Thank you for your help all. I added a 470 pf capacitance from the gate to the drain on the n-channel which seems to have stopped the resonance and the burst resonance Before I had connected the capacitor between gate to source which secured a very stable oscillator. Other changes done which didn´t affect the behaviour were
- a replacement of the wirewound 0,22 ohm to 5 parallel 2W 1 ohm
- 68 nf over the bias adjustment resistor
regards
Örn
- a replacement of the wirewound 0,22 ohm to 5 parallel 2W 1 ohm
- 68 nf over the bias adjustment resistor
regards
Örn
Another useful approach here is to insert a ferrite bead in the gate circuit right at the MOSFET itself. This works very well with bipolar emitter followers as well.
The cap gate to drain is also a good solution.
Output stages: not as simple as they look.
I'd still look at putting some decent compensation around the VAS stage as described by Wahab. This would then really clean this amp up.
The cap gate to drain is also a good solution.
Output stages: not as simple as they look.
I'd still look at putting some decent compensation around the VAS stage as described by Wahab. This would then really clean this amp up.
That should have changed it as you are seeing, though I'm not convinced that adding pure reactance has eliminated it from all your operating conditions.
Thanks
-Antonio
Hi Örn
Adding 470pf to Cgd is too much. This may have a negative effect on the driver stage, and may affect the stability of the global loop. The typical type of local mosfet oscillation is 20+Mhz so 47pf would be more in the range required. Placing a 150R resistor in series with the 47pf cap gives a place for the resonating energy to dissipate and creates a zero (as in pole/zero analysis) ~22MHz and should be sufficient to dampen the ‘oscillator’. F=1/(2pi*RC) At audio frequencies this zero has no effect. It is also important to do the same thing to the P-ch device even though Ciss of the P-ch is generally larger as a reflection of larger die size, so it is usually the N-ch that is more likely to oscillate. This is a simple cheap thing to do to insure unconditional stability from the mosfets. By using this simple cheap technique to stabilize the mosfets, the gate stopper resistor can be reduced. Because the two devices have different Ciss, the optimum gate stopper resistor will not be the same for N-ch and P-ch. It is perfectly acceptable to have different Rg values. You can play with the value of the zero and lowering the value of the gate stopper resistor and test the limits of stability. This increases the effective Ft of the mosfet significantly as it is mostly determined by the LP filter created by Rg and Ciss, which is a variable dependent on Vds, and thus is not fixed. This means that the oscillation bursts are aggravated by Ciss becoming larger at lower Vds due to reduced Ft.
Thanks
i did lower to 300 pf the Cgd but you suggest 47 Pf with dampening resistor. This is what I read from you
-- 150R -- -- 47 pf -- Drain
!
-- 220 ohm -------gate
The output is Very clear and free fom almost any oscillation, much better than before. i have also added 22pf in the driver stage between base and collector on the cascode
i did lower to 300 pf the Cgd but you suggest 47 Pf with dampening resistor. This is what I read from you
-- 150R -- -- 47 pf -- Drain
!
-- 220 ohm -------gate
The output is Very clear and free fom almost any oscillation, much better than before. i have also added 22pf in the driver stage between base and collector on the cascode
Its really a neat effect you might consider it a great time to investigate it as much as possible rather just making it go away.
You can try by adding small values of C g-d until you can see some effect(possible having it start later etc) then you know what sort of impedance you are workiong against (the inherent dg capacitance which should be failry low ~15p but climbing as you get near the rail).
You can try increasing the base resistor until the oscillation stops that indicates how much negative resistance you need to cancel.
You can try addding the rc from g to ground instead.
Then you can add some capacitance to the output load and see if the oscillation comes back (~100-1000p)
Thanks
-Antonio
You can try by adding small values of C g-d until you can see some effect(possible having it start later etc) then you know what sort of impedance you are workiong against (the inherent dg capacitance which should be failry low ~15p but climbing as you get near the rail).
You can try increasing the base resistor until the oscillation stops that indicates how much negative resistance you need to cancel.
You can try addding the rc from g to ground instead.
Then you can add some capacitance to the output load and see if the oscillation comes back (~100-1000p)
Thanks
-Antonio
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.