Class A stereo amplifier

[Pawelr98]

Hello

I wanted to share my project of class A stereo amplifier.
The design was an experiment at first but later I transformed it into a fully working stereo amp.

KD502 is a vintage Czechoslovakian power transistor.
60V,20A, 150W NPN transistor.
Similar to a western 2N3772 but with better gain and ~2MHz Ft.
In reality it seems to be able to handle much higher power, the case is fully copper (which gives low 0.866K/W Rjc).
Power is mainly limited by 155°C Tjmax (lower than most modern transistors).
The die is huge.With 200°C Tjmax it would be a 200W transistor.

BD139 is just a medium power NPN transistor used for gain boost.
Nothing fancy. Doesn't require cooling(bit warm).

BDX18 is a MJ2955/2N2955 under a different name.
This one was made by Hungarian TUNGSRAM back in the Eastern Bloc days.

Design is as following:


The symmetrical +-15V is mainly for opamp.
Rails are fully regulated, +30V is created by uA723+BD139+BDX18 combo(BDX18 + BD139 create NPN Sziklai pair).
+15V is barely loaded so I used a simple 7815.
6800uF 50V capacitor, 30V transformer, KBPC5010 rectifier.

Why PNP BDX18 for regulation ?
Collector is case, this way I don't need any isolation.
Isolation increases thermal resistance which is a big problem for class A amp.
Case is +30V potential.

10A fuse is used for short circuit protection.
Current limiting had to be removed in order to allow isolation-less connection.

Why the capacitor on output ?
I had trouble keeping the DC offset low enough.
Opamp offset gets amplified which creates too large DC offset.
I tested DC coupled output but the distortion was horrible.

Fan cutout is not pretty. Simply drilled a lot of holes in a circle.
Will have to correct this when I purchase some dremel-like device.

Fan reduces the temperature of output stage.
Without the fan it was 95/100°C first/second transistor (case temperatures).
Fan lowered this to 75°C/80°C which is much more comfortable.

BDX18 stays at 85°C, high but there's still bit of headroom.

Improvised high power resistor may be noticed (floating radiator).
Two 1W resistors in parallel, glued(thermally conductive glue) to a heatsink.
This powers the fan(standard 12V PC fan) from 30V rail.

Testing stage on bench power supply (DIY, 2x2N3055 + uA723,0-30V 0-6.5A).

The backround noise is from camera.







"TO DO" list:
-better fan cutout
-new resistor for the fan
-better power cord
-VU meters (for looks)
-possible bigger radiator for BDX18

[John Heath]

I had a similar idea but more brute force simple with a go big or go home design. 300 watts class A and cheap as dirt. Straight off the lines with no isolation transformer. Maybe you can use some of it if it works for you. 

[Gyro]

I like the 'balls-out' engineering! 

You might find something of interest on the Class A amplifier site (John Linsley Hood design based...  http://sound-au.com/tcaas/index.htm

or Nelson Pass's site... http://www.firstwatt.com/articles.html

[Conrad Hoffman]

I've done something similar, but with power MOSFETS and replacing the 8 ohm resistor with a constant current sink. The sink can be really simple and have a pot so you can adjust the idle current. These simple circuits can do a remarkable job if you don't mind the inefficiency.

[Zero999]

Is it really necessary to have a regulated PSU? It's quite likely the power supply rejection of the op-amp is sufficient.

There's DC in the potentiometers, which can cause noise, when they're adjusted. It's better to AC couple signal through the potentiometers.

[Leo Bodnar]

I have played with something similar 20-25 years ago as well.
AD audio grade opamp, MOSFET, current sink (passive or MOSFET in current source mode - with a few thermally coupled diodes in gate bias to stop it from running away, also a lot of paralleled J-FETs.)
It's great fun and very rewarding when it actually works.
You can almost convince yourself that it has pure sound but years later reality catches up and it belongs on a skip.
Leo

[John Heath]

I've done something similar, but with power MOSFETS and replacing the 8 ohm resistor with a constant current sink. The sink can be really simple and have a pot so you can adjust the idle current. These simple circuits can do a remarkable job if you don't mind the inefficiency.

Yes and I am liking it. Constant current source not 8 ohm resistor. It is tempting to tweak the constant current source a bit for higher current at low voltage then lower at high voltage difference but hey it is a class A amp so leave it alone. Just make it a constant current source as you said. It will need a fan to keep it at a reasonable heat but that is the price if one wants to say true class A amp. 

[Pawelr98]

I have played with something similar 20-25 years ago as well.
AD audio grade opamp, MOSFET, current sink (passive or MOSFET in current source mode - with a few thermally coupled diodes in gate bias to stop it from running away, also a lot of paralleled J-FETs.)
It's great fun and very rewarding when it actually works.
You can almost convince yourself that it has pure sound but years later reality catches up and it belongs on a skip.
Leo

I did test it on my usual home theater system.

Sound quality was very good.
I also did a frequency sweep(full audio frequency range) to test for any noticeable distortion(8R resistor for load).
Oscillograms were clear.

I may mod the amplifier into standard Class A topology later.
I just need a good way of radiator/amplifier case isolation.

It's easier to isolate entire radiator from the case than to isolate the transistors.
In early stage I used 2N3055 for voltage regulation which required the output stage radiator to be isolated.
It was messy and there was trouble with mounting (radiator is heavy) so I purchased BDX18 just for this project.

In general I prefer bipolar over mosfet. Less components needed for basic operation (two resistors for minimalistic class A amp, no work point stabilisation) and can operate on low voltages.
TESLA KD502 output characteristics look like this:


This transistor was used in audio amplifiers but in power supplies as well (it's high power after all).

The opamp I use is LM833. It's similar to NE5532 in performance and uses the same pinout.

Regulated PSU has noticable advantage.
Some of the heat is moved away from output stage.
With the PSU you have +30V supply, without you would have ~36V average.
Few more volts times current and you get plenty of extra heat on output stage.

Main filter cap(6800uF) shows ~4Vpp ripple, quite high if you ask me.
Voltage drops is what keeps the BDX18 cool enough.
Normally I would use 10000uF cap for such load but the electronics shop didn't have one.
I had to use a trick in PSU in order to make it operate properly, uA723 has separate capacitor connected through diode.
Voltage on this capacitor is not dropping as much which gives uA723 some extra voltage for Ube drops.

I really like vintage technology. A part of my collection:


TG70 were the only power transistors in Poland for quite some time.
TEWA TG70 is germanium PNP, 10W,30V,1.5A, Ft=100KHz.
It was even used in new devices in early 70's because early silicon power transistors had poor reliablity("fixed" by reducing their rating).

Currently working on AB class amp using USSR germanium transistors.
I have MP10(NPN),MP37(NPN),MP42(PNP).
They are not complementary so I have to pick a good MP10/37 and pair it with a weak MP42.
MP42 has higher gain,higher current and slightly higher power (150mW vs 200mW IIRC).

USSR had a long story of germanium.
My 1992(USSR collapsed in 1991) C1-107 oscilloscope still has one or two germanium transistors in deflection circuitry.

[John Heath]

The first transistor radio made TR1 in 1955 employed a class A output stage running on battery power. Almost a carbon copy of a 5 tube radio of the same time. I imagine the battery did not last that long but then again how long will a cell phone go on one charge.

[Pawelr98]

Those are NPN ?

In Poland there were no NPN transistors until the introduction of silicon transistors in 1967.
Mesa transistors.

Push-Pull amplifiers at the time used transformers or imported NPN transistors.


"Tramp" portable radio, Japanese transistors were used for radio frequency stage.
TG5 is a small germanium transistor, 75mW,30V,10mA Ft=800KHz.

Fast germanium transistors came with new technology.
TG5 is Alloy-junction transistor.
TG39 is Alloy-junction/diffused and has "big" 40MHz Ft, it replaced imported transistors in radio receivers.

TG70 is an Alloy-junction transistor yet it was still in production in early 70's.
USSR produced alloy-juction transistors pretty much until collapse.

Polish engineers actually managed to design a PNP-only Push-Pull amplifier which had no transformer.

It was a concept which was never actually built.

Another example built and tested:


Works but has unsymmetric output.
http://www.forum-trioda.pl/viewtopic.php?f=8&t=15635&hilit=german+tewa+pnp

[mikerj]

I'd like to see actual testing to show the inefficiency does actually lead to improved performance over a good class AB design.

[Gyro]

I think you're talking a different performance, rather than improved... but one that sounds nice. The figures certainly won't say 'improved'.

Personally I'm a fan of sounds nice.

[Leo Bodnar]

From my fleeting experience with this design it sounds really good at lower levels with music that does not have a lot of dynamic range but distorts easily and unpleasantly (unlike tubes) at [subjectively] normal sound levels  where cheapo AB keeps chugging along.
If you don't crank it up it easily beats your average Yamakai.  It is DIY built in one evening so it always sounds nicer than it measures.
Leo
P.S. I am probably a bit random but the facts are - I built mine, I thought it sounded awesome, 20 years later I think it's nothing special and it's waiting to be chucked out.

[John Heath]

From my fleeting experience with this design it sounds really good at lower levels with music that does not have a lot of dynamic range but distorts easily and unpleasantly (unlike tubes) at [subjectively] normal sound levels  where cheapo AB keeps chugging along.
If you don't crank it up it easily beats your average Yamakai.  It is DIY built in one evening so it always sounds nicer than it measures.
Leo
P.S. I am probably a bit random but the facts are - I built mine, I thought it sounded awesome, 20 years later I think it's nothing special and it's waiting to be chucked out.

The class AB , more B than A , with two transformers was so good for battery transistor radios that it lasted 25 years unchanged. Class AB is the best of both worlds. What happened to class D ? The switching strategy for audio is great theoretically but it never seemed to get off the ground. 

[Audioguru]

One earlier post said there was no visible distortion on a 'scope. Oh yeah? I can't see 2% even harmonics distortion on a 'scope but I can hear it. Less than 0.05% distortion sounds good.
This amplifier like many vacuum tube class-A amplifiers produces plenty of even harmonics distortion at high levels that some people say makes it sound "warm" and "musical".
I think that any kind of distortion is bad. Also I can smell the massive heat produced by this amplifier.

[Conrad Hoffman]

There's a trick that happens when you use a power MOSFET and the circuit is set up right. Since the turn-on of the MOSFET will be several volts, the opamp zero output point isn't at zero. It doesn't go through its output crossover point at low signal levels, and by the time it does, it doesn't matter. I don't know about the design belonging in the skip. Mine had excellent measured performance and made a dandy headphone amp or office amp.

[cvanc]

Great to see all this homemade stuff.  It's both great fun and a great opportunity to learn at the same time.  Well done, Pawelr98!

[Pawelr98]

One earlier post said there was no visible distortion on a 'scope. Oh yeah? I can't see 2% even harmonics distortion on a 'scope but I can hear it. Less than 0.05% distortion sounds good.
This amplifier like many vacuum tube class-A amplifiers produces plenty of even harmonics distortion at high levels that some people say makes it sound "warm" and "musical".
I think that any kind of distortion is bad. Also I can smell the massive heat produced by this amplifier.

I purchased analog scope instead of digital one(high-end analog VS chinese digital) so measuring THD won't be that easy.
I agree that distortion is a bad thing.

At the same time I often prefer "modified" sound instead of original.

I prefer headphones over speakers due to prices and portability.
FiiO E07K Amp+DAC combo.
I have set slight treble bump because many headphones boost the bass bit too much.
At some point I experimented with my UNITRA FS032 hardware equalizer.
Music can be more enjoyable if you are able to customize it.
Lowering bass levels is certainly one of the things I like, especially on tracks with good vocals as it makes them stand out.

My only comparsion at the moment is Panasonic SA-HE70, class "H+" (that is AB with adjusted supply voltage).
At some point I had Unitra WS432(exported to west as Dynamic Speaker SA 940) which is a classic AB class ampifier.
<0.2% THD@1KHz at rated output (2x35W).

[John Heath]

Before starting I like the scope in the picture posted. 150 MHz and it is real not a digital simulation therefore at 150 MHz what you see is what it is not software connecting dots.

The amplifier can only do so much. What about the speaker? A negative resistor could be used to make most speaker close to perfect within a power range , theoretically. To say why is a long story , apologies in advance.

Cassette tape motors used to have mechanical weights with contacts to govern the speed , centripetal force , like the old steam engines. One bright spark came up with the idea of making a theoretically perfect electric cassette motor that will in theory will turn at a RPM that equals the volt to the motor regardless of load. A perfect motor is too expensive so he put a op-amp simulated negative resistor in series with the motor. This causes the motor to act like a perfect motor without the cost of making a perfect motor. The copper wire resistance of a motor is 10 ohms so you set the negative resister to 10 ohms in series and it becomes a perfect motor where voltage equals RPM regardless of motor load. This idea worked so well that most cassette cap stand motors were negative resistor type for the last ten years of cassette players.

With this in mind could the same be done for speakers? No one has tried it to my understanding but it seems plausible.  An 8 ohm speaker having a DC resistance of 2 ohm would require a op-amp simulated negative resistor of 2 ohm in series to make it a perfect speaker. A tiny 3 inch and 12 inch speaker both Z = 8 should act the same. If the little 3 inch speaker does not care for bass notes then the 2 ohm negative resistor will rip the voice coil right out of it trying to make those bass notes, ha. 

[NiHaoMike]

What happened to class D ? The switching strategy for audio is great theoretically but it never seemed to get off the ground. 

It didn't become common until digital audio became mainstream. Much easier to justify the cost of a DSP to convert PCM into a bitstream when you need the DSP for other purposes anyways.

Chip scale digital amplifiers even started appearing in USB Type C headphones - apparently logic can scale down to the point where it ends up using less power than precision analog circuits that do not scale well. They more or less made a Delta Sigma DAC that connects directly to the headphones.

[Pawelr98]

Before starting I like the scope in the picture posted. 150 MHz and it is real not a digital simulation therefore at 150 MHz what you see is what it is not software connecting dots.

Schlumberger 5228
It's a 250MHz scope made by French Sefram.

Which is why I purchased it over a chinese digital oscilloscope.
Some RF guy was measuring signals close to 500MHz on it.
All three channels have full bandwidth, third channel has "crippled" functionality.
0.1V/div and 1V/div, no 50R termination and no custom V/div.
It's not my only scope.

C1-107, 5MHz single channel.XY + external trigger
It was my first scope before I gathered money for a proper high-end scope.
There's a built-in multimeter. Needs some repairs and adjustments but some functions work correctly.
ADC,display is all discrete TTL + analog.


Do you guys purchase NOS devices instead of new ?
I usually do because the quality is better and you don't get counterfeits.

There's Czechoslovakian TESLA, Hungarian TUNGSRAM or MEV and Polish Unitra-CEMI.
Plenty of NOS hardware at low pricing.
A pack of 4xKD502 + shipping costs 17PLN.
With current 3.35PLN/USD it gives ~5USD.

Occasionally I get some western Motorola or even RCA.
Additionally I take apart electronic junk and strip it from all components.
They go into big storage wall for components.

Too bad that I mainly get ATX pc power supplies. I mainly get low voltage caps and high voltage transistors from them.

My first amplifier was 2xTDA2003 amp for a TV set around 2 years ago.
Only the chips and the PCB was new.
All capacitors and resistors were used components.
Works to this day.
Sound quality is good for just watching TV.
Main problem is that the switching supply noise comes through sound card of my HTPC (DVB-T tuner installed).
Semi-fixed by adding 27R resistors on amplifier input (low but does the job).

Switching power supplies are a big problem.
I constructed a DIY AM radio.
AA144 germanium diode for detection, BC516 + LM386 for audio amplification.

Poland has 1MW Long-wave (225KHz, Polish Radio Channel 1) transmitter in Solec Kujawski(it has enough power to be received in Japan or USA as most "competing" transmitters got turned off in recent years).
This city is right next to Bydgoszcz which I live in.
Over here the signal is excellent.
Once time I decided to take my radio to Gdansk(~140KM north) where I study at Polytechnic(Technical University).
Station could be picked up but I had lots of noise which covered the audio signal(broadcast could be heard in backround).

At one point in history we had 2MW long-wave transmitter (which had effective power output of 3MW) in Konstantynow.
Highest construction on earth (646meters) until collapse in 1991.
https://en.wikipedia.org/wiki/Warsaw_radio_mast

Now I want to build a more advanced radio.
TDA2822M for audio amplification. It can run down to 1.8V which means higher volume with the same 18650 cell(LM386 is barely running at such voltages).
BC516 for RF amplification this time.
AM detection using the same AA144.

I learned radio technology from russian DIY youtube channels along with USSR radio-magazines (mainly schematics).
I know cyrillic (self learned) so I can understand plenty of stuff just from Polish-Russian language similarities.

Circuit #8 is what inspires me

Two-stage RF amp replaced by a single BC516 (PNP darlington) and audio amplifier replaced by TDA2822M.

[John Heath]

I am liking diagram # 6. One resister , wow. Is that a complete radio or just the last IF and audio out?

[Leo Bodnar]

I am liking diagram # 6. One resister , wow. Is that a complete radio or just the last IF and audio out?

This chap seems to be an awesome educator.
https://translate.google.com/translate?sl=auto&tl=en&js=y&prev=_t&hl=en&ie=UTF-8&u=https%3A%2F%2Fwww.litmir.me%2Fbr%2F%3Fb%3D284795%26p%3D66&edit-text=
Translation is a bit off but you get the idea.
Leo

[John Heath]

I am liking diagram # 6. One resister , wow. Is that a complete radio or just the last IF and audio out?

This chap seems to be an awesome educator.
https://translate.google.com/translate?sl=auto&tl=en&js=y&prev=_t&hl=en&ie=UTF-8&u=https%3A%2F%2Fwww.litmir.me%2Fbr%2F%3Fb%3D284795%26p%3D66&edit-text=
Translation is a bit off but you get the idea.
Leo

Nice link. This is old school thinking at it's best. The foundations that got us to where we are today. I especially liked the positive feed back to increase Q factor. A tricky balancing act if it is a little too much positive feed back but it can not lock up as it is a AC coupling not DC. The hand drawn art work is a nice touch as well. 

[Pawelr98]

I did construct a modified #8 circuit.

Two stage RF amp-> single BC516
Audio amp-> TDA2822M

Lower diode in the AM detection got replaced by low value resistor.

I had trouble with audio distortion. Connected oscilloscope in 50R(integrated 50R terminator) mode over the lower diode.
Distortion was gone and the waveform was looking very good compared to usual 1MR mode.

Replaced the diode with 56R resistor and it works very well now.
The radio seems to be much more noise-proof.

Lowering the impedance does wonders sometimes.
I guess it's because of the BC516 gain. Small-signal darlingtons have insane gain.
BC516 has a minimum of 30K.

The AM modulated waveform on analog scope looks beautiful.


Old books/magazines are a good source of knowledge.
In eastern bloc the semiconductors were scarce and expensive.
Which is why many designs are minimalistic.

Very often the available devices were "second-rate" or to be honest "factory trash".
Example would be transistors with single digit gain.

Purchasing an entire device was often impossible.
You had people building entire TV's from parts.
"Factory Trash" CRT was often used as a base for a working TV.


I learned my radio stuff from here: