Ultrafast NoCap-ClassAB

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hi Guys !

I thougt, try something different, a ClassAB amplifier without ANY
feedbackcompensation-cap. This is my result...
I had to use tripledarlingtonoutput because of the low current (2ma)
through the folded cascodes, higher current would give too low gain.

Some specs: (out of sim)

THD:
0.003% 20khz, 1watts into 8ohm
0.006% 20khz, 9watts into 8ohm
0.05% 20khz, 64watts into 8ohm (full power)

Distortionfigures for 1khz are "identical".

Frequencyresponse:

-3db: 1.66hz to 820khz
-1db: 3.25hz to 413khz

Prototype is running without oscillation, sounding very impressive !
The sound is completely uncolorized and clean, but pleasant.
It's the most detailed amp i've ever heard, the reproduction of
voices similar to the DoZ.
The music played with this amp is somehow "live"...

What do you think ?

Mike
 

Attachments

  • focasym1.gif
    focasym1.gif
    15.1 KB · Views: 2,886
I think you should test it thoroughly on the bench. Have you a function generator and a scope? You may find it is unstable into capacitive loads - change the input cap from 100p to 470p, then try driving it with a 1kHz squarewave into a few capacitors ranging from 1nF to 10uF. Vary the amplitude. Be prepared for oscillation!
 
That's an interesting schematic Mike. Maybe I'll try it if have some spare time, I can fully test it with 'Audio precision system one'. Although I doubt that there is no need in ANY frequency compensation. By the way are you sure that collector loads of differential stage are shown correctly? I may be mistaken but from the first look, with 330 Ohm there amplifier is in shutdown mode, have to be closer to 2,2kOhm :confused:
 
Hi Mike,
I'm glad it sounds really good. The zippy response of the feedback is helping to mitigate the non-linearities I'm just warning you that, in my experience, uncompensated circuits of this topology have a tendency to oscillate at several MHz if their load becomes too reactive and this can lead to the power transistors overheating, and short circuiting and then your speakers receive dc and you have a sense of humour failure at this point (yes, I've done this myself!). Even with an output inductor I imagine the amp would become excitable or just plain oscillate into highly capacitive loads. Speakers present a variable reactive load to the amp - and the extent of this will depend upon the speaker. Just be careful not to damage anything by accident.

Without a scope it is almost impossible to see what is going on. You could try attaching an AC voltmeter across the output, with no input, and apply caps across the output and watch the reading. If it rises then there is some oscillation. Unfortunately, the slow response of the voltmeter means you may not realise it is oscillating for a few seconds by which time the output devices MAY overheat. So best when doing a capacitor test to limit the current to the output transistors in some way - like with series resitors of 10-ohm or so. You may burn some resistors this way, but not the output transistors.

If you don't have the test gear I'd just listen carefully to it and listen out for signs of instability. Sometimes you get a roughness on some sounds - usually treble instruments or strong women's voices, sometimes snare drums. This can often be a pronounced crackly sound. Sometimes you get instantaneous dips in volume. Without any music playing try sharply tapping the woofer cone of your speaker with your finger and listen for chirping. The extend of the symptoms will be slightly different in each amp you build because the transistor parameters and parasitic capacitances are different.

If all is well then don't worry about it. But be careful to reassess if you change the speakers for a different pair or change the speaker cables in case the new loading sets it off.

Happy listening :)
 
Hi Alme !
Yes, the 330ohms are correct, this is folded cascode, behaving very
different ! The lower these resistor, the higher the current in "vas"...
With the folded cascode i was able to speed up the whole amp
in a way that the 100pF at input seems enough compensation.
Yes, would be great if you have sophisticated equipment for testing !

Hi traderbam !
Thanks for warning, i will observe closely. I did not recognize any
symptoms yet, from other amps oscillating i collected some of the
typical symptoms.
I already had different speakers connected, no change in quiscent
current or strange noises triggered by some instruments.
I had not tested yet moving the woofer with no input, i will try now !
I expected the circuit to oscillate, but was surprised seeing it work.

Mike
 
Hi Mike,

Place an AM portable close to your construction and tune up and the MW band. Any quietening means oscillation.

Try your sims with 50% pot setting, distortion may increase. This is why we have pre-amps with buffered pots.

It could be the choke that is preventing oscillations. Chokes are normally damped with a parallel carbon resistor.

Good luck with the investigations.


Cheers .......... Graham.
 
I'm just curious :

If you haven't got an oscilloscope, how do you know that your circuit doesn't oscillate?

Only severe oscillations at zero crossing are noticed as an increase in bias current. Some oscillations are so small in amplitude that they may only be noticed measuring the VAS current
 
Eva said:
I'm just curious :

If you haven't got an oscilloscope, how do you know that your circuit doesn't oscillate?

Only severe oscillations at zero crossing are noticed as an increase in bias current. Some oscillations are so small in amplitude that they may only be noticed measuring the VAS current

Typically, there are much more symptoms indicating oscillation, but
you are right, i can't be absolutely sure. In sims it doesn't oscillate,
but phasemargin was not very big. So i changed it to doubledarlington,
this should remove any doubt about oscillation. (in theory)
You really shouldn't underestimate the speed of folded cascode !
Shouldn't an oscillating amp sound bad ? (muffled)
The other problem is, sims showed "oscillation" with 0.5uv at 64mhz,
i doubt that a normal scope could show oscillation at this freq ?
At least i did not find any typical symptom of oscillation, but
still missing the test with AM-radio...

john curl said:
I would put a 5-10 ohm resistor in parallel with the output inductor (2uH). Everybody does, for good reason.

Yes, i will do that, but i never understood what it is really good for ?

Mike
 
I own a Hameg HM407-2 [40Mhz] and I've succesfully detected up to 65Mhz waveforms with it [undamped diode ringing from SMPS]

At such high frequencies you can no longer rely on amplitude or phase, but you know that there is a RF waveform here

I would start gathering money for an oscilloscope if I was you ;)
 
re: Ultrafast NoCap-ClassAB

MikeB

Does the circuit need a resistor or current source from the
negative node of the green led to ground (or neg rail)?

This would provide bias current for the led and bias (base current) for
the two associated pnp 2n5401 transistors.


Tom

Mostly lurking than posting.
 
resistance in parallel with output L provides damping of reasoance, cable complex Z and inductor L can have multiple high frequency reasonances and a low value R can soak up some of the energy - ideally the rising Z of the output L "disconnects" the amp output from cable C at high frequencies and the amp only sees the low value R ("grounded" by the cable C) as its high frequency load - making high frequency stability easier to assure
 
MikeB,

I working something same. But I think to use current source in the second stage as well. It's improve the balance, and the PSRR in the second stage too.
I think, there is capacitor in Your amplifier. But this cap built into the transistors. And if You use higher power transistors in the second stage, instead of 2N5401s, You will get more phase margin.
You can get more phase margin, if You increase the local feedbacks, and decrease the global.

sajti
 
Hi guys !

Eva, as i make amps only as hobby it would be too much money
to buy a scope, i will try to borrow one somewhere...

Tom2, i just skipped this resistor in sim, the voltagesupply
in sims is 100% stable. On prototype the green led is of course
fed with a resistor.

Alme, it seems impossible to get a "standard" topology stable
without feedbackcompensation, with folded cascode it's different.

jcx, thanks for the explanation, i will add a 10ohm...

Sajti, welcome to the FC-club !
But for the case of the folded cascode, "slower" transistors reduce
phasemargin as they introduce delay. But the inputcapacitance
of the drivers connected to the vas do in fact "replace" these caps.
As i switched to doubledarlington these caps got very big, this has
the advantage that the vas follows exactly the voltage inside the
drivers, beeing syncron with the outputstage. This result in perfect
behaviour for fast transients, means squarewaves are reproduced
perfect, not a single sign of overshooting/ringing in the whole circuit.
As i use sligthly higher current through the FC than the diffamp
puts out, the vas-devices can never saturate because of slewlimits.
So, only diffamp saturates, but it does not enter reversebias to
have recover from...
I was inspired by the JLH/DoZ, also having no caps at all, and i had
the thought that this could be the reason for their sound.
I seem to be right, this amp shows the same sonic reproduction
for voices and dynamics.

I checked with a version using ccs instead of this, but it showed
far inferior transient speed (slewrate), using same currents.
A 200khz signal was fully triangulated, where the other version
was still sinus with reasonable thd. As i use diffampoutput
balanced followed by a currentmirror, this amp is fully balanced.

Mike
 
updated schematic

Here is updated schematic using doubledarlington.
The current in vas is a bit small for this, but the amp sounds
slightly better this way...
The 2 100k in vas to gnd were necessary to extend openloopbandwidth
again to cover audioband again. Without, the sound was too dry...

Mike
 

Attachments

  • focasym2.gif
    focasym2.gif
    14.3 KB · Views: 1,327
AX tech editor
Joined 2002
Paid Member
MikeB,

Interesting circuit. I have a question: I think your oscillation-fredom (in the sim) is very much due to the fact that you have set the gain quite high, some 45 times. What happens when you set the gain to 26 or 28, which is more common, it may be different. All feedback amps get more stable if you increase the closed loop gain.

How does the freq response look? Can you post that? That would also give an indication of the stability margin. You wouldn't want it to burst into oscillations when you connect a diffferent speaker!

What happens when you use a LED + resistor in the sim instead of the 1.8-something battery, that may be also more realistic also.


Jan Didden
 
Hi Jan !

I checked your suggestion, i replaced the 1.87 voltage with 3 diodes
in series fed with 15k, (I had no adequate LED-model)
there are only subtile changes.

I checked with different closedloopgains, with gain of 26 results
in acsweep look identical, with gain 1:10 the acsweep gets critical,
but no oscillation yet. I decreased timestep from 10ns to 2ns to
be sure.

Attached is acsweep, blue is output before coil, green left base
of diffamp, red right base of diffamp. Feedback keeps always
below input, so should be safe.

Mike
 

Attachments

  • fs_ac_1.gif
    fs_ac_1.gif
    9.4 KB · Views: 1,065
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.