Here is the latest creation (re-creation ) from the "badger hole" .
- A really stable triple OPS - I've swapped outputs , hooked up without a zoble ...
all the abuse.
- This amp is 32 years old and I'm still abusing it daily (just needed a re-cap).
-My LT simulation worked first time , needed to tweak the unconventional compensation scheme - it never exhibited any instability even without the "tweak" .
- Since I actually own the real live amp (and listen to it a lot) , I was able to
compare LT's output to the real world operating points - 99% accurate.
Below are the "goods"
The schematic and the .ASC (spice file)
PS - the PCB is next , I REALLY like this little amp. (try to guess who designed it ? )
OS
- A really stable triple OPS - I've swapped outputs , hooked up without a zoble ...
all the abuse.
- This amp is 32 years old and I'm still abusing it daily (just needed a re-cap).
-My LT simulation worked first time , needed to tweak the unconventional compensation scheme - it never exhibited any instability even without the "tweak" .
- Since I actually own the real live amp (and listen to it a lot) , I was able to
compare LT's output to the real world operating points - 99% accurate.
Below are the "goods"
The schematic and the .ASC (spice file)
PS - the PCB is next , I REALLY like this little amp. (try to guess who designed it ? )
OS
Attachments
Member
Joined 2009
Paid Member
I see you are back in the saddle on this one
Looks like a good solid performer from the schematic. All those LTP's - could include a balanced input option ?
I'm not , but sansui sure was in 81' .
Here's some specs :
- .02% THD20 on the simulator, .03% rated in the OEM manual.
.002% full power @ 1K
- -107db noise. My real amp is noiseless (ear to tweeter-full volume).
I'm surprised they put such a good amp in a "mid-fi".
With my digital source(SB live -PC) , this really sounds wonderful on the Dan Richie audiophile 2-ways in the workshop.
(below link and attachment) X-LS Classic kit
The circuit - I've found that by loading the VAS (R25 -47K) you can control
the H2 / H3 balance in the Fourier results. 22k gives H2 dominant at the
expense of total THD. 33K seems to be the sweet spot , at least for these models.
The real amp uses 2sa992 /2sc1845 almost exclusively , 50VDC rails.
The longevity of this circuit is amazing .... found it in the college of Saint Rose dumpster
( many beer- soaked "enhancements" of "higher education" ).
It worked at lower volumes even with dry electrolytics -amazing.
The compensation is "weird" . I know C17 (lead comp.) and it's effect/purpose.
I'm curious about R7 /C5. I don't know what this network is called , but it greatly contributes to this amps stability.
The amp also exhibits the best clipping behavior , I would guess most of
32 years of clipping .
OS
Attachments
Could the board occupy just 1/3 of the UMS so we can make a 4-6 channel GLA in a store 4u case? That would improve my system's SAF.
I would just go for a couple RSA's. (really sarcastic acronyms)
At least I gave a circuit , unlike the "revolution class a/b thread" ... with
78 replies and over 3000 views.
Too many EF2 OPS's around here.
If I scale this up to a mje340/350 - njw0302/0281 - mjl21193/94 output stage,
perhaps builders will be a little less apprehensive of triples.
OS
Member
Joined 2009
Paid Member
The Pioneer amplifier I moded into my TGM6 amplifier (there's a thread with schematics somewhere around here) which also has this 'R25', same value in fact. The load on the VAS was a key part of the compensation (like your R7 C5) as it also had no Cdom. The lead comp does seem a little large at 22pF, but not by much.
I love the speakers - red, great colour. I hope these weren't punished with 32 years of clipping !
I love the speakers - red, great colour. I hope these weren't punished with 32 years of clipping !
Ahhhh .... while others are asleep and fussing over the "revolution" , I
continue to dissect this cool circuit.
Attachment 1 shows the clipping behavior. At least it is symmetrical , but
shows a small amount of "sticking" . The edges are rounded , not as bad as
"blameless" clipping.
Attachment 2 is driving a capacitive load. I had to really coax this amp to
give me a damped ringing response. Bridged the OP inductor , lowered the
load resistor. VERY stable ... I've not simulated too many like this.
All this is reflected in the real amp , 32 years with original semi's is a pretty good track record .....
This must become a project !
OS
continue to dissect this cool circuit.
Attachment 1 shows the clipping behavior. At least it is symmetrical , but
shows a small amount of "sticking" . The edges are rounded , not as bad as
"blameless" clipping.
Attachment 2 is driving a capacitive load. I had to really coax this amp to
give me a damped ringing response. Bridged the OP inductor , lowered the
load resistor. VERY stable ... I've not simulated too many like this.
All this is reflected in the real amp , 32 years with original semi's is a pretty good track record .....
This must become a project !
OS
Attachments
The Pioneer amplifier I moded into my TGM6 amplifier (there's a thread with schematics somewhere around here) which also has this 'R25', same value in fact. The load on the VAS was a key part of the compensation (like your R7 C5) as it also had no Cdom. The lead comp does seem a little large at 22pF, but not by much.
I love the speakers - red, great colour. I hope these weren't punished with 32 years of clipping !
That was the "stock photo" of the AV123's , I have piano gloss black ones "made in Cali,Columbia"
(checked the boxes for contraband - found these
outside a local tavern - perfect condition).
I'm nice to my AV123's - I use them as a reference to hear the imaging and
subtleties of my amps , for clipping I use my "bombed out" realistic 3-ways
(10"/5"/3") - only $20 , even with new woofers.
I haven't done much amp stuff lately , I have been increasing my loudspeaker
"collection".
OS
Too many EF2 OPS's around here.
If I scale this up to a mje340/350 - njw0302/0281 - mjl21193/94 output stage,
perhaps builders will be a little less apprehensive of triples.
OS
Way too many. Especially when they parallel a dozen or so outputs and think they can drive an ohm. And then wonder why the VAS runs out of gas at normal current levels. I hate to run VAS's at 50mA because they burn up the PCB and drift all over the place. Much better to use EF3 with a couple miserable mA in the predriver.
A lot of builders have trouble with triples (tribbles?) because they insist on doing weird things with them. A straight complementary EF3 is pretty darn stable with either high or low fT parts in the driver/predriver.
Way too many. Especially when they parallel a dozen or so outputs and think they can drive an ohm. And then wonder why the VAS runs out of gas at normal current levels. I hate to run VAS's at 50mA because they burn up the PCB and drift all over the place. Much better to use EF3 with a couple miserable mA in the predriver.
A lot of builders have trouble with triples (tribbles?) because they insist on doing weird things with them. A straight complementary EF3 is pretty darn stable with either high or low fT parts in the driver/predriver.
I came to these same conclusions after reviewing some of the more popular
threads that use those EF2's with 5 or more OP's and claim 2R loads.
In the attempt to not "run out of gas " VAS wise , some are using drivers
to do the job. Besides the much higher Cob of these devices (mje15032/33 or
2sa1837/2sc4793 -not too bad) , you also have a limited choice of a 1A SOA drivers to run 5 or more pairs of "hungry" outputs.
Moon audio and other overpriced high-end amps use outputs as drivers. You can't (or shouldn't)
do this with the standard EF2.
I have the opinion that 3 pair of outputs is the maximum for the EF2.
OS
Very nice and refreshing amplifier considering the "current"
subjectivists tendencies , it should allow very good perfs
that might well surprise some esoterics design afficionados...
No doubt that i ll give it a try in the simulator...
Stability would be greatly improved with say a 22pf from
collector to base of Q7.
Also , what happened to Q7/Q11.?.?..
subjectivists tendencies , it should allow very good perfs
that might well surprise some esoterics design afficionados...
No doubt that i ll give it a try in the simulator...
Stability would be greatly improved with say a 22pf from
collector to base of Q7.
Also , what happened to Q7/Q11.?.?..
Sansui 3900/5900
Here is the "truth" .
Attachment 1 is the amp section. The input stages (I tested them) are run by
a +/- 18VDC supply. It is strange they regulate the positive 18v from the amps main
V+ rail and derive just the negative 18V for Q1-2 from an auxiliary supply.
I might "clean up" the differential supplies and create two zener regulated
ones instead , it will make for a "standalone amp".
This is really a Sansui Z3900 , the PDF is too big.
You will notice the Sansui had a DC servo - I changed this into it's passive
equivalent.
OS
Very nice and refreshing amplifier considering the "current"
subjectivists tendencies , it should allow very good perfs
that might well surprise some esoterics design afficionados...
No doubt that i ll give it a try in the simulator...
Stability would be greatly improved with say a 22pf from
collector to base of Q7.
Also , what happened to Q7/Q11.?.?..
Here is the "truth" .
Attachment 1 is the amp section. The input stages (I tested them) are run by
a +/- 18VDC supply. It is strange they regulate the positive 18v from the amps main
V+ rail and derive just the negative 18V for Q1-2 from an auxiliary supply.
I might "clean up" the differential supplies and create two zener regulated
ones instead , it will make for a "standalone amp".
This is really a Sansui Z3900 , the PDF is too big.
You will notice the Sansui had a DC servo - I changed this into it's passive
equivalent.
OS
Attachments
Last edited:
Zener supplies
Very good !
Lower distortion .... .008% THD20K 8R
Used 15V zeners with 2 simple follower trannies (below).
One pair of differentials (Q1-2) sees 12v and the other (Q3-4) 35V.
Many choices here !
Sansui made some economic choices and just "stole" existing supplies
for the differential to save $$$$ .
Much more stable , I can change the Q1-2 Ic over a much wider range without
oscillation now ,as well.
PS - scaled down R8-9 to 3K3 to keep things "vintage".... (-14.4VDC instead of -18VDC)
OS
Very good !
Lower distortion .... .008% THD20K 8R
Used 15V zeners with 2 simple follower trannies (below).
One pair of differentials (Q1-2) sees 12v and the other (Q3-4) 35V.
Many choices here !
Sansui made some economic choices and just "stole" existing supplies
for the differential to save $$$$ .
Much more stable , I can change the Q1-2 Ic over a much wider range without
oscillation now ,as well.
PS - scaled down R8-9 to 3K3 to keep things "vintage".... (-14.4VDC instead of -18VDC)
OS
Attachments
Last edited:
OMG -Q7 and 11
But it actually simulated right -
I fixed this - and wow ! .001% 20k.
- Tiny glitch in the stages is gone , and the loop gain is "more normal"
(typical) 500k UG with 80+ margin....
-Q1-2 nearly superimpose (balanced)
-Q1-2 are 1ma , Q2-3 are 2ma , the VAS is 3ma.
The actual amp measures real close to this.
This is almost ready to FLY! (pcb)
PS- this amp's topology was actually able to compensate for 2 swapped E-C's !
New .ASC (below)
OS
But it actually simulated right -
I fixed this - and wow ! .001% 20k.
- Tiny glitch in the stages is gone , and the loop gain is "more normal"
(typical) 500k UG with 80+ margin....
-Q1-2 nearly superimpose (balanced)
-Q1-2 are 1ma , Q2-3 are 2ma , the VAS is 3ma.
The actual amp measures real close to this.
This is almost ready to FLY! (pcb)
PS- this amp's topology was actually able to compensate for 2 swapped E-C's !
New .ASC (below)
OS
Attachments
Last edited:
Hi, OS!
Very nice amp, indeed.
Probably I did not get it right, but it isn't a little too much 50 V rails for a single pair? I see that you have used good robust ONSemi outputs, but you can be sure that sooner or later somebody will hook some 4 Ohm speakers on it.
I would be tempted to put 3 pairs on output and call it a good 250 Watter!
Very nice amp, indeed.
Probably I did not get it right, but it isn't a little too much 50 V rails for a single pair? I see that you have used good robust ONSemi outputs, but you can be sure that sooner or later somebody will hook some 4 Ohm speakers on it.
I would be tempted to put 3 pairs on output and call it a good 250 Watter!
Hi, OS!
Very nice amp, indeed.
Probably I did not get it right, but it isn't a little too much 50 V rails for a single pair? I see that you have used good robust ONSemi outputs, but you can be sure that sooner or later somebody will hook some 4 Ohm speakers on it.
I would be tempted to put 3 pairs on output and call it a good 250 Watter!
I'll design a single pair MJL21193/94 version first. Using the triple OP stage
demands it. That should give about 80W/channel (8R) ,the MJL's soa would also be good enough for a 4R load at 40V rails.
Solved the compensation issue. Standard miller comp. with a B-C capacitor
won't work with this amp !
Phase compensation -(R7/C5) of the first differential plus a little lead compensation from the VAS (R6) let's you "shape" the phase margin and unity gain point of the amplifier response curve. Harder ! , but a more elegant
approach. Original "vintage" (47R and .0022uf) values still gave a stable amp.
(below attachment)... this amp is compared to the standard "blameless".
Latest .ASC below - (with the "vintage" parts modeled) the performance came out real close to the OEM specs - but just a little better.
OS
Attachments
GLAv1.2 (final)
All is wonderful ...
V1.2 is SO close to the Sansui 3900 , both in compensation and actual voltages.
(Attachment 1) is the final work.
My tweaks to C6 (lead comp.) and the "lag" comp. (R7/C5) are within range
of the original specs.
The simulation also allows for higher Hfe Zetex devices , 2n5401's , 2sa992's
or any other to-92 device for Q1-4.
Any to-220 (mje15032/33 , 2sa1837/2sc4793) as driver , and any OP device
..including the low gain/low Ft mjl21193/4... work very well with this amp.
R25 and 26 load the VAS , which gives a monotonically decaying harmonic
output.
The differentials and VAS are at such low current (no heatsinks),
I can squeeze this into a 76 X 102 mm board for the 70W version (attachment 2).
The main board will use njw TO-3P's as drivers !!
For those who want the "Frankenstein" version , attach the "GLAeX" PCB
(attachment 3).Run your 80V rails, you will have a 400W true triple OPS.
Also , the final .ASC is also below.
PCB's + the BOM will be done in the next few days.
OS
All is wonderful ...
V1.2 is SO close to the Sansui 3900 , both in compensation and actual voltages.
(Attachment 1) is the final work.
My tweaks to C6 (lead comp.) and the "lag" comp. (R7/C5) are within range
of the original specs.
The simulation also allows for higher Hfe Zetex devices , 2n5401's , 2sa992's
or any other to-92 device for Q1-4.
Any to-220 (mje15032/33 , 2sa1837/2sc4793) as driver , and any OP device
..including the low gain/low Ft mjl21193/4... work very well with this amp.
R25 and 26 load the VAS , which gives a monotonically decaying harmonic
output.
The differentials and VAS are at such low current (no heatsinks),
I can squeeze this into a 76 X 102 mm board for the 70W version (attachment 2).
The main board will use njw TO-3P's as drivers !!
For those who want the "Frankenstein" version , attach the "GLAeX" PCB
(attachment 3).Run your 80V rails, you will have a 400W true triple OPS.
Also , the final .ASC is also below.
PCB's + the BOM will be done in the next few days.
OS
Attachments
- Home
- Amplifiers
- Solid State
- GLA (good little amp) "survivor"