Another 35 watt Solid State Amplifier

Introduction

Another SSA? Yes, against all HiFi-religions, I offer a class-B amplifier with overall feed back which performs excellent (< 0.001 % distortion) with a 4 Ω resistive load.
It permits a capacitive load of up to 10 uF!
It is intrinsic short circuit proof.
Without the input filter the power bandwidth is >1 MHz and the G.B-product = 13 MHz.
With two of them in balance one could make over 120 WRMS

The Main Diagram

Explanation of the Main Diagram

The amp contains just a few components: an op amp and four tran- sistors with their biasing components. Henk ten Pierick measured the OPA134 as the best sounding op amp in high input impedance applications. With his research method, active components can be ranked in sound quality. Many listening sessions approved his method.
The op amp drives the NPN-PNP-pair BD139-BD140 which has a fabulous reputation for many years. The eventually power transistors: 2SA1943 and 2SC5200 have been chosen because of their flat beta-character as function of the collector current.
Because of the OPA134, the power supply voltage is limited to
+18/-18 volt. To keep the full output voltage (34 Vpp) available at the loaded output, the power stage should be fed with +24/-24 volt (mind R3 and R4). These limit the maximal available power to



35 watt into 4 Ω.
The amp is short circuit proof, simply because the op amp can't supply a current which could drive the 2SA1943 and the 2SC5200 into a region which should harm them. For all sake another BD139/BD140 have been added which 'eat' the current from the base of the 2SC/2SA's if the shortcut curent should arise over 15 A.
R7 and R8 serve the quiescent current in the BD's (~42 mA), R5 and R6 serve the 2SA1943/2SC5200's (~15 mA).

This type of power stages hardly suffer from cross over distortion. Without overall feedback it performs with <0.05 % distortion at 28 watt, 0.03 % at 0.5 watt and 0.02 % at 0.25 watt. There is no reason any more to flee into class A with its huge power consumption.

The feedback

As I stated above, the power stage does not really need feedback, so that the feedback could be restricted to the op amp itself (the 'local feedback'). The consequence is that the power stage should be corrected for the output offset voltage (preferably < 5 mV). With the 'overall feedback' the loop gain for DC is equal to the open loop gain of the op amp (120 dB) because of C2! The output offset melts away to some micro volt, so the small amount of distortion does!



It is reduced to 0.0007 % even with less quiescent current (R7, R8 = 2.7k and R5, R6 = 15 Ω).
R13 and R14 dictate a gain of 11x. The input circuit R11 and R12 reduces the total gain to 10x. R12 - C2 keeps out input > 60 kHz
(if the source impedance is adequate).
The feed back is >110 dB@0Hz, 70 dB@1kHz and >40 dB@20kHz!

Driving into a load of 0.1 Ω



The load is 0.1 Ω (nearly short circuited),
the input voltage is 1.5 V (yellow),
the collector current of the 2SA1943 (green) is nearly
15 A,
the output voltage (red) is a block of 1.5 V, and
the voltage at the base of the BD139 (blue) proofs that the OPA134 is overloaded, even without the protection transistors.

Distortion

The cross over distortion of the SSA is very small, but could it be of help to put C44 parallel to R14? Looking at the pictures, there seems no difference in distortion within the audio reagion (be aware of the different levels!).
Mind that the noise flour elevates with the value of C44.




Output > 25 watt with C44 = 0


Output > 25 watt with C44 = 47 pF


Output > 25 watt with C44 = 100 pF

But what about the distortion outside the audio spectrum? Mind that this distortion earlier or later could fold back into the audio spectrum if there is a large component there or a 'foreign signal' which could enter via the cables connected.
For this test the amp is oversteered a bit to get more distortion. The input signal is 2 volt while the maximum permissable input is 1.8 volt.
Because of the changing noise floor again it is not simple to judge.

Eventually C44 has been left out for stability reasons. This had been decided during extensive testing of the hardware.
By the way, these figures have been simulated with the 'local feedback'.


Input = 2 volt with C44 = 0


Output > 25 watt with C44 = 47 pF


Output > 25 watt with C44 = 100 pF

The AsSymm

If two 30 watt SSA's will be assembled to a 120 watt amplifier, one needs a circuit which transforms an unbalanced (asymmetric) audio input signal into a balanced (symmetric) one. This could be established with the diagram to the right. The two outputs are connected to the two input connecrors of the SSA's so that they are driven in anti phase.
Henceforth the speakersystem is balanced driven. This means that it has to be connected between the two hot connectors of both SSA's. The output ground connectors will not   be used.

Remarks

The two diodes 1N4148 should be fixed into a small hole in the heatsinc.
Do fix the capacitors C3 and C4 because of less distortion and more power output.
This amp is able to drive 60 watt into a 2 Ω load! If more current is wanted (eg. in a smaller load impedance) the 2SA1943/2SC5200-s with R3-R4, simply could be doubled. There is no need for any update of the rest of the circuit.

There is already a printed circuit layout of the SSA in: The Printed Circuit Board.
There are no critical parts or trics if the prescribed components are used.
For safety reasons some diodes and fuses should be added to the circuit as shown below.

The Finishing Touch?

Tentative Conclusions

MicroSim has proven to be a nice rather easy to learn simulator, at least for audio applications. The developed amplifier circuit is quite simple and certainly not new, but did the very good performance be common knowledge?
The crux of the matter is: Does the model MicroSim cover the design in all aspects? One of the problems: 'oscillations' (which occurred in practice) had NOT been predicted. A small 10 nF-capacitor directly in parallel with the output terminals, transforms the amp into a rather strong transmitter!
The remaining question that occupies me is: whould it sound well? The OPA134 is said to be an excellent sounding op amp (Henk ten Pierick) but is this still the case if it is driven so high? The data sheet depicts:

.... in combination with high output drive capability and excellent dc performance allows use in a wide variety of demanding applications. In addition, the OPA134’s wide output swing, to within 1V of the rails, allows increased headroom making it ideal for use in any audio circuit.

However, the ultimate answer will come with the realisation of the SSA and the listening to it in a well behaving known environment.
I will keep the reader informed...... but before this, the amplifier must be put together...

The Printed Circuit Board

With the PCB layout the diagram of the SSA has been drawn again with the component numbering as on the circuit board.
The dimensions of the board drawings is not exactly 1 : 1. The length of the board should be 143 mm and the width: 73 mm.
This should be corrected with the Adobe Reader.
Two components in the diagram (2x 100 µF) have been added after preparing the layout. They should be added at the copper side.


Later on some changes upon R6//R8 and R5//R7 have been executed as shown in 'Some pictures' below.

  1. Circuit board, component side
  2. Circuit board, copper side: mirrored
  3. Circuit board with numbered components

Get down to work!

With the pictures and diagrams, the printed circuit board and 'Some Pictures', an experianced technician will be able to build the amp. Do not forget the ground connection between 'ground' (in my case: the brass bar) and the tag on the PCB positioned close to "+ IN -".

Far different from the simulations, the value of R7 and R8 should be about 4k7 for the desired quiescent current, as long as connected to the 24 volt rails.
The only problem I met, was the sensitivity of the quiescent



current to the power supply. Between the bases of the BD's and the output the differential gain is about 23 times! This means that R7 and R8 should be connected to the 18 volt stabilisers. Of cource the value of R7 and R8 should be changed to about 3/4 of that found with the rail tension. This only solves the problem partly, so a ferm power supply should be used!
This needs some board changes, as for the large elco's which should not be placed behind the fuses! (see the pictures below).

Results

The used power unit supplies plus and minus 26 volt unloaded and 24 volt with a load of 10 A.
R7 and R8 should be carefully tuned to assure a quiescent current of 42 mA through the BD's (for the least cross over distortion) which corresponds with a current of about 15 mA through the 2SA/2SC's. This depends also on the trade mark of the 1N4148.... See 'Temperature stability' below.

With the values I have found for R7 and R8 (3.3k < R7 = R8 < 3.9k) the output of the OPA134 became 31.6 Vpp, so the power output into a 4 Ω load is 32 watt. The 32 watt power bandwidth is at least 20 kHz.
I did not measure at higher frequencies.
The output offset < 10 mV with 'local feed back' and < 0.4 mV with 'overall feed back'. During the measurements no other differences could be found in the behaviour of the amp between local- and overall feedback.

Temperature Stability

The configuration of the 1N4148's and the BD's with Re = 0.5 Ω, is rather temperature stable: with about 8 mA through the diodes, the quiescent current through the BD's changes from 40.12 mA@20°C to 40.43 mA@50°C. The quiescent current through the 2SA/2SC's however


changes from 6.5 mA@20°C to 33.6 mA@50°C! This means that R5/R6 should change from 13 Ω@20°C to 11,15 Ω@50°C to keep the current unchanged. 18 Ω parallel to a 100 Ω NTC of 0.5%/°C will fix the job.

Oscillations!

As stated before, C44 (parallal to R14) should be left out. An RC-combination of 6 Ω in series with 100 nF over the output terminals is of help, but even then the 4 Ω loaded amp shows instabilities at the peaks of the full power output signal independent of the local or overall feedback.


The remedy should be: put a dominant pole in the amp, somewhere over 20 kHz. To make a long story short: 220 pF parralel to the output of the op amp solves the problem. (This update has not been shown in the pictures.)

Thus.......

Harmonic Distortion



Listening Tests

Somewhere else on this site one could find the equipment used.
As a signal source: a highly upgraded CD-player Philips CD624 (with CDM 4) with a very low jitter clock (< .3 ps), a software digital filter in an FPGA, and PCM1704's with LT1028 op amp I/V-convertors.
The loudspeakers used, are those discribed as: ESL + MFB = the best of 2 worlds.
Also conventional speaker systems have been connected as well as Paul Vancluysen's Quad ESL-63 system.
A number of self-made recordings have been listened to. But for the judging of the feedback ('local' or 'overall') served Shubert's String Quartet in c, D703 played by the Artemis Trio + Truls Mork cello.
With the ESL + MFB the amp is hardly loaded. Only at frequencies over 15 kHz the capacitive impedance is in the order of 4 Ω.
How it sounds? Just one word: brilliant without ifs and buts!

Listening with the conventional two-way system: Ensemble PA1 the sound is blameless (as Douglas Self calls it) with the 'overall feedback'. However, with the 'local feedback' the strong peaces became blurred, as Bart van der Laan calls it.

Listening with Magneplanar SMGa offered a different image: the strong parts remained correct with the 'local feedback'. With the 'overall feedback' the week subtleties came through in more detail, sometimes called: transparency.

Listening with the Quad ESL-63's gave the same impression.

Conclusions up till now

In most cases 35 watt will satisfy if the speaker system is not too insensitive and you do not want ear-splitting levels in your room. I will see if some bootstrapping of the power connections to the OPA134 could enhance the power output without loss of quality.
Undoubtly the 'overall feedback' is superior over the 'local feedback'. Have in mind that THRESHOLD advertises with this 'local feedback' (and writes it on the front panel of their amplifiers!!). I agree with Douglas Self that there are no "inexplicable influences on audio quality". In other words: less distortion simply sounds better!

This rather simple design yields an excellent amp, no more and no less.

The power consumption of the amp is that low that, even after playing uncompressed music for a long time, the heat sinc hardly rises in temperature if 'measured by hand'.

Bootstrapping

As I stated before, the amount of output power could be managed by bootstrapping the OPA134. Lifting the supply-terminals of the op amp with a part of the output voltage, allows a larger output swing.
Three levels of bootstrapping have been examined, obtained by changing the R25/R26-ratio: resp. 1/11 (0.090909), 33/183 (0.1803278) and 0.4000 times the output signal.
The maximum obtained power is resp. >40, >50 and 100 watt.
The main power supply should be matched accordingly to >24, >27 and >32 volt.

Bootstrapping and Sound

In this design the gain is 20 dB, 10 times. Apart from the input circuit the gain is 11. With '1/11 times bootstrap' the op amp supply-pins are 'modulated' with 1/11 of the output signal, which is exactly the input voltage! In case there is no AC-signal over the unlinear parasitic (Miller) capacitors of the input transistors of the op amp! With a high impedance input source this is of great importance. With this design the impedances are not realy high so that the gain in sound will be restricted.

The Bootstrapped Amp

Final Listening Tests

The 1/11 bootstrap sounds a little bit better than without bootstrapping, but there is a BIG BUT: noise! If jumper J2 has been lifted, the noise of the stabilisers (7818 and 7918) comes through.



The listener has to judge himself if this is acceptable, or find a solution.... Luckily the noise is independent of the strength of the bootstrapping.
I leave further experiments to the diyer....

Thanks to ......

Guido Tent for the 'difficult components' and Bart van der Laan for making the PCB layout and both for the support with the listening tests at TentLabs.


januari 23 - 2010
Herbert Rutgers.
last update: februari 8, 2012

Some Pictures