-
Studio Grade Headphones on the cheap - A DIY 3D Printable Project.
Posted by JamesLynton on 01 Feb, 2021 14:08 -
This post is a copy of one I made elsewhere on the Alibre CAD forum following some trouble shooting, then I found out they have some competition on there right now, so decided to go all out in documenting the project, rather than just lazily post a few files. This post contains a complete break down of my research and (half cocked) analysis, hopefully this is useful to other people here too.
https://www.alibreforum.com/forum/index.php?threads/studio-grade-headphones-on-the-cheap-a-diy-3d-printable-project.22335/
This post mentions prices in £ and $ - these refer to UK Pounds and US Dollars.
(I will refrain from linking to particular parts listings, as if folks happen to all rush to the same place it might cause them to jack the price up, or run out of stock.)
In Short:
Spend as small an amount of money (and time) as possible and get awesome, robust, repairable studio performance headphones that should have cost you in the region of £300-600/$400-800 off the shelf, for a mere £75/$100 or so.
(and as small as £20-25/$28-35, if you compromise on drivers, and therefore accept certain limitations)
Ooooh, shiney. (also available in discrete plain black, white, pink (?) & so on)
Being +/- 5 dB of baseline is pretty typical of good headphone drivers.
Distortion measured at respective levels in dB SPL as shown on the right.
(Note: limitations of the test microphone also seem to also inflate measured distortion at lower levels)
** Files attached to this post for opening and editing in Alibre, and also the STL files for both default model versions for sending directly to printer without any further editing. **
Assembly Notes:
Just go with modifying your preferred cosmetic choice of PowerA Fusion headphones with aliexpress (etc) imported 'fostex' 50mm (32 ohm version) microfibre drivers installed in the provided 3D printed insert and cotton + pillow stuffing in the rear chamber, if you *just* want some best bang for the buck headphones that will blow your metaphorical socks off, and aren't too bothered about, or interested in digesting several pages of design intent and analysis.
Also I strongly advise removing the internal headphone volume control in the left hand ear cup, bypassing it electrically (its kind of silly, and gets in the way); followed by trimming the plastic supports which held it in place with wire cutters, or your weapon of choice. Then, fill the hole the shaft engaged with with glue to stop the earcup shim from rotating.
For Best Results:
Use a sound source, or dedicated headphone amp which is known to deliver decent power/current to them. The in-built headphone amps in some professional sound cards are pretty decent, and some mixers too.
I tend to use a portable xDuoo - XD05 headphone amp/dac - this is an affordable state of the art XMOS/AKM powered soundcard capable of playing back DSD material (if you have it), and can support 384 kHz at 32 bits (in theory, and quite needlessly excessive (you are a bat, right?)), and most importantly has massive output power (more than you'll need, but the transient dynamics will thank you). Also it can connect to PC/Phone via USB, Optical, or a lossless Bluetooth adaptor. Also, it is a mere £200/$289 for this top shelf product; lesser models, even from xDuoo, are available for around £50-120/$35-155 with various levels of compromise.
You will however, with headphones like these, realise just how irredeemably awful some sound sources/amps are. However, you can expect them to produce (more or less) the best sound any given device is capable of producing.
For sub bass (in particular) that will absolutely rattle your head off, natural dynamics, and clarity up top; be sure to invest in a proper headphone amp, or a decent source.
I like having the option of ramping / experiencing already prominent extremely low bass sometimes, and having the kick drums and bass drones rattle my head; it has a curious way of engaging you with the material, without annoying the neighbours (we're still talking headphones, right ?), distorting the overall tonal balance, or presenting any real risk to your hearing - unlike just turning up everything really, really loud.
Why would you want/need a pair?:
Well designed headphones are the cheapest and most effective route to truly high quality, private listening experiences which negate the need for spending thousands on a studio monitoring grade speaker setup, acoustic room treatment, and sitting in that one optimal sweet spot the entire time to get a similar result. (somewhat) recent advances in driver technology mean that this is both affordable too, exists in the realm of reality (not marketing-land); oh, and fits in your bag, and can go everywhere with you.
Also, the market is absolutely flooded with garbage headphones (even in the professional realm) which fail to reproduce material faithfully, or satisfyingly. The solution from there on out is to either know the good from the bad (whilst not getting sold snake oil by deluded salesmen) and purchase one of the few commercially available, faithful models at considerable expense... Or, build your own using known good parts for close to the materials cost the manufacturer of such pays.
A great test of system dynamics is Eivør Pálsdóttir's album Slør - It is a curious mix of varied bjork'esue (but much better) Keltic/tribal sounding music which mixes acoustic instruments, female vocals, massive drums and pins the lot together on a bed of minimal electronic synthery, and general weirdness.
(She is good, and varied - give her your money)
Find a copy in *uncompressed* format Flac/CD, whatever ...and then compare it to what the version on youtube sounds like, the difference in detail and dynamics is absolutely shocking.
(I prefer the version *not* in English)
Whatever your music tastes, you might find something to like about it, its an interesting album.
Use Responsibly:
With all of the headphone drivers mentioned later, you will find optimal listening levels between 10 o'clock and 12 o'clock on the xDuoo XD-05's volume dial on the first gain switch setting (yes, this headphone amp has a silly name). This translates to a rough median output range (source material dependent) of 60-80 dB SPL at the ear. Which in the background 'silence' of headphones, is perceived as louder than you might think. Generally, its quite safe, particularly if you stick to around that average of 60-80 dB SPL
Official guidance on safe SPLs whilst limiting the possibility of long term hearing damage is as follows.
Permissible exposure times:
82 dB - 16 hours
85 dB - 8 hours
88 dB - 4 hours
91 dB - 1 hours
94 dB - 30 mins
97 dB - 15 mins
100 dB - 7 1/2 mins
106 dB - 3 3/4 mins
(yay, fractions - remember those ?)
For content that isn't heavily compressed, this is my general observation for listening level comfort:
50-55 dB SPL - somewhat quiet.
60-65 dB SPL - normal quiet.
70-75 dB SPL - normal clear.
80-85 dB SPL - getting loud.
By that measure, i assume my personal headphone, and other sound exposure tendencies to be generally quite responsible. Which explains (in part) why despite not having been a teenager for some time now, I can hear roughly 16 khz or so with no apparent attenuation just fine.
All of these headphones have been measured as generally capable of 100-106 dB SPL (although some might burn out quickly), and remarkably some of them handle that with no appreciable increase to distortion (the fostex like 50mm driver for one). In as much, consider using (overall) volume controls responsibly.
Also, consider measuring typical listening levels once in a while with a reputable spl meter (not your iPhone - sorry), if you want be 'sure' accuracy, play a 1 kHz tone with the headphones on something other than your head, and observe the measurement for the meter jammed in the cup. regardless of exposure weighting settings of the meter, this is at a frequency all good meters give zero calibrated attenuation to.
All in, if it sounds uncomfortably loud, or you observe a reduction in hearing threshold after use, consider your life choices.
Curiosity Note 1: The entire dynamic range of a CD / typical (universal baseline) 16 bit audio file is a 'mere' 96 dB - go figure on how that factors into all such relating considerations
As for film soundtracks, its higher at 24 bits, at a theoretical 144 dB - for heavily impactful transient reproductions of key scenes at high levels when played at 'full' volume. Given that most 24 bit or more DACs can actually often struggle to keep the noise floor in control below 20 bits - a clean 120 dB; theres plenty of interesting audio engineering discussion rabbit holes to disappear down.
Curiosity Note 1B: The total useful dynamic range of human hearing is around 120dB, the threshold of pain is around the same. As for perception of quieter sounds against louder ones at different frequencies, it is much narrower (and at last count currently actively debated). In as much, the critical perceptibility of distortion is lower than you might expect, depending on the source material, and where in the frequency range it occurs.
Curiosity Note 2: Due to reduced direct sensitivity of the ear to (low) bass (you feel the majority of it at high spls), you can safely tolerate it at much higher levels without need to reduce exposure. Quite why clubs and bars 'insist' on loud consisting of the distorted mid-high mess comparable to the mating calls of serenading angle grinders is quite beyond me (its not fun, or healthy frankly - and as for the poor (bar) staff, well).
Note 3: For more information of this sort (if you love details, and whatnot) and my working method, see the final sections of this post.
** PART 2 of 7 Follows.. **
-
** PART 2 **
The Details - (the meaty stuff):
Project Background and Design Considerations:
This project started late last year when I became aware of high quality headphone drivers finding their way onto aliexpress (kind of a Chinese amazon for industrial parts) for sensible, and sometimes amazing prices. In relation to this, I was also looking for an affordable, tasteful looking, comfortable and high performance way to house these drivers.
Having owned and used various high end studio headphones over the years and found most lacking in certain areas - comfort, durability, sometimes aspects of sound reproduction and all too often charging far too much; the motivation was taken to make something that could enable housing of such drivers cheap, easy, and facilitate future repair - such that in cases of damage or loss its never a huge amount of money or time down the drain. These headphones would be designed with firm knowledge of acoustics and measurements of results applied, alongside listening to the results, of course.
The results I got in the end I greatly prefer to my previously owned Denon AH-D2000's (made under an oem contract by Fostex), so much so; that I sold them.
Initially it was assessed whether to do a fully 3d printed version. This was ruled out due to a need to process and form some metal parts (for headbands, etc), and some fairly tricky leather/textile work. Given that (otherwise) good quality headphone shells with rubbish drivers in are fairly easy to come by for less than the cost of the diy materials needed (and all the time involved); a commonly available, exceptionally nice model was selected to base this on.
The model selected as the base is the fairly sensible, minimalist looking PowerA Fusion headphone set. These can be found everywhere for around £8-25/$10-35 new and used; and are available in numerous colours to suit your tastes. The basic look of this model is comparable to dj & studio headphones, however they are aimed at the gamer market and have a detachable mic (which you might also find useful). The frame and finish is robust and comfortable to wear, and they feature a fully modular 3.5mm headphone cable; meaning you will never have to worry about accidentally catching or breaking the cable.
It is also possible to buy very nice bare wooden shell headphone shells on aliexpress for around £50/$60, though I don't like the aesthetic and functional design intent of any of them so far; and of course, you'll 'need' to protect them from cosmetic damage - that adds extra cost to the project at no strict, true benefit to sound quality (the core aim).
From there a 3D printable insert designed for it which allows the universal use of almost all headphone drivers (40-50mm - or anything with modification + with, and without preinstalled metal grilles).
This piece also includes acoustic design improvements which share similarities to those seen in the likes of the excellent Fostex and high end Denon headphones. The main improvement over most headphones here (aside of decent drivers) is the inclusion of baffled vents that couple the rear headphone chamber to the ear pads, the benefits of doing so are many:
When properly (low pass) filtered with baffling materials and pads, this results in sub bass the drivers would not otherwise produce being coupled to your head via the ear pads over a larger area than the drivers themselves. It also serves to help reduce noise pollution coming from the headphones with any sound from the driver at normal levels trying to escape the ear pads being cancelled out by the (upper) frequencies coming from the rear chamber there being out of phase (in reverse polarity) with those from the front. Also, with this you get all the benefits of both open backed headphones and closed ones - and none of the disadvantages (its a super munchkin card, okay?). Unlike open backed ones, you get incredible, realistic bass response and great noise isolation from the environment; and unlike typical closed backed ones, you get an open, natural sound due to effective elimination of unwanted chamber resonances and reduced back wave pressure stress on the driver.
Another notable improvement (not seen in Fostex, or any professional headphones i've seen so far) is to tilt the drivers axis slightly off the baffle plane, this takes the driver out of parallel plane relationship to rear chamber surfaces, greatly reducing the chances/effects of strong standing wave resonances those geometries would normally define. Also this serves to better aim sound into the ear and likewise reduces parallel surface issues on the head side as well.
Finally, the other mandatory modification is in adding a modest amount of damping to the rear chamber, this includes one cotton pad on the rear chamber surface, polyfibre stuffing as used in pillows (its basically the same as acoustic fibre which is sold at inflated prices) placed lightly stuffed in the chamber, and another cotton pad cut to shape to fit the rear rim of the rear chamber vents (serves as a slight flow restrictor and low pass filter).
In fact, if any headphones (especially big ones) are lacking a fibre filling, you should add it by default; it reduces/eliminates chamber resonances and slightly increases apparent chamber volume to the driver, therefore slightly deepening bass response for 'free'. See below for how it improves even the stock, unmodified Fusion headphones. It flattens and deepens the bass hump, eliminates the 300-500hz chamber resonance dip, reduces overall HF losses and virtually eliminates some dips in response in the 10-20k presence region. Also, harmonic distortion (THD) is *greatly* reduced throughout the frequency and output range.
Performance scope of the original fusion headphones before replacing drivers:
Whoops forgot to change that dB SPL scale:
(-90)20 dB/30/40/50/60/70/80/90/100/110(0)
If you are feeling super, super cheap - you could just stuff the stock headphones and call it a day, whilst accepting that they have a frequency peak at 8k, a slightly softened presence region (10-20k) and are limited to 70-80db for clean sound reproduction - this limitation also tends to noticeably 'flatten' the sound of music/material with good dynamics (not compressed to death like almost all mainstream pop music from the late 90s onwards). For once, the stock drivers in some cheap headphones weren't actually entirely terrible, but still have the limitations that costing a probable mere $1 to 50 cents brings.
(still, you can gut those drivers with wire cutters, and make amazingly effective fridge magnets with them)
** PART 3 - Coming Next... **
-
** PART 3 **
Choosing and Testing Driver Candidates:
After considerable searching, 3 models of drivers were selected and tested as worthy of consideration.
The first two of these are often listed on aliexpress (and elsewhere) as being Fostex/Foster drivers, or alluding to as such. Whether these are the genuine article or not, is unknown, and I will not assume as such, but the performance is no less impressive (no obvious branding is present on these import drivers - oem?).
Notes: it would be nice to have an official source for the certified original article at a fair price, though I have yet to see one.
The reason for selecting these is the use of a microfibre driver cone, like real speaker cones; is that it imparts time tested, excellent damping properties. It reduces/eliminates driver resonances and harsh bell modes often present in bad/average drivers - this lends itself to an extremely natural sound and incidentally massive power handling capabilities with negligible distortion increase relative to output levels; great for dynamic music/material and for fatigue free listening at 'all' volume levels. Incidentally, they also do bass/sub bass far better than any pure plastic/metal driver i've ever heard, or expect to.
Fostex like 40mm Drivers - 16 ohm Version - MODIFIED Vents:
The first of these, the 'fostex' like 40mm, is the cheapest available at a remarkable £14/$20 or so per a pair delivered. output level is impressive for such a small driver. The sound is natural, clean and handles dynamic material in a revealing manner, which will impress if you are not used to it from other sound sources. The drawbacks are (still excellent) more limited power handling capability compared to the 50mm version.
As this models default configuration is bass light, I open up two of the rear vent holes on the rear flow restrictor (carefully!) with a screwdriver by default. This functions to loosen the manufacturer imposed damping on the driver which normally constrains the bass output of this driver at the cost of overall power handling. I do this such that the overall tone balance of the driver is more natural, and reaches far deeper. The cost of modifying the drivers out of design spec in this case, is an increase in measured harmonic distortion from the bass region; which is remarkably only particularly evident, and measured, when pushing these headphones hard.
If the use case isn't for listening to mega thumpy wumpy bass at rather high levels, and is more talky/general use content focused, then these will probably be just fine for you.
Whoops forgot to change that dB SPL scale:
(-90)20 dB/30/40/50/60/70/80/90/100/110(0)
Fostex like 50mm Drivers - 32 ohm Version:
The second of these is the driver I currently, and somewhat confidently consider the absolute daddy.. The 'fostex' like 50mm comes in at roughly £50/$70 a pair; and includes a preinstalled metal grill. Similar story to the 40mm versions, except absolutely massive power handling, and it is capable of the deepest sub bass i've heard out of headphones yet. Ripe for eqing slightly for preferred sound profile in certain instances.
Distortion barely perceptibly (and measured) increases no matter how much you spank these within levels/durations which will leave you with ears good for listening with.
Note 1: This appears to be the same (unbranded) driver as I had in my Denon Fostex made OEM headphones, and which also features in their line-up - or is at any rate, something that behaves very similarly from personal experience.
Note 2: There is also a 'gold' plated metal version of this driver; I have not tested it, and the seller provided frequency plots aren't too encouraging (soft up top) ..Avoid it (or send me some to test), and go for the black glass reinforced, injection moulded plastic versions instead.
P7 like 40mm Drivers - 24 ohms:
The third is a curious beast, it is allegedly the driver used in the B&W P7 headphones, it comes in at around £34/$45 a pair delivered, it also looks kinda cute; and it is yet another microfiber driver which remarkably manages almost the same sub bass output as the 50mm 'fostex', but seems to run out of steam when pushed really, really hard.
It has a similar tonal profile and lack of distortion throughout most of its range, but is let down by what looks like one final step of driver design polish production didn't go thru. It has a roughly -12db dip at around 3 kHz, and a harmonic distortion level of around 0.6% there; Which then spikes up from the average level, by another +12db or so at around 8 kHz - the prominent 2nd and 3rd harmonic distortion components of 3 kHz are 6 and 9 kHz respectively. when material gives cause to excite these harmonics and peaks the result can sound rather shouty in this particularly sensitive part of the hearing range; I first suspected this when I found listening to a female news broadcaster on them at normal levels somewhat uncomfortable. This is likely the product of unusually strong cone breakup/bell modes resulting from the choice of construction which could of been changed just slightly in the design phase to eliminate/tame it. Unfortunate, as otherwise they are quite excellent drivers. Possible mitigations for these are narrowly notching out the 3 kHz'ish excitation frequency, and moderate eqing; but when the budget fostex like 40mm is dirt cheap, and the 50mm is a mere £15 a pair more - why bother ?
** PART 4 - Coming Next... **
-
** PART 4 **
Mystery Coated Cone 40mm Drivers - (not very good):
Tried these weird looking things for a laugh, on closer inspection on receiving them it looked like someone had taken a 40mm fostex like driver (cost about the same) and sprayed the cone with stone finish plasticote paint. The result sounds absolutely awful. the extra mass on the driver dulls the hf response greatly, and disproportionate increased rigidity of the cone results in some pretty harsh bell harmonics and time smear ringing. If I could be bothered to reinstall these, i'd measure them properly. Short of it is don't buy these, you could of had a decent, cheap meal for the price - or bought the actually decent budget fostex like 40mm drivers for the same expenditure.
Tesla like 45-50mm Drivers - (I would try them):
Seriously considered sampling these - Allegedly Tesla (not the car company) drivers, the magnets are absolutely bfg massive, and powerful - implying possible high sensitivity, fast transient response and ferocious damping. The mottled cone design seems to indicate an effort to break up potential for cone bell modes and therefore permit using heavier plastic without excessive cone breakup distortion. Though the seller provided graphs (they can be totally wrong) appeared mediocre (although key versions of the real deal drivers are actually good); and well, given that these are just a shade more than the 50mm fostex like drivers, I wasn't in a rush to buy them just for testing - however, I will gladly test a pair at someone else's expense, if it interests them.
Outside of this, so far I have not found much of huge interest from budget sources that would seem like it'll be available over the longer term. Most of the basic plastic and metal drivers are kind of mediocre by nature, or more commonly shockingly bad. Though you might find some surprise gems if you go digging.
Bonus Content:
Denon AH-D2000 Teardown:
Heres whats inside the genuine article Denon headphones made by Fostex under an OEM contract back in the day. You can see from here various aspects of the construction.
Kicking myself for not getting current, proper measurements of these before selling them on, so I could post some more concrete measured analysis of how they compared; but oh well.. that opportunity passed.
Update:
Borrowed this from an old headroom post, illustrates the frequency plot measurements they took of the Denon's and some others.
On a personal level, I prefer the results I've been getting and theres still further scope for tweaking/adding to the hardware of the diy ones here later, of course
** PART 5 - Coming Next... **
-
** PART 5 **
Sony MDR-V150 - (for comparison):
Not very good headphones frankly, posting them as a yardstick of what £15 gets you elsewhere. They sound muddy, have poor power handling and are rather uncomfortable. And frankly the 30mm driver they contain is barely fit for the dollar store (imho). The measured results are however quite revealing, and surprising, its a lesson in taking even reliable frequency plots of drivers with a pinch of salt, as whilst they do go 'flat' up to 17 kHz or so, it doesn't sound like it. The reason for this muddiness masking the sound lies elsewhere, in this case the THD vs Frequency plot gives away the likely cause, a massive amount of harmonic distortion below 200-250 Hz *at all listening levels* and particularly bad at normal ones. And when the fundamental bass response below 150 Hz is dropping off like a stone, all of that 'bass' content is getting turned directly into harmonics filling up the mid range (and more) with garbage.
In final conclusion on that one, its also interesting to note how whilst modifying the 40mm fostex like drivers for more bass did result in a rise in distortion which looks superficially similar to this, its not on the same order of magnitude & remains controlled/virtually non existent at normal listening levels.
Note: Graph above also shows the original Fusion driver plot (they are way better than these things
** PART 6 - Coming Next... **
-
** PART 6 **
Bonus content - round 2:
All Drivers Comparison:
Not the easiest to read, but here it is.
Whoops forgot to change that dB SPL scale:
(-90)20 dB/30/40/50/60/70/80/90/100/110(0)
Now all drivers compared at respective output levels side by side:
Typical normal 'loud':
Actually loud, and irresponsibly so:
For responsible listeners, this is most useful in showing if the drivers run out of steam and colour/compress the sound of transients in good recordings when played at normal levels.
Good performance here indicates spacious, dynamic, natural sound in general use.
Discussing Distortion:
A certain amount of distortion actually adds warmth or presence to material.
When it happens in the bass range, it can add warmth/punch to the bass - too much and it smears and confuses the midrange. When it happens with mid range, it is less tolerable due to how sensitive the ear is there, and can result in vocals sounding shouty; in the high mid range, the extra harmonics will find themselves up in the presence region of 10k+, which as the name implies increases perceived presence - sometimes this can be pleasant.
Note: With low bass however (90 Hz and below'ish) the ear is dramatically less offended by distortion, it gets perceived as bass punch instead, so you can get away with far more there. 30 Hz has 2nd and 3rd harmonics at 60 and 90, 50 Hz is 100 and 150, and 90 is 180 and 270. All more or less in the kick region. 1-2% percent there is quite common, and is often perceived as just warmth. 10% though, unless thats a guitar pedal, you don't want that
When discussing the perception of distortion, its interesting to translate THD to the decibel difference between the Fundamental signal and the distortion elements and look at it that way. A conversion chart to that effect follows below:
THD to dB:
0.001 = -100 dB
0.01 = -80 dB
0.1 = -60 dB
0.2 = -54 dB
0.3 = -50 dB
0.4 = -48 dB
0.5 = -46 dB
0.6 = -44 dB
0.7 = -43 dB
0.8 = -42 dB
0.9 = -41 dB
1.0 = -40 dB
2.0 = -34 dB
3.0 = -30 dB
5.0 = -26 dB
10 = -20 dB
15 = -26 dB
30 = -10 dB
100 = 0 dB
Given that typical 16 bit CD spec audio files have a dynamic range of 96dB, and that you'll likely listen to that at 80 to 90 dB max; then accounting for the presence of background noise in an actually quiet room, even with headphones on being 25-10dB, you are then looking at 55-70 dB of dynamic range space which within you are hearing your audio material. This means that at those listening levels you'd be hard done to hear 0.1 to 0.2% distortion above the noise floor at peak ear sensitivity at all, and for the most part all of the good drivers keep their distortion well below this target point at those levels.
All in this means most of the drivers mentioned here get '5 stars' for keeping mid-high distortion below the theoretical perception limit at 0.1 to 0.2%, and low bass distortion below 1-2% (added warmth rather than mud)
Noise floor in test room vs noise floor of soundcard with no mic attached:
Ambient noise floor in room is 25-30db below 200hz, and around 10-5db in the midrange and up.
Pretty decent, all in.
Time domain smear:
Outside of that though, theres another often not discussed or considered distortion, in the time domain. Think of a cymbal, speaker and headphone designers spend a lot of time and effort trying to prevent them behaving like one when 'excited' ..if the speaker keeps ringing after its been 'told' to stop, then this smears the sounds presence out thru time muddying things up. A little is fine, however.
One example of 'good' / tolerable time smear is with tweeters, think of how cymbals sound on different speakers ..a little smear adds presence and sparkle to them, particularly when a peppering of cone breakup generates extra harmonics in that region too. Only thing is, it really throws you when a speaker/headphones don't do that.. and you might mistake its absence for less response up top - even though measurements will show you otherwise. Perception of sound is strange like that.
No driver is (clinically) perfect - but they can be downright amazing:
No driver is perfect, but if the design engineers have done a good job, they will have those compromise tolerances within the range of adding a pleasant, mild colouration where it does occur, rather than an offensive one. and with such drivers, all thats left to do is add a tiny bit of eq to bring the tonal presentation into your preferred range.
Theres many other potential talking points like this, however I consider these the most important to consider.
** Part 7 next...**
-
** PART 7 **
Testing Methodology:
Equipment Used:
* MOTU - 96KHz Ultralite mk2 Firewire soundcard - (affordable used, and seriously robust)
* Behringer - ECM8000 calibration microphone - (been around forever, cheap, and work well)
* xDuoo - XD-05 Headphone Amp - (Optimal for testing all headphone loads I'm likely to encounter)
* Ymec's - DSSF3 audio test suite - (I keep trying others, though this still works better than all of them)
* An AZ8929 SPL meter.
* A diy foam jig for testing headphones.
* Anycubic - Photon Mono SE - (for printing the mod parts)
* A ready supply of coffee.
Measuring equipment characteristics:
MOTU Soundcard Noise floor with nothing attached:
Pretty decent, that.
All frequency plot measurements taken at 96 khz / 24 bit and plotted out to 50 Khz so extended 20 kHz+ response could be seen too.
Turns out the measurement mic extends out to at least 40 kHz just fine, though not certain of its accuracy there.
MOTU Soundcard THD vs Frequency with a loopback cable:
Measuring THD vs Frequency:
Distortion level across the measurement range is sufficiently low to not influence achievable distortion levels real drivers are capable of. All THD measurements taken at 96 khz / 24 bit so that distortion elements at 10-20k could be reliably tracked.
MOTU Soundcard THD vs Output level with a loopback cable:
A more challenging test, showing particularly how ADCs and DACs struggle at the bottom end of their range, and very mild saturation sets in at the top end.
Measuring THD vs SPL:
Due to this, when measuring distortion below -60dB on the soundcard (50 dB at the mic) the measurement clearly becomes invalid, as ADC/DAC distortion is starting to approach influencing the distortion by as much as that of the device under test. That said, at levels that quiet its quite likely that small signal non-linearities in the measurement mic, and speaker are also setting in at a similar rate. However, it doesn't matter that much, as you'd be hard pressed hear those distortion elements above the noise floor of the room.
(1% at 40 dB SPL is -40dB down for the distortion elements, and therefore more or less cancels out
Measuring Frequency Response:
Frequency response is therefore measured at -20 dB out from the soundcard (90 dB at the headphone driver) ..as this is in the sweet spot where sound card amplitude distortion is lowest, distortion is only just starting to creep back up at the driver (within drivers linear range), and it correlates as a loudness typical of of maximum expected use.
Also aligning the output level of the headphone amp as such (gain structure) insures a full realistic and useable driver output of 110dB, whilst allowing a 120db peak measurement headroom (by setting the mic channel gain to +10) ..also this keeps all parts of the equation in the sweet spot for making reliable measurements within the range of typical use volumes (and just outside of them too).
Calibrating the measurement software for SPL at mic:
Done by jamming SPL meter in the headphones cup whilst on the test rig and using its calibration feature to match the measurements seen (at -20dB / 90 dB SPL at the mic at 1 kHz) seen on the meter.
Headphone Test Jig:
To save poor ear drums and arms a 4 inch block of foam, and two books behind it, to support and isolate the second driver from the measurement was selected. this was found to be equivalent to a head measurement when the mic was jammed thru a hole in the middle of the foam, such that it faced the headphone face on in the middle. The foam acts almost like an infinite baffle, preventing any reflection modes, in addition a small doughnut of the aforementioned pillow stuffing was placed in the headphone cup - this prevents reflections within the empty cavity from forming strongly there either - and also served to maintain correct bass response in the chamber. I'd photograph this, however it was so quickly thrown together to get the job done, i'd rather not
Setting the measurement point:
This was done in fast FFT mode, with the headphones gently moved around a tiny amount in the centre origin. Despite best efforts, cancellation modes from reflections and hf driver beaming, the microphone can in close quarters get an inaccurate read. when it was in the right spot to avoid those remaining peaks, the measurements commenced. The other option was to take dozens of measurements all over the axis shop and add them up for an average; thats silly time consuming and has its own problems. so, I went the quick way about it.
heres what happens when you measure off centre intentionally on such a hunt, it gets messy !
Also, wear those headphones properly, otherwise you'll likely be dramatically outside of the drivers central sweet spot
Also, see attachments for the open office / libre office spreadsheets used to crunch and combine some of this data.
It was an absolute mare to get it half way working with logarithmic graphs in a halfway useable manner, but got close enough in the end
Printing Equipment Used:
Anycubic Photon Mono SE resin printer and their resin - its affordable and fast curing stuff.
Affordable kit and amazing performance, recommended.
Alibre Keyshot Renders:
Think that is everything, thanks for reading along and have fun with this information, and this project
-
Part Design & Test Files - Part 2/5:
Rename to: Headphone_project_files.z01
(spanned zip archive)
-
Part Design & Test Files - Part 3/5:
Rename to: Headphone_project_files.z02
(spanned zip archive)
-
Part Design & Test Files - Part 4/5:
Rename to: Headphone_project_files.z03
(spanned zip archive)
-
Part Design & Test Files - Part 5/5:
Rename to: Headphone_project_files.z04
(spanned zip archive)
DONE
Now, anyone know why inline images from uploads is not working ? i'd rather not rely on external hosts to preserve the work that went into this project.
Ta
-
This looks like a rad project. I'm excited to check it out. Can you provide any more information on the AD_PKG file?
Based on the link you posted to the other forums I'm guessing it's for "Alibre" (I've never heard of this before so it's brand new to me). Reading it as "A...Libre" I was stoked to hear about a new "libre" cad tool! Alas though it seems like it's proprietary and solely for Windows, thus any information you can provide on how others open/use/convert the file to other CAD formats would be super!
Regardless, thanks SO much for also showing off the noise + ThD stats too. Makes me even more excited to kick the tires and explore. -
Thanks.
Plenty of cad apps can edit/work with the STL mesh files provided for 3D printing, and also many apps used for preparing models for 3d printing. so theres some latitude there for editing them in that most basic form.
That said I can also export to Solidworks from Alibre, and some other formats as well if folks need it.
Went for Alibre here as its pretty much Solidworks (but actually affordable, perpetually owned, and *not* some cloud or subscription program thing) ..Also looked at Siemens Solid Edge (seriously nice software), and its still far too much money for what it does; and well, I really don't trust Autodesk and the direction things like fusion 360 are going in. The problem in the CAD industry atm is none of the long term big players are interested in the literal need for serious CAD from hobbyists, and small businesses with tight budgets. -
Interesting project, but I'm not entirely convinced that this midrange dip at 900Hz is a good idea
There is way more to enclosure design than just mechanically fitting something around the driver. I have no clue about it, by the way
But if you look at FR plots of commercial $400+ headpones (which you compare yourself with), they are significantly flatter. -
the first dip is at 500-600 hz, this is a standard function of cone breakup inherent in drivers relative to their dimensions, mass and material properties. variations of a few db due to such are pretty standard and not of consequence. getting a single driver to stretch out cleanly all the way from 20hz to more than 20,000hz is quite a feat of engineering/materials science ..and certainly, its not something that single driver, full range speakers working at higher output volumes do particularly well. the lower apparent dip (we're talking a few db here) in that region is could be a function of their means of measurement, as much as other factors of colouration boosting the apparent measurement - what i can say for sure is they didn't measure using exactly the same means as i did, so its not an apples for apples comparison on the fine details
In any case, three things.. one i'm more concerned about spikey, irregular responses/drops further up, that messes things up in an area where you notice it far more; and factors like distortion/ringing have far more impact on the sound (and are less passively correctable). the base design of this mount was done such that it is in the universal sweet spot to work well with most drivers. there is latitude to tailor and tune it (and things like rear cavity volume and stuffing) to bring specific drivers even further into their sweet spots.
What I can say, is the old Denon d2000's plainly didn't sound as nice or balanced, there was noticeable excessive mud in the lower mid-range from the design choices made, and a duller response up top. Still, for what they were, they sounded quite remarkable against the competition of the time
All in, consider this revision 1 of this project, its a solid base to build minor refinements on top of
Also, for reference, here is the response curve of one of the better designed compact 'full' range (no sub/low bass of course) loud speakers on the market faital pro's 3" 3fe22 ..which is of similar dimensions (but designed to run significantly louder of course - this illustrates again the typical minor baseline response curve variations inherent in well designed sound transducers, and particularly in the best available (stretching it to the limit) full range drivers. 
https://faitalpro.com/en/products/LF_Loudspeakers/product_details/index.php?id=401000100
Regards.
-
I encourage you to play with an EQ plugin in your media player before stating that something is of no consequence
A huge reference compilation of measurements of commercial headphones from all price brackets is available here:
https://www.stereophile.com/content/innerfidelity-headphone-measurements -
I know all about that site, and yes.. flattening the response like a ruler remains an option (doing so excessively can mess up things in the time domain instead though - impacting stereo imaging and transients)
In any case, this is a solid starting point for packaging up quality headphone drivers affordably and robustly into something considerably more than just useable; I encourage you and others to take this open sourced design and improve on it; particularly when dialling in the last few drops out of a specific driver is in mind.
Thanks, and I look forward to your future contributions to tuning the base design of these headphones
Having thought on it some more with regards to tuning the last drops in for the 50mm fostex specifically, increasing the apparent chamber volume thru packing the stuffing more heavily (which also further improves hf response via extra damping of cancellation/summation dips there) will probably do it - that will broaden, and flatten the first part of the response curve. Or if needed, the actual chamber volume will be increased slightly. I'll v1.1 the details for that one when I have more time to play around again, so probably in a few weeks
-
Big thankyou for sharing this here. I would never have found it on the other forums and I can't view any of the pictures there without registering.
I'm now contemplating making my own, along with some cobbled together rough hardware EQ. I should probably do the latter first, it would make my existing headphones a lot nicer to use for extended periods. -
And now for an affordable, high performance headphone amplifier to go along with this project:
A £8 Chinese special headphone amp board loaded with Texas Instrument's TPA6120A2 current feedback topology amplifier chip arrived today. This thing absolutely spanks it out, as clear as the xDuoo XD-05 headphone amp I have; However it handles that low sub bass even better.
..Also, yes ..you can really tell when the DACs in your source device are garbage with this - In as much, invest in a good soundcard/dac board, and this setup will sing.
(a number of good quality boards with AKM's AK4493, TI's PCM5102a, ESS's ES9038/28 DACs on for example, are available cheaply as diy boards)
Threw in a tiny £4 +/- 15v switching converter and the whole lot runs off USB, quite capable of running off a single lithium ion cell.
Whats more, the converter does not introduce any noise of its own into the signal path, as the power supply decoupling involved throughout is quite adequate.
TPA6120A2 Documentation:
https://www.ti.com/lit/ds/symlink/tpa6120a2.pdf
-
Excellent thread! You inspired me to try a DIY headphone project.
There are many headphone amplifiers with the TPA6120A2 chip. However, I am confused. Some with identical PCB layouts are listed as VAC while others are VDC. For example, the listing for this one (https://www.aliexpress.com/i/2251832821951317.html?gatewayAdapt=4itemAdapt) is VDC (12 & 24 VDC), but the board clearly states VAC. It seems identical to the one in your photo. As you mention you are using USB for power, I assume that the amplifier has VDC input. Can you clarify? I am looking for one of these amplifier variants for portable use with battery power. -
A really interesting thread, thank you, as the owner of a pair of BOSE Quietcomfort headphones and a recently replaced set of RODE NT-100 headphones I look forward to taking some time to read all that you've done.
-
For anyone interested, apparently these amplifiers accept VDC or VAC power sources. Based on the OP's recommendation, I have ordered a couple.