LM3886F, discreet with MOSFET outputs

[Hayk]

The output stage of LM3886 and its predecessors, is unique, not mentioned or described anywhere I know, even not in the evangils of St-Bob or St-Douglas, amplifier design handbooks.
Many pretending experts described it as ordinary quasi complimentary, Linn topology, so no one dared to contradict since.
MOSFETs intended of switching purposes, nead 250ma bias to overcome the crossover region, with this topology, only 65ma is sufficient, see bellow the open loop output for 2vp and 20vp with 8 ohms load.
It eliminates the drivers, with higher open loop gain, it doesn't need beta multiplier.
The rise slew rate with temporary compensation, is over 30v/us.

 

[solderdude]

You mean the quasi complementary output stage ?

Compound vs Darlington

ESP - Compound Pair Vs. Darlington Pair

sound-au.com

Why would it not be considered a quasi complementary output stage ?
Can you explain the rationale behind the output stage of 3886 differing from quasi complementary ?
That is.. other than R1 in the schematic below going to the feedback point instead of EQ2 and R4 being in EQ4 ?

 

[Hayk]

My internet access is slow, I need to post in several times to avoid loosing my massage, uploads.
Read again please.

 

[solderdude]

Can you explain why the 3886 output stage is very different from, or can not be called a Darlington + Sziklai pair (quasi-complementary output) ?
Is it because of the different emitter resistor/current limiter config ?

 

[Hayk]

For quasi complimentary as name suggests, it needs complimentary drivers. In LM3886, the drivers aren't complimentary. The lower driver is linked via diode to the emitter of the VAS, not in Linn topology. The 1k resistor value on the VAS emitter is very sensitive.
Once, I wanted to convert the JLH 1969 classA, into class AB and arrived to a similar circuit. I didn't check with my circuit yet, with LM3886, the upper output doesn't cutoff, this why it has very low crossover distortion.
Edit.
In DS, you can read, If most amplifiers are classified as class B, the lm3886 is quasi class AB, not quasi complimentary classB.

 

[Hayk]

To get à decent rising slew rate, I increased the VAS current to 6ma, the transistor is its 250mw limit of 500mw max. The slew rate is 20v/us using IRFP450 190watt, slightly better than IRFP240 150W.
The IRFP450 is very commonly used in SMPS.
You can see that the two outputs have the same on off speed with very little shoot through.
The amp at this stage doesn't need any compensation, just a lead capacitor of 15pf to get nice square waves without input filter.

 

[Hayk]

Mater of THD, the LM3886 with+/-35@8 ohms has 0.002% THD+N for 27W.
For now, with 2SA970 low noise inputs, biased 0.5ma each, no CCS, I have 0.003% THD for same condition.

 

[Hayk]

I did reach on simulation as good as lm3886 in THD with ordinary transistors and bootstrap. The optimal bias is 50ma. No compensation.
I will built as is to hear how it sounds.
added.
Output z is 0.0003 ohm resistive in whole audio range.

 

[Hayk]

Ready for PCB. Although it simulates perfectly stable without compensation, I don't believe in such miracle. I added a 100pf compensation on the current mirror via the Zobel that filters out the low frequencies and the Lead capacitor reduced to 5pf.
The curves are NFB of the amp.
You can observe by comparison with and without compensation, that my compensation technique by current mirror doesn't effect the audio band, 60db NFB@40khz in both cases. The 2n5551 has 10db less gain but I need to compare by listening as the VAS transistor is the sound of this amplifier.

 

[Hayk]

VM1 is the voltage across the upper 0.22ohm and VM2 the lower.
As shown, the upper transistor doesn't cutoff and continues active.
Always 27w 8 ohms 1khz.

 

[solderdude]

For quasi complimentary as name suggests, it needs complimentary drivers.

No it needs to have the same output devices (in this case NPN) and one of those devices basically works as a complementary device by adding the PNP (T3 in your case).
That's the 'quasi' part.

The fun part of the 3886 is that all active components are on 1 chip with shared temperature.
Requires a few external components.
Has a mute function
Has over current limiting protection.
Under voltage protection. (No pops when turned on or off)
Requires no idle current adjustment.
Wide supply voltage range 20-94V
Output protection for inductive loads.

The fun part of your circuit is:
DIY.
Good performance (it seems).
Fun project.
Needs more components despite being being smaller in active component count.

Will you be offering PCB's Gerbers or KiCad files to interested people ?

 

[audio_tony]

Ready for PCB. Although it simulates perfectly stable without compensation, I don't believe in such miracle. I added a 100pf compensation on the current mirror via the Zobel that filters out the low frequencies and the Lead capacitor reduced to 5pf.
The curves are NFB of the amp.
You can observe by comparison with and without compensation, that my compensation technique by current mirror doesn't effect the audio band, 60db NFB@40khz in both cases. The 2n5551 has 10db less gain but I need to compare by listening as the VAS transistor is the sound of this amplifier.

C2 (1m = 1000uF?) is going to cause issues at power on, as the LTP is going to be starved of current until that cap charges via R11.

I see a similar issue with C6 (C8? - I can't see the number properly) - at 10000uF the feedback cap is huge as well.

 

[Hayk]

These values are only to help the simulator short transient time to get precise measurements. I'll post the kicad schematic tomorrow.

 

[Hayk]

The LM3886 is not ordinary silicon, it is aerospace quality silicon carbide of 90's. The outputs as you see in the picture are 4 units. They are intended to function safely up to 250°C. The low level, for long term reliability reason, can safely function at operating average temperature of 150°C instead of 120°C for normal silicon junction.
At high temperature, the thermal feedback decreases and allows use of small
heatsink.
I need first to validate the circuit and the PCB, always needs retouching. When fully tested, I'll post the final circuit along the gerber file for free of use as I did lately with the 3 transistor class A amp.

3 transistor classA 50w/25w with variable output impedance as SE triodes.

Function description is here. https://www.eevblog.com/forum/projects/can-this-3-transistor-classa-amp-50w4ohm-with-1-3a-bias-still-be-called-classa/#msg5701985 This amp initially I designed to function with constant current, where the power CCS modulates itself to keep a constant current from...

www.audiosciencereview.com

 

[Hayk]

Preliminary PCB coming soon.

 

[audio_tony]

Preliminary PCB coming soon.

With C3 = 100u charging via R6 (24k) there is going to be some instability at power on (probably a large 'thump' as the LTP current stabilises).

It'll take ~12s for that 100u cap to reach 99% charge via 24k.

I would suggest that the maximum value for C3 would be around 10uF but would prefer 4.7u or even 470n.

 

[Hayk]

The VAS stage needs few uA to be biased and the current mirror will gives the difference. The bootstrap capacitor also is in the same story. This why trial mode is best way to design although the Tina simulator can do it.
I over dimension the capacitor footprints so that I can try different values. Once determined on trial, I redimention them. You can not imagine how values determined by engineering end up sounding wrong. The 220uf DC blocking on the feedback is also the same.

 

[Hayk]

I simulated for you the start ups with sudden power supply switched on.
With 2uf and 100uf and no bootstrap.

 

[audio_tony]

I'm basing this on my experience rather than measurements.

Some years ago, I built a modified version of the Elektor 70/100W darlington amp. A series resistor and zener diode are used as a CCS for the LTP.

I put a cap in parallel with the zener - the series resistor is only 5k6 - and a 10uF or higher cap resulted in an almighty thump at power on (which of course would be masked by a protection / muting relay) - even though the charge time constant was less than 0.5 second. I suspect that it's possible for this to cause other issues - i.e. gross DC offset and potentially O/P failure.

The bootstrap capacitor is not typically an issue, as it's sole purpose is to increase the drive current as power output increases, and the amp will even work fine without it (but will have asymmetric clipping at higher output due to lack of current drive).
As mosfets typically require lower drive currents, the bootstrap may not even be necessary in your circuit.

 

[Hayk]

With zener 11v and 4.7k instead of 24k it works the same and good start up.
No need to capacitor across zener, the output noise is the same, 18uv.
With CCS instead of bootstrap, the THD is doubled.