ZMF Caldera - New Planar Magnetic from ZMF!

[Resolve]

Well that’s why I had to try it just to be sure I knew what I’m talking about. I was more surprised the Harman curve didn’t completely change the headphones. I still very much enjoy them with the Harman curve and have now found some songs even sound better. It’s just the same albums I had trouble with on the Expanse also have trouble with Caldera and Harman curve. I wouldn’t consider the Harman curve beige or vanilla (I love vanilla). It more what I’d call a pop sound. Elevate the bass and mids and try to make it really exciting. This is why Armim said the Caldera sounds boring in comparison.

So I won’t say you moved away from the Harman curve but I like that the final sound is more subtle and takes into consideration the enclosure and air flow. The Harman curve does still work and even some may actually end up preferring after trying it. But if all headphones were delivered with this sound, it would definitely then be more about the look and feel and could see people sharing their favorite EQ settings more. I personally like to try to find a headphone that already sounds great than to EQ it after.

Important to remember that Harman is smoothed to about 1/2 octave, and based on his measurements I'd expect Caldera to not be totally off with the same smoothing applied (not as much ear gain, so maybe a bit warmer). A common mistake made by folks over on that forum is to over focus on reading the tea leaves with the measurements, thinking a perfect match to OE2018 is going to sound best to them. This ignores the significance of HRTF differences and HpTF variation for individuals, as well as the other well established preference groups from the research.


The best way to think of it IMO is as a way to evaluate headphones relative to common preferences, not necessarily how all headphones *should* be tuned. I've said this in other places as well but I tend to think the Harman research has been done a disservice by target cults who cherry pick pieces of it to suit their narratives, completely missing the point of the research altogether and the incredible service it is to the community.

Edit: I should clarify, I do not have experience with this headphone so this is pure speculation on my part.

 

[goldwerger]

What I DO know without a doubt, scientifically, existentially, exophilistically - is that the new amp that @L0rdGwyn delivered to ZMF yesterday defies all laws of human existence, I literally can't believe how amazing it is.

I went with a color called "Caldera:"


And wow! It's a volcano exploding in my brain every time I've used it so far. I've need a wet towel and a bath half a dozen times in the last day from this thing.

But really @L0rdGwyn you're a legend - and for those in awe of his work as I often am we are working on a Caldera driving commercial version of the "Aegis" which should be released in Q1 this coming year. The Aegis is defined by ample power (it even drives the Tungsten) and a mind-bending-blowing large stage while keeping a firm grip on the low end. I'm beyond excited for it and it's this kind of stuff that keeps me going at tough times of the year like this when things are just busy and stressful!

Aegis thread link: https://www.head-fi.org/threads/aegis-diy-tube-headphone-amplifier.965530/

Here's some quick pics of the "Air-Universe; Caldera Edition." I will have pics of the Aegis in a few weeks as well which will be widely available.

Seven!!!!!! Count em seven Monolith Magnetics transformers. Holy moly...yes it's heavy.


Sauce shot of the cool transformer covers.



This one my be a little out of focus, volume knob on the left and impedance selector on the right.

Holly crap man!!!!!!!

I never get jealuos off anyone's pleasure. Well, I never did...

 

[robo24]

There are so many issues with the lack of understanding of sampling and science over at ASR. A number of problems I see with just a quick review of some of the Harman research.
1. Not a simple random sample and not really representative of any real-world population given the demographics they report. Therefore their findings cannot be applied to a general population.
2. Whatever median or average preferred curve they find means just that. It definitely doesn’t mean the majority prefer that exact curve, though could mean a majority prefer within a certain level of variation within that curve, which gets more narrow the more people you sample. You can sample 100 for some measurement scale that goes from 1 to 100 and find 50 prefer 1 and 50 prefer 100 and get an average of 50.5 which is literally no person’s preference. Now let’s say it was a simple random sample and actually 100% of people preferred the exact same curve. All that means is that under these exact same conditions with the same equipment and same workers running the study, adjusting the headphones used on the rig, using the exact same headphones all shown to measure exactly the same that indeed that is the preferred curve which may be applicable to the general population only if demographics were representative of that population. Any manufacturer using a different rig or even slightly different placement on the same rig would likely have slightly different results.

Even if you could alter all those variables so that everyone always used the exact same measuring equipment perfectly and pads never wore down and everyone always put headphones on to fit exactly the same, so what? To get that average or median curve, actual individual preferences would still be all over the place among those sampled, again meaning that only a minority prefer the exact average or median curve. Given the sales of Beats and other consumer headphones that don’t follow Harman it could be argued in the real world that people actually prefer crappy, very bass heavy headphones.

Good scientists are deeply skeptical of even their own findings, actively try to find ways to break them, and if they can’t still are very up front about the limitations of applying findings to a general population, especially if there are major sampling deficiencies as in the Harman research. We also know that for individuals, preferences can vary by the day, time of day, amount of sleep, how much noise exposure that day, caffein, having a cold or allergies or not, etc. Even if every person had exactly the same ears, inner ears, etc. so that everyone experienced sound exactly the same, there would still likely be cultural influences based on all kinds of things. One culture might love high pitched female opera singing, while another prefers music that is very bass heavy, etc. As a former scientist myself, those are the things I had to deal with in anything I got or tried to get published.

I like Zach’s approach in that Harman has some utility and it’s a baseline of sorts for making comparisons. Unfortunately the way it’s generally discussed and touted by proponents is quite unscientific. ASR is the site for those folks! As a consumer I’ll just use my own ears and go from there thanks!

 

[Pharmaboy]

sanding and finishing is my place of zen

(smiling)

 

[jlbrach]

I think we should have just one sound for headphones in different physical housings

 

[Zachik]

Haha I think a cool feature would be to put a rotatable outer "11" ring in place so that you can make sure wherever the volume is point that you're always at 11. But maybe a lot of engineering for a cerebral satisfaction.

My Glenn EL3N amp volume already goes to "11"

 

[Leto Dal]

Not sure if its the right or a clear analogy but I had to think of Bill Hicks several times reading that ASR thread in the last days

 

[goldwerger]

My Glenn EL3N amp volume already goes to "11"

with you, buddy, I'm sure it's craked high with some classic rock and lots of electric

 

[zach915m]

Important to remember that Harman is smoothed to about 1/2 octave, and based on his measurements I'd expect Caldera to not be totally off with the same smoothing applied (not as much ear gain, so maybe a bit warmer). A common mistake made by folks over on that forum is to over focus on reading the tea leaves with the measurements, thinking a perfect match to OE2018 is going to sound best to them. This ignores the significance of HRTF differences and HpTF variation for individuals, as well as the other well established preference groups from the research.


The best way to think of it IMO is as a way to evaluate headphones relative to common preferences, not necessarily how all headphones *should* be tuned. I've said this in other places as well but I tend to think the Harman research has been done a disservice by target cults who cherry pick pieces of it to suit their narratives, completely missing the point of the research altogether and the incredible service it is to the community.

Edit: I should clarify, I do not have experience with this headphone so this is pure speculation on my part.

Haha that is something I totally forgot about - the smoothing! My general understanding has been that unless listening to a specific frequency, we as listeners are going to hear music, or full frequency responses as blended, is it 1/9th or something that's supposed to be closer to how we actually hear? I know you're doing some kind of admirable research on the "ranges" that different types of listeners might appreciate, I look forward to reading more about it.

So anyways - being in the shop all day every day this time of year I took the opportunity to grab an Atrium Closed (more harman tuned) and my caldera and run them with the 1/3 smoothing on on my rig. And yes I used the 711 coupler / AECM 206.

The Atrium Closed with no smoothing looks like this with suede pads. and keep in mind, while the suede pads are closest to harman, I find them too bright for my tastes, I know plenty of owners who love them.

EDIT: We have titan mesh in perf and solid config now for the Atrium Closed and other closed models, I'll post it for sale in the DIY section of the site for those who like suede but want to tamp down that treble as I would.


With 1/3 smoothing:


My preferred lambskin pads on Atrium Closed 1/3 smoothed:



And Caldera ofcourse with 1/3 smoothed:



So anyways - am I here to use these graphs to make some sort of argument, no! This is all just information that is really fascinating. The end goal for me is always to use this stuff to help people find the right headphone for them since as illustrated above we have ones that do and don't follow harman more or less closely. I don't use smoothing much in my R/D and every day use for obvious reasons since I am more in the practice than theoretical area. But this all is worthwhile stuff.

I guess my main question with this for you @Resolve is should measurers be smoothing the headphones they measure to match the smoothing of the harman target? Or should we always view both to see a blended and non blended view of the frequencies? More information is better right?

 

[Benesyed]

Sometimes when I look at a recipe and the ingredients within it I can make a conclusion that I won't like it. Sometimes I'm right, however some of my favorite foods such as Chicken Tikka Masala contain ingredients that I generally don't like. Pizza is an other example of this. The ingredient list isn't great for me, but who doesn't like pizza!?!

So basically, LOOKING at the ingredient list of what I plan to eat gives information, but it isn't until I actually hear taste it that I can determine if I like it or not. My 7 year old will often refuse to try something just based on what is in it. But, he's 7, and I'd think that adults would be different in that respect. Also, my 7 year old doesn't post on the internet his opinion of something he's never tried.


And speaking of never tried...THAT AMP! I've already been drooling over it in the DIY threads from @L0rdGwyn. I would be very excited to try that meal!!!

I think it's a little bit more than this. I would say it's like the ingredient list and the recipe. There is still wiggle room but especially if you've eaten a lot of different foods and know a lot about cooking, you can guess. But nothing beats actually tasting.

 

[Resolve]

Haha that is something I totally forgot about - the smoothing! My general understanding has been that unless listening to a specific frequency, we as listeners are going to hear music, or full frequency responses as blended, is it 1/9th or something that's supposed to be closer to how we actually hear? I know you're doing some kind of admirable research on the "ranges" that different types of listeners might appreciate, I look forward to reading more about it.

So anyways - being in the shop all day every day this time of year I took the opportunity to grab an Atrium Closed (more harman tuned) and my caldera and run them with the 1/3 smoothing on on my rig. And yes I used the 711 coupler / AECM 206.

The Atrium Closed with no smoothing looks like this with suede pads. and keep in mind, while the suede pads are closest to harman, I find them too bright for my tastes, I know plenty of owners who love them.


With 1/3 smoothing:


My preferred lambskin pads on Atrium Closed 1/3 smoothed:



And Caldera ofcourse with 1/3 smoothed:



So anyways - am I here to use these graphs to make some sort of argument, no! This is all just information that is really fascinating. The end goal for me is always to use this stuff to help people find the right headphone for them since as illustrated above we have ones that do and don't follow harman more or less closely. I don't use smoothing much in my R/D and every day use for obvious reasons since I am more in the practice than theoretical area. But this all is worthwhile stuff.

I guess my main question with this for you @Resolve is should measurers be smoothing the headphones they measure to match the smoothing of the harman target? Or should we always view both to see a blended and non blended view of the frequencies? More information is better right?

My understanding is that beyond 1/6th stuff stops being all that perceptually relevant, but this could also depend on the individual. So maybe it's like as the sample size increases you lose granularity in the process, and certainly there are bound to be some narrowband peaks that can be more perceptually relevant than others where they don't look particularly egregious when smoothed to 1/6th. So I typically go for 1/12th or 1/24th because it's a good balance of readability completeness of data representation.

To answer your question, I don't think we should be smoothing the data beyond 1/12th because I think it matters that all the data be visible, but when showing measurements against the Harman target it makes sense to also show it smoothed to the same degree so people can see the apples to apples result, or as I like to call it "sound signature". It's also bound to be a more useful jumping off point for people getting into EQ than trying to do all the fine-grained adjustments to match a target. I see this all the time where folks will hate on a given target because they tried to EQ to match it, when in reality any EQ above 4-5khz needs to be done by ear anyway.

The other reason not to smooth the data beyond 1/12th though is for the benefit of seeing it relative to the ear transfer function, which is a more fine-grained reference point. Now I should be clear that there's some debate abouts the usefulness of some of those fine-grained features, but in my testing so far with the B&K 5128, for over-ear headphones it does yield some interesting results. I'm not sure how well-known this is but the calculated DFHRTF for the 5128 (not the one supplied with it) is the most clear ear transfer function out of all the rigs we use, since the one for GRAS KEMAR is based on the large format 0065 pinna and the 4128's is smoothed. We do have the data for the KB5000, thanks to Oratory1990's work on that but the calculation is still in progress - Blaine goes into more detail on these here.

With all of that said, it can be difficult to really determine the benefits of fine-grained FR vs fine-grained ear transfer function on person-specific basis since each headphone is bound to behave slightly differently on each head. At the moment I'm erring on the side of "more data is better", and I'm quite confident we can learn a lot more about the subjective characteristics that people love in headphones by improving the analysis of the data - in fact I think we already have, certainly in the case of in-ear headphones (it turns out acoustic Z and pinna effect really matter).

All of that is to say that if we were content with the older paradigm of showing a fine-grained result against a coarse-grained target, we wouldn't need to bother with all of this.

There is something else though that I may as well put here since some of this is a discussion about Harman after-all, and I saw this come up in that thread on ASR as well and it has to do with the circle of confusion. This is just my personal opinion on things, and not necessarily backed up by anything in particular, but I strongly suspect that the current discourse surrounding the topic would be very different if Dr. Olive hadn't been as focused on solving that problem.

What I mean by this is that if you look at how the preference research is done, it doesn't really line up with the idea that some folks are running with - the idea that Harman OE 2018 is the one true curve that everything must match. In fact it lines up a lot better with the kind of cluster analysis that gets done with preference research in general. Simply put, if we discard the goal of solving the circle of confusion, the established preference groupings would be better reflected as 'targets' rather than 'target'. This is how it is with preference research in other industries as well.

While my own viewpoint on 'fidelity' is slightly different from Blaine's, I suspect we're aligned on this point, and its one of the reasons we want to encourage the use of preference boundaries or ranges to provide a more complete picture of a headphone's performance relative to known segments. When you go to the grocery store you don't typically see just one flavor of anything, and that's precisely because there are different preference groups that the market is serving. And, when you look into the Harman research beyond the headlines or what some folks say on ASR you realize that while yes there is a dominant group, there's a good case for having different preference groups accounted for as well. The paper on segmentation in particular is useful for this for anyone wanting to get into that.

I'm also encouraged to see that Dr. Olive is working with the 5128 on some new research, so it'll be interesting if anything changes based on that. Either way, new preference research is bound to be conducted (or maybe we get a different manufacturer to make their preference research public?), and we can always incorporate that into what we're doing too.

TL;DR - Targets, not target - but maybe we'll get there with new research.

 

[zach915m]

My understanding is that beyond 1/6th stuff stops being all that perceptually relevant, but this could also depend on the individual. So maybe it's like as the sample size increases you lose granularity in the process, and certainly there are bound to be some narrowband peaks that can be more perceptually relevant than others where they don't look particularly egregious when smoothed to 1/6th. So I typically go for 1/12th or 1/24th because it's a good balance of readability completeness of data representation.

To answer your question, I don't think we should be smoothing the data beyond 1/12th because I think it matters that all the data be visible, but when showing measurements against the Harman target it makes sense to also show it smoothed to the same degree so people can see the apples to apples result, or as I like to call it "sound signature". It's also bound to be a more useful jumping off point for people getting into EQ than trying to do all the fine-grained adjustments to match a target. I see this all the time where folks will hate on a given target because they tried to EQ to match it, when in reality any EQ above 4-5khz needs to be done by ear anyway.

The other reason not to smooth the data beyond 1/12th though is for the benefit of seeing it relative to the ear transfer function, which is a more fine-grained reference point. Now I should be clear that there's some debate abouts the usefulness of some of those fine-grained features, but in my testing so far with the B&K 5128, for over-ear headphones it does yield some interesting results. I'm not sure how well-known this is but the calculated DFHRTF for the 5128 (not the one supplied with it) is the most clear ear transfer function out of all the rigs we use, since the one for GRAS KEMAR is based on the large format 0065 pinna and the 4128's is smoothed. We do have the data for the KB5000, thanks to Oratory1990's work on that but the calculation is still in progress - Blaine goes into more detail on these here.

With all of that said, it can be difficult to really determine the benefits of fine-grained FR vs fine-grained ear transfer function on person-specific basis since each headphone is bound to behave slightly differently on each head. At the moment I'm erring on the side of "more data is better", and I'm quite confident we can learn a lot more about the subjective characteristics that people love in headphones by improving the analysis of the data - in fact I think we already have, certainly in the case of in-ear headphones (it turns out acoustic Z really matters).

All of that is to say that if we were content with the older paradigm of showing a fine-grained result against a coarse-grained target, we wouldn't need to bother with all of this.

There is something else though that I may as well put here since some of this is a discussion about Harman after-all, and I saw this come up in that thread on ASR as well and it has to do with the circle of confusion. This is just my personal opinion on things, and not necessarily backed up by anything in particular, but I strongly suspect that the current discourse surrounding the topic would be very different if Dr. Olive hadn't been as focused on solving that problem.

What I mean by this is that if you look at how the preference research is done, it doesn't really line up with the idea that some folks are running with - the idea that Harman OE 2018 is the one true curve that everything must match. In fact it lines up a lot better with the kind of cluster analysis that gets done with preference research in general. Simply put, if we discard the goal of solving the circle of confusion, the established preference groupings would be better reflected as 'targets' rather than 'target'. This is how it is with preference research in other industries as well.

While my own viewpoint on 'fidelity' is slightly different from Blaine's, I suspect we're aligned on this point, and its one of the reasons we want to encourage the use of preference boundaries or ranges to provide a more complete picture of a headphone's performance relative to known segments. When you go to the grocery store you don't typically see just one flavor of anything, and that's precisely because there are different preference groups that the market is serving. And, when you look into the Harman research beyond the headlines or what some folks say on ASR you realize that while yes there is a dominant group, there's a good case for having different preference groups accounted for as well. The paper on segmentation in particular is useful for this for anyone wanting to get into that.

I'm also encouraged to see that Dr. Olive is working with the 5128 on some new research, so it'll be interesting if anything changes based on that. Either way, new preference research is bound to be conducted (or maybe we get a different manufacturer to make their preference research public?), and we can always incorporate that into what we're doing too.

TL;DR - Targets, not target - but maybe we'll get there with new research.

Hey thanks again for the detailed response!

The only thing about EQ and that the idea of preference ranges makes me realize, is that I personally would rather own different headphones that suit different responses rather than continuously EQ a single headphone based on the music, mood, listening volume, my ever changing HRTF etc. It's also more convenient for my life style.

Maybe this is more singular to zmf type owners or not, but most of us own more than a single headphone right for different moods/situations?

I guess that's kind of the idea of the various ZMF models and I'm slowly remember why I make the headphone range I do again. I guess it all comes full circle and I feel like I understand everything a little better now, including why I'm sitting at a sanding bench on a Sunday night making headphones I love.

Anyways thanks all I appreciate the discussion.

 

[Resolve]

my ever changing HRTF

I'd be curious what would cause an HRTF to change hahah. But yeah I was going to say different pads are a good way to accommodate different preferences as well - also just how you feel like listening to music at any given time. That's part of the fun of it all.

 

[zach915m]

I'd be curious what would cause an HRTF to change hahah. But yeah I was going to say different pads are a good way to accommodate different preferences as well - also just how you feel like listening to music at any given time. That's part of the fun of it all.

Oh yeah I mean things like mood, listening volume, what I'm listening to, what chain, what pads feel good, humidity, allergies etc not that like, my anatomy or brain is changing hahah, at least as much that I know of.

Edit yes I know I'm using the idea of hrtf incorrectly sorry!

 

[iamoneagain]

My last post about ASR. So looks like it’s basically an easy formula for the review. How closely does said headphone follow the Harman curve, low distortion, easy to EQ, and cost. So just look at the DCA E3 review. All the forum member ratings are just based on that formula. Has nothing to do with actually hearing said headphone. So it’s no wonder the Caldera scored poorly.

I also learned the surrogate headphone used in the original Harman study was the Senn HD800. They just EQ’d to several other manufactures EQ settings for the first round of testing. After general curve was agreed upon, they did more testing with bass and treble controls. I wonder how closely the end results matched the stock HD800 frequency response.

But if you wonder why the Harman curve should only be used as one part of your reference and not the final result, look how different the Caldera is from the HD800. Also people were asking why should they even buy a Stealth anymore if the E3 measures the same and is half the price. That’s where Dan had to explain bigger and smoother sound, things the graphs don’t show.

Ok back to enjoying my Caldera. And somehow in all of this, I’m back to using the suede pads and loving it. Ended up sounding the best on those albums the Expanse or Harman curve had trouble with.