awb50 ,a simple power amp

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
after being asked to provide a simple design, here one i designed
using a very classical topology....some minor tweaking would be
needed to adjust the power stage quiescent current..
the best is to select diodes for the strip used to bias until
adequate voltage gap is found to provide about 40 mA/power device;
this is the solution i used...
the rating is 60W/8 OHM ....it can be used with 4 ohm speakers..
i can t provide a pcb layout, as it was directly build using a veroboard...
 

Attachments

  • AWB50 B1.zip
    10.3 KB · Views: 1,347
Marshall Leach style of amplifiers... the double symetrical differential was very

used dear Wahab, good to factories as they need matching...not very good to build at our home.

The symetry has enormous appeal in our souls..we love symetrical designs, symetrical things, and beauty is something symetrical too.

I really do not like them, the sound is strange, seems to me it is clogged, alike a plumbing where the water does not pass, obstruction!.... but there are folks that loves.

In Brazil, the majority of professional designs use this circuit, there are hundreds of brands, and all them using the topologie.

Once i have tested the Leach amplifier, was a 100 watts unit, connected to 2 15 inches woofer, and the bass was there, strong and precise....when i replaced to a chip amplifier, a TDAXXXX, the bass came strong..the 20 watts chip eated the Leach into the breakfast!

Observe the schematic from Leach amplifier, double symetrical input served by CCS and double VAS, and the output is traditional output...can you perceive they are almost the same..the main frame is almost the same.

PMA is publishing something alike, from John Curl (inspired), and for sure will play the same type of sound...topologie sounds!

Be happy with your amplifier.

regards,

Carlos
 

Attachments

  • ckt.pdf
    121 KB · Views: 1,062
Last edited:
hi carlos, thank you for the feedback...
i understand your point of view, and i d like to add some comments
for the sake of the technical discussion...
so you listened to a leach amp (who is in no way the inventor
of this topology) and found it lacking, not even worth of a ,say
TDA2040....then you blame the topology rather than the one that
implemented the design you used...
at first analysis, i would say that mr leach implementation is
overdesigned... i think the half cascode he use is a mistake, as the
fuctionning point will be stable with no signal, but it s unbalanced in
dynamic conditions...
rather i prefered to make the things simple,so i refrained from using a
cascode, be it with the differentials or with the vas..not even current
mirrors and current sources where used..
now, while there s some benefit in some of these artefacts,
there are serious ill effects as well, not counting the increased complexity..
also, i used a soft compensation that have a 3dB/ octave frequency roll,
thus allowing a slightly higher gain at high frequencies..
sonics are simply very good, and i think far better than those allowed
by assymetric differential (LIN topology).
the LIN has serious drawbacks, like an assymetric slew rate,
more instability, the mainly third harmonic distorsion residual,
contrary to the symetrical differential, although as PA systems for instruments,
they are largely enough, but i would not use them where high definition
audio is needed...a special case is the one of two serial differentials, that is
superior to the classic lin topology, and more stable, at least in my experimentations..

i would like some more feedback about your own experiences, as many
brains will (almost) always be more smart than a single one..


regards,

wahab
 
Let's listen other opinions about...i think i am a rarity, almost all forum loves

The Leach amplifier, and they do not perceive strange things into the sound..the opposite, they love that sonics.

All double symetric amplifier i have built sounded that style..them i give up to build those ones..if double symetrical..them i am out!

Well...maybe mine had errors... it is possible..... why not?

Let's listen other folks.

regards,

Carlos
 
C3 and C9 position...a missed opportunity?

I'm wondering about your choice of value and position for C3 and C9...many would choose smaller values, and using them as miller compensation caps, each
from collector to base of Q1 and Q12. It typically offers better stability and lower distortion, but you may have a better reason for their current position...

Your schematic is so neat, perhaps it is in a simulator already, and you could experiment...

Just curious...

Best Regards,

Dan
 
Wahab,

A nice design, points of interest:

1. Moderate emitter degen on input pairs
2. Shunt Cdom
3. Diode biasing
4. Heavy VAS degeneration
5. Rail decoupling for higher PSRR
6. Drivers switch off at high output
7. Large output degen resistors
8. Relatively low OLG design with low loop gain
9. Low value of Zobel cap, only 10nF.
10. Very high corner frequency for LP input filter (1.6MHz)
11. Rather old VAS and driver transistors!
12. Not sure you need R3 and R26, they do not set LTP balance
13. You designed this in 1984? You must be as old as me!!

I have not yet simulated it but with a symmetrical topology like this the even order harmonics will be far lower than the odds, which is not the best arrangement for easy listening. But it's a good, stable design and no adjustments.

Cheers,

Hugh
 
Last edited:
I'm wondering about your choice of value and position for C3 and C9...many would choose smaller values, and using them as miller compensation caps, each
from collector to base of Q1 and Q12. It typically offers better stability and lower distortion, but you may have a better reason for their current position...

Your schematic is so neat, perhaps it is in a simulator already, and you could experiment...

Just curious...

Best Regards,

Dan

hi, djoffe..

as AKSA noticed it,the compensation is shunt...this allow a 6dB/octave frequency roll off, instead of the 12 dB/octave provided with a lag compensation, i.e , a capacitance from base to collector of the vas...
the goal is to have a higher open loop bandwith so the negative feedback
keeps being efficient at the higher end of the audio spectre..

the design is old, but as you say, i simulated it recently, as the power
devices i used at the time didn t have a FT as high as those in this version,
so i wanted to see if it was stable before making a publication..

all the best,

wahab
 
Wahab,

A nice design, points of interest:

1. Moderate emitter degen on input pairs
2. Shunt Cdom
3. Diode biasing
4. Heavy VAS degeneration
5. Rail decoupling for higher PSRR
6. Drivers switch off at high output
7. Large output degen resistors
8. Relatively low OLG design with low loop gain
9. Low value of Zobel cap, only 10nF.
10. Very high corner frequency for LP input filter (1.6MHz)
11. Rather old VAS and driver transistors!
12. Not sure you need R3 and R26, they do not set LTP balance
13. You designed this in 1984? You must be as old as me!!

I have not yet simulated it but with a symmetrical topology like this the even order harmonics will be far lower than the odds, which is not the best arrangement for easy listening. But it's a good, stable design and no adjustments.

Cheers,

Hugh

hi hugh, thank you for the (positive) feedback, i can only resound at it !!

you have described the philosophy of this amp better than i could
myself do....i d like to comment on the points you smartly made, so
we can confront our point of view, although i m 100% agree with
your words, i tell the reasons of these choice and i will do no more
than expliciting yours views..

1. allow low noise, good dc offset cancelation,
2. less agressive roll off, extended OL bandwith..
3. i hate those VBE multiplier, they are suited to mass
produced amps, a dyier have the time to tweak..
4. to linearise the vas, as low OL gain means low loop gain,
so distorsion cancelation by use of neg. feed back is not as efficient
as in high OLG designs....
5. improving PSSR so the LG is not wasted in correcting
the supply nonlinear behaviours under load..
6. it give better results on the onset of clipping ...
7.originally, there s 4 power devices by rail, beside, this high
degeneration improve the idle current/temperature behaviour and
reduce tendencies to oscillation by the final stage..
8. in the goal of tamming instability and improved OL linearity
allow less feedback for the same distorsion figures..
9. 100 nF + 10/22R as in most schematic is inadequate..
since the time constant of a let say 100nf/10R is 1 uS, and that
at least 3T are necessary to load the capacitance, it s way to
high for amplifiers with rise time of 1 to 2 uS...overshoots can occurs,
and as pointed in the preceding notes, there are other ways for
preventing oscillations that are less "destructive" for the signal...
10. high frequency corner with low imedance source..
i had to take account of the reduced bandwith if one was to
insert a potentiometer at the amp input.
11.yes !! but at the time, they were almost miraculous !
beside, the design has heavy local feed back, so
these bjs are not pushed to their limits..
12. true, i just added them cause it simplify the layout, and it make
the two transistors having similar frequency response..
13. surely! i m 45..
i discovered this topology in 1977 in a review that was discussing an
1975 application note of SGS ATES (now part of ST micro) along with
the ideas of a finnish engineer named matti otala...(no date misprint,
i was in my 14, i began being a diyer at 9, no joke )...
since i was playing music, i did design this one in 1984 for use in stage..
reliability was the main factor, so i made it as light as possible...
i moded it a few years later to adapt hitachi 2SJ48/2SK133 lateral fets..
this is the version i use for more than 15 years, i designed it in a way
that no cmpensaion capacitor is needed, and it s rock stable...
i will later publish this design, but i started with my older part,
as a lot of people (a majority?) prefer the bjts sonics....

if you want to simulate the thing, i can send you an SXSCH file
as i used symetrix to do the drawing and check some behaviours,
so you don t have to bother redrawing it...

thank you for your constructive post,
best regards,

wahab
 
www.hifisonix.com
Joined 2003
Paid Member
This is a good stable topology.

.02% at 20KHz with as little as 20db feedback is easliy acheivable at 20 KHz.

What I love about this topology is that its really easy to create a non slew rate limited design - just run the front end LTP's at c. 10mA with some heavy degen. Result is a very clean, open sounding amp.

Wahab, my feedback to you is try a Hawksford VAS - you will find it gives you about 20 to 30db of lower closed loop distortion for the same loop gain.

Good luck!

:cool:
 
Actually...Miller Compensation is 6 dB/octave

The Miller compensation alternative has the same phase characteristics, mostly. There is a Right Half Plane Zero that complicates things a bit, but that's typically at a rather high frequency.

Did you simulate this in LTSpice? If so, can you make your LTSpice file available so we could play with it, without having to re-enter everything?

Thanks...

Dan
 
This is a good stable topology.

.02% at 20KHz with as little as 20db feedback is easliy acheivable at 20 KHz.

What I love about this topology is that its really easy to create a non slew rate limited design - just run the front end LTP's at c. 10mA with some heavy degen. Result is a very clean, open sounding amp.

Wahab, my feedback to you is try a Hawksford VAS - you will find it gives you about 20 to 30db of lower closed loop distortion for the same loop gain.

Good luck!

:cool:

hi, bonsai..

thank you for your point of view..
yes, there s some benefit running the LTPs at high current..
at the time, i designed versions with cascoding, and they
seemed to work well, i will publish those designs as soon as
i ll make some simulations, as there s serious ill effects on
a cascode , particularly on transient signals..
beside, the more transistors on the chain, the less stability,
and going beyond two voltage gain stage usually do
more harm than good...just curious, how many stage is
your amplifier?..

all the best,

wahab
 
The Miller compensation alternative has the same phase characteristics, mostly. There is a Right Half Plane Zero that complicates things a bit, but that's typically at a rather high frequency.

Did you simulate this in LTSpice? If so, can you make your LTSpice file available so we could play with it, without having to re-enter everything?

Thanks...

Dan

just send me your mail, and i ll send you the file..
be aware that i used simetrix, the file is a SXSCH, and it
seems that it doesn t work on LT spice, although i must admit
i don t know the functionnality of this simulator....otherwise, you
can download the free version of simetrix...
 
Hi Wahab,

I have done the simulations and found excellent results, here:

Gain 26.8dB, -1dB point at 282KHz
Loop gain 48dB at 1KHz
Unity Loop Gain -63 degrees at 1.6MHz (WOW!!)
-3dB dominant pole at 4.3KHz
Phase shift at 100KHz is -13.8 degrees

These are good results, although phase shift is a little high. I have substituted A1360/C3423 for the VAS devices, A1837/C4793 for the drivers, and 4148s for the diode string. I have only 15V zeners, not 18V, so the feed resistor is 3K9. This all gives around 420mA quiescent, which is too high, very deep AB, but the results are very good with very low distortion: -108dB at 20KHz for H5, not bad at all! A pot across one of the diodes would reduce this bias level to a more reasonably 80mA or so..... then I will take more measurements.

Thank you for sharing this circuit,

Hugh
 
Hi Wahab,

I have done the simulations and found excellent results, here:

Gain 26.8dB, -1dB point at 282KHz
Loop gain 48dB at 1KHz
Unity Loop Gain -63 degrees at 1.6MHz (WOW!!)
-3dB dominant pole at 4.3KHz
Phase shift at 100KHz is -13.8 degrees

These are good results, although phase shift is a little high. I have substituted A1360/C3423 for the VAS devices, A1837/C4793 for the drivers, and 4148s for the diode string. I have only 15V zeners, not 18V, so the feed resistor is 3K9. This all gives around 420mA quiescent, which is too high, very deep AB, but the results are very good with very low distortion: -108dB at 20KHz for H5, not bad at all! A pot across one of the diodes would reduce this bias level to a more reasonably 80mA or so..... then I will take more measurements.

Thank you for sharing this circuit,

Hugh

aksa, my friend, a big thank you for giving your time
and your measurements...
my simulator is a free version and it has limited resolution
for distorsion measurements..
you have done the good thing by using better transistors,
as my own spice library is very limited, and i ve not the models
for those high end japanese transistors.

yes,the phase shift is high, although the amp seems very stable
under capacitive load in my tries, so i didn t look to
reduce it...
to adjust the quiescent current, the better is to mix
1N4148 and 1N4004 in the strip of diodes...they are serial,
so it will have no influence in the dynamic behaviour...

by the way, which simulator do you use?..

best regards,
wahab
 
Wahab's Amplifier Design

Here it is, in LTSpice. I have attached it as a txt file; rename to asc, and run in LTSpice IV.

Remove the short from input signal generator for transient, replace and remove from the AC source in the fb loop for AC analysis.

Cheers,

Hugh
 

Attachments

  • Wahab1.gif
    Wahab1.gif
    124.4 KB · Views: 961
  • Wahab1.txt
    10.2 KB · Views: 169
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.