Abomination! or Painting a Mustache on the Mona Lisa meets the Island of Dr. Moreau

[roender]

Adjust the OPS bias in order to obtain 18mV over one Re, or 36mV over two opposite Res.
Don't connect drivers to the output node and bias them to at least 80mA or even better at 120-150mA.

 

[sajti]

I agree with Roender. My current amplifier sounds better if I increase the bias on the driver. I use MJE15032/33, with about 100mA, and the predrivers (KSA1381/KSC3503) with 10mA. The output is 4 pairs of FJL4215/4315, biased to 150mA.

Sajti

 

[megajocke]

0.1 ohm Re and 5 ohm Rb at +-70V is a bit too thermally unstable for my taste if the transistors are to be insulated from the heatsink. Any mismatch and overbias after signal is removed will be amplified about three times by positive thermal feedback. The mounting needs to be perfect, there is almost no margin for errors. A little too high thermal resistance on one of the transistors and it's instant destruction.

0.15 ohm might be acceptable though.

 

[AndrewT]

0.1 ohm Re and 5 ohm Rb at +-70V is a bit too thermally unstable

0.15 ohm might be acceptable though.

use Cordell's formula to check for output thermal stability.

 

[MJL21193]

Adjust the OPS bias in order to obtain 18mV over one Re, or 36mV over two opposite Res.
Don't connect drivers to the output node and bias them to at least 80mA or even better at 120-150mA.

I agree with Roender.

I've been trying different scenarios with the simulation, mostly to see the level of distortion with the changes. Without a doubt, the high bias gives the lowest THD but at the cost of serious heat...
I will build it, test it and listen to it with both high and medium bias on both the drivers and the output stage to see what works best.

I've pretty much made up my mind on the MJW21193/94 for the outputs and I've even downloaded the simulation models for these from OnSemi (GASP!). They are not giving me THD numbers as good as andy_c's MJL4281A/4302A models but the results are still within reason - .025% THD. As good as it needs to be I think.

 

[MJL21193]

0.15 ohm might be acceptable though.

Hi mega,
I see what you mean and I'd rather not buy a quantity of outputs large enough to match so I've upped the emitter Rs to .15 with no apparent ill effects.

 

[sajti]

I would try the faster NJW series as output devices. I have good experience with them, especially the NJW0302/0281.

Sajti

 

[MJL21193]

Hi sajti,
The NJW's don't have a high enough SOA for my voltage. The MJW's are 3A at 70V. I chose these over the MJL's because they are (1) cheaper and (2) physically smaller - TO-247 as opposed to TO-264.
I have enough of the NJL3281D/1302D to build these but I don't want the in-package diodes. Using 2 for them CFP drivers gives me the tracking (plus the Vbe multiplier transistor).

 

[roender]

MJL,

You make a big mistake thinking that the diode inside slave transistor will follow the CFP driver Vbe. You have to monitor the master Vbe, not the slave

 

[MJL21193]

MJL,

You make a big mistake thinking that the diode inside slave transistor will follow the CFP driver Vbe. You have to monitor the master Vbe, not the slave

The master and slave will be side by side, mounted on the main heatsink. I did that mostly for novelty - the Vbe multiplier (U19) will effectively track the temp.

 

[sajti]

Hi sajti,
The NJW's don't have a high enough SOA for my voltage. The MJW's are 3A at 70V. I chose these over the MJL's because they are (1) cheaper and (2) physically smaller - TO-247 as opposed to TO-264.

Yes, but You can use 6 pairs of 0302/0281, due the lower internal capacitances. The only advantage of the 21193/4 is the SOA. But the speed, and the B linearity is worse than X302/x281

Sajti

 

[MJL21193]

The speed yes but why do you say the 0281/0302 is more linear than the MJW's?
When it comes down to it, the 2SA1943/C5200 are equal to the 0281/0302 and more accessible - available at Digikey.

 

[sajti]

If you check the Hfe vs Ic diagramm, you will find why the 0281/0302 more linear. Even the 3281/1302 more linear (up to 5A).
I get them from Farnell.

Sajti

 

[MJL21193]

How about 5 pairs of these: 2STC5242 and 2STA1962

Half the price of the MJW's.

 

[sajti]

Looks nice. Low internal capacitance, and the hfe looks linear up to 3A

Sajti

 

[MJL21193]

I'll order 20 of each and change my schematic for 5 pairs or possibly 6 pairs.
No spice model available for these. I'll use the 2SA1943/C5200 for an indication.

 

[roender]

Looks nice. Low internal capacitance, and the hfe looks linear up to 3A

Sajti

I'll order 20 of each and change my schematic for 5 pairs or possibly 6 pairs.
No spice model available for these. I'll use the 2SA1943/C5200 for an indication.

I'm sorry to say that, but IMHO you have no idea how to read and interpret a datasheet.
How did you come to the conclusion that ST transistors are more linear than 5200/1943 or 3281/1302?

 

[sajti]

Not more linear, but cheaper....

What I told, is that NJW1302/3281 is more linear.

Sajti

 

[MJL21193]

How did you come to the conclusion that ST transistors are more linear than 5200/1943 or 3281/1302?

What's your opinion of the ST devices? I think they look pretty good for the price - these are actually a little more expensive than the Fairchild 2SA1943/C5200.
I have had good results with the 2SA1943/C5200 in my Patchwork amps, but I want a smaller package for this one with more devices.

 

[sajti]

You can get 2SC5242 from Fairchild as well. The direct order from the manufacturer means no problem with the fake transistors.

Sajti