H
HAYK
Wrong compensation. The LM3886 doesn't use Miller but Chery. The 68pF should 10pF be linked to the output. The open loop output impedance changes considerably.
Why bootstrap? why input buffers are eliminated? The original current mirror is not good? Or these are simplification?
Why bootstrap? why input buffers are eliminated? The original current mirror is not good? Or these are simplification?
Attachments
Last edited by a moderator:
D
Deleted member 550749
68pf is for frequency bandwidth, original schematic doesn't have this value and 10pf is used inverting feedback with feedback resistor, this types of circuit u can find cordell amplifier book, class b tipe, same think with some tweeks CLASS B TO AB,
D
Deleted member 550749
Bootstraping because diod configuration biasing simple and easy , just find correct resistor value,
D
Deleted member 550749
Ok, compensation simplified just removed auto adjust current mirror circuit, and it's simple and working
D
Deleted member 550749
Removed bootstrap, added 1 more transistor (bd140) for biasing current of 1.1ma , now bass improved but high frequency sound not responding like bootstrap, but my offset voltage is 0 👍🥰
Is the schematic at post # 9 a reverse engineering circuit of LM3886 IC?
MrAK87, I admire your spirit, but I think it may be necessary to have matched transistor pairs for good repeatability.
https://www.digikey.in/en/products/filter/transistors/bipolar-bjt/bipolar-transistor-arrays/277
https://www.digikey.in/en/products/filter/transistors/bipolar-bjt/bipolar-transistor-arrays/277
D
Deleted member 550749
Katiy , yes it's lm3886 reverse final circuit, working superbly, with 0 offset , doing to make propper pcb from jlc and will post another thread about it, love the simple circuit and performance 😎 and also u can achieve more wattage with some extra powerful transistor and easily 30-0-30 gonna survive 👍
H
HAYK
You don't need to repeat (20k+100n+2xdiode) just link the new CCS base to the base of the LTP's CCS .
TI's spice model for the LM3886 has(had?) serious problems. Clearly the "equivalent circuits" is incomplete. Besides the output protection, the VAS emitter resistor and unlimited VAS current creates a nasty phase inversion and rail sticking on the negative clipping. I don't have any chips to confirm this, but I assume the real chip has some kind of VAS current limit, probably a Baker clamp.
https://www.diyaudio.com/community/...nner-to-advanced.260627/page-161#post-6916593
But if you are making a discrete amplifier, there is little reason to repeat the LM3886 quasi output design. It's easier and provides better results to do a complementary output. Chip amps are quasi because NPN and quality PNP transistor do not live on the same chip substrate. A Quasi output is like having two left feet.
https://www.diyaudio.com/community/...nner-to-advanced.260627/page-161#post-6916593
But if you are making a discrete amplifier, there is little reason to repeat the LM3886 quasi output design. It's easier and provides better results to do a complementary output. Chip amps are quasi because NPN and quality PNP transistor do not live on the same chip substrate. A Quasi output is like having two left feet.
That's only true if the ~20K provides more than the VAS current, so that when you clip and the VAS CCS saturates, it does not feedback to the LTP. It may be easier/safer to just use two bias sources. A series resistor on the base of the VAS CCS is usually the solution to limit the amount of bias it can steal.