GC SuperSymmetry pt II

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
jh6you said:
Confused:

The amp output is disconnected from the output of the chip. How come??? I hope this is a drawing mistake. Otherwise, I have to admit that I should have studied myself more.

JH

I agreed with you when I first looked at it... But I decided to bite my tongue and defer to the wisdom of NP and look at it in a different way.

Try working backwards. Firstly, we'll assume this is some kind of amplifying device right? If you note that the GCs are NOT connected to the outputs, so what is? It looks a lot like a complementary CE amplifier in bridge mode (well besides the X feedback). So working backwards from there, if you were to take the upper left transistor and assume it's a CE amplifier, you'll get gain by modulating the base voltage. How would the base voltage get modulated? Well, the only way is by varying the current through the associated base resistor... or in otherwords through current draw in the GC...
So, the amplification action happens when the GCs push/pull current at eachother and thusly take it from the supply rails.

One thing I notice is that you need the chipamps to pull quiescent current to set a bias but the more current the GCs pull, the lower their rails go which limits the amount of current they can output... it's pretty likely that ideally you would want very little current to shift between the GCs which also helps stabilizes their rail voltages and most of your gain is going to come from a complementary VAS action. It's already a high-gain configuration.

If I am right then actually any decent audio OPamp would likely do the job because 40mA swing would probably be enough. Although the higher bias current of a GC would be easier to work with.

Just my thoughts...
--
Danny
 
Well... either I'm working in the right direction or gonna go down burning so lets add more fuel..

3 more thoughts.

Firstly, if I was right about the opamps, then a suitable replacement might use something that was unity gain stable instead and not necessarily need the input attenuator. A higher feedback factor so less stage distortion. The downside is that the PSRR of the GC is so good.

Secondly, since the pushpull action of the 2 opamps puts a virtual ground in the middle of the resistor, why not split it to gnd on both ends... because it would require 2 matching elements instead of one unmatched device... ok answered my own question there..

Thirdly, say we use a 8-ohm load, same virtual ground applies to the complementary VAS so in reality, each gain pair sees a 4-ohm load. That means that their gain is actually pretty low. In order to compensate for this, we'd have to use more input signal. So it seems that the real gain mechanism is current pull over the VCC/VSS resistors of the opamps... in which case we would really want 4 nicely matched devices there.
And then following this observation, since the gain of the BJTs isn't critical, perhaps choosing a fairly high Re (so more local feedback) might help reduce distortion in that stage...

The question that remains then is... if the BJTs are more of a current gain device, why not use an EF instead of a CE?
--
Danny
 
I see that differences in bias current and DC offset which is essentially non-symmetric bias current would require that the VCC/VSS resistors be adjustable, but once you do that you would then be disrupting atleast some of the gain, much more signifcantly than any minor stage distortion... very well matched devices with very low DC offset is a requirement...
 
It would hurt except that I've had lots of episodes with friends where we would come up with some totally absurd idea that would almost seem like it would work... talk about it for a while and then dismiss it.
one example would be using a couple of diving tanks to make a water cannon on the back of my Jeep to siphon water from the river and shoot it back at passing rafters...
something like "lets take the current draw from the supply rails to modulate the gain of an amplifier" seems a lot like that, I'm conditioned to the shock...
Just one problem, NP actually wrote it down...
--
Danny
 
Actually, on my way in to work this morning, I was thinking about it again and I see a potential problem.

I previously stated that you would perferably want minor current variations around the rails because as the opamp drives higher output voltages, it draws more current therefore lowering the availible output voltage to drive.

However, I also later on concluded that the main gain mechanism must be from the modulated current on the supply rails.

I suspect that the first effect would provide a absolute maximum voltage swing of 1/2 VCC limiting the gain... perhaps if one could squeeze some gain out of the output like 2 or 3 then this would be mitigated, but now we're talking about using a speaker as a gain setting device.
--
Danny
 
The one and only
Joined 2001
Paid Member
If you forget the "SS" parts of the circuit, you'll see that we
really have just added a standard sort of current booster -
as the op amp delivers positive current at its output, it draw
current from the positive rail, which increases the drive to the
PNP power transistor, which send the voltage and current
positive at the Collector of the transistor, which is attached to
the output load. :cool:
 
Disabled Account
Joined 2002
You is still silly:

Honestly, I still can’t understand well. Conjugate complementary...??? Means that the DC bias current represents the real part while the AC current the imaginary part of the conjugate complex number?

When there is a life of AC current in X- shape, the central point sees the symmetric resistance if there is no speaker impedance. If so, the output AC voltage is the one we see across the speaker impedance (i.e. Vac, say, by transimpedance)?
 
The one and only
Joined 2001
Paid Member
When I say conjugate complementary I am using the (probably)
antiquated term which decribes the output stages of amps
like the early Motorolas or TigerSauruses and such. The drivers
are NPN on the positive side and PNP on the negative side and
their Emitters attach to the output of the amp through resistors.
They are driven at their Bases with a bias network - standard
stuff. Coming off the Collectors of the drivers we meet the Bases
of the actual output devices, a PNP on the positive side, and an
NPN on the negative side. The Collectors of the outputs look
at the amp output, and their Emitters look at the supplies.

You know the one I'm talking about. :cool:
 
I'm still not convinced this is a really feasible idea... where does the gain come from? Modulating current over VCC_R and you shoot yourself in the foot with supply voltage. Use the BJT and you're using a non-constant impedance to set the gain, not to mention that you have to raise bias current very high in order to get a respectable gain so that feedback is useful.
Or maybe that's the trick........ 1A bias and 2-ohm pessimistic guess, and you'll get slightly under 80V/V openloop with degeneration, I guess that's not so bad... What about thermal runaway?
--
Danny
 
The one and only
Joined 2001
Paid Member
The use of supply draw of an op amp to drive output devices
is a pretty old idea - I seem to recall seeing it in the old
National app notes, and I was also taught that technique by
the quasi-legendary Charlie Button.

Yes, it works. It does require an output load (thoughtfully
provided by the SS connection between the outputs of the
two chip amps) and a good PS rejection ratio on the part of
the chips (of which they have a surfeit).

You will also see this technique appearing in designs under the
thread "Monolithic SuperSymmetry", where it has been
successfully constructed.

:cool:

Remember, there's a cool prize if I'm mistaken.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.