Audeze LCD-X Review (2021 Edition Headphone)

[Jonne Haven]

Here's how one random filter compares across sample rates, once as PEQ once as GEQ:
View attachment 227852 View attachment 227853

Some notes:

• How much a PEQ filter varies with Fs depends a lot on the specific parameters (Frequency, Gain, Q). Some designs are pretty agnostic.
• The GEQ shows some variability below 300Hz or so that I can't explain.
• The one shown here is an almost 600 band GEQ exported straight from REW.
• GEQs with fewer bands will vary more with sample rate, but still much less than PEQ and only at low frequencies.

You should match sample rates when designing or deploying PEQ ideally.
Similarly, any PEQ you find online should specify the intended sample rate. AutoEQ for example is 44.1k, Maiky76 does all his PEQs at 96k.


Here's the frequency response of @Jonne Haven 's PEQ design on the LCD-X 2021, using both oratory's and amirm's data:
View attachment 227878 View attachment 227879

Here's how one random filter compares across sample rates, once as PEQ once as GEQ:
View attachment 227852 View attachment 227853

Some notes:

• How much a PEQ filter varies with Fs depends a lot on the specific parameters (Frequency, Gain, Q). Some designs are pretty agnostic.
• The GEQ shows some variability below 300Hz or so that I can't explain.
• The one shown here is an almost 600 band GEQ exported straight from REW.
• GEQs with fewer bands will vary more with sample rate, but still much less than PEQ and only at low frequencies.

You should match sample rates when designing or deploying PEQ ideally.
Similarly, any PEQ you find online should specify the intended sample rate. AutoEQ for example is 44.1k, Maiky76 does all his PEQs at 96k.


Here's the frequency response of @Jonne Haven 's PEQ design on the LCD-X 2021, using both oratory's and amirm's data:
View attachment 227878 View attachment 227879

Whoa! That's way enlightening. Thanks for doing that, I have to study that more in depth!

Blessings
-J

 

[Robbo99999]

Here's how one random filter compares across sample rates, once as PEQ once as GEQ:
View attachment 227852 View attachment 227853

Some notes:

• How much a PEQ filter varies with Fs depends a lot on the specific parameters (Frequency, Gain, Q). Some designs are pretty agnostic.
• The GEQ shows some variability below 300Hz or so that I can't explain.
• The one shown here is an almost 600 band GEQ exported straight from REW.
• GEQs with fewer bands will vary more with sample rate, but still much less than PEQ and only at low frequencies.

You should match sample rates when designing or deploying PEQ ideally.
Similarly, any PEQ you find online should specify the intended sample rate. AutoEQ for example is 44.1k, Maiky76 does all his PEQs at 96k.


Here's the frequency response of @Jonne Haven 's PEQ design on the LCD-X 2021, using both oratory's and amirm's data:
View attachment 227878 View attachment 227879

Re PEQ which I'm interested in, not a huge difference between 44kHz and 48kHz then, which are the ones I use, although it does make a difference higher up the frequency range, and big time if you choose those highest sample rates. But a question for you - in REW when you create the the filters, how do you know what sample rate they're set at, how do you change what sample rate they're set at? Interestingly though all the filters I've created in REW have measured as predicted on both speakers and headphones when I've measured the response after EQ......I have a miniDSP EARS for measuring headphones and following is the predicted response of my K702 after EQ and the actual measured response after EQ (it's not the same seating, predicted is based on an average of measurements and actual measured after EQ is left & right channel one measurement each, but not significantly different):
Predicted:

Actual Measurement after EQ:

So going back to my question, how do we know which sample rate REW uses for it's filter creation?

 

[Robbo99999]

Here's the frequency response of @Jonne Haven 's PEQ design on the LCD-X 2021, using both oratory's and amirm's data:
View attachment 227878 View attachment 227879

Re @Jonne Haven 's EQ, it doesn't look too bad overall in terms of what it might sound like (if you love bass!), but the massive treble over 10kHz could be unsettling or painful.

 

[Jonne Haven]

Re @Jonne Haven 's EQ, it doesn't look too bad overall in terms of what it might sound like, but the massive treble over 10kHz could be unsettling or painful.

Yeah, I had no idea there was that drop at 700 to 1500 Hz. I may have to do something about that if that's accurate. And the treble is a personal taste sort of thing and I agree that it's not for everybody. That's why I offered those adjustable bands to taste in the highs.

Thanks again for the data!

Blessings,
-J

 

[Robbo99999]

Yeah, I had no idea there was that drop at 700 to 1500 Hz. I may have to do something about that if that's accurate. And the treble is a personal taste sort of thing and I agree that it's not for everybody. That's why I offered those adjustable bands to taste in the highs.

Thanks again for the data!

Blessings,
-J

Well I think I knew that you had a shortfall at 1kHz back in the day (a few months ago) when you showed us your track you'd created, because I think you'd showed us your initial headphone EQ at the same time, and it had a lack at 1kHz when I eyeballed it back then, which led me to expect your mix might be too 1kHz rich.......which was why I paid attention to that aspect when I listened to your track......and I felt your track did indeed have a bit too much 1kHz in it....which I mentioned to you and you agreed and reduced the amount of 1kHz in your track. Kudos to you re your track you showed us!

 

[Jonne Haven]

Here's how one random filter compares across sample rates, once as PEQ once as GEQ:
View attachment 227852 View attachment 227853

Some notes:

• How much a PEQ filter varies with Fs depends a lot on the specific parameters (Frequency, Gain, Q). Some designs are pretty agnostic.
• The GEQ shows some variability below 300Hz or so that I can't explain.
• The one shown here is an almost 600 band GEQ exported straight from REW.
• GEQs with fewer bands will vary more with sample rate, but still much less than PEQ and only at low frequencies.

You should match sample rates when designing or deploying PEQ ideally.
Similarly, any PEQ you find online should specify the intended sample rate. AutoEQ for example is 44.1k, Maiky76 does all his PEQs at 96k.


Here's the frequency response of @Jonne Haven 's PEQ design on the LCD-X 2021, using both oratory's and amirm's data:
View attachment 227878 View attachment 227879

Your graph says that the red curve is my EQ plus oratory's. What numbers of oratory's are you using? I'm interested in doing this kind of graphing myself. I have REW but I'm a noob at using it. Any help you could give me would be wonderful!

Yeah, I'm not sure I totally understand how you got the final combined red curve. Is it oratory's measurements? Where did you get the data?

Thanks ahead of time,

-J

 

[KeithPhantom]

Your graph says that the red curve is my EQ plus oratory's. What numbers of oratory's are you using? I'm interested in doing this kind of graphing myself. I have REW but I'm a noob at using it. Any help you could give me would be wonderful!

Yeah, I'm not sure I totally understand how you got the final combined red curve. Is it oratory's measurements? Where did you get the data?

Thanks ahead of time,

-J

I think AutoEQ has a way you can take a curve picture and transform it into coordinates and then use autoeq.py to generate a summed signal.

 

[staticV3]

What numbers of oratory's are you using?

The ones from his LCD-X PDF:

Which you can download as raw .csv here.

Yeah, I'm not sure I totally understand how you got the final combined red curve.

I used oratory's raw frequency response of the LCD-X and applied the frequency response of your Peace PEQ filter to it.

 

[jae]

What are with these crazy EQs? At this rate It's time to start taking in-ear measurements!

 

[Jonne Haven]

Here's how one random filter compares across sample rates, once as PEQ once as GEQ:
View attachment 227852 View attachment 227853

Some notes:

• How much a PEQ filter varies with Fs depends a lot on the specific parameters (Frequency, Gain, Q). Some designs are pretty agnostic.
• The GEQ shows some variability below 300Hz or so that I can't explain.
• The one shown here is an almost 600 band GEQ exported straight from REW.
• GEQs with fewer bands will vary more with sample rate, but still much less than PEQ and only at low frequencies.

You should match sample rates when designing or deploying PEQ ideally.
Similarly, any PEQ you find online should specify the intended sample rate. AutoEQ for example is 44.1k, Maiky76 does all his PEQs at 96k.


Here's the frequency response of @Jonne Haven 's PEQ design on the LCD-X 2021, using both oratory's and amirm's data:
View attachment 227878 View attachment 227879

Thanks for your help so far @staticV3! I had some trouble downloading the .csv file from GitHub because I couldn't find a download link so I just copied and pasted the raw data into a text document and made the extension .csv.

I would really like to be able to do what you did above (applying my PEQ to the raw measurements and then overlaying the harman curve over it) but I can't seem to make heads or tails of it in REW. There doesn't seem to be a specific guide to do this particular kind of task around that explains it simply so a nub like me can understand. I've tried to figure it out on my own but REW is less than intuitive to work with.

If you would be willing to do me a favor and show me how to do it step by step in REW or whatever software I need I'd be much obliged. If it's too much hassle I understand completely, but I thought I should at least ask. I don't want to be the guy who is always asking someone else to do his measurements for him.

PM me if this is something you're willing to take on. I don't want to hijack this thread.

Thanks ahead of time!
Blessings,
-J

 

[staticV3]

@Jonne Haven Here's my workflow:

Btw, I just noticed that I made a mistake while copying your PEQ values into Peace the first time around.
Filter 10 should have a gain of -1dB, but I entered +1dB by accident. That's why the graphs I posted before have that dip around 1kHz. I'm sorry.

 

[Jonne Haven]

@Jonne Haven Here's my workflow:

Btw, I just noticed that I made a mistake while copying your PEQ values into Peace the first time around.
Filter 10 should have a gain of -1dB, but I entered +1dB by accident. That's why the graphs I posted before have that dip around 1kHz. I'm sorry.

Most excellent!!! Thank you so much, @staticV3!

And don't worry about the "clerical" error, I do it all the time. But it is amazing what a simple + or - might do to a curve. I've posted my EQ in a few forums so I'm glad I don't have to redact anything or what's even worse, have to retrain my ears to my reference tracks. I was kinda curious how my @oratory1990 based mids could have been off by that much. I'm just relieved that my ears aren't as off as I initially thought they might be, so phew!


Thanks again! And so quick!
Blessings,
-J

 

[Jimbob54]

Here's how one random filter compares across sample rates, once as PEQ once as GEQ:

Can I just clarify a couple of things here as I try and wrap my head around this. Just looking at the PEQ chart. When you say "one random filter" do you mean one set of eq parameters (so a whole set of adjustments for one headphone, as one might find from oratory etc) or just ONE single filter (eg one freq, amplitude and Q)

And are the lines the eq profile adjustments (as I'd see for eg in PEACE /APO or the predicted FR after adjustment?

Ta

 

[staticV3]

When you say "one random filter" do you mean one set of eq parameters (so a whole set of adjustments for one headphone, as one might find from oratory etc)

Yes.

And are the lines the eq profile adjustments (as I'd see for eg in PEACE /APO or the predicted FR after adjustment?

They show the frequency response of just the EQ.

 

[Robbo99999]

@staticV3 , how were you able to model the differences between filters with the same parameters between different sample rates? I'd want to do this too so I can play about with different filters and the effect of sample rate changes, and it might have some fringe practical benefits between the different sample rates I use of 44kHz & 48kHz. Ideally I'd want to be in a position to ensure that my Total EQ Curve is identical between 44kHz and 48kHz.

 

[staticV3]

@Robbo99999 I used EQApo for GEQ, EQApo+Peace for PEQ.
I played REW's measurement sweep via foobar2000 Default output, then recorded it with Audacity's Wasapi loopback function (captures post-APO desktop audio), then let REW calculate the FR using the sweep import function.
Finally I changed sample rates using the Sound Control Panel, rinse and repeat.

Fyi, PEQ behaving like that is not unique to EQApo. Mathaudio Headphone EQ for example does the same. It's inherent to every PEQ.
That's why miniDSP runs theirs always at a fixed sample rate with a resampler in front, to avoid different source Fs affecting the end result.

 

[Robbo99999]

@Robbo99999 I used EQApo for GEQ, EQApo+Peace for PEQ.
I played REW's measurement sweep via foobar2000 Default output, then recorded it with Audacity's Wasapi loopback function (captures post-APO desktop audio), then let REW calculate the FR using the sweep import function.
Finally I changed sample rates using the Sound Control Panel, rinse and repeat.

Fyi, PEQ behaving like that is not unique to EQApo. Mathaudio Headphone EQ for example does the same. It's inherent to every PEQ.
That's why miniDSP runs theirs always at a fixed sample rate with a resampler in front, to avoid different source Fs affecting the end result.

Cheers, thanks for that info. Ah, so you were actually measuring the differences. I'd be looking to simulate the differences in a program so that I can manipulate the filters to ensure that the Total EQ Curve is the same at both 44kHz and 48kHz - any ideas on how to do this?

(Good to know about miniDSP, I have the 2x4 (non-HD), but I figured sample rate wouldn't matter with that as it's analog in & analog out......and the 48kHz rate that it uses internally matches the Generic 48kHz mode of REW in terms of working out the PEQ required).

 

[staticV3]

any ideas on how to do this?

unfortunately not. I can use other people's python scripts via CLI, but I'm a long way off writing my own.

Edit: maybe try asking REW's developer if he could add that function?

 

[Robbo99999]

unfortunately not. I can use other people's python scripts via CLI, but I'm a long way off writing my own.

Edit: maybe try asking REW's developer if he could add that function?

Yeah, that would be good if REW had an option for that. I do find it strange that EQ Player Software (EqualiserAPO, Roon, Neutron Player) I'm assuming doesn't account totally properly for sample rate changes. Afterall it's surely just a mathematical problem that can be written into the software. I mean I know from talking to Oratory that current EQ software like EqualiserAPO & the others will change the coefficients of the biquads automatically based on the sample rate change, so it does attempt to ensure that filters (along with their parameters: filter type, Q, Gain) are somewhat interchangeable between different sample rates, but I don't know why it won't calculate it perfectly so that the Total EQ Curve will be exactly the same when flipping between sample rates - just doesn't make sense to me that it's not done perfectly. I'm not a member on REW's forum, but it might be worth joining to see about adding that function!

 

[staticV3]

I mean I know from talking to Oratory that current EQ software like EqualiserAPO & the others will change the coefficients of the biquads automatically based on the sample rate change

That's news to me actually! What I've measured sure doesn't look like that's what's happening.