blink, not sure what's up here.

3. A little confused by the info given with the results. I expected a before and simulated after, and a target. What it gives you is simulated after and target, and you can adjust the magnitude of each corrected band in terms of deviation from the target? Am I getting that right? I don't feel super confident adjusting these settings without seeing the before correction measurements.

As far pro's this is probably the most "colorless" correction I've heard if that makes any sense. I've tried most of the software suites out there like sonarworks, IK arc, both of those never sounded quite right to me, something off about the mid and top end. I had dirac back in version 3 when there was a desktop app and I feel like that was the best that I had heard before this dsp. I've only run it once and I was happy with the results, sounds great. It was nice to be able to add a shelf in the top end. Tossed on father john misty pure comedy and it was just magic.

 

[DJBonoBobo]

Got my ma-1 mic in, or is it the p48? it says both so who knows. The cost of this thing is insane for what you get. It comes in a thin cardboard box in some foam with no case, I would've expected some way of safely storing the mic for $250. The mic itself is very small, half the size of the emm6. It's build doesn't instill confidence but it's probably going to be a use it and put it away mic so whatever.

p48 means it has to be connected toi 48V phantom power. "Insane" cost... well, i paid 350 EUR for a Dirac stereo license only + 120 for a UMIK. Here you are getting a mic plus software for 250 EUR. I consider this "sane".

1. Setting output level is weird, it mutes the speakers and plays what I assume is an internal loopback measurement to determine output gain? I'm assuming it wants the monitor set to the lowest input sens. Mine were on stock settings so 100db, at the level the software picked this would likely have damaged the speakers, at this point I'm very aware of the relationship between my interfaces volume knob and how it relates to the speakers output so I didn't bother running it anywhere the softwares desired output. It's weird because later on it has you set the mic gain and speaker output together in real time and I find that's far more useful.

I don´t know what you were doing, but i think it´s supposed to work like this: You set input (mic) and output gain to some value (i use a middle position), then the software does a test to make sure measurement volume is always the same regardless what your gain settings are. Then you leave those settings alone. Next step is a short burst measurement around 85db that checks noise level. If that´s ok you don´t have to change anything else.

2. When identifying monitors I have them selected properly but in the actual measurement it shows the opposite speaker being measured? Left makes the left monitor logo blink, right makes the right one blink, not sure what's up here.

Red means "muted".

3. A little confused by the info given with the results. I expected a before and simulated after, and a target. What it gives you is simulated after and target, and you can adjust the magnitude of each corrected band in terms of deviation from the target? Am I getting that right? I don't feel super confident adjusting these settings without seeing the before correction measurements.

You can see the "before" result if of you uncheck "use automatic calibration". It takes a few seconds. Not very intuitive, i admit.

As far pro's this is probably the most "colorless" correction I've heard if that makes any sense. I've tried most of the software suites out there like sonarworks, IK arc, both of those never sounded quite right to me, something off about the mid and top end. I had dirac back in version 3 when there was a desktop app and I feel like that was the best that I had heard before this dsp. I've only run it once and I was happy with the results, sounds great. It was nice to be able to add a shelf in the top end. Tossed on father john misty pure comedy and it was just magic.

Good to that you had good results after all.

 

[Deleted member 12642]

p48 means it has to be connected toi 48V phantom power. "Insane" cost... well, i paid 350 EUR for a Dirac stereo license only + 120 for a UMIK. Here you are getting a mic plus software for 250 EUR. I consider this "sane".

Yeah but with software suites you can use the corrections you generate on any setup, this one is mostly locked to the neumann ecosystem. If you take into account the fact that you've already paid for the actual dsp portion of things when you buy the speakers and the software is a free download, it doesn't seem like the best value. They could've at least given you a case, what am I supposed to do just throw it in a drawer? Dayton can include a case with the emm6.

I don´t know what you were doing, but i think it´s supposed to work like this: You set input (mic) and output gain to some value (i use a middle position), then the software does a test to make sure measurement volume is always the same regardless what your gain settings are. Then you leave those settings alone. Next step is a short burst measurement around 85db that checks noise level. If that´s ok you don´t have to change anything else.

Just following the steps, it runs the test then tells me how much to turn the gain up on my interface. If I left it there it would've destroyed the poor speakers.

Red means "muted".

Lol what, man they need to maybe redo this part and make it so the speaker that is changing it's visual state is the identified speaker. Why would anyone assume the speaker that isn't flashing is the selected one. Some googling shows I'm not the only who got thrown off by that. I don't know if it affected measurements at all but I'm moving things into another room and have to redo them anyway.

You can see the "before" result if of you uncheck "use automatic calibration". It takes a few seconds. Not very intuitive, i admit.

I'll check that out

Good to that you had good results after all.

Only thing I didn't like is that it gave me a big boost at like 60hz that didn't need to be there.

 

[Deleted member 12642]

Move stuff into my treated studio. They're on a desk now which kinda sucks but they're k&m 26772 stands, nice stands I have to say.

Anyways went ahead and redid measurements. Was much easier this time around as I'm familiar with it's process and quirks. First set I followed directly the neumman instructions in terms of distances with the mic. I found that the results were really bright, way too bright for me. Tried applying some filter adjustments and nodded my head "that's better".

Welp turns out that the filters don't apply unless you disable then re-enable correction, so I was a victim of placebo in that moment

Here is the before. Pretty tubby sound and stuff in the mid bass just kind of bounces around, hard to just stereo placement.

Here's after. Way too bright for me. Feels like way too much energy from the tweeter and the woofer isn't scaling with it with volume.

And here's he high cuts I messed with to show the dsp in action a bit more. Thanks to motu's excellent driver my m4 supports multi asio applications at the same time so I can measure with REW while playing with MA-1. I just used the fixed points and will probably go back and adjust them but I have -2db at 1k and 3k, and a -5db slope for the top shelf filter. I may adjust it a tad more with more targeted frequency points. They need to add the ability to adjust the currently applied correction without having to remeasure.

Subjective junk - They sound so good now, detail is unmatched with anything I've ever heard. The stereo image is much more stable now with the mid bass cleaned up. Area of most improvement for me is how audible distortion is in music now, both creative intentional uses of it for say vocals and unintentional mess ups. They sound a lot less dry now too. Dayton Sub1000 is handling what these can't, just dialed and placed it by ear.

 

[Deleted member 12642]

Anyone have an idea what that ripple in the mid range is from? desk? ceiling?

 

[hyperplanar]

Anyone have an idea what that ripple in the mid range is from? desk? ceiling?

Desk reflection?

 

[hyperplanar]

I wonder if Neumann will ever release a free-to-use version of Neumann Control for Mac/Windows. It’s infuriating that I have to borrow an iPad in order to tweak my EQ.

If they’re not planning on doing so, I wonder how much work it’d take to reverse engineer the communication protocol from the software to the KH750 and reimplement it in order to be able to tweak the EQ from a computer for free. The MA1 app appears to be written in C# and ILSpy can cleanly decompile it... hmm... I’ll start taking a look this week if I have time. From a first glance looks like mDNS is involved for discovery and communications are AES encrypted but I found the key already.

EDIT: did some more digging - haven't tested yet but looks like the mic serial validation routine simply checks to see if the serial starts with a 6 and is ten digits long. The requested "microphone code" actually encodes the calibration; it contains the gain values for a series of preset frequency/Q bell filters. In other words, an individual mic's calibration is not downloaded off a server somewhere. This seems to be overlaid on top of a reference "golden mic" calibration:

More details on the microphone code: the first five characters are the encoded gain values for one of the filters each, while the sixth character is a checksum. If you want to try this software out (with the "golden mic" reference curve and no additional correction) you can use any 10-digit serial starting with 6 and microphone code 777773.

The five filters the gain values apply to, in order, are these:

The formula to derive the gain values:
Convert the first to fifth characters individually from hexadecimal to decimal (0-9 stay the same, A=10, B=11, C=12, D=13, E=14, F=15).
Subtract 7 from each result.
For characters one and two: multiply by 0.125.
For character three: multiply by 0.18.
For characters four and five: Multiply by 0.5.
This gives you the gain value for the corresponding filter in the list above.

This info should help anyone that wants to calculate their individual mic's calibration curve - take the golden mic curve and add the five filters encoded in the microphone code to it.


So, curiosity got the better of me and I decided to give the software a shot, but man, what a mistake... sounds much worse than my manual calibration. Basically gave me a much bumpier version of my own calibration. I know that the results with the actual MA1 mic could be different, but my measurement mic is pretty flat under 5 kHz which is where all the corrections are, so I wouldn't expect much of a difference other than a shelf of a dB or two under 200 Hz due to the preset "golden mic" calibration curve. At least it's good to know that my target response is pretty similar, but now I'm going to need to either borrow an iPad or reverse engineer the communication protocol to redo my manual calibration.

Ignore comb filtering artifact at 10kHz, this was because both speakers were measured simultaneously and the mic wasn't perfectly centered. I accepted the target curve the software suggested, other than adjusting it >1 kHz so that it would leave the treble untouched. Huge peak at 41 Hz was not addressed, and the crossover/delay settings I chose in my manual calibration filled in the 100-200 Hz dips, while the MA1's did not. On the plus side, knocking off the peak at 1 kHz improved midrange naturalness and clarity.

Interestingly, seems like the MA1 software decided to invert the polarity of the subwoofer, since it stays at about 180 degrees phase. This does net less phase rotation, but...

The group delay is not any better whatsoever. As can be concluded from this graph and the FR, the bass sounds much sloppier now.

Overall, for someone that doesn't have a good understanding of how to set the delay and generate their own EQ, the software can get you about halfway there in much less time, so I see the appeal. I don't know why, but I was expecting a result better than my manual calibration, which didn't turn out to be the case. It's still much better than how my setup sounds uncorrected though.

 

[DJBonoBobo]

This info should help anyone that wants to calculate their individual mic's calibration curve - take the golden mic curve and add the five filters encoded in the microphone code to it.

Thank you, very interesting. I imported the curve in REW and set the EQ how you suggested based on the code on my mic. It looks plausible. But how do i create a cal-file from the "Predicted"-curve in REW?

Overall, for someone that doesn't have a good understanding of how to set the delay and generate their own EQ, the software can get you about halfway there in much less time, so I see the appeal. I don't know why, but I was expecting a result better than my manual calibration, which didn't turn out to be the case. It's still much better than how my setup sounds uncorrected though.

FWIW in my case the automatic correction was way better than my manual attempts. But i surely don´t have a good enough understanding of what i tried before. I consider myself kind of an "enthusiast hobbyist" without special technical knowledge, but some trial and error experience i collected over a few years. The software was a bit better than me in integrating the sub and a lot better in smoothing the area above 200 Hz. But it is good to read another perspective from someone with obviously more technical knowledge.

BTW I observed quite significantly different results depending on how the speakers and subwoofer were positioned. Perhaps the software manages differently depending on the situation.

 

[DJBonoBobo]

Thank you, very interesting. I imported the curve in REW and set the EQ how you suggested based on the code on my mic. It looks plausible. But how do i create a cal-file from the "Predicted"-curve in REW?

I think i figured it out (export filters impulse response as wav, import impulse response, all spl -> trace arithmetics -> A*B, export measurement as txt). I´ll try to use it later.

 

[hyperplanar]

FWIW in my case the automatic correction was way better than my manual attempts. But i surely don´t have a good enough understanding of what i tried before. I consider myself kind of an "enthusiast hobbyist" without special technical knowledge, but some trial and error experience i collected over a few years. The software was a bit better than me in integrating the sub and a lot better in smoothing the area above 200 Hz. But it is good to read another perspective from someone with obviously more technical knowledge.

Yeah, it did a better job in the 200-1000 Hz range too, but the bass response... :/
I suppose the software struggled with the 41 Hz peak in my room because it's extremely sharp. Maybe the algorithm doesn't consider it to be very audible of an issue - in terms of frequency response, perhaps not, because only a very narrow pitch is excited. But such a high Q peak makes the temporal performance terrible around that range, causing the dreaded booming one note bass sound quality. Also many electronic kick drums fundamentally consist of a pitch sweeping down, which can trigger the FR peak.
I'm nothing more than an enthusiast hobbyist as well I got into being an audiophile ten years ago when I was 16 looking for studio monitors to make music and being mesmerized by how good they sounded compared to my hand-me-down Logitech computer speakers and my Honda Civic's speakers. Basically learned everything from the valuable resources on this forum.

I think i figured it out (export filters impulse response as wav, import impulse response, all spl -> trace arithmetics -> A*B, export measurement as txt). I´ll try to use it later.

That's basically the same way I would do it, yes. However I'm not sure if mic calibration curves are presented in this format, or inverted/upside-down. Basically the final resulting curve from A*Bing the golden mic curve and the hex mic curve is what needs to A*Bed with any measurement in order to correct it - it is the inverse of the mic's frequency response. If a mic calibration file represents the frequency response of the mic, and not the inverse, then you'll need to invert it as well.

 

[DJBonoBobo]

MMM-measurements (average of (R+sub) + (L+sub)) with a individually calibrated UMIK-1 (calibrated by hifi-selbstbau.de) vs. MA 1 uncalibrated vs. MA 1 calibrated with the method described by @hyperplanar (pls do not overthink small differences between UMIK 1 and MA 1 because of the nature of MMM measurements). I don´t know which one is "correct", just wanted to show the difference the calibration makes:

Same, but zoomed in (20 dB scale), no smoothing, thinner traces:

PS: I aligned the traces manually for best looking fit.