50W, Class A, another one... :)

[keantoken]

Oh, God.

I am having problems.

I just welded a light bulb to my hemostats.

The sparks were pretty.

I should do this more often, to strengthen my heart and improve cardiovascular health.

I have been trying to recharge my lead-acid bench batteries, but they may be completely discharged by the time I get them connected again.

Okay, I just got my Hemostats unwelded. Luckily it just snapped off. Wow, if I get enough practice with this, maybe I will never need solder. I'll just weld the wires directly together.

That is an interesting idea. Connect an inductor (+ parallel diode) in series with a battery to prevent sparks, and perhaps you now have an instant solder device! Nah, it would have been done before...

Well, batteries are charging yet again. Tomorrow I might be able to resume work on my project.

- keantoken

 

[kenpeter]

I've discovered a possible problem in most of my designs...
Voltage clipping problem may be handled elegantly, but
slew rate clipping still seems to be very ugly.

You had spotted it before I ever did. When you tried the
96Khz dummy loop as an input source, turned garbage.
Good clue that I might have needed faster slew to avoid
transient intermodulations. It should have handled that
fast a pure sine, and it didn't.

Yeah, its true we don't hear much if anything above 20KHz.
But we rarely listen to just one pure sinewave by itself.
And I'll illustrate a test case where this might be relevant:

If one were to play 1KHz, 3KHz, 5KHz, 7KHz and all the
other odd harmonics below 20Khz that sum to a square
wave as one example. I think the worst case slope is a
lot faster than a simple 20KHz sine all by itself.

And my circuits weren't handling even 20KHz rail to rail
without some strangeness. The slope of the amplified
error at the collector of the backward Allison transistor
needs to be many times faster than the music signal.
When it can't keep up, it goes sorta nutzo instead....

We can kind of get away with it not being fast, as it
dumbs itself down to an autobias circuit. Even when
it can't react quick enough to behave as Allison/Aleph
error correcting base followers, the output transistors
should revert to simple uncorrected emitter followers.

But the mess mine makes trying too hard to be fast,
still somehow not keeping pace, is whats killing me...
I think I need to work on how to make it transition
more cleanly: From fast base follower below 20KHz
to slower emitter follower above. But not getting all
weird on me inbetween...

 

[keantoken]

That design with the current mirrors+Allison+MOSFET has some issues, is what I noticed the most. It would be best to give the current mirrors some gain, so they can self-correct. This is jus off of memory, I might be completely wonky here...

Well, I just wired my first mains transformer. Now I have low-current 60, 20, and 80V pk-pk outputs (WTF? I got his out of a radio receiver, for crying out loud... Brand name is Sansui)

I hooked everything up, covered the dangerous parts with heatshrink, checked connections, turned it on, and waited for something to explode.

Thank God it didn't. Now I am happy. Now to test out with a rectifier and caps.

What would I need 80V low-current for? Avalanche pulser?
I wonder how much HV power I could tease out of it with a large voltage multiplier?
Well, I suppose I could power some tubes if I hooked it up to a multiplier... That's an idea.

I will be ordering a toroidal 15-0-15 50VA trafo, but nothing can really happen until it gets here... Besides some demolition/mad scientist experiments.

- keantoken

 

[kenpeter]

Everytime I tried bootstrapless mirror drives with any more
current gain, it broke into oscillations... Never quite fixed that.

Actually every time I try bootstrapped Alephs with Sziklai or
Darlington followers to enhance the slew of the gate drives,
it goes just as weird for me.

I can't seem to gain faster local slew after the comparators,
at least not with any stabilty. I need to drive more current
swing directly into the comparator emitters, no way round it...

But I've also been playing with JFETs for the comparators
and diamond buffer. I'll show you here in a minute. They got
some of the same issues, but less tail current if you should
accidently drive one into cutoff (as happens in clipping).
They also sim far more stable, but that may just be LTSpice...

But the main diff, I'm pushing up the quiescent current of my
comparators from 4mA to 8mA...

 

[keantoken]

That's neat!

I tried using Jfets in just the normal Allison style but it didn't work and oscillated. I need to experiment with FETs more.

Note: Those diodes actually decrease the gain of the Allison feedback loop, which should increase stability at the expense of distortion.

I just ordered two 12.6V, 3A transformers, so I'm ready to go once they get here.

- keantoken

 

[kenpeter]

Ken,
I like your bootstrapped current source for the Allison bias transistors.

That particular style bootstrap circuit cloned straight from
Nelson Pass' Aleph. I can't claim any original thinking there...

Though its quite unusual to abuse an Aleph as a voltage
follower, much less two of them in compliment. Though
even this idea (with BJTs) seems to have be Allison's.
And just who exactly is/was Allison?... I can't come up
with anything new, been invented before I got there.

 

[kenpeter]

For JFETs as comparators, you need to make sure the threshold
of turn off is less than VGS-ON for the MOSFET to be controlled.

The ones I specified are fairly low threshold, and available in
matched pairs (by a different number 2sk389/2sj109). This
makes the sums across the diamond equal in both branches.

But if the MOSFET thresholds aren't as well matched, you still
need jigger with the little resistors in the middle that set the
offsets. There is nothing "automatic" like a bipolar threshold.

The JFET models are Toshiba I think.

.model J2sj74 PJF(Beta=92.12m Rs=7.748 Rd=7.748 Betatce=-.5 Lambda=4.464m
+ Vto=-.5428 Vtotc=-2.5m Cgd=85.67p M=.3246 Pb=.3905 Fc=.5
+ Cgs=78.27p Isr=129.8p Nr=2 Is=12.98p N=1 Xti=3 Alpha=10u Vk=100
+ Kf=26.64E-18 Af=1)

.model J2sk170 NJF(Beta=59.86m Rs=4.151 Rd=4.151 Betatce=-.5 Lambda=1.923m
+ Vto=-.5024 Vtotc=-2.5m Cgd=20p M=.3805 Pb=.4746 Fc=.5
+ Cgs=25.48p Isr=84.77p Nr=2 Is=8.477p N=1 Xti=3 Alpha=10u Vk=100
+ Kf=111.3E-18 Af=1)

 

[kenpeter]

Those diodes are not there for stabilty, they are for class AB.

You can't have smooth AB if the voltages at top and bottom
of the pincers have a constant offset, and ours certainly do.
This only allows class A till one or the other clips, and then
the transition will be a hard one.

You can look at my Schottky as current variable voltage drops,
and thus pincers that "give".

 

[keantoken]

I was just writing my thoughts on the diodes, nothing more. I probably should have clarified.

The affect of making it Class AB is nice. However I don't know if it is possible to reduce distortion apart from component selection, however. The Allison works well assuming we use perfect resistors. If the the resistors aren't perfect, then their VI curves must cancel. The VI curves of the diodes don't cancel perfectly, and should leave dominant 3rd harmonics.

We could probably reduce distortion if we bypassed the diodes with caps. Perhaps we could leave in some "desirable" 3rd harmonics in the bass region

Please don't mind my usage of "we". I'm not implying anything. I just write that way...

And just who exactly is/was Allison?... I can't come up
with anything new, been invented before I got there.

...
I know the feeling.

- keantoken

 

[keantoken]

Okay. I got my trafos today, and have been trying to rig up my power without exploding something. After a few times of accidentally discharging my 8800uF filter caps, I decided to go the safe route and safely discharge them beforehand with a 4 ohm power resistor.

My cardiovascular health is slowly improving.

I was thinking one day and thought of this. It's extremely simple and probably unnecessary, but I thought it was interesting enough to post.

Normally, we prefer to set R1 and R2 at the highest possible values so as to not waste any more power.

Assume that we make R1 and R2 only as large as is needed to push 20V into an 8-Ohm load. Because of the roughly 45mA contributed by the driver transistors, the output transistors will actually turn off at near max output, which causes a disturbing spike in distortion. To fix this, we add R23, which gives enough bias current to prevent this. Because of the constant voltage across R23, there will be constant current through it also.

I don't know why you wouldn't just decrease R1 and R2, unless you were concerned with precision. But just to throw this out there. I like it because it looks elegant.

By the way, I read about how you can reform electrolytics to operate at a higher voltage. Are there any bad consequences if you take a bunch of 12V caps and reform them to work at 24V?

- keantoken

 

[AndrewT]

By the way, I read about how you can reform electrolytics to operate at a higher voltage. Are there any bad consequences if you take a bunch of 12V caps and reform them to work at 24V?

I asked this about a year ago and got no response.
Can you give a link to the information you found?

 

[AndrewT]

Normally, we prefer to set R1 and R2 at the highest possible values so as to not waste any more power.

this confuses me.
I thought that making these two resistors of lower value saved wasted power.

 

[keantoken]

It was in a book called "electronics the easy way". I can't find it because the author mentioned it seemingly on a side comment.

Here is a link.

http://www.coilgun.info/mark5/capacitor_bank.htm

- keantoken

 

[sawreyrw]

High voltage electrolytic capacitors that have not been used for 1 year or 2 need to be "reformed" to condition the dielectric. It is absolutely not intended to allow the caps to be used above their rated voltage. Don't try it, because when high energy caps explode they can be very dangerous.

 

[kenpeter]

Regarding R23, I also have that in parallel with my AB Schottkys
for the same reason. But I'm abusing 2SK170 as an 11mA CCS.
Actually 4 in parallel, gates strapped directly to source. Or with
2.2 or 3.3 ohms under each source.... Higher dynamic impedance.

Its enough to keep the gate drive signal from dropping below
the 4 or so volts needed near pinchoff. And keep my comparator
from saturating, if its a bipolar type... I can get away with only
one such 11mA CCS, if I pump bootstraps from a sidechain amp.

Also got Zenier from collector to base (or gate) of my comparator
to handle the other extreme. The leaking Zenier then keeps the
comparator from ever being able to turn fully off. The gate of the
power device wouldn't respond to more than 8 Volts anyway...

I don't usually post my versions cluttered with extra junk that
keep transistors from misbehaving. I can if you want, but makes
harder to illustrate the core topolgy.

 

[keantoken]

this confuses me.
I thought that making these two resistors of lower value saved wasted power.

Making them lower will increase bias current, increasing output power, but if don't need that output power then it just takes up space. Lowering current draw lowers unused power in this case. Since voltage without current is not power, we are not saving any energy that was wasted otherwise.

Regarding R23, I also have that in parallel with my AB Schottkys
for the same reason. But I'm abusing 2SK170 as an 11mA CCS.
Actually 4 in parallel, gates strapped directly to source. Or with
2.2 or 3.3 ohms under each source.... Higher dynamic impedance.

Its enough to keep the gate drive signal from dropping below
the 4 or so volts needed near pinchoff. And keep my comparator
from saturating, if its a bipolar type... I can get away with only
one such 11mA CCS, if I pump bootstraps from a sidechain amp.

Also got Zenier from collector to base (or gate) of my comparator
to handle the other extreme. The leaking Zenier then keeps the
comparator from ever being able to turn fully off. The gate of the
power device wouldn't respond to more than 8 Volts anyway...

I don't usually post my versions cluttered with extra junk that
keep transistors from misbehaving. I can if you want, but makes
harder to illustrate the core topolgy.

The protection circuits are just as important as the real thing IMO. I welcome any ideas you have to keep it from "misbehaving". That said, I usually don't have a problem wading through such circuits to get to the core. So feel free to post any schematic, no matter how convoluted. I can redraw it in LTSpice.



IT'S ALIIIIIIIIIIIVVVVVEEE!!!!!!!!!

Just got the output stage working. The bias resistors are quite out of tolerance, so I have 200mV out put offset. No oscillations so far, and Allison Vbe reads 1.33V, exactly as specified in the simulator.

I will hook it up to a 16-Ohm speaker and a compact CD player and see if it plays well.

+-20V unregulated rails, 8800uF cap per rail.

- keantoken

 

[AndrewT]

"Normally, we prefer to set R1 and R2 at the highest possible values so as to not waste any more power. "

Making them lower will increase bias current, increasing output power, but if don't need that output power then it just takes up space. Lowering current draw lowers unused power in this case.

that makes much more sense.
You were referring to ClassAB optimum bias setting and thus Pq of the output stage.
What has that to do with a ClassA amplifier?

 

[keantoken]

I wasn't thinking AB at all. I've not read enough theory to know what Pq is (but I'm pretty sure I can guess).

In my latest schematic, R20 and R21 put about 45mA of bias current through the drivers, AKA Q10 and Q11. If we reeeaally scrounge on the bias resistors, then the output transistors (Q8 and Q9) will turn off when the output current gets within 45mA of the max. To correct this, R23 can be used and will cause a constant 45mA to be drawn, ensuring that the output transistors never turn off.

Hmm. That's an interesting idea. Bias the outputs at class B or AB, and have the drivers be the only class A part. I wonder if this can work? Unfortunately, it would take some tweaking of BE resistors.

I hope this clears things up,
- keantoken

BTW, I hooked up the LTP and VAS to the Allison. I got some sound, but bad oscillations ensued. I will redo the layout tomorrow if possible, with some changes, and try again. Too bad I only have ceramics to use for Cdom.

 

[AndrewT]

if one or the other of the output transistors stop controlling the output current then the stage is not ClassA.

A bit like a single ended ClassA stage transistor always controls the output current, then similarly the push pull stage requires both transistors to always control the current turning one either off on to a low near constant current while the other is left doing all the work is ClassAB.

 

[keantoken]

if one or the other of the output transistors stop controlling the output current then the stage is not ClassA.

A bit like a single ended ClassA stage transistor always controls the output current, then similarly the push pull stage requires both transistors to always control the current turning one either off on to a low near constant current while the other is left doing all the work is ClassAB.

Don't worry, it's fully class A.

When I get his working on breadboard, I will try to work out a PCB layout. From there is history, depending on what the measurements are, impressions, etc.

- keantoken