Diamante -a discrete medium power opamp

Hello and happy New Year!

I wanted to post something for the forum on the first day of 2008 in the hope that we all have happy and enjoyable new year, and hopefully some of you can use it in your audio pursuits.

I have designed a discrete operational amp/headphone amp I have named "Diamante" mostly because it has a diamond buffer type output stage.

The design is pretty simple and not at all expensive to build. The cost/performance ratio is very high. The prototype was made with all BC550 and BC560 transistors but others with the same pin configuration can also be used. I chose BC550/BC560 because of their low noise and high gain. Many transistors have reverse pin configurations and they can also be used but you will need to put them in reversed of course.

The basic schematic as shown is configured as a 6db voltage gain headphone amp. But you can configure the board any way you like. The amp is unity gain stable with 10pf at C5. Choose RF and RG so as not to overly load the output stage.

The output R (R26) is a 3W type with magnet wire wrapped around it (15-20 turns is fine) to form an output inductor. This will prevent ringing into capacitive loads, but it is not required. The amp is stable without it. if you wish to omit it you can jumper it with wire or take your output at the resistor pad.

Rbias sets the bias for all stages. I should be a 1K pot, when initially setting up the amp set it to full 1K resistance. Then slowly adjust it downward until you get about 25mv across the emitter resistors (R1-R6). This is a safe bias point.

Once you have the bias properly set, then adjust Roffset to get 0V offset from output to GND.

Wait around 5 minutes and re-check bias and offset.

This amp is pretty simple to scale. As it is presented it should be about to do 100ma or so without much problem. If you change the output devices you could get even more output.

It drives headphones with ease and that's really the purpose I had in mind. I will easily drive even 32ohm cans. 300ohm types such as HD650s are cake for this amp.

Noise floor is very low with the transistors specified. I can't hear any noise through the cans - nada.

Distortion is very very low, though I don't have any pretty graphs to show you yet.

So happy new year to all. I am providing PDFs and the Schematic for those who want to etch boards for DIY use only. The amp can actually be built pretty easily on prototyping board too.

Here is a picture of a prototype (slightly different CCS) in use. Special thanks to Brian Donegan who etched stuffed the first boards and encouraged me to finish the project.



  • ha1 012.jpg
    ha1 012.jpg
    95.1 KB · Views: 8,057
Nordic said:
Thanks Russ, will try it this year....

Do you know the song Diamante? It is by Zucherro... realy nice song...

The word allthough pronouced diffirently is the exact spelling for our (Afrikaans) word for diamonds...

Hi Nordic,

I am not familiar with the song, but I will listen to it if I can find it.

The diamond structure is integral to the amp, that's why I used the name. :)

Some notes...

RBias = 1K multi-turn pot in rheostat mode. Adjust it to 1K before initial power up. If If you wish you can note the resistance of across Rbias and replace the pot with a metal film resistor after you have the bias current you want.

Roffset = 10K multi-turn pot.

Roffset can actually be omitted, the offset should still be very low, especially if you match transistors.

Transistor matching should not be required at all.

When choosing Qs especially for the input and VAS use high gain types (200 or better hfe if possible) for best results.

BC560, and BC550 B and C types will work very well.

The LEDs we used on the prototype were green LEDs which have a forward voltage of about 2V.

Design note:

R24 is used to set the current flowing through Q25 and D1.

The value for R24 is fine as long as you don't bias the amp very high.

If you want to run more bias you should change the value of R24 to something like 47-221R. 100R is a good compromise and will work at all practical bias levels.

I have run the amp with up to 20ma bias, but I would not advise it long term as the output Qs get very warm. 10ma per device is easily achievable, and the amp sounds very very good biased there. To get 10ma bias per device you will adjust for about 49mv across any of the output emitter resistors(R1-R6). At 10ma per device (30ma bias total) you will find the amp hardly ever (never some some headphones) goes out of class A operation. :)

This amp is nice and quiet, with great excellent low distortion performance. I have been listening to it for hours on end.

Re: Hi Mr White !

zeonrider said:
Everything is clear but Vcc & -Vcc is ?

Or I, miss something.

Regards zeoN_Rider


+/- 12 to 15volts will work just fine. It will depend largely on the transistors you choose to use. You could actually go higher, but you may need to run a lower bias.

I am using +/- 15V rails and the instructions presented so far relate to that power supply.

Re: Hi!

zeonrider said:
Mr.White thanks.

I, have many MD8003, about 40pcs( dual NPN).
Can I, put MD8003 in place of Q22-Q23?

Regards zeoN_Rider

Sure, it will work fine. :) I am not sure of the part foot print or pin-out, but as long as you put the pins to the correct pads on the PCB it will work just fine.