Head-Fi.org › Forums › Misc.-Category Forums › DIY (Do-It-Yourself) Discussions › HCC86, Low Voltage Hybrid Headphone Amp for DIY Beginners
New Posts  All Forums:Forum Nav:

HCC86, Low Voltage Hybrid Headphone Amp for DIY Beginners

post #1 of 9
Thread Starter 

HCC86: hybrid-capacitor-coupled-86 headphone amplifier, my first offical PCB headphone project. The design goal is "an easy to build, good sounding and powerful all-purpose hybrid headphone amp for electronics DIY beginners". I have never done an DIY project using PCB layout, might as well do it the fancy way this time.

This is still work in progress, the design itself has pretty much settled, just have to order some boards, do a test build and fine tune the component values.


Some "not so usual" features:

* Designed for DIY build and DIY build only. The boards are layout to be as "DIY beginner friendly" as possible. All parts are DIP, pads are spread out, easy to solder and hard to bridge cross by mistake.

* Two input options: nornal CD input and LOD (Imod, Sansa fuze) input. The LOD input has DC blocking caps so there is no need for an external cap box. Due to the high input impedance of tube circuit, the internal DC blocking cap value can be fairly small, allowing the builder to use fairly high-end signal caps at reasonable build cost.

* Volume control is of shunt configuration, the sound is less influenced by the potentialmenter, allowing the use of reasonablely priced parts (default is Alpes RK27, "alps blue") and still obtain good sound. Does require good quality input resistors, though.

* Tube is loaded using CCS and the cathode is biased using diode (s), the operating point will be fixed by CCS and diodes, no manual adjustment required. Just build and run.

* Output stage is biased into class-A by about 30mA using CCS ic's. This helps to open up the sound stage.


The "not so un-usual" features:

* low voltage throughtout. No worry of getting zapped by the high B+.

* gain stage tube is ECC86, which gives good performence at low voltage since it was specifically designed for that kind of operation.

* The output stage utilizes 4x BUF634 (or LME49600) in parallel, per channel, powerful enough to drive most dynamic headphones with authority (maybe not the orthos, though).

* Output is capacitor coupled. No DC offset to worry about.


This is not a cheap build by any stretch of imagination. Parts cost will be in the $300~$400 range. Some mid-level boutique parts will be on the standard BOM, there is not much need to upgrade once the build is complete. All parts will come from Mouser and Partsconnexion.


* Skill level requirement: Very low. The builder does need to know how to do basic soldering and can use a drill to drill the aluminum enclosure. A multimeter will be helpful but not necessary.


There are two boards involved, each is 4x3.1" (the max size allowed by eagle-lite).  One is the HCC input/output board containing input/output section and the tube; the other is PSBU86 containing power supply and buffer section. These two boards will be linked together by short jumpers. They will fit into a Hammond 1455N1601 or 1455L1601.









Edited by AudioCats - 6/29/13 at 2:30pm
post #2 of 9
Thread Starter 

some HCC circuitry details:

* the tube is loaded using LM234 CCS ic set to about 1.5mA (actual mA depends on which standard value resistor is avaialbe).

*The cathod is bias using small signal diode, either one 1N4148 or two 1N270. The Diodes themselves are loaded using CCS at about 8mA, this reduces the bias voltage fluctuation (due to the changing tube cathode current) when playing music. This should set the cathod voltage at about 0.65~0.75V.

* The above circuit should set the ECC86's operating point at around 14V.

* Cathode biasing diodes are by-passed using small film caps. These caps are optional, the effect is clearly audible. Only high quality caps should be used, or it can worsen the sound (if in doubt, do not by-pass). Capcitance value is not critical.


PSBU circuity details:

* Power input will be from a 24V/2A AC wallwart.

* Once rectified, the main power is filtered by four Nichicon Muse KZ 1000uf/50V reservoir caps. The voltage is then fed into three LM317 voltage regulators: Reg-BL and Reg-BR provide power for left and right buffer stage, Reg-T provides power for the tube stage and optional/add-on features (power-up timer, for example).

* Filament supply has its own rectifier and reservoir cap (3300uf Muse KW). Filament current is regulated by a LM317 (configured to work as CCS), set to about 310mA. This LM317 is to be mounted on the underside of the board and bolted to the aluminum enclosure (using the enclosure as heat sink).

* each channel has four TO-220 package Buf-643 in parrel. The class-A biasing is done using four LM234 (each set to about 8mA)

* Each BUF634 ( or LME49600) has its own bandwidth resistor (the initial value will be 150R, I will probably do some experiment to try to find the sweeter spot)


All in all a pretty simple circuit, just a lot of parts. smile.gif

Edited by AudioCats - 5/12/13 at 9:37pm
post #3 of 9
Thread Starter 

ECC86 selection:

According to the internet rumors, all ECC86 were made by Amperax. However they don't sound the same, if anyone is interested in doing a HCC86 build..... probably want to obtain the right tubes first. There are basically three different types of ECC86 construction.


A) Tall A-frame with dimpled disc getter

The A-frame is tall and get "notched" at the top. Disc getter has 8 dimples.

This type is my favorite, especially the Valvo. As usual, Germen tubes sound a little sharper and Holland tubes a little warmer/thicker. 


B) short A-frame, smooth disc getter

The A-frame is shorter. Getter is a solid disc, no dimples.


C) Halo getter

The getter on top is a ring.



I have tried 7 different kind of ECC86, the type-A (tall A-frame dimpled disc) Valvo is by far my favorite. They can usually be found on ebay, I bought mine from euroKlang, shipping was fast, tubes arrived in about 10 days (from Germany to US).


Occasionally ECC86 are available at AntiqueElectronicsSupply at around $7 each. I bought some recently, they were all Holland type-C,  IMO they are a bit dull sounding, I use them mostly for testing purpose.

Edited by AudioCats - 6/29/13 at 4:15pm
post #4 of 9
Thread Starter 



Got the boards and finished the first build. Found some problems (resistor values for LM317 regulators were backwards, tube loading CCS resistor value needed adjustment, and I didn't leave enough space on the HCC board for larger boutique caps).


The cost for the build was about $100 per board (there are two boards, the HCC and the PSBU86), plus $50 for the enclosure and transformer -->  $250 total for the standard build which already uses boutique parts at the critical spots and there is no need to upgrade (and any further component upgrade might be just waste of $$).  


For even higher output, up to five PSBU86 boards can be paralleled though I don't see why anyone would need 20 buffers per channel, at a certain point the output capacitors will become the bottle neck, not the buffers.


The sound is enjoyable, especially when playing the voice. It drives smaller ortho's with no problem. I tried a pair of Paradox (modded Fostex T50rp) with it, at the Denver meet and really liked the sound, I could listen to that combo all day.


There are some problems to be worked out, though. For one, the case temperature ( I am using the case as heatsink) is much higher than I would like, I will have to either re-arrange the filament supply -- which generates most of the heat -- or use a larger case.

Edited by AudioCats - 6/29/13 at 3:04pm
post #5 of 9
Thread Starter 

reserved, for parts list

post #6 of 9

This may be a stupid question - I'm not an expert on this stuff by any stretch of imagination - but:

In most of the solid state amp designs, a single BUF634 per channel seems to be accepted as being powerful enough to drive most headphones. Is there a tube/hybrid specific reason why you have four each in there? Those seem to be the major cost factor in this design.

Also, does the filament supply really benefit from such a large and high quality cap? Tube filaments don't really strike me as being parts that would profit that much.

Anyways, kudos for sharing this!

post #7 of 9
Thread Starter 

* The reason for paralleling buffers is to obtain low enough output impedance to achieve good dampening factor.

I want to get a dampening factor of at least 10. This is designed as a "general purpose" amp, it needs to be able to drive most common phones well enough, and that includes lower impedance stuff like 32 ohm Grado's. To get the dampening factor of 10, the output impedance of the amp need to be 32/10=3.2.  Since each BUF634 has a 10 ohm output resistor built-in (it is mentioned in the data sheet), four buffers in parallel can only go as low as 2.5 ohms, (and this is just from the buffers alone, before the output coupling cap. The cap will add even more ohms). Four buffers per channel is not an over kill once the low-Z phones are involved. 


As of whether one can hear the difference when stacking more buffers.... In my previous experiments, the benefit of paralleling 3 buffers was audible even with mid-ohm phones (200~300 ohm), the low-ohm phones can benefit from stacking even more buffers. Ideally I would want 6~10 buffers per channel, but this design is geared toward "easy build" and "good enough for general purpose", and there is only so much room on the PSBU board.... I only put down 4 per channel.


* Using muse KW cap for filament supply is not an overkill. The KW is inexpensive and is not considered as a boutique part, it is just the basic model of the muse line. The requirement for that caps is voltage rating at least 50V and capacitance at least 2200uf, the KW is actually one of the cheaper caps that meets the requirement. Muse KW 3300uf/50V is $2.3 each from Mouser, other caps that meet the requirement are not much cheaper if not more expensive. So, might as well keep it all Muse.      

Edited by AudioCats - 6/30/13 at 9:28pm
post #8 of 9

I'm finding the schematic awfully difficult to read.  T-L on the plate goes to TO-L on the power schematic?  So plate and and cathode has a ccs each?  Why?


Edit: nevermind, I see what you are doing.  Why not just use LEDs for cathode bias?


Edit2: I found an example.  Look at the torpedo amp for what I'm referring to.  CCS plate (using transistors and LEDs), and transistorized LED cathode bias.

Edited by holland - 7/1/13 at 11:19am
post #9 of 9
Thread Starter 

In order to set the operating point at somewhere near the middle of B+, the cathode needs to be at around 0.7V. I don't think any LED can go that low, yet. Maybe there are other ways to do it, but using diodes (1 diode or 2 diodes) is the simplest and the end result is good. The 8mA CCS loads the diode so the forward voltage doesn't vary too much when cathode current changes, might not be absolutely necessary but it helps to avoid having to match diodes.


The key thing here is "no parts matching and adjustment necessary, just solder and play".  

Edited by AudioCats - 7/2/13 at 3:31pm
New Posts  All Forums:Forum Nav:
  Return Home
Head-Fi.org › Forums › Misc.-Category Forums › DIY (Do-It-Yourself) Discussions › HCC86, Low Voltage Hybrid Headphone Amp for DIY Beginners