300W BJT AMP
FOR HIGH CURRENT INTO LOW LOADS.
A guy asked me to rebuild his Phase Linear 700 amps.
These old bangers were
famous for being "flame linear", and "failing with smoke" when the party got
I made the case larger, increased the fins on the
heatsinks to what they should
have been originally, and used the following schematics below.
I apologize about the hand drawn and scanned schematics, but I doubt there is
great need for me to spend 2 days redrawing Figs 1, 2, 3 , 4, 5 in MS paint for
everyone to enjoy better. I doubt anyone would want to build the schematic shown.
I don't want to waste too much time telling everyone
exactly what the finer points of
this amp are. It sounded OK at completion after 3 long weeks of hard slog. My mental
conclusion after the work was "NEVER AGAIN !!" I just don't like power transistors
in output stages.
The 6 x MJL21194 and 6 x MJL21193 power transistors
were chosen because they
were cheap, available, have high voltage & current ratings and have been used now
in many high powered amplifiers. They are flat pack types needing only one bolt to
I started off with the supposedly wonderful output
stage with as shown on the schematic
and thought that the darlington pair arrangement wouldn't be too hard to drive with
the MJE340/350 Q4,5 gain pair. But the output stage Q9, Q10 base input resistance
was still quite low and the NPN and PNP output devices have such a huge variation
in hfe for each half positive and negative wave cycle that thd was about 40% at the driver
stage so when a lot of global NFB was added I did not get the low distortion measurements
which are so routinely easy and probably necessary with solid state. This is because the
MJE340/350 Q4/5 pair have high collector output resistance and the darlington output
stage base Q9/10 input resistance is low by comparison to the collector resistance of
Q4&Q5 driver stage. I thought perhaps an arrangement as used in Crown amps with a
darlington triple output stage could be used, but I don't like copying other designers.
So I added the buffer emitter follower Q6 with dc
supply from CCS Q7. The low Z output
resistance of Q6 drives the base inputs to Q9&Q10. Q8 is the VBE transistor on the
heatsink to control bias current in the output stage.
The buffer reduced the effect of the low output stage
input resistance and the open
loop distortion dropped about 30dB and all was well. I was able to measure less than
0.01% at near full power with the thd reducing linearly with reducing output voltage
right down to very low power, and at a few watts I couldn't easily measure the thd.
There was no sign of any crossover distortion.
main problem I encountered was with cross conduction when
at highish power at 27kHz. The output transistors appeared to be so sluggish about
turning off that the devices on each side of the PP circuit tended to stay on and seemingly
ignoring the directions of the driver amp. This phenomenon leads to considerable current
flow from rail to rail through the output the transistors and not all going through the load.
What a horror! I finally settled for the above schematic which seemed to work fairly well to
reduce the very poor behavior of cross conduction. Even without a load connected at 27kHz,
the amp drew 10amps from each rail at Vout high levels!!!! Mosfets never display this
sort of poor operation, and they don't need to be set up as darlington pairs and they
don't need a buffer stage.
Douglas Self mentions this phenomena at his website
and gives only limited advice
about how to avoid cross conduction. If there is/was a solid state guru, it might be Douglas.
I wondered how disastrous it could be if cross conduction occurred at lower F, but even
his pages are largely free of "what to damn well do when smoke threatens." Anyway,
there was little trouble at under 20kHz, and the amp has not given trouble after 8 years
( its now 2011). It happily and effortlessly drives a subwoofer with 15" driver and the
recording of the Space Shuttle launch I heard was fraught with the fear that the windows
might break walls would shatter, and the roof tiles could fly away, but the amp soldiered
onwards without any smoke.
Although it sounded fair with wide range music, tubes
and mosfets are better.
The amp serves its owner well as a sub-woofer amplifier where it can easily churn out
300 watts in bridged mode into 8 ohms and a lot more right down to 2 ohms.
Power output with +/- 55V rails and without allowing
for slight rail sag for sine wave
testing due to high PT winding resistance and some choke resistance in power supply:-
8 ohms, 180 watts per channel,
4 ohms, 360 watts per channel,
2 ohms, 600 watts per channel.
Bridged, 8 ohms :- 360 watts, 4 ohms:- 600 watts.
The power supply for the "Turner Linear" bjt amp used
the original Phase Linear
power transformer. Unfortunately, the tranny hummed badly like many PA amps do
and I didn't want to use the 62V windings to make +/-86 volt rails because the possible
400W of power into 8 ohms from such high rails and from both channels wasn't wanted
or needed, and may not have been reliable. I built two carefully designed swinging chokes
which are 0.35Henrys on 32mm stack x 25mm tongue GOSS cores. These were varnished
and potted in mild steel boxes with roof pitch. This made two cubic boxes with about 90mm
along each side. The the original 10,000 uF caps and power transformer were retained with
the bridge rectifier. Noise with the choke input set up was much less than with a cap input
supply since there are not such huge switching currents to excite motion in the transformer.
I added rectifiers from the 62V-0-62V CT windings to directly charge up a CRC filtered input
amp rails for the driver amp and added simple emitter follower regulators and the final
technical performance was quite pleasing. There is some sag in the rails when continuous
power is used, but with rock and roll taken up to occasional clipping the rails don't sag much.
There were no bad resonance effects due to the chokes and 10,000 uF being series
resonant at about 3 Hz.
For anyone trying to
build an amp like this :-
You DON'T need to use choke filtering for +/- 57Vdc amp rails. I only used LC because the rail
voltages would have been too high for this project. Anyone building the 2 channel amp would
simply use a 1K2 VA power transformer with windings as follows :-
Primary = To suit the range of national mains voltages 250V, 240V, 230V, 220V, 110V, 115V, 120V,
and suit a frequency of 50Hz, with Bac max of 1.0 Tesla, to keep noise and heat low.
Secondaries for +/- 57Vdc = Two windings of 43Vac x 20A rated.
Best practice is to make TWO MONOBLOC amps, and have
a 600VA PT in each with same
two 43Vac secondaries rated for 10Amps. These are most easily used to make a center tapped
winding using ONE 35A rated bridge.
To make the +/- 57Vdc, the diode bridge will directly charge the rail caps shown on Fig 1A as
30,000 uF, which would in fact be 3 paralleled 10,000uF rated for 63Vdc.
With 2 ohm load on each channel at At full Po of 600Watts, the B+/- rails will sag to
about +/-55Vdc and total PSU power per channel will be around 800W and Idc to each rail = 7.7A.
Therefore ripple voltage at 30,000uF for each channel = 0.57Vrms. A stereo amp using one PT
would need 2 x 60,000 uF caps. In my 300W stereo amp using mosfets I used 2 x 100,000 uF
caps rated for 75Vdc, and made by Sprague with 5mm threaded terminals to take the high
peak charge currents at full Po.
To make the +/- 70Vdc rails for input stages, the two
43Vac windings can used to feed a
C + Diode + diode + C type of voltage doubler to make +/- 116Vdc. The same parts as shown
in Fig 2 may be used but with 220r in series from 16Vdc to regulator transistor collectors.
The 2k2 feeding 15V+56V zener diodes will need to be increased to 8k2.
This shows the arrangement for switching the amp from
stereo to dual mono to bridged.
This is the dc detection schematic for the large bjt
amp. Its an essential item because
if ever there is a failure of a device and the fuses don't blow, 50V dc fed into an
8 ohm woofer will have it in flames withing 30 seconds!
This is my improvised meter schematic using
existing meters. They tended to stick
a bit, and I didn't waste time trying to fix 30+ year old meters designed by accountants.
Originally they were horribly bouncy and very badly calibrated. They were never
designed to indicate real power regardless of load. They are very useful though,
like a fashion statement by a fish swimming around in the sea with a pink umbrella.
To build a circuit that indicates power rather than just voltage takes a little more work.
The above schematic has some partial ability to indicate voltage logarithmically.
Amplifier voltage meters should be logarithmic. This means that they are calibrated
for full output voltage level into 8 ohms for full scale, so then for 1/10 of the maximum
power the meter needle should be about 1/3 of the way across the dial, and for
1/100 of max power the needle is at 1/10 the way across the dial.
After all the work I did on the Phase Linear I
concluded that I would never again use
BJT output devices. And I will never again reform such a recalcitrant amplifier such
as one of these ancient horrors that would ideally be better used as a boat anchor.
So if you have
a bunch of dead Phase Linear amps, don't ask me to fix them;
is what I said in 2004. But I retired in 2012, and if you ask me nicely before 2010,
I could find time to build you a new and better amp instead, using my special tube
powered Time Machine.
there was more work involved with reforming a wayward and fault
Phase Linear than building a brand new amp, and because it was a repair job,
I could not charge much for my work.
Back to Index