will it work?


2009-03-08 3:36 am
i have been teaching myself how to use Eagle for the past 2 weeks and recently started to teach myself LTspice. its a steeeeep learning curve as i know nothing but its fun and i'm learning a fair bit more.

ok to the point, its more than likely rubbish and nonsense but i drew this up for a bit of practise.

I'd say this stands a decent chance of working, though it's not particularly optimized yet. 75 ohms of output impedance is a bit high, the output buffer is not in the main feedback loop, and I wouldn't rely on the offset trim giving enough range to correct for buffer offset.

What are you planning to use for PNP1/NPN1 since you're running those at a whopping 57 mA (250 ohm emitter R)? Sure seems a bit high to me, I'd go for about an order of magnitude less than that. They could also use some degeneration resistors depending on what kind of output transistor currents you are planning to run.

Getting the buffer inside the main feedback loop:
Aside from the obvious of routing the 10k to buffer output, I would suggest a series RC combo where it is right now. 6k8 in series with, say, 100p or so?

BTW, if your power supply is supposed to be along the lines of the first try, you should be able to put a little electrolytic across the regulators' upper feedback resistor (or lower, for the negative supply version). This drops AC gain to unity and accordingly reduces noise. Seems to work pretty reliably in these simple regs as they don't have that much open-loop gain.


2009-03-08 3:36 am
thanks for your feedback sgrossklass,

input AC will be 12vac and the +/- DC rails should end up around 10vdc. the rectification i'm planning on using are mouser # 512-DF005M which are 4 pdip rated at 50v so keeping everything contained on board. regulation transistors are BD139/BD140 or should i look for something else?
if you know of something that has better regulation and compact that will fit on the board i'm all ears :)

output impedance, R17 (75r) and C4 (.82u) go to make a low pass filter to cut the output frequency to about 25Khz, i can change this about and use say a 0.22uf capacitor and 29r. or have i got that wrong and stick with a zobel?

the output bias on the opamp is a jfet cascode to bias the opamp into class A. the jfets will be graded so J1 will have a higher idss than J2. or have i got that wrong?

opamp trim, i'm using the OPA228PA at the moment and it takes very little to trim compared to the OPA134. i'm using 12K to get the output at 0.7mv at the headphone socket. my thinking is 50K variable trimmer should be plenty to play with.

all PNP/NPN will be BC337-25TA/BC327-25TA so should be ok putting 40-60ma through them. with the DC rails only running at 10vdc will this not drop the output to around 30ma and 1-1.3v?

i'm still confused about the buffer output and i'm thinking of removing the two extra transistors for the Sziklai pair.

i'll admit that its a copy of RJM's sapphire with things moved about. i added regulation and a bias cascode.

i'm planning on building it on one of those protoboards that you just plug components into , its practise to help me learn, that and i'm bored and this is fun so keeping me entertained.

i have tried getting to grips with LTspice but as yet i'm still confused.

some pretty pictures, values of components have not been configured yet.


the output buffer is not in the main feedback loop, and I wouldn't rely on the offset trim giving enough range to correct for buffer offset.

The Sziklai pair has loop gain and can stand alone as a low distortion buffer in a two stage amplifier for this reason. It's been done commercially (Nakamichi Stasis) and does offer a marginal benefit when driving "difficult" loads.

However, this benefit is questionable in this application, primarily because of DC offset consideration. Enclosing the buffer into the global feedback presents a new set of problems though, which will require optimization to circumvent. The Sziklai pair will probably need a Miller capacitor for stability, and if the output stage isn't a lot faster than the op amp then the op amp will have to be slowed down or the feedback will have to be compensated. TI.com has a couple of tutorials addressing this topic.


2009-03-08 3:36 am
Thanks Eddie

there is a miller cap there but as per i put it in the wrong place.

i have been messing with all different combination's in LTspice which is fun.

funny you should mention the TI tutorial pdf's as i have been reading them over the past few nights and it looks like the BUF634 is pretty much the same configuration but contained within one nice compact package.

using the BUF634 i can squeeze a dual setup onto one board. its a bit crowded though :)
using the BUF634 i can squeeze a dual setup onto one board. its a bit crowded though :)

BUF634 does all your heavy lifting. It's almost silly to build a discrete buffer.

Look here http://www.diyaudio.com/forums/headphone-systems/289557-old-phono-pre-amp-headphone-buffer.html

Edit: I build a NE5532/OPA2134 + 2xBUF634 single ended headphone amp on that Radio Shack protoboard, complete with all capacitors, that I cut down to 2 1/2" by 2 3/4". You could make it even smaller without the bulky (4x500 uF 50 volt) output capacitors. Single ended has more signal capacitors and a bunch of anti latch up/ anti thump diodes, too; plus an active bias circuit and slow start. All on a 2 1/2" + 2 3/4" board, plug and play. I buy super tiny (and expen$ive) CMF50 resistors, small package electrolytic capacitors where I can get away with it. I use DIP-8 package BUF634s. It's all through hole parts.
And I'm legally blind in one eye, can's see jack out of the other and I designed and built that board.


2009-03-08 3:36 am
here is my next iteration. this time i have gone with a stereo single board. C1-L/C1-R are one on top of the other to save board space. i'm going to add a zip-tie base on the board so i can zip-tie both the capacitors together.




eagle files are now attached for those with bad vision.


  • Psychosis Headamp.zip
    77.9 KB · Views: 6