Opamp amplifier with TIP120/125 power stage
I do remember very well when my parants did no longer accept that I listened to music with loudspeakers late at night. Since those days I am a headphone addict, and used to hear music by headphones. I started with a Sennheiser HD414, the whole number of cans I owned over the years is unknown. However, there is one headphone which I thought I could affort for my passion: the AKG K1000. It was about 1500DM at that time, and it set new standards in sound quality for me. This headphone, which is more or less a pair of small but special loudspeakers, is phantastic. But it works only connected to a real power amplifier, you cant use it with a pocket CD player or a Walkman-type MD player.
The second part of the story is that I am a bass player who has to practice a lot, even if it is quiet in the night. To keep a long story short I happened upon this great site named Headwize, built various of those amps and ended up with a Apheared 47 (short form: A47) and a AKG K240 for listening to music and practicing. One day I got the idea to plug my AKG K1000 into the A47. What a surprise, the sound was amazing, even if the volume was a little bit too low. fa-schmidt in the Headwize forum gave me the advice to use something like the Earle Eaton amplifier, so I realized this project in a few days.
The original design can be found here in Headwize. After several discussions I decided to change some parts or to modify values. So the final schematics for my amp (called RBHead now :-)) looked like this:
Main changes are:
- MJE transistors replaced by TIP120/125
- Opamp in preamp stage now OPA2132
instead of NE5532/NE5534
- Output cap 4700uF, not 2200uF
- Emitter resistors 3.3, not 10Ohms.
The design is simple and straight forward. The input signal is preamplified and fed into a class AB output stage. The output cap is not really necessary but as I am careful with DC into my headphones I used it.
The only critical component in this design can be the LED that is used for creating the necessary Vbe for the output transistors, realizing the AB class of the output stage. More details later.
The power supply is a my standard 7815/7915 version with 2200uF caps, comparable to the original Eaton design. In some designs you will find 1N400x across each regulator. Add this after your personal taste.
One additional part is a small circuit for speaker protection. The headphones are only connected to the amps outputs after the power supply provides full operating voltage. A rather primitive design but it works. The circuit is connected to the +24V unregulated path of the positive supply. The 470uF capacitor is charged via the resistor combination 47k/250k. When the voltage between Vbe of the 2N2222 reaches 0.7V the transistor activates the relais and the headphones are connected to the amplifier outputs. Powering down the amp enables the cap to discharge and the relais becomes inactive, disconnecting the headphone lines.
Power-on delay time is:
(R * C) * (0.7V / 24V)
With the given resistor values this makes an adjustable delay of 0.7 ... 4 seconds.
All three circuits, power supply, amplifier and power-on delay, were built on separate boards to be able to replace them later (when or if a better circuit will be available).
The most obvious change is replacing the MJEs by TIPs. This should have a direct effect on the B-to-B LED as Vbe multiplier, with the MJE243 Vbe(on) = 1.5V and the TIP120 with Vbe(on) = 2.5V according to the data sheets. An example to calculate such a Vbe stage can be found here or, with more basics, also here. Another very complete overview for amplifier design can be found in this site.
Following the design principles with TIP120/125 we should have 4 x Vbe = 2.8V (2 x 0.7V per darlington) across the LED, giving a quiescent current Iqc of 20mA. As the LED gives only 2V the resulting Iqc is only 2.5mA. This would mean that the TIPs are not really in on-position as they should be for class AB.
Using solder posts for the LEDs makes it possible to try different LEDs without burning the board, it's very easy to exchange the LEDs. You can use single components or a single board, connected directly or with short wires.
So I built the little circuit called «amplified diode» and replaced the LED by this active control. In fact, now it was possible to adjust Iqc to 20mA. Connecting a load and an input signal to the amp showed no effect in sound but the TIPs got very hot, Iqc now raised to more than 60mA which is too much. Thermal runaway, this is what happened now, could be avoided if the 2N2222 was directly mounted on the TIP's heat sinks. But why making it more complicated than necessary? With 20mA Iqc we have to use a much bigger heat sink for both TIPs and an isolating layer.
One solution would be: raising emitter freedback resistors to 10ohms. This would help to keep thermal runanway under control, but the additional resistance in the output line would not help making the sound better but worse.
Next try: replacing the LED with 2 x 1N4148. Voltage across these two diodes was 1.4V now, not a really surprising value. Iqc went down to 1.4mA, much lower, too low, so the LEDs came back. Using a standard small LED as Vbe supply now, like the original design said, works fine. Using darlingtons and not standard transistors has a great effect on the behaviour of the amplifier.
It seems that even if the theoretical design rules say we should have an Iqc of 20mA this is not really relevant for the sound. The TIPs are definitely not in a proper on-position for class AB, I think. Nevertheless, even when using very small input signals there is no noticable distortion.
So this small amount of distortion seems to be visible on an oscilloscope but not perceived by our ears, or what happens? . The K1000 are very precise headphones, you can hear smallest details in music or sounds. You can even hear the drummer's breathing and the conductor fissling with his papers. No distortion, it's a mystery. Or not. The solution is simple: as you can see there is a 33k resistor leading from the output back to the opamps inverting input. The opamp's feature is sending an output voltage so it keeps its two input terminals at the same voltage. By using the feedback path to the emitters of the output transistors, the opamp senses any dead zone where no transistor is conducting. So it sends an appropriate signal to the bases of the output transistors to drive them into conduction and keeps up with the input signal waveform. The opamp compensates the dead zones below 1.4V for the darlingtons. Still, the LED keeps the darlingtons near the conducting range (1Volt for 1.4V Vbe(on)). Nad this is also the answer why there is no real difference in sound when using LED oder two diodes. The opamps avoids distortion.
Opamp preamplifier stage
In Earle Eaton's design a NE5534 was used. I had some NE5532 left, so I spent 2 x 1/2 of them for an easier layout. This also gave me the possiblity to use very different types of opamps. So i tried:
- TL072 C:
No sound difference compared to NE5532. Both NE5532 and TL072 are inexpensive opamps and provide a good sound quality.
My impression was that the OPA2132 provided a better sounding treble range. So I left the OPA in there, until I need them for the next pocket amp.
The NE553x opamps are sufficient for use as a preamp stage. All opamps do not produce any audible noise.
The original layout used MJE transistors. These are hard to get here, the direct replacement types are BD235/236. I was curious about the difference between the standard bipolars and the darlingtons, so I tried and replaced the TIPs in one amplifier by BDs to compare the sound and output volume. There was no real difference in sound or volume. Perhaps somewhat esoteric, the TIPs with their higher hFE had a minute better bass response, a little bit more drive. So they stay.
Some more mods
An additional 0.1 or 0.22uF cap across the 4700uF output cap gives a slightly better treble response, as it reacts faster than the big electrolytic cap. The 1k resistor in the input path can be left out. Even the power supply can be simplified, using an unregulated version with at least 2x4700uF as buffer caps.
As expected the output volume is very different with the varying headphone impedance:
|AKG K240 Monitor||600Ohms||Sufficient, but not really loud|
|AKG K141 Studio||55Ohms||Head splitting|
|AKG K500||120Ohms ?||Ear splitting|
The second factor is the output level of the audio source. When building the amp I used my Sony MD player as a test source, and the K1000 sounded well and nice. But the volume was not more than acceptable. Connecting the amp to the Terratec 24/96 in my PC later showed a very different picture. The Terratec card has a +4db output for studio line level, not only -10dB like most soundcards. With the Terratec there is even enough headroom for more output level, at the moment the out-level is set to -4.5dB. So if your amp does not provide the volume you expect you may blame your audio source, not the amplifier.
Some technical details
I use veroboards for my circuits. The layout is very easy and symmetric. If some of your components are bigger there is enough room for modifications.
There is also a PCB version for EaglePCB, with an optimized design and for a single channel:
What is right and what is wrong?
||The amplifier produces nearly no noise, provides enough power even for the K1000 and is easy to realize. It sounds very well, with a very tight bass and not over-brilliant but clear. However, compared to a CMoy or A47 there is a lack of ... how to describe? ... let's say: harmonic sweetness. On the other hand it sounds better with the K1000 than my ancient Onkyo power amp.|
||I will try the same design with standard transistors like BD237/BD238 and see how the Iqc runs in this case. I will also mount all transistors on a common heat sink, including the 2N2222 in the Vbe section.|
|No, there is nothing really wrong with this amp.|
At last: some pictures
||The case seems ok but is not the best slice of bread|
Right side: amplifier
Left top: power supply
Left bottom: power-up delay
||Less details ...|
||More details ... the amplifier ...|
||... power supply and power-up delay.|
Due to the great results with my first version I decided to build a second version of this amp. The main reason was that I needed a set of monitor speakers for mixing or transcribing songs. So the second version for headphones and speakers came to life.
The following changes were made:
- A power supply able to provide 2 x 20V @ 1.2A, non-regulated this time
- Switch for output to headphones or to speakers
- Slightly higher gain with a feedback resistor of 47k instead of 33k
- 7K/W heatsink for the TIPs, 3.3 Ohms emitter resistors @ 9W
- Single NE5534A replaced dual NE5532
- No XLR4 socket for the K1000
Power supply and amplifier + time delay went on single boards, so it looks like this now:
||There is one headphone socket on the front and one on the rear panel, plus a switch on the front panel. Connection to the speakers happens with a little breakout box with plug and speaker terminals (not shown here).|
I used clips and a silicon foil to mount the darlingtons. This method is very simple and easy to handle. Highly recommended!
||And a better case this time ...|
This amp seems to sound a little bit better than the first version, but I didn't make a real A-to-B check. The more expensive NE5534A? The different layout of the board?
There is one minor flaw: with headphones the amp is dead quiet, no noise, no hum. When connected to speakers (Behringer Monitor 1C) there is a very little hum. You can only hear it when you put your ear very near to the speakers but it's there. It's definitely not a ground loop problem and no shielding problem. However, there are only few occasions when I use the monitor speakers, so I dont care.
The amp provides a lot of power, more than I will ever need. And the darlingtons and the power supply do not get really hot, I think 2 x 15W can be expected with low distortion. The sound with speakers is very transparent and defined, using really good monitors could show what this design is able to produce. This is my final version of this amp which does a good job. Overall costs were around 80€.