DIY Headphone Amp - part 2

Rroff

Rroff

Rroff

Man of Honour
Joined
13 Oct 2006
Posts
94,846
Since building a "final" headphone amplifier which I was largely happy with I've kept dabbling with the subject as its an enjoyable hobby I can easily dip in and out of as time allows (and kill time researching stuff almost anywhere that I have a smartphone/tablet and an internet connection).

So I built a somewhat modular prototype that I could easily add and remove different components to see what the effect is.

Power supply has dual headers for split power sources which can be upto 2x DCV01 type chips or a virtual ground splitter.

Signal input has a header for trying different dc blocking capacitors

The gain and 2x buffer op amps are socketed and additionally have headers so I can either plug in components i.e. bypass capacitors or anything else circuitry wise around the op amps themselves or to easily connect adaptors for op amps that don't come in DIP8 form.

I9RKLmr.jpg


I had a couple of goals in mind with this:

(1) I wanted to try out the DCV010512D as a power source - this gives a nicely isolated 0v ground as well as handily producing positive and negative 12v rails from any ~5v source i.e. USB - which means no worries like with a virtual ground where you can accidentally end up frying your audio equipment due to the ground being at half supply voltage instead of 0v - however due to the chip using basically charge pumps which are looked down on in audio I was interested to see how effectively that could be filtered out.

(2) I've never been entirely convinced by the combination of buffer op amp I'd settled on (JRC4556) and the HD-600 headphones combination which is also the buffer commonly used in the O2 headphone amplifier - and I'd noticed a few posts online since questioning the combination which got me interested in seeing if there was something that sounded better.
 
Last edited:
Regarding the DCV010512D:

To filter the power from this I setup an LC filter with a 100uh inductor followed by 0.1uf ceramic capacitor (I'm not really well versed in LC filters so these could be bad values or improved on) followed by LM317TG/337TG regulators with additional ripple suppression which finally outputs into a bank of 470uf Panasonic FM capacitors and Nichicon FG closer to the op amps. This seems largely successful from my rather basic abilities to test - even from a noisy USB connection anything getting through is below what my equipment can test and not noticeable audibly in the headphones to my ears.
 
While I was at it I decided to (subjectively) test some of the film capacitors I have on a level playing field - these block DC from the audio source to protect the headphones.

For this test I left in the OPA2132 gain op amp and JRC4556 buffers that I'd put in for testing the circuit was operational:

Jellyfish Audio metallised polyester 0.22uf (yellow)

Fairly generic capacitor - compared to the others it doesn't quite have as good bass but otherwise largely similar - fairly good with vocals but higher frequency instruments can be a touch grainy.

Panasonic ECWF2474JAQ 0.47uf (small dark red/orange)

Compact size which can be useful - a touch lean bass wise and sometimes treble can be a little grainy but otherwise works fairly well.

Cornell Dubilier 930C1P47K-F 0.47uf (white)

To my ears it sounds nice and neutral throughout the frequency range - compared to the 2 above there is more punch and separation to the sound (this is probably due to the capacitor having less impact on the sound quality than it doing anything extra to the sound).

Nichicon QXP2J474KRPT 0.47uf (big orange/red ones connected on the amp)

Like the CD 930 there is better separation and punch to the sound than the first 2 - not quite as punchy as the CD 930 but slightly better bass extension.
 
Last edited:
For the next step I intend to take a look at the follow opamps used as the output buffer while using an OPA2227 as the gain opamp:

JRC4556 - as used in a few mainstream amps and the O2 due to its current capabilities
LM4562 - recently become popular in a lot of soundcards
AD8066 - one of my personal favourites - a little lacking on current drive but good slew rates and other factors relevant to driving headphones
LMH6643 - (this one doesn't work at the voltage range the amp is configured for so will require additional circuitry)

If anyone has any suggestions for others to try I'd be interested.

Once I've found a buffer IC I like I intend to try these in the gain stage:

OPA2132/2134
OPA2227
OPA2228
AD8066 (will also be testing this one with the AD8066 used as buffers due to some specifics related to its bass performance)
OPA2211 (the OPA1612 often found in higher end consumer devices is the "audio" version of this op amp but some claim the 2211 has more spacious sound).
OPA2140
LM4562

The ultimate goal to find something I like the best with the HD-600s, but also going to explore the AD8066 options a bit more as it has some interesting properties in regards to bass.
 
Last edited:
Thing that has interested me the most so far - swapping my headphones between this with the 2132/4556 combination and the previous amp I built with with 2228/4556 combination - I specifically choose the 2228 due to its "mellow sound" but female vocals are night and day different - on the 2132 so sterile on the 2228 lush, smooth and pleasing to the ears - given how popular the 2132/2134 was/is in mid-range audio equipment I'd imagine it is responsible for driving a lot of people to tube amps heh.

Will have to try the 2228 in this project to see if its more than just the op amp affecting the sound as the power supply setup is very different between the 2.
 
Updated:

Testing (subjective listening) some of the options for a buffer op amp with the OPA2227 in the gain stage:

JRC4556 - compared to the LM4562 they sound "slower" and really don't get the HD-600s moving like the 4562.

LM4562 - not getting a "this is the one" feel but faster, punchier and a lot more open sounding than the 4556 - grown on me a bit after awhile - the combination of 2227 + 4562 sounds awesome with some kinds of music like hardstyle.

AD8066 - bass has more thump while still sounding controlled compared to the 2 above but lacks the "openness" of the 4562 (some call it veiled) - the veiled nature is really apparent swapping straight between it and the 4562.

Out of the options tested so far the 4562 is winning by a mile the only thing that gives me any pause for thought is the bass on the 8066, however the 4556 has significant current driving capabilities the other 2 lack which isn't a problem with the 600s but could be cutting it close for more demanding IEMs, etc.

Would be interested if anyone has any suggestions of other op amps to try as the buffers.


As an aside swapping the OPA2132 out of the gain stage for the OPA2227 in this test straight off I notice a decent increase in articulation and detail over the OPA2132 which I expected anyhow but if anyone is considering making their own CMoy I'd highly recommend considering it over the 2132/2134.
 
Last edited:
Started to compare different op amps in the gain stage while using the LM4562 as buffer.

OPA2132 - meh while its largely transparent and accurate I have no idea why its used other than being extremely tolerant of poor/basic circuitry. It sounds sterile like someone playing the same tune for the 1000th time - accurate enough but no passion to it.

OPA2227 - similar sound to the 2132 but more life, more detail, thicker bass, treble is inoffensive at best - huge upgrade to the 2132 while still being almost as tolerant on the circuitry around it (and handles noise, etc. better).

OPA2228 - overall similar to the 2227 - same thick solid state bass and treble is nothing to speak of (not good or bad) but the mid-range and vocals are smoother, mellower, while still sounding detailed - low fatiguing sound with really lovely vocals that sound closer to tube than solid state. Got a feeling I'm not going to find something to top this for my preference in sound. (Seems slightly more susceptible to RF/EM interference than 2227 for some reason).

OPA2211 - this one is good - not had a lot of time with it but more than any other op amp I've tried so far you get a real sense of position. (EDIT: 2211 is awesome IMO while it doesn't have the "tonality" of the 2228 it has a sharp, clean, uplifting sound and you can really hear where the instruments and vocals are coming from in properly made music which is somewhat lost on the op amps above).

LM4562 - crisp and punchy but in the gain stage the overall sound is a bit cold and metallic at times. (I've read that can be due to HF noise which "could" be an issue with the power supply).

OPA2140 - bit of a wildcard as I've not seen it used much in audio but heard 1-2 mentions and wanted to check it out - has a darker, "bassy" sound, reasonable detail, fairly good positional reproduction. I actually really like this one - has probably the best sounding piano of any op amp I've heard but a bit limited in what music it does and doesn't sound good with. (EDIT: A good contender with the 8066 as while not quite as detailed bass it has the same kind of 3D deep thump to the bass without being so noticeably veiled).
 
Last edited:
While not very scientific but I only intend it as my subjective ramblings anyhow the results so far aren't really that unexpected.

The OPA2211 + LM4562 combination is the closest to far to how I expect the music to sound - which given that the "audio" version of the OPA2211 (OPA1612) is popular in higher end consumer audio and the LM4562 underpins TI's high end audio op amps (LME range) hardly shocking.

Nothing else I've tried has the tone of the OPA2228 in the gain stage but I think the 2211 wins overall but sometimes its nice to have that mellow sound of the 2228.

I think the problem with the 4556 and HD-600 possibly comes down to the slew rates - the HD-600 aren't particularly current demanding (which is the strength of the 4556) but do like lots and lots of voltage gain and the ability to swing that efficiently - while 1.4-1.6V/µs is generally considered as much as you need for audio and the 4556 is rated for 3V/µs typical my guess is that either it doesn't actually attain that or given its low price there is a huge variation in what each example of the IC will do - regardless compared to the 4562 it sounds a little smeary (for want of a better way to describe it) driving the HD-600s and I couldn't recommend the combination now based on my experience.

EDIT: Removed stuff about the OPA2140 as there might be another reason for the issue I heard (loose connection) - will have to retest.
 
Last edited:
In case anyone saw my earlier comments on the OPA2140 - when I came to taking it apart after I was done for the day (the 2140 was the last I tested) I noticed one of the power connections had worked a bit loose from swapping op amps in and out multiple times - retesting today after fixing that I'm not noticing the issues I thought the OPA2140 might have had (still has a fairly hefty on/off thump compared to the other op amps though).

EDIT: Do really like the sound on the 2140 - gorgeous dark, gloomy texture with deep "3D" bass but struggles a lot with complex treble and bass at the same time with a bit of mushiness creeping into the high frequencies.
 
Last edited:
Was talking to someone who happened to be something of an expert on the subject... was quite critical of the way I was using the AD8066 and strongly recommended actually using some old school techniques with bypassing the op amp, etc. for the AD8066 specifically.

ZAR09r2.jpg

(Near the middle of the image 3 capacitors soldered directly to the breakout board for the small IC).

Completely transforms it - treble veil is largely lifted.

He is also of the opinion that driving headphones like the HD-600s with anything less than 10V/us slew rates in the output stage is probably less than ideal.

EDIT: Actually quite a stunning difference treble now soars where before it was quite mundane - seems to have come with a bit of an increase in dc offset but still in the order of lower 10s of mV so not significantly concerning.
 
Last edited:
So I did the rail to rail bypass on the LM4562s I've seen a couple of people rave about on head-fi and sat back stunned - I don't think I've ever heard any setup that sounds as resolving - the sound is an order of magnitude more controlled and punchy just by adding a good quality 10uf capacitor over the power rails (I've already good fairly good rail to ground decoupling).
 
hah :( I'm actually blown away by how good this 8066/4562 setup sounds (with some input from people who clearly know far more than me) - but it isn't the sound. The 2211/4562 combination sounds bad with nothing and really good overall but also isn't grabbing me.

Really I want the mid-range of the 2228 (just has really great tone), treble of the 2211 (its just lacking "tonality" overall even though it sounds good but has a nice open/airy treble) and bass of the 8066 lol.


EDIT: Seen some very interesting comments (posted a couple of days ago) by the guy who designed the new OPA1622 which looks like potentially a great headphone buffer op amp but sadly not in a DIY friendly package:

"I was very adamant that I didn't want to create a part with excellent specs but terrible sound quality in the real world. I did not want to create just another op amp with mediocre linearity hidden under 140dB of loop gain. Or one with 100V/us slew rate achieved with a crude slew-boost circuit (not all slew boosting circuits are crude though). These sins tend to be exposed in demanding applications."

The OPA1622 coincidentally is also a (true) 10V/us slew rate device - further puts the nail in the 4556 for me as its specifically designed to work optimally with demanding headphones and the slew rates weren't just by accident.
 
Last edited:
Got some LM6172s in the post - even as buffers they aren't the easiest to get stable but have a good mix of ability to swing lots of voltage and current output so going to atleast have a play and see how they perform/sound.
 
LM6172 in - one nice feature - they seem to suppress any power on/off transients. Sound is "warmer" and a little slower than the LM4562s (I suspect I don't have them quite 100% stable) - vocals are awesome - lush and intimate don't quite stand out from the rest of the sound as well as the 4562s but benefit from being less cold sounding.

Slight weird one - one channel I've always had a little more dc offset than the other with - with these in that channel is pushed right upto the most I'm comfortable with -25mv while the other channel which was always good is even better 0.1mv.

EDIT: Not really gelling with this one - seems to have a slightly off colour not dissimilar to the 4556 in higher pitch vocals, etc. and overall not removed enough from the slightly smeary sound I get from the 4556s (might be due to not having it fully stable as its not a chip you can just drop into any old circuit - its designed for ADSL and other high frequency applications).

A complete guess but in light of the comment about the 1622 the 6172 might be using some kind of slew rate boosting circuit to get its ridiculous 3000V/us which would be perfectly acceptable for stuff like ADSL but impacts on the sonic performance.
 
Last edited:
Following on from my other post - did a rapid prototype of the O2 topology with a different power supply design and some slight changes to accommodate an AD8066 gain and LM4562 output op amps just to see how it performed/sounded:

N98atOa.jpg


Had some problems as it turned out one of the regulators was faulty so some nasty bodge was done to bypass the power supply with a working one from another project:

W8ualxM.jpg

Unfortunately while this sorted the serious issues I first encountered with the gain required to stabilise the AD8066 it was still very easy to drive the amp into distortion if you didn't pay attention to the input levels - when under carefully controlled conditions it did show just how brilliant the original O2 design is and prototype board doesn't come close to doing it justice. I suspect the changes required to make it play nice with that combination would undo much of the advantages it has over the more traditional topology.
 
Last edited:
This looks awesome. Anyone got any suggestion on a learning kit for soldering and basic electronics the ones on ebay seem a bit sparse.
 
TBH jumped in the deep end myself so not something I have much experience of - for learning the sunfounder kits look fairly decent for the money though largely orientated around the Arduino platform (which could be quite useful to have experience of).

I started with a generic 9 piece or whatever soldering kit off Amazon but ended up fairly quickly getting myself a temperature controlled workstation as the basic irons aren't entirely ideal for more serious use.
 
You must log in or register to reply here.
Back
Top Bottom