Magic - My first Lateral MOSFET Push-Pull amplifier

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Last year I made a SE amp using these FETs. They did an amazing job. I was impressed.

Still, I had a lust towards Project-101 amp on Rod's site. The one amp that changed his view on FET power amps and by himself considered to be THE best amp on his site. I was like "MAN I'M DEAD!" after seeing that most component values were hidden. But it was in the past. Things have changed and I made my own version of the amplifier. With component values that are almost standard for this kind of amps, there is in fact zero contribution from me into this design. I thank the great guys who created this configuration for the first time.

I increased the LTP bias a bit higher than usual. Inputs are equipped with a CCS and a mirror with degeneration resistors of standard values. VAS is based on BC550C instead of a bigger transistor. AND lastly, the design is inverted as a whole from the original, with PNP LTP and NPN VAS. Because I liked it this way and also because Mr. Self used this config as reference in his book. :)

I used a Vbe multiplier instead of a plain pot for the output bias, coz I don't like current more than 1mA flowing through the pot. With the 2k2 value of the bias pot the quiescent current is 88mA.

And I added a 0.22ohm/2W resistor in the feedback path to ground. This is for increasing the amp's output impedance to around 2.2ohm when connecting the speaker return wire to the positive side of the resistor. It DID produce excellent results with my 8ohm speakers!

Power supply is much lower than original P101. This was done because I need no more than 20watts from it to drive my speakers which are made for high-pass only. For the same reason the bootstrap cap is limited to 33uF; same as the feedback cap. The feedback cap is paralleled with a 1uF/63V film type on the board, but it's missing in the schematic. :eek: :p The (big) input capacitor is a 1uF/630V metallized polypropylene type (MPTA).

I didn't connect any capacitor to the G and S pins of the N-FET. It looked like the amp is stable and quiet without it and my FM receiver sounds clear without any strange noise. The gate stoppers are two 100ohm resistors for each FET; one on-board, another on-FET. I like this config.

Zeners are 3v3. Well, I'm targetting for 2.5A max output current, but am not sure if the 3v3 Zeners are okay or not. Thanks for any help on this matter.

Finally, it sounds good, natural... neither too bright nor too warm. Especially with the increased output impedance. Rod was right. I hope it will be with me for decades to come.

Laterals rock!
 

Attachments

  • DSC05565.JPG
    DSC05565.JPG
    735.8 KB · Views: 2,626
  • DSC05573.JPG
    DSC05573.JPG
    894.1 KB · Views: 2,029
  • DSC05566.JPG
    DSC05566.JPG
    692.1 KB · Views: 1,752
  • DSC05577.JPG
    DSC05577.JPG
    853.1 KB · Views: 1,702
  • magic.png
    magic.png
    24 KB · Views: 2,534
Last edited:
The only slight objection I have is the inclusion of the Vbe multiplier to set quiescent current. It's actually not needed, lateral MOS compensates itself for standing currents over ~100mA per pair. You can use a simple trimmer and cap in parallel. That being said, a Vbe multiplier can be set up to guard against loss of wiper contact, something the simple trimmer arrangement cannot (but the Vbe transistor must not be on the hetasink with the LMOSFETs). Still, getting the right standing current is so easy with laterals you can use trimmer then replace it with say, two series connected fixed resistors of the required resistance.
 
Thanks ilimzn. I already know this. The Vbe multiplier saves one great headache about wiper contact and pots available here are not too reliable. In my config the multiplier is on-board, not on or near FETs.

Can you please help me with the Zener value?
 
Last edited:
The cap is there to 'overwhelm' the nonlinearity of the input capacitances of the MOSFET, so it should be about 10x worst case (Ciss || Crss) x 2 (x2 because of two MOSFETs in a pair). For a very small value of the trimmer, and correspondingly large value of the current in the driving stage, you can completely omit it as the trimer itself will have a low enough impedance to shunt the input capacitances.
I should mention that some different topologies for driving MOSFETs must NOT have this capacitor. In these, the MOSFETs are never turned off. However, in a simple amp like this, turning one MOSFET on fully requires Vgs of nearly 10V, while the Vgs required for ~100mA standing current is around 1V. This means that while one MOSFET is turned on fully, and has Vgs 10V, the other has -8V and is in a deep off state. If one looks at the Vgs/Vds versus Ciss/Crss curves, it becomes apparent that the P and N MOSFET are somewhat different so just charging and discharging these capacitances with the same constant current produces a crossover notch because of the difference in the Ciss curves. The notch itself happens with a 'delay' which depends on the charge/discharge current. Because of this 'delay' characteristic, the notch appears on different places in a waveform depending on frequency and amplitude, so you get a nonlinear form of distortion. The cap across the bias setting element (be that a trimmer or a transistor etc...) acts as a local 'buffer' so that much more current is available to circulate between the gates, essentially making both MOSFET gates look like one less complex capacitance to the driver, by 'buffering' the above discontinuity in Ciss. In theory, you could make it 'infinite' but in practise there are problems with frequency response of very large caps. A HF non-inductive cap is the right choice here, 10nF-1uF range.
IT should be noted that the discontinuity of Ciss between P and N CH types is still visible regardless of the capacitor due to the gate stoppers, because they limit charge/discharge current. Unfortunately, there is little that can be done against that, as the stoppers are a necessity. On the other hand, there is a different approach, where the crossover notch is 'modified' by the careful choice of gate stopper values. National Semiconductor has a very nice write-up about this in the documentation of one of their LMExxxx amplifier chips, as these use external power semiconductors (including MOSFETs).
 
Calculate your peak current requirement based on the lowest impedance the amp is going to drive, and by that I mean the lowest impedance of your real-world speaker. For 8 ohm speakers, 5 ohms used for calculation is common practice and also leaves some headroom.
 
In the end of the day it`s up to you, but why pay for something futile and detrimental?

For the protection of the goodness. Didn't I say I want those Zeners for short circuit protection? And please don't come with discouragement like this. I know well what I am doing. And I asked for what I don't know. If you cannot help me then Thanks, no problem. But please please don't ask me back why I want to know.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.