Rebirth of the Phoenix

[PHEONIX]

Hello Edmond

The quirk is at my (my error) end Q27 and Q29 are in the right package now.

I loaded the Fairchild models 2SC5200 and 2SA1943 and they give a similar performance (Ibias to Vs) that the NJL3281 and NJL1302 do.



Regards
Arthur

[Edmond Stuart]

>As for TT-trannies: How large is V_BE of the TT-trannies WRT V_FWD of the TT-diodes?

I just simmed a simple test circuit and it seemed that V_FWD of the NJL4281/4302 diodes tracks V_BE NJL1302/3281 trannies reasonably well. Let's hope the same applies to the NJL4281/4302 trannies.

This is one of the things that has to be confirmed by real measurements.
Arthur, could you do this, please?

Cheers,
E.

[PHEONIX]

Hello Edmond

For completeness I will attempt to do it for both NJL4281/4302 and NJL1302/3281.


Regards
Arthur

[PHEONIX]

Hello Edmond

Is it ok to build it with one pair (1*NPN and 1*PNP ) or do I need 2 pairs as drawn above.

Regards
Arthur

[AndrewT]

1pair gives the two diodes you need to check the bias voltages.

[AndrewT]

Edmond's bias is quite different from Roender's
Roender shows 5 diodes with a low current through them. This forced us to adopt a 3pair output stage.
Whereas Edmond shows 2 diodes with a high current through them. {Suits a 1pair build, but not 2pair nor more.}
becomes {Suits a 2pair build, but not 3pair nor more}

Edit.
Edmond's next post explains why 2pair suits his biasing topology.

[Edmond Stuart]

Originally Posted by PHEONIX

Hello Edmond

Is it ok to build it with one pair (1*NPN and 1*PNP ) or do I need 2 pairs as drawn above.

Regards
Arthur

Hi Arthur,

Regarding the test circuit, of course that's okay. However, regarding the final version, you need two pairs for two reasons:
1. In order to not exceed the SOA.
2. We need a 2nd pair of TT-diodes (D13 & D14) in order to avoid cross-conduction of the O/P devices. Please see my previous post on this: Rebirth of the Phoenix
In one of my next comments I will elaborate on this and make clear why a simple R and C between the bases of NPN and PNP output devices is essentially totally wrong.

Cheers,
E

[PHEONIX]

Hi Edmond

As its drawn the bias current is 103 ma per device, with one pair of output the bias current is 113 ma per device according to the simulator.

>We need a 2nd pair of TT-diodes (D13 & D14) in order to avoid cross-conduction of the O/P devices.

I follow the reasoning for this namely to maintain low impedence between the emitters at high frequencies so these drivers can operate in push pull. This is a nice use of the additional TT diodes.

Regards
Arthur

[Edmond Stuart]

Now, let's fiddle with the stuff between the bases of NPN and PNP output devices and see what happens if we apply a 200kHz sine wave. Admittedly, a rather high frequency, but it makes perfectly clear that essentially there are cross-conduction issues.

1. If no C13 (100nF), no diodes (D13 & D14), only R49=220 Ohms: After only cycle the bias rose from 100mA to 2A, see 1st pic. Needless to say that this is totally unacceptable.

To remedy this nasty habit of BJTs (and sloppy drivers), almost invariably a speed-up capacitor is placed between the bases of NPN and PNP output devices. It cures the trouble, but not for long, as the cap get overcharged after a few cycles. As a result, the bias of the O/P devices increases and equally worse, the bias of driver decreases. See 2nd pic (C13=100nF & R49=220 Ohms).
BTW, in some cases, the decrease of the driver bias may even lead to HF instability (probably, this was the case with a precursor of syn08's YAP).
This is the reason why in some of my designs I'm using two separate push-pull drivers, one for the top O/P devices and one for the bottom O/P devices and no cross-coupled cap at all.

However, a far more elegant and simple solution is to make clever use of the matched TT diodes, which provides a low and frequency independent impedance between the emitters of the drivers. Now both of them can sink and source enough current to turn off the O/P trannies more quickly. With C13=100nF, R49=2.2R and the TT-diodes installed, the increase in bias current is far less, only 150mA. See 3rd pic.

Apart from more power, that's why I strongly recommend to use two pairs of NJL4281/4302, i.e two TT-diodes for biasing and two TT-diodes for fast turn-off (btw, no 3281/1302, as these trannies have different and less suitable diodes).

Cheers,
E.

edit: apparently we cross-posted.

[AndrewT]

Thanks Edmond.
That's the first time I've seen that explanation and the accompanying plots.