Audiolense XO Stability Experience

[mccarty350]

I can't find a way to take a poll but given some of the other threads I'd like to gauge what the average stability experience is so that we have an over arching picture.

Here's my ask for those who participate (and if my questions are poor I'm game to revise them):

Question: On average how many issues to you experience with Audiolense XO. Issues being vaguely defined as over arching crashes, correction procedures that bomb out, measurements that crash or any other unanticipated result that falls under the genre of stability?
A. I never have stability issues.
B. On average 1 in 10 sessions I have with the software will result in an issue being experienced.
C. On average 1 in 5 sessions I have with the software will result in an issue being experienced.
D. On average 1 in 2 sessions I have with the software will result in an issue being experienced.
E. I have an issue every time I use the software.

My experience given my environment is D

My environment:
1. Windows 10 Pro
2. Early Generation i7 with 16 gigs of RAM
3. Asus Xonar U7 Mk II sound Card using Default Drivers
4. 8 channel convolution files being generated in almost every case.
5. ASIO4ALL being leveraged.
6. Minidsp UMIK for measurements.


Probably many other variables to my environment that I'm missing that are pertinent but perhaps we can find some commonalities on things that don't work well.

 

[Mitchco]

Just a quick comment. I have been using Audiolense for over 12 years. And over the past 3 years pretty much everyday. As part of my DSP calibration service, I get to see client measurements from a wide variety of systems, both consumer and professional from around the world. No-one has reported stability issues with Audiolense.

I personally have not run into any instability issues over the past 12 years. Sure there have been a few hiccups, but Bernt has been very responsive in fixing any issues that are repeatable.

The number 1 issue I find is getting high quality measurements, specifically related to timing. Timing issues meaning when a measurement session is repeated 3 times in a row that the delays recorded in the measurement window, delay column, are the same. It is all about consistency. This issue sometimes can be exacerbated when the number of channels is increased.

These timing issues can be traced to some DAC's and more specifically the audio drivers for those DAC's seem to be not implemented very well. And it can be random. For some folks the combination of the same DAC in their system works fine but in others not so much.

There are so many variables in "computer-based" audio it is hard to determine what the actual cause is. But if you find a solution, Bernt already has a thread about "Sound card and microphone fixes" that one can post a solution there: https://www.avnirvana.com/threads/sound-card-microphone-fixes.10099/

 

[sshd]

I have used Audiolense XO since 2010 and it has been some journey from B with AL4 to D with AL5-6.

My environment:
1. Windows 7
2. Intel Xeon with C226 chipset
3. Layla 3G, replaced with Motu 8A
4. Calibrated microphone bought from Bernt

You might need to tune your Windows, update/downgrade drivers, disable certain services and make sure you don't have any latency issues. This is relatively easy in Windows 7 with the dpclat tool, which doesn't work in W10+.

You also need to understand your audio interface and microphone amp. Certain noise issues on my Layla 3G were caused by 100% input level on the interface.

Audiolense 4.x never failed when measuring. It crashed from time to time when generating filters, but that was acceptable.

Audiolense 5.x (I tried several versions) had about 50% chance of completing the measurements and all of them were broken. According to Bernt, it was the more aggressive noise removal algorithm in AL5 than AL4 than broke them. There is a way to fix the measurement by looking at the graph, understanding what you see, clicking the right places, and undoing the noise removal. I don't like this. What else might be broken?

Audiolense 6 beta was unable to perform any measurements.

I am currently using AL6.7 filters with a measurement from AL4.12. The sound is wonderful and Bernt offered excellent advice and support. I am completely satisfied and have not bothered meassuring for three years.

One thing to note is that AL4.12 is a 32-bit application and meassures via 32-bit ASIO drivers.
Measurring with 64-bit AL will use 64-bit ASIO drivers. These drivers might not be as stable as the 32-bit.

 

[whoareyou]

I have had issues, but none of them showstoppers or what I'd call stability. And by that I mean app just shutting down.
Some, I believe are threading issues... i.e. Buttons that can be clicked when processing is in progress and these result in unhandled exceptions. Shouldn't happen, but nothing horrible happens.
I've had UI scaling issues in my environment(s). Other issues are with input fields. Sometimes I think I've entered a value but AL will "disagree" that the data was input.
Other hiccups are with target designer, particularly when I've tried multitargets. Sometimes data just disappears and necessitates it be reentered.

In all fairness, some of these I've reported and some I have not, but when reported they've been fixed / investigated if root cause can be determined/replicated, and Bernt has indicated he is working to resolve some of this stuff.
Overall, I've found most issues to be relatively minor, and I now probably work around them without realizing it, since I rarely encounter them.

Biggest difficulty was getting measurements to be reliable and/or just work. That took a while, but once I had that sorted out I seem to be mostly good to go whenever I've used AL.

 

[Omid]

I just want to start by saying that Audiolense is an awesome piece of software that achieves great results. There are some annoyances but with patience, one can get the desired results.

My issues are similar to @whoareyou, but the one that stands out for me is measurement sweeps. I frequently get knocking sounds midway through a sweep. I stop the sweep right away to prevent speaker damage, so I admit I haven't saved the measurement to look for clues. If I run a sweep again, the same happens, unless I play 1-2 seconds of an audio track in Roon (I figure it resets the buffer? I might be wrong). I can then repeat the measurement without changing anything, and get an excellent measurement.

I find that longer sweeps are more likely to cause knocking sounds. If the playback is too loud, causing clipping, I think it might cause knocking too.

This happens even on a simple 2.0 setup, both on my Nuc PC as well as my Dell laptop. I think I also tried through a MAC using the AVB Ethernet connection to the MOTU without success. For a Dac, I use a MOTU 8A (tried both USB 3 and 2 cables & ports), although a Xonar U7 USB DAC had the same issue. I 've tried a USB mic (umm6) and a powered mic (emm6), same issues. I've used the MOTU driver, WASAPI, Directsound, WinMME, as well as ASIO4All, not all worked in my setup, but those that did had the same problem. I've tried with and without drift correction, same deal. I don't recall experimenting with noise removal though.

That said, sticking to a 3-5 second sweep, I do manage to get successful measurements after a couple of trials.

Again I want to echo @whoareyou's comments, Bernt has always been very receptive to troubleshooting issues, continually improving the software and generally helping out. I would not want anyone who is thinking about trying Audiolense to be discouraged by this discussion. The results are really worth the effort.

 

[mccarty350]

I have used Audiolense XO since 2010 and it has been some journey from B with AL4 to D with AL5-6.

My environment:
1. Windows 7
2. Intel Xeon with C226 chipset
3. Layla 3G, replaced with Motu 8A
4. Calibrated microphone bought from Bernt

You might need to tune your Windows, update/downgrade drivers, disable certain services and make sure you don't have any latency issues. This is relatively easy in Windows 7 with the dpclat tool, which doesn't work in W10+.

You also need to understand your audio interface and microphone amp. Certain noise issues on my Layla 3G were caused by 100% input level on the interface.

Audiolense 4.x never failed when measuring. It crashed from time to time when generating filters, but that was acceptable.

Audiolense 5.x (I tried several versions) had about 50% chance of completing the measurements and all of them were broken. According to Bernt, it was the more aggressive noise removal algorithm in AL5 than AL4 than broke them. There is a way to fix the measurement by looking at the graph, understanding what you see, clicking the right places, and undoing the noise removal. I don't like this. What else might be broken?

Audiolense 6 beta was unable to perform any measurements.

I am currently using AL6.7 filters with a measurement from AL4.12. The sound is wonderful and Bernt offered excellent advice and support. I am completely satisfied and have not bothered meassuring for three years.

One thing to note is that AL4.12 is a 32-bit application and meassures via 32-bit ASIO drivers.
Measurring with 64-bit AL will use 64-bit ASIO drivers. These drivers might not be as stable as the 32-bit.

Good advice, perhaps I should drop down to 32 bit AL and might gain some stability?

 

[mccarty350]

I just want to start by saying that Audiolense is an awesome piece of software that achieves great results. There are some annoyances but with patience, one can get the desired results.

My issues are similar to @whoareyou, but the one that stands out for me is measurement sweeps. I frequently get knocking sounds midway through a sweep. I stop the sweep right away to prevent speaker damage, so I admit I haven't saved the measurement to look for clues. If I run a sweep again, the same happens, unless I play 1-2 seconds of an audio track in Roon (I figure it resets the buffer? I might be wrong). I can then repeat the measurement without changing anything, and get an excellent measurement.

I find that longer sweeps are more likely to cause knocking sounds. If the playback is too loud, causing clipping, I think it might cause knocking too.

This happens even on a simple 2.0 setup, both on my Nuc PC as well as my Dell laptop. I think I also tried through a MAC using the AVB Ethernet connection to the MOTU without success. For a Dac, I use a MOTU 8A (tried both USB 3 and 2 cables & ports), although a Xonar U7 USB DAC had the same issue. I 've tried a USB mic (umm6) and a powered mic (emm6), same issues. I've used the MOTU driver, WASAPI, Directsound, WinMME, as well as ASIO4All, not all worked in my setup, but those that did had the same problem. I've tried with and without drift correction, same deal. I don't recall experimenting with noise removal though.

That said, sticking to a 3-5 second sweep, I do manage to get successful measurements after a couple of trials.

Again I want to echo @whoareyou's comments, Bernt has always been very receptive to troubleshooting issues, continually improving the software and generally helping out. I would not want anyone who is thinking about trying Audiolense to be discouraged by this discussion. The results are really worth the effort.

Ditto for the results being worth the effort. I haven't found anything this good from a results perspective out there. Period.

Thank you for the advice. So part one is I'll try switching to 32 bit, part two I'll try sticking to relatively low sweeps. I might even be receptive to buying a new microphone/upgrading mine since apparently the UMIK's aren't great and the heart of audiolense operation relies on good measurements. I also have had the situation regarding the knocking occur before and indeed that one scared the crap out of me. It sounds like your success rate with measurements matches mine. If there is a certain piece of hardware that would work far better I'd absolutely consider it.

 

[spalmgre]

E. I have an issue every time I use the software.

I have wanted to buy XO Audiolence but I get serious stability problems using the trial versions.

I'm a software entrepreneur for over 30 years and I could never be able to run a business with this level of problems. I have really tried several times during the last 7 years to get the different versions to work. During this time I have had several different Win computers and have still not been able to find a working setup. At the same time, I have used REW, rePhase, Impala, jRiver, Voicemeter, and more with very few problems.

I can see here a pattern that is not new to me. Small software companies have trouble financing such a level of product development that is needed to make a commercially successful software product.
But even small budgets can produce working solutions and that gives hope for the better. REW and Impala are stable so why can not Audiolence be also?

The core innovation is there and does work but only for some. So many of us can not see that it is worth putting down 400€ for software that most of the time does run into trouble.

I really like to have an answer on how to go fix the problem. One solution could be letting us import REW or Impala measurements and only run the filter calculation in Audiolence. This would solve instability that is related to different audio drivers or programs.

 

[whoareyou]

I really like to have an answer on how to go fix the problem. One solution could be letting us import REW or Impala measurements and only run the filter calculation in Audiolence. This would solve instability that is related to different audio drivers or programs.

So, your issues are mostly with measurements?

 

[spalmgre]

So, your issues are mostly with measurements?

I have not bought the XO license as I have had great problems with the stability. So I have not used the software as a regular user. But I have used the trial version that actually let me use all features but the saving of the calculated filters.
I have experienced most of my trouble with the measuring "module". This as I have had problems getting to the point where I would use the filter design features.
So I don't have enough experience to know how stable the filter design module is. But with extensive experience testing a multitude of software as CTO I have a hunch that there is some kind of conflict with windows and the audio drivers.
I have not tried to use a clean Win installation. Maybe using Oracle VirtualBox and installing there a clean window could do the trick. But the Win audio drivers will still be there so don't know how they interact in between the multitude of software layers.

 

[juicehifi]

I have wanted to buy XO Audiolence but I get serious stability problems using the trial versions.

I'm a software entrepreneur for over 30 years and I could never be able to run a business with this level of problems. I have really tried several times during the last 7 years to get the different versions to work. During this time I have had several different Win computers and have still not been able to find a working setup. At the same time, I have used REW, rePhase, Impala, jRiver, Voicemeter, and more with very few problems.

I can see here a pattern that is not new to me. Small software companies have trouble financing such a level of product development that is needed to make a commercially successful software product.
But even small budgets can produce working solutions and that gives hope for the better. REW and Impala are stable so why can not Audiolence be also?

The core innovation is there and does work but only for some. So many of us can not see that it is worth putting down 400€ for software that most of the time does run into trouble.

I really like to have an answer on how to go fix the problem. One solution could be letting us import REW or Impala measurements and only run the filter calculation in Audiolence. This would solve instability that is related to different audio drivers or programs.

At the moment I am not aware of any show stopping bugs in Audiolense.

And I haven't seen stability issues for many years.

Besides myself, Mitch is probably the one single person who has the most extensive user experience. And he seems to see it the same way.

Most of the problems I see are related to usb microphones and sound cards. Sometimes they refuse to work together, and other times one or the other or both produce glithes in the audio stream, which leads to measurements with significant timing errors. There is room for improvement in Audiolense in handling exceptions that arise when e.g. the recording file only contains noise or is empty, and an attempt to extract impulses fails. But fixing these issues would not fix the real problem, which are audio devices that do not deliver a steady audio stream even though they have given a green light for the exact settings being used.

FYI, Audiolense is using PortAudio for the measurements. PA is an open source software used in several applications.

Audio device problems have come and gone through the years. There were a lot of problems wihen Microsoft launched the current audio engine with Windows Vista. All of them should have and probably would have worked as good as before if the drivers had been made 100% according to spec. Then, gradually these problems became fewer and farther apart as old devices were phased out and new drivers were introduced. Then as now, it was mostly USB devices that were difficult. For the last few years, a high number of USB converters and professional usb sound cards have been launched. Many of them are using 3rd party drivers, which suggests that the scope of testing and debugging has been somewhat limited. Furthermore I suspect that USB streaming is generally inferior to a dedicated channel with regards to minimum latency and stability. Bottom line is this: Most of the recording problems that is experienced have causes that are outside of my reach. Audiolense exposes problems that does not surface during more regular use. The only thing I can do is to provide a variety of streaming option in the hope that at least one of them will work reliably. I have almost implemented any workaround that I can think of by now.

If stability is your main consideration you should opt for a professional sound card that uses a PCI(e) slot and a measurement microphone that you can hook up to one of the sound card's inputs.

 

[mccarty350]

Bernt, after all my attempts to stabilize the measurements (that is where I have issues, the rest of my issues are just very occasional) I believe that I'm going to take your advice and instead of using my USB sound card and USB microphone I'm going to purchase an analog calibrated microphone with an external phantom power supply (I assume I need this) and then try all of my measurements via the Asus Xonar U7 II's microphone input which (I believe) should allow me to disable clock drift correction. Does this sound like a sound direction to you? I have a fully modular system where I swap in drivers and build correction filters and take measurements probably weekly as I switch things in and out and experiment and I'm batting under 50% success rate on getting a decent measurement without a crash, IR extraction error, or file open error. I have created a batch file to clean out the measurement file detritus that lands in the measurement director in an attempt to eliminate file open errors as well. I use AL enough to invest in making it stable and I love the result.

 

[juicehifi]

Yes this solution have very good odds.

I can't give you any guarantees, but I have been using the U7 in the past for measurement as well as playback, with a microphone plugged straight in. I never had any measurement problems. Unfortunately my U7 has gone to hifi-heaven so I am not able to repeate the test..

 

[mccarty350]

I'll close the loop on if this helps me or not, I'm willing to invest in a solution. It's funny Bernt, from what I've read the hardware failure rate is pretty high on those U7's. From what I've read the PCIE versions of most cars have a poor SNR due to PC noise carrying over into the sound card after reading reviews, that's what pushed me towards the USB versions.

 

[mccarty350]

I am cross posting this here from the sound and microphone fixes sticky - because I find it immensely important to our topic. Feel free to delete from anywhere if I'm creating an annoying scenario Bernt:


Update: I know that there have been tons of reported issues reported with the measurement module. This is early feedback:
1. Prevous/present configuratin:
Asus Xonar U7 Mk II usb sound card, works great with 8 channel, reasonable price although discontinued for playback. I use the ASIO drivers.
2. Minidsp Umik-1 using WASAPI exclusive mode for measurements.
Result: Less than 50% success rate in completing measurements without erroring out, have to use clock drift correction since they are both USB devices.

Tested Configuration:
1. Asus Xonar U7 analog mic input with no clock drift correction configured. ASIO playback.
2. WASAPI exclusive mic input on ASUS U7 mic.
3. Phantom power supply has NOT arrived so I knew that my actual results would be unviable but I wanted to check for stability.
4. Used my Emotiva EMM1 uncalibrated mic for proof of concept, no phantom power so again just a proof of concept, ordered a Dayton analog EMM6 that I believe arrives tomorrow. Using mini to female XLR cable to feed it.

Result: I have taken no less than 6 measurements with NO CRASHES. None. I'm talking even with 20 second sweeps. This is showing massive affirmation (as someone that has had ENDLESS crashes) that if you go with one source without clock drift correction i.e. all in one device like one audio interface or sound card shared for playback and recording that the stability difference is literally night and day. I may be flush with victory right now but I will report back once my phantom power supply comes in and I can take real measurements with a calibrated mic. All of my captured data is 'unsuccessful' due to the lack of phantom power.

Quote Reply
Report Edit

 

[whoareyou]

Sorry, I'm not sure what you mean by "reasonable price although discontinued for playback". Do you mean that you only use the U7 for taking measurements, but for playback you have a different DAC? Just curious.

Thanks

 

[mccarty350]

Sorry, the previous poorly written email was mangled as a result of me multitasking heavily. When I read it back myself it sounds ridiculous, I was literally presenting on a conference bridge as I tried to complete that post. It was an epic fail.

To clarify, I meant that I was using one USB soundcard the ASUS Xonar U7 II - which is a recently discontinued (sadly) model - and a USB UMIK for recording - which requires clock drift correction enabled because it's a separate USB device. It also made the measurement module INCREDIBLY (under 50% success rate) unstable. Bernt has said many times that separate devices doing recording and playback, especially if they are USB, introduce clocking issues and unexpected behavior. As of yesterday I had a spare analog mic that leveraged an XLR connection and hooked it into my sound card's microphone input as a stability proof of concept. At the time I didn't have the calibrated mic that I ordered on hand so I introduced a microphone that I had on hand that required phantom power in to the equation - without having a phantom power supply. So in effect I could test taking measurements that I know would show up as unsuccessful due to the lack of phantom power to see if they would crash during measurement or not. I tested this a half dozen times with no failures - none. I would have crashed and burned in my prior configuration.

As of about 4 hours ago I received a calibrated Dayton EMM-6 and a 48v phantom power inline power supply unit into the equation. This allowed me to:

1. Take 96khz measurements for the first time (which per Bernt have little if any value over 48khz measurements but I wanted to try it out)
2. I was able to take them using the same ASIO driver for both the measurement and playback at the same time (no WASAPI even needed where as before I had to use WASAPI with my mic)
3. I was able to turn off clock drift connection.
4. I turned off play/record streaming.
5. I set the time to 10 seconds which often introduces the infamous 'knocking' noise.

I was able to take no less than 5 consecutive measurements in this manner successfully, each of them (after playing with the recording levels + my output volume on my preamp) topping out as 'excellent'.

I wish I had done this ages ago. For a sub-$100 investment my ability to use AL just became SO much easier.

Hopefully this message that is currently occupying 100% of my focus makes much more sense. It affirms that audiolense has the potential under the right circumstances to be much more stable.

Now...I am no longer getting the IR extraction error but I AM able to (I believe) mitigate the file open error. I have a batch file on my desktop that deletes the detritus from previous measurements from the measurement directory that I execute before my measurements and this seems to have helped.

Bernt, this one is for you and may explain some of the horrible microsoft incurred issues that are plaguing us. Windows 11 ships trying to back up the contents of Documents and Settings into One Drive by default. This means that (and I've watched explorer as it happens) that it tries to back up the new mpf and wav files that AL creates as part of it's measurement process and tries to back them up IMMEDIATELY after they are created, in effect locking these files while it synchs them with one drive.

I - in order to eliminate the whole issue - moved the default AL directory to another hard drift upon which my OS does not reside to eliminate any Windows incurred shenanigans and it in addition seemed to incrementally improve things.

 

[whoareyou]

Bernt, this one is for you and may explain some of the horrible microsoft incurred issues that are plaguing us. Windows 11 ships trying to back up the contents of Documents and Settings into One Drive by default. This means that (and I've watched explorer as it happens) that it tries to back up the new mpf and wav files that AL creates as part of it's measurement process and tries to back them up IMMEDIATELY after they are created, in effect locking these files while it synchs them with one drive.

I - in order to eliminate the whole issue - moved the default AL directory to another hard drift upon which my OS does not reside to eliminate any Windows incurred shenanigans and it in addition seemed to incrementally improve things.

Thanks for the clarifications. Very helpful.

Interesting about One Drive. I've never enabled this feature so it's possible that I've had better luck with my measurements because my Documents are not backing up to OneDrive. It would be interesting to see what others are doing with One Drive.

Perhaps whatever is going on is similar to issues I seem to resolve by configuring AL exclusions in Defender.

I also don't want to give the impression that my measurements (ASIO Dac, UMIK-1) are glitch free. Even with the Defender exclusions, I still encounter the strange pops and clicks but I don't take frequent measurements, so it's not that big of a deal for me to work around them.

 

[mccarty350]

I bought the first new machine in an age and it's running windows 11. All that garbage was on by default and I didn't spend the time to disable it in the OS so basically I just ignored it until the light bulb went off and I went to see if one drive was trying to synch the files as they were created - and indeed it was.

I followed your thread and made the same exclusions (and then a few more) but I was still plagued by issues.

The pops and clicks in my new scenario have very very rarely re-occurred so indeed, using ASIO and the same device to take measurements and perform playback is soooo much better.

Let me ask you a question. Do you know how to get measurements to use the same channel orders, settings, measurement settings, etc. without having to plug them in every time?

Every time i create a new setup I have to re-enter those settings I wish I could just make them defaults, maybe there is the ability to do such and I'm just not aware. The amount of time I spend on audiolense a week is pretty extensive. I'll always tweak something by changing the crossover points or octave width, etc to optimize and since my system is modular it I'm always changing driver modules, positioning, etc. As a result I burn a lot of time having to put in my defaults over and over.

 

[whoareyou]

With new setup, I'm not aware of way to avoid reenter parameters. Pretty sure each time I've done this I had to reenter everything.

But when reusing same setup, parameters seem to be cached when I select "New Measurement".

I don't know if there's some leeway in what can be edited/saved in existing setup (so that you are reusing the "same" setup) that would allow the last settings i.e. channel order to be carried over.

You have an interesting use case

 

[juicehifi]

The whole setup is saved in the measurement. If it doesn’t stick you can first choose a different setup …. e.g regular stereo …. any setup that doesn’t match with yours will do. Then, when you open your last measurement, you will get all the settings attached, including the re-routing.