Since Quasi has to move his web page, has some one Quasi configuration table that was on NMOS350's page. It's handle on R11 and R18 value function rail voltage.
Regards Marc
Regards Marc
Well I still haven't figured out what is causing all these oscillations. I have managed to measure the frequency of the oscillations, 500kHz. They only occur when the output current rises above 400mA peak-peak, and it only ever happens on the positive cycle. It seems to be causing the ground plane of the amp to oscillate, as the oscillations seem to drift back into whatever device is connected to it.
I'm using 2n5551 as the transistors for the LTP stage, could that be causing an issue? I cannot get access to anything else. That is the only thing that differs from the original design.
Can anyone give me some pointers as to where the oscillations might be starting from? LTP? VAS? output fets? power source? (see previous posts for more info)
Thanks -canca87
I'm using 2n5551 as the transistors for the LTP stage, could that be causing an issue? I cannot get access to anything else. That is the only thing that differs from the original design.
Can anyone give me some pointers as to where the oscillations might be starting from? LTP? VAS? output fets? power source? (see previous posts for more info)
Thanks -canca87
surely the cause of oscillationscompensation is not very accurate...
the 100 pf capacitor must beconnected between base and collector of Q4,
you can reduce it to47/68 pf...connected as you made it, between collectors
of the two input transistor , it won t work..
you can get rid of the two 100pf connected from base to emitter of
Q7/Q8...if the amp is correctly compensated, they are useless..
also, Q1 and Q5 bases must not be tied together, cause the vas is
modulating the ltp in this way...put a 10k resistor from the reference
diode to base of Q1, and a 22uF Capacitance from Q1 base to negative
rail...
the 100 pf capacitor must beconnected between base and collector of Q4,
you can reduce it to47/68 pf...connected as you made it, between collectors
of the two input transistor , it won t work..
you can get rid of the two 100pf connected from base to emitter of
Q7/Q8...if the amp is correctly compensated, they are useless..
also, Q1 and Q5 bases must not be tied together, cause the vas is
modulating the ltp in this way...put a 10k resistor from the reference
diode to base of Q1, and a 22uF Capacitance from Q1 base to negative
rail...
ok, thanks for the pdf...
the MOS gate resistance are somewhat low, 27 ohm..try as much
as 100 ohm, and perhaps even 220....
C4 can be increased to 68 PF to tame down instability...
10k RESISTOR from R11/D1 to base of T4, and a 22uF capacitance
from T4 base to negative rail, this will prevent ltp modulation by the vas..
last, put a 1 to 4.7 ohm , 5W , non inductive resistance in paralele with L1....
hope this will help a little..
the MOS gate resistance are somewhat low, 27 ohm..try as much
as 100 ohm, and perhaps even 220....
C4 can be increased to 68 PF to tame down instability...
10k RESISTOR from R11/D1 to base of T4, and a 22uF capacitance
from T4 base to negative rail, this will prevent ltp modulation by the vas..
last, put a 1 to 4.7 ohm , 5W , non inductive resistance in paralele with L1....
hope this will help a little..
Many of these amps have been built, and stability is very good. However, with global fb amps, every active component is important from the POV of loop gain, and I suspect that if you return to the correct input devices, you should see perfect stability.
I know Con, the designer, personally, and believe me he is very thorough and has covered all bases. Use the components he suggests; if necessary get them from Mouser/Digikey.
The only other option would be to play around with lag compensation (39pF) and phase lag (10pF), but you are essentially modifying the design when you do this, so all the hard work done by Con and others to establish proven reliability will be undone.
Hugh
I know Con, the designer, personally, and believe me he is very thorough and has covered all bases. Use the components he suggests; if necessary get them from Mouser/Digikey.
The only other option would be to play around with lag compensation (39pF) and phase lag (10pF), but you are essentially modifying the design when you do this, so all the hard work done by Con and others to establish proven reliability will be undone.
Hugh
I used component came from farnell, all everything is fine. yes AKSA is correct try to vary only the value of c4 and c8, I think this is only the way to omit the oscillation. modifying the value of gate resistor is not advisable because the rds on of the trannies is extremely 27R 47R is enough 🙂
rlg
what have the rds to do with the gate resistance???
the modification i talk about are proved guidelines in almost every
design well studied...
the only thing i di not talk about is the phase comp, as
i think that this 10pf should better be adjusted in real situation,
not simulating...through my simulator, i noticed that with this topology,
it was sometime not enough, and as much as 22PF was needed in
some designs..but before tweaking this cap, the best is to increase
the cdom capacitance , first try 47pf , and if not enough, 68 pf....
beside, in many simulation with powermos, i did found that
27 r is too low,..myself, i use 100 R...anyway, increasing this
value wont make the amp slower, as i highly doubt that the driver
has such low output impedance...
Wahab,
You are right of course, however, changing Cdom will profoundly change sonics, so you must be careful to use the minimum possible, and supplement it with careful choice of phase lag cap (or, for capacitive loads like ESLs, phase lead from VAS to fb node).
I believe 27R as a gate stopper is very low, normally you would expect to see around 100-220R in this position.
Hugh
You are right of course, however, changing Cdom will profoundly change sonics, so you must be careful to use the minimum possible, and supplement it with careful choice of phase lag cap (or, for capacitive loads like ESLs, phase lead from VAS to fb node).
I believe 27R as a gate stopper is very low, normally you would expect to see around 100-220R in this position.
Hugh
time to charge the mosfet Cgs (gate/drain capacitance INCREASE proportionaly...since this capacitance is about 1 nF in an
hexfet, this will increase this time from about 27 ns (27R)to 100nS (100R),
and logicaly 220 nS for a 220 R....
keep in mind that this is the time to reach 63 % of the source voltage,
so if the latter is say 50V, it will take less time to make the hexfet
conduct, as the threshold voltage is reached when the vgs is about 2/4V,
depending on the device...
increasing the value to 100 get rid of the gate inductance influence...
Marc,
Wahab is right.
I suspect you know what it does, but in short, increasing the gate stopper impedes the passage of charge to and from the mosfet gate, slowing it down, and preventing it from self-oscillation, often a problem with these devices. In a Class AB, the mosfet is turning on and off constantly, so this charge must move back and forth. Typically gate capacitance in an audio amplifier is around 600pF or more per device.
Hugh
Wahab is right.
I suspect you know what it does, but in short, increasing the gate stopper impedes the passage of charge to and from the mosfet gate, slowing it down, and preventing it from self-oscillation, often a problem with these devices. In a Class AB, the mosfet is turning on and off constantly, so this charge must move back and forth. Typically gate capacitance in an audio amplifier is around 600pF or more per device.
Hugh
That's what i suspect, but not enough experience and knowledge to confirm...Yes it's in mind
Marc
The mosfet is a seductive device, because superficially it looks so easy to drive. This suggests you can drop one current amplification stage and simplify the circuit.
However, the movement of this gate charge back and forth requires a low impedance source for good speed, even at audio frequencies, and the tendency to oscillate at around 5-10MHz makes them hard to tame. In Class A mosfets are outstanding, since charge transfer is less, and they are so thermally robust.
OTOH, bipolars have improved a lot in the last 15 years, and my personal choice is always for these tighter, cheaper devices such as C5200/A1943. In my opinion, bipolars have higher transconductance and are more stable as current amplifiers. Mosfets are ideal as switchers in smps and Class D amps, nothing better.
Horses for courses.....
However, the movement of this gate charge back and forth requires a low impedance source for good speed, even at audio frequencies, and the tendency to oscillate at around 5-10MHz makes them hard to tame. In Class A mosfets are outstanding, since charge transfer is less, and they are so thermally robust.
OTOH, bipolars have improved a lot in the last 15 years, and my personal choice is always for these tighter, cheaper devices such as C5200/A1943. In my opinion, bipolars have higher transconductance and are more stable as current amplifiers. Mosfets are ideal as switchers in smps and Class D amps, nothing better.
Horses for courses.....
That's what i understand. I asked this question beacause i need a quick (time building) amp and i have beside 2 Nmos 200 board. I reserve my MJL bipo for brother of quasi and need to match them (it take time for 100 pieces) with your procedure, and perhaps will i try toshibas C5200 in ths position to. For Nmos 200 the mosfet choice is more difficult then in 350 due the fact that their is only one couple of output device. The right compromise must be found between SOA possibility and low Ciss. IRFP460 seems to be one...
Marc
hugh, globaly,you are right, i will just give a few remarks based on my
experience..
usually, bjts work very well, specialy those you named, which are
nothing else than replacement of the famed 2SA1302/2SC3281...
i also use some magical darlingtons from sanken, MP1620/MN2488, truly
fantastic devices which simplify the circuit and layout..
but my prefered amp is mosfet using 2SJ50/2SK135 paraleled ,now obolete as
this is TO3 devices, but they are replaced by the 2SJ162/2SK1058 which
are exactly the same chips, i checked the datasheet, but in TO3P case..
i m using a symetrical differential, i designed it in a way that there s no need for compensation capacitance, be it the lag or phase, and it s rock stable; i used it even as monitor amp on stage with my sets of keyboards..
the wide open loop bandwith allowed by the mosfet render the
negative feedback very efficient at high frequency ,even if the open loop
gain is not high, as the OL -3 db point is vastly higher than with a
classicaly compensated amp..i don t use hexfets, as the hitachi fets
i named are uncomparably better for audio, although for class A ,
hexfets are relevant..
regards,
wahab
There was some discussion about the problem with R21 that sits on the collector of T10. This resistor was an attempt to equalise the voltage swing of the output stage given that there is insufficient voltage available for the positive side driver transistor to fully turn on the positive rail FETs. This of course is not a problem on the negative side.
The problem with this resistor is that it creates a variable voltage drop instead of a fixed voltage drop. I thought the there was insufficient current change in the driver stage for this to be a real problem but there are some undesired effects that may or may not affect some builds.
This link takes you to the discussion.
http://www.diyaudio.com/forums/solid-state/43331-power-amp-under-development-270.html#post1428765
The discussion continues with the final solution on post #2709.
The effect of the diodes is to provide a fixed voltage drop rather than a variable one. I'm not sure if this will solve your particular problem but it is a worthwhile improvement for all constructors of the Nmos series. The Nmos MkII (built by a few) has this feature built in.
The other thing to try is replacing all the 0.1uF capacitors to a different brand. Suprisingly some brands have difficulty in bypassing high frequency effectively, causing the ringing problem you are experiencing. You probably already know this and may have changed these already.
Remember that you must back off the bias completely when you substitute R21 for the diodes (or LED) and then set the amp up again as per the guide.
Cheers
Quasi
The problem with this resistor is that it creates a variable voltage drop instead of a fixed voltage drop. I thought the there was insufficient current change in the driver stage for this to be a real problem but there are some undesired effects that may or may not affect some builds.
This link takes you to the discussion.
http://www.diyaudio.com/forums/solid-state/43331-power-amp-under-development-270.html#post1428765
The discussion continues with the final solution on post #2709.
The effect of the diodes is to provide a fixed voltage drop rather than a variable one. I'm not sure if this will solve your particular problem but it is a worthwhile improvement for all constructors of the Nmos series. The Nmos MkII (built by a few) has this feature built in.
The other thing to try is replacing all the 0.1uF capacitors to a different brand. Suprisingly some brands have difficulty in bypassing high frequency effectively, causing the ringing problem you are experiencing. You probably already know this and may have changed these already.
Remember that you must back off the bias completely when you substitute R21 for the diodes (or LED) and then set the amp up again as per the guide.
Cheers
Quasi