The JC2-HPA Headphone Amplifier

I stumble upon a headphone amplifier kit on Aliexpress, quite by accident. It’s based on John Curl’s JC-2 line stage circuit with an output buffer added, it costs peanuts, and it’s so attractive that keilau asked me to create its own dedicated thread. I duplicate some of the previously written stuff here to avoid all the jumping back and forth.

The original Mark Levinson JC-2 line stage is made of a complementary differential FET input stage followed by a bipolar voltage gain stage (VAS). John Curl (JC) published the schematics in the Audio Amateur in the late 70’s, but it’s not really for the DIY-ers/home builders because it requires matched unobtanium FETs. And this is the main event as far as this kit is concerned: it comes with FETs matched to… 1% in my case!

The catch? These are not the much praised 2SJ74/2SK170, but the 2SJ103/2SK246: they’re noisier, but good enough for line-level duty. I must admit the quality of the matching blew me away, so I ordered another kit to double check and as a spare.

First I was a bit reluctant to start this thread now before checking the second unit, but EUVL assured me the J103/K246 are still plentiful in China, and Weiliang, the supplier, is known for its quality work.

This will be a 4-part article, a work in progress, and a journey. What I have in mind is:
  1. Abstract (this here one)
  2. A quick start guide: what/where to buy, for those who know their stuff, minimum verbiage
  3. A bit more details on what to do, as the original kit comes with no instruction
  4. The tweaks
Disclaimer: I’m not related to any of the sellers, no way.
 
JC2-HPA: Quick Start

The kit:
My 1st kit comes with a main PCB, all parts including a ¼” jack, 4 heatsinks, metric standoffs & screws, but no potentiometer. The schematics are in the links
  • The seller of my first kit is no longer in business
  • This is where I ordered my second kit
  • If worst comes to worst, do a search with E4, fet, headphone amplifier, etc.

The transformer:
  • 2x15V 30VA if you want to run the kit without any mods
  • 2x18V 30VA if you want to add a pre-regulator: more later
  • My choice (220V!)

The box:
There’s a dedicated aluminum box, pre-drilled and silk-screened. It comes with a fused IEC connector, a power switch, a volume knob, stick on feet, and some metric screws:
Misc. parts:
You’ll need:
  • A potentiometer (maybe)
  • 2 RCA jacks
  • Wires

Also included here, the current version of my simulations, along with the supporting files; just put them in their respective folders:

C:\Program Files\LTC\LTspiceXVII\lib\sym for the .asy
C:\Program Files\LTC\LTspiceXVII\lib\sub for the .lib & .sub and you should be good.

This asc has most of the tweaks I’ll get into in a bit, including a DC servo that may or may not make it to the physical amp. I couldn’t find models for the 2SJ407/2SK2381, so I use the IRF710/9610 instead; these are also those which will probably find their way into the physical amp, at least in its V1, but I’m way ahead of myself.

Mind you, the frequency domain is not optimized yet, so the poles may need more tweaking.

If someone knows where the “.tran 0 {simtime} {dlytime} {timestep}” magic comes from, please let me know, I’d like to thank the guy.

More later
 

Attachments

  • JC2-HPA-210f-2-5-100-servo-OK2++.asc
    10.4 KB · Views: 317
  • Zung.zip
    9.7 KB · Views: 217
Last edited:
JC2-HPA: a bit more detail

OK, I never said there’s gonna be a step-by-step guide; just a bit more detail, because the various bags that come with the kit can be confusing.

Fets
They come in a plastic bag, with a hand written label that says 10.22 in my case, whatever that means. And so I tested them, using EUVL’s jig, Idss at 9V and 25°C:
  • 2SK246: 9.48, 9.42, 9.42, 9.50mA
  • 2SJ103: 9.28, 9.38, 9.32, 9.29mA
Even though my calculus is a bit rusty, it looks like a 1% match!
The 10.22 is likely 10.22mA Idss when tested in Siberia (negative tempco)
But there’s a catch – more about it in the tweak section, stay tuned

BJTs
They come in 2 separate bags:
  • 1 with the “helper” parts for biasing: 4 x C546B (=BC546B)
  • 1 with the active VAS parts: 2 x C546B and 2 x C556A (?)
That’s strange, and the only goof I’ve found so far: they took the trouble of precision matching (to the last digit) the hfe of the same sex BJTs: N-ch 289, P-ch 184 😊. Of course, matching the opposite sex would have been much more profitable!
Not too critical for me, I’ve got loads of small signal BJT's around – more about it in the tweak section

MOSfet
The kit comes with 2 pairs of 2SJ407/2SK2381 output buffer. They’ll work, but there are better options in the tweak section.

Rectifiers
There’re 4 BYV26E Ultra-Fast Avalanche Sinterglass Diode for the main supply, and 3 1N4007 for the protection. These will do unless you want to do the regulator tweak.

Resistors
1/4W 1% 5-ring resistors. Interestingly, they’re beige instead of light blue as is customary for China. I spot checked a few for accuracy, and I’d say they’ll do.

Electrolytics
2 x 3300u 25V 105°C, bright blue, labeled Elna
2 x 330u 35V 105°C, dark grey/gold, labeled FC
2 x 100u 16V 85°C, black/gold, labeled Silmic II (!)
4 x 100u 35V 105°C, brown/grey, labeled Teapo
3 x 10u 50V 105°C, black/grey, labeled Lelon
Are they genuine? I’d say the Teapo and the Lelon: yes, the rest: no
See my substitutes in the tweak section

Other capacitors
6x .1u 63V ceramic bypass caps: these go to the trash immediately. I use the cheap Chinese CBB (polypropylene) instead
Same with the compensation and the input filter ceramic’s: I replace these with silver mica

Protection
A nice bonus is the on board uPC1237 protector with DC detection, delayed turn-on and quick turn-off.

Transformer
My preference is R-core: it has less stray magnetic field than the E-I core, and less coupling between primary and secondary than the toroidal. The Chinese ones are cheap and well made: I’ve yet to hear one buzz. I chose 30VA because it’s the biggest size that’ll fit comfortably inside the box.

Heatsink
The provided ones look wimpy. I think it’s better to bolt them to the chassis: more holes to drill, but better thermal resistance.

The box
Well made, perfect fit and good looking. You need to drill the holes for the main PCB, the transformer, and an optional pre-regulator. Also, the hole for the potentiometer is too small for most man-sized pots and needs to be enlarged.

Operating conditions
The seller advises running 1.7-2.3mA per JFet for the 1st stage set by “LW2” in the original schematic, or P2 in the asc, and 60mA for the MOSFet, set by “LW1” or U1.
After messing with different combinations, I settle for 2mA for the JFets, 5mA for the VAS, and 100mA for the MOSFets. The models for JFets that I found give higher distortion above 2mA.

More later
 
JC2-HPA: Tweaks: semiconductors

Fets
The ugly part: we’ve seen the Idss are well matched. Good. What about Yfs/Gm? Here’s what gives, measured at Idss:
  • 2SK246: 3.76, 3.78, 3.74, 3.76 mS
  • 2SJ103: 5.18, 5.28, 5.12, 5.18 mS
While the same-sex matching is spectacular, there’s a sizable difference between opposite sex. In other words, the lower part will always have higher gain than the upper part. It’s a question I raised with JC long ago, and his view is: “it’s still better than no complementarity at all”. EUVL advocated the use of source degeneration resistors in the higher gain part. It’s doable without butchering the PCB, but it’ll not look good. Is it worth it? I don’t know, you guys tell me.

BJTs
Long ago, I measured the THD in some small signal BJTs, and found pretty large differences between brands, the results are here. For this project, I’ll use some of the better ones I already have, like the Philips or Siemens; of course, they’re long gone. Hopefully, you have variations in your junk box to sort from.
Again, I don’t think it’s a big deal: in a BJT, Gm = Ic/Vt, where Vt is a constant and = 25mV. Avoid the exotics like Ferranti/Zetex or uR, or those that are squarish and thin, stay with known American or Japanese brands. For matching, a $10 transistor/component tester will do nicely (not for the Fet’s!!!).

MOSfet
The kit comes with 2 pairs of 2SJ407/2SK2381 output buffer. They have some strange thermal hysteresis issues reported by EUVL here .
The IRF710/9610 are cheap, widely available, praised by Jay, and Nelson Pass called them “good enough”, but they have shortcomings: high Vgs, >4V for .1A Id, positive tempco, and non ideal Crss. I think I can live with that, but I still keep looking.
EUVL prefers the 2SK1058/2SJ162 and the Exicon’s, but they’re not pin-compatible.
Any other options?
 
THE HPA: Tweak: power supply & electrolytics

Power supply
The original kit feeds the MOSFets directly from the smoothing caps, the rest goes through a BJT/Zener stabilizer. There’s a picture from the seller showing a couple of offboard big caps, which is a valid solution.
Because the operation of the circuit has some sensitivities, both static and dynamic, to the power supply, I want to play it differently and add an offboard regulator based on the ubiquitous LM317/337, replacing the ‘lytics and the ceramic’s that come with with known good stuff (I know, I could have wired it on a perf board, but hey, it’s only 5 bucks!). Then I remove the onboard Zener turning the stabilizer into a cap multiplier to buff out the nasties from the regulator. These are all in the attached asc's and the BOM, and the various pin spacings are accounted for.
There’s also an option to add Elvee’s D-Noizator, maybe someday.
Finally, I was kind of expecting a fight with the snubbing business, but a couple of caps deals with it nicely, see the attached sim, but don’t worry about the DC, I know the turn ratio is off.

Electrolytics
I still don’t know if the current Chinese ‘lytics are good or not, and I’m too lazy to do listening tests. From a reliability standpoint, I’ve yet to see 1 single failure out of the countless computers I’ve had over the last 30-40 years, in spite of all the horror stories. Even the 2 original main smoothing caps in my 50 years old McIntosh MC240 still test better than new, with dissipation factors of .04 and .07. The fatalities I can recall are mostly from audio gears: ARC D-90, Dynaco PAS-3x, some big and expensive Rifa caps in one of my SS preamps, and a single instance from a HP3400E RMS voltmeter.
Anyway, I decide to play it safe and stick to the known good stuff, mostly Elna Silmic’s and Nichicon KT/UKT’s. The constraints are pin spacing on the PCB. I include my BOM with the part numbers, they’re all available from Digi-Key.
 

Attachments

  • HTB1_PS.jpg
    HTB1_PS.jpg
    51.7 KB · Views: 1,804
  • JC2-HPA-BoM.zip
    9.5 KB · Views: 93
  • HTB1aHX_QpXXXXclaXXXq6xXFXXXb.jpg
    HTB1aHX_QpXXXXclaXXXq6xXFXXXb.jpg
    174.7 KB · Views: 1,824
  • Xformer-002.asc
    4.6 KB · Views: 67
Last edited:
Thumbs up for your effort in summarising everything.
But you should not call this JC-2.

It is a E4 headphone amp from Weiliang.
Part of the circuit has similarities to the JC-2 topology.
But that is where it stops.
Thr original JC-2 does not have an output follower, for one.

Cheers,
Patrick
The Weiliang E4 has the JC2 dual differential jFET input topology, and a MOSFET output. John Curl recommended only 2SK170/2SJ74 for the JC-2 input. I am very glad to see the E4 coming back to the DIY market. It is a very cost effective design. Its current price is about half of its first e-bay appearance.

The E19 I have is almost identical to Weiliang's E5. It has the Marantz HDAM input. The jFET are marked as K170, but I doubt that it is genuine stock.
My E5 Class A hifi headphone amp
 
Last edited:
Thank you for the links! Those predrilled chassis appear to be a good bargain. I wonder whether I can shoehorn a Noir or an ACP+ into that box; it won't cost much money to find out.
I chosen the T-2205 enclosure because of the size of my E19 headphone amp board and the use of a regulated PSU.
T-2205 headphone Enclosure Full Aluminum amp chassis DAC box Premplifier Case | eBay

It cost a little more, but the extra internal volume is very good for running the amp in class A. Otherwise, you will have to use the case as heatsink. You do have to drill out the front and back panel with relatively large holes. A hand drill will be difficult to use. I have a 8" tabletop drill press, $54 from Harbor Freight. We usually do not have drill bit large enough in our drill bit set. I use a Titanium High Speed Steel Step Bit which works up to 3/4" diameter holes and good enough for most headamp project. Aluminum panel is relatively easy to drill.
Steel Step Drill Set - 3 Piece, Titanium Nitride Step Drill Bits
 
I'm jealous because in Switzerland, these go for at least $250.
I'm disappointed because it seems HF no longer give out those Super Saver coupons: we used to have a blast getting those DMM's in exchange of $1 and a coupon! I still have one in its original packaging.
The performance of this Harbor Freight drill press is nothing to write home about. It wobbles. But the main inaccuracy comes from my workmanship. I just drill bigger hole to compensate for it. After a year, the hole oversize need is much reduced. But its low price allows me to justify getting a drill press for occasional hobby use. I use it to drill the mounting hole for a customerize hardwood tonearm board too. I can drill the PCB mounting holes at exact diameter now. :) :nod: