What do you get when you cross...

...a balanced line stage with X- feedback with a pair of direct coupled Class A followers using dynamic current sources (or not)?
I asked myself that question, and came up with what I think is a neat little circuit. It eliminates the capacitors on the output of the followers, and uses the balanced line stage to provide voltage gain, and also to bias the followers. Just another variation on the X stuff, but I thought it was interesting. Let me know if anyone wants to see it, and I'll get a schematic together.
A couple of downfalls- it has about 20 volts on both speaker terminals, and it requires major heat sinking, but that's ok, right? We all know by now where to get heat sinks from, correct? ;)
Of course, I am obliged to give credit where credit is due- NP and Grey, for sharing the x concept, and Andrea Ciuffoli for the power follower idea.
I don't think it's really an improvement on the AlephX, but it is at least something new to talk about.
-NS
 

Nelson Pass

The one and only
Paid Member
2001-03-29 12:38 am
20 volts is not a problem at the speaker terminals as
long as both terminals have it.

People need to get over their fear of absolute DC values
on differential circuits. It's no less reliable than an ordinary
amplifier, and if you're that afraid, you'll need to cap couple
your speaker with everything....
 
Thanks for pointing that out, Nelson. I really wasn't seeing it as a problem, but there are those here that do.
BTW, I did get the variable current sources re-configured to work off of current sensing instead of the voltage sensing like I had it before.
Right now I have a design that I might try to build- here are the specs:

Power consuption: about 160W
Output power @ 8 ohms: almost 30 watts
It will do about 20W into 4 ohms (I plan on doubling the total bias later for better low impedance performance)
+30, -12 volt rails (trying to keep dissipation down on the current source- may have to increase the neg. rail)
output stage biased at about 1.8 amps/side.
I plan to use 2 followers and 2 current sources per side.
There dosen't seem to be a lot of interest in this, so I guess I won't post the finished schematic... maybe on my web site later.
 
Use a choke or efficiency's a joke.

nobody special said:
Power consuption: about 160W
Output power @ 8 ohms: almost 30 watts
It will do about 20W into 4 ohms (I plan on doubling the total bias later for better low impedance performance)
+30, -12 volt rails (trying to keep dissipation down on the current source- may have to increase the neg. rail)
output stage biased at about 1.8 amps/side.

With my class A fet source follower I use an 85 mH choke in source leg and with a single +30v rail and 3.5 amps (= 105 watts dissipation) I can *easily* swing plus/minus 28 volts into 8 ohms. That's 50 watts output. I reckon it's worth looking into. Hardly anyone seems to do it this way. It's all current source stuff which of course works great but efficiency is awful. You do have the option though of making the current source non-linear whch can be useful.

GP.

PS. Please post your schematic.
 
Thanks for the interest, guys. I will work on getting a decent schematic together. Look for it this weekend.
I didn't know that you could use a choke like that! What kind of efficiency could you attain with that configuration? It looks like my circuit is about 29%, which I didn't think was all that bad considering it was class A.
NS
 
Chokes are cool.

Well, theoretical maximum efficiency for a class A is 50% at full output. The chokes I am using, in the best tradition of diy ;) are very much overkill to the max. Therefore the losses in it then are very low, just over a watt each at idle. Not sure how much at full blast. The peak mosfet current with an 8 ohm resistive load for me at 28.28 v peak is 3.535 (peak) amps into the load + 3.535 (almost constant) amps into the choke = 7.07 amps total. The Rdson of the fet is 0.13 ohms so it will drop .92 volts. Sooo... a really rough efficiency estimate is 28.28/(28.28+.92)*50% = 48.4%. The DC choke winding loss would bring it down to about say 47%. Have a look at my thread here. http://www.diyaudio.com/forums/showthread.php?s=&threadid=6338 I'm busy constructing and drilling and cutting and suchlike at the moment so I haven't posted much for a couple of days but I may put up a progress picture soon.

GP.
 
...Follower with current sources?

I wouldn't do it .For a follower with no voltage gain i.e. 1 or 0.99 current sourcing probably causes non linearities between variation of voltage on the gate and output variation at the source(through resistor). .....

All because perfect speaker never shows in real life :bawling:

:att'n:

A friend of mine is trying this config right now.I will tell the sonics when he phones me I guess ;)
 
Follower distortion

In this type of topology distortion comes from the followers becase the Ic vs. Vbe (or Is vs. Vgs) relationship is not linear (exponential for BJT & square law(ish) for FET).
So, as the current drawn by the load - even if it is a perfect resistance - changes with signal level it will modulate the Vbe or Vgs and generate distortion.

However, the follower must be biased in class A so the distortion at low output currents is also low.

If you have overall NFB then that will compensate for the follower non-linearity by predistorting the base (gate) voltage.
Alternatively there is the Halcro method (or whoever actually invented it) of simply measuring the follower Vbe/Vgs and subtracting it at the base/gate (with a unity gain amp). This works well in simulation but I haven't built one yet.

Regards
13th Duke of Wymbourne
 
Re: Follower distortion

13DoW said:
Alternatively there is the Halcro method (or whoever actually invented it) of simply measuring the follower Vbe/Vgs and subtracting it at the base/gate (with a unity gain amp).


Hi 13DoW, have a look at this tricky little cct that I re-invented recently. Is that the sort of thing you mean?
http://www.diyaudio.com/forums/showthread.php?s=&threadid=4541
GP.
 
Yep, that's the idea

I didn't read your thread but I recently saw the idea being kicked around in a Halcro amp thread on Solid State. Poster's claimed the idea pre-dates Halcro (hence my earlier comment). It's a beautifully simple idea - so I suspect it's been around for a long time.
I added an idealized correction component (a vccs) to the simulations of a PA design I've been kicking around for a while.
If, say, the 3HD component at the follower output is -60dBc but is -80dBc at the input (ie follower raising the harmonic by 20dB) then with the correction added the 3HD at the follower output is now -80dBc and at the input it is -60dBc (becasue its been pre-distorted). One cool thing is that you could easily switch this mechanism on & off to listen for a difference.

Regards
13DoW