MOSFET amp, oscillations and protection circuitry

Status
Not open for further replies.
Jump to Latest
D

danny92

Member
Joined 2012
#1
Hi everyone,

I'm trying to build an amplifier using two pairs of BUZ901/906 but the problem is that the amplifier could be unstable in the real world, and can blow up the expensive output devices, I'm thinking about incorporating a low pass filter at the input, a Zobel network at the output in conjuction with an inductor, I'll also use gate stopper resistors to reduce the bandwidth of the MOSFETs, I'll also incorporate the amplifier with a multislope VI limiter (I'm thinking about making one based on Michael Kiwanuka's designs), a thermal protection circuit, a DC protection circuit and an output relay for soft start.
About the compensation itself I don't know if I should use a simple Miller compensation, IMC, TMC, TPC or even MIC.
I don't know if I should use a triple or a only two predrivers.
It was told me that I should use complex oscillation dump circuitry between MOSFETs drain and gate, including inductors, resistors, capacitors, otherwise the MOSFETs will blow up.
As about the IPS I don't know if should use unipolar or bipolar IPS, BJTs seem to have lower distortion on simulations.
The VAS is the same, I don't know if I should choose a single ended current loaded VAS, push-pull VAS or a differential VAS.
I've been reading Bob Cordell's and Randy Slone's books but I'm getting overwhelmed and I don't know were to start. 🙁
Can anyone help me here?

Best regards,
Daniel Almeida
 
#2
Hi

The info about a small gate to drain Zobel filter as Bob recomends is a very good way to protect against local oscillations and still maintain the very high bandwidth of the mosfet. As for multislope VI limitation, I thought that was to protect against secondary breakdown in BJT's. Mosfets do not have a secondary breakdown region and are thermally limited. This makes them more robust and popular in larger power amps. It doesn't take much driving current (50ma) for the mosfets and one driver stage is probably plenty for two pair. If you go with an HEC mosfet output stage😉 like Bob did in his book, then the other driver stages are needed because they have multiple purposes.
 
#4
CBS240 said:
Hi

The info about a small gate to drain Zobel filter as Bob recomends is a very good way to protect against local oscillations and still maintain the very high bandwidth of the mosfet. As for multislope VI limitation, I thought that was to protect against secondary breakdown in BJT's. Mosfets do not have a secondary breakdown region and are thermally limited. This makes them more robust and popular in larger power amps. It doesn't take much driving current (50ma) for the mosfets and one driver stage is probably plenty for two pair. If you go with an HEC mosfet output stage😉 like Bob did in his book, then the other driver stages are needed because they have multiple purposes.
Click to expand...

Slowing down the output stage is also good way, but not best. There is better way by reducing the HF feedback and fasten the OPS response, it may reach the maximum output stage capability and solid stability, but it become very complex. For simple, just reduce the HF feedback.
 
Last edited:
#6
danny92 said:
Hi everyone,

I'm trying to build an amplifier using two pairs of BUZ901/906 but the problem is that the amplifier could be unstable in the real world, and can blow up the expensive output devices, I'm thinking about incorporating a low pass filter at the input, a Zobel network at the output in conjuction with an inductor, I'll also use gate stopper resistors to reduce the bandwidth of the MOSFETs, I'll also incorporate the amplifier with a multislope VI limiter (I'm thinking about making one based on Michael Kiwanuka's designs), a thermal protection circuit, a DC protection circuit and an output relay for soft start.
About the compensation itself I don't know if I should use a simple Miller compensation, IMC, TMC, TPC or even MIC.
I don't know if I should use a triple or a only two predrivers.
It was told me that I should use complex oscillation dump circuitry between MOSFETs drain and gate, including inductors, resistors, capacitors, otherwise the MOSFETs will blow up.
As about the IPS I don't know if should use unipolar or bipolar IPS, BJTs seem to have lower distortion on simulations.
The VAS is the same, I don't know if I should choose a single ended current loaded VAS, push-pull VAS or a differential VAS.
I've been reading Bob Cordell's and Randy Slone's books but I'm getting overwhelmed and I don't know were to start. 🙁
Can anyone help me here?

Best regards,
Daniel Almeida
Click to expand...

if you can show a schematic that would be useful. Solutions for oscillation can range from simple things to complex depending on where the problem lies. For me, these have ranged from PSU/layout issues, too small gate stoppers, issues with paralleled OP fets, feedback compensation etc. Some folks have used small ferrite bead on the gate as a cure-all and you can try that if you like.

Some feel that it can degrade the sound but when you have major incurable osc, it may be worth a shot.
 
#7
danny92 said:
Hello ontoaba and thank you for your help,

For reduce the HF feedback I should use the Zobel from gate to drain?

Best regards,
Daniel Almeida
Click to expand...

There is many ways, one example is using inductor in differential inputs:
http://www.hparchive.com/Manuals/Barney_Oliver_Amplifier_Manual.pdf
These 220uH inductors help to reduce the HF gain.

You could use what is best that fit your amp. Just try to do "this and that", I usually do "this and that" until oscillation stop.🙂
 
#10
The stability of the mosfet is not affected so much by the closed loop of the amplifier circuit. It has less to do with the amplifier circuit and more to do with the mosfet itself. Taking the mosfet out of the circuit, the AC model of a mosfet shows a Hartley oscillator, or a Colpitts depending on how you want to look at it. This internal circuit is comprised of the internal capacitances and lead/bonding wire inductance/resistance. Bob covers this in his book, but I don’t remember the page, my copy is in storage at the moment. The typical frequency this internal circuit oscillates is in the 10’s of Mhz range. It can show up as a hash on the audio frequency signal and cause distortion and cross conduction resulting in failure of the mosfets. It is important to dampen this oscillator so that this does not occur. The things you need to properly use a power mosfet are gate to source protection Zener diodes (lateral type fets usually have this built in), a gate stopper resistor, a gate to drain Zobel (zero set to ~30MHz), and high frequency local decoupling, such as a 100nf film cap || electrolytic decoupling cap from drain to GND (this GND path leading back with a separate trace to the star GND). The frequency of the zero would be 1/(2*pi*R*C); so 1/(2*pi*120R*47pf) = 28MHz. The zero should be set above the oscillation frequency of the mosfet. This filter should go as close to the device package as possible so as to cut out as much lead inductance as possible. It is more complicated than using a BJT, but if done properly Mosfets are great output devices and can be quite reliable in delivering the peak power output into reactive loading due to the lack of secondary breakdown.

Bigun said:
I vote for keep it simple - single ended VAS.

I heard a rumour that HEXFETs do suffer from something that has the same symptoms as BJT secondary breakdown. LatFETs are a better story.
Click to expand...

This is partly true. There are certain types of vertical power mosfets that do have secondary breakdown like features. These are the Trench-fets, Uni-fets, and others with similar architecture. These types are to be avoided for linear output stages. Hex-fets and similar cellular types are OK and do not show this characteristic. Also the planer stripe type devices are immune from secondary breakdown and can take quite a surprising amount of peak power. Vertical fets are designed for switching. This means the datasheets are geared toward that purpose. You have to make sure you don’t have the wrong type of vertical mosfet if you are making an analog output stage. Rds on is irrelevant as is matching Vds. Gm is the most important thing that should be ‘matched’ or rather N-ch and P-ch devices with similar Gm.
 
Last edited:
#11
Hi everyone and thanks for all your posts,
I'm sorry for not answering earlier,

I'm simulating this simple amp architecture that I've found in Bob Cordell's book, it's a very simple topology using lateral MOSFETs (I've used Cordell's models for 2SK1056 and 2SJ162) I don't have good models for BUZ901/906 (and I've lots of them). This amplifier suffers from problems related to high distortion, I've a thread talking about this amplifier and their high THD values.
This amplifier still need some improvements:
- A DC servo??? (I don't have the models of TL081 to simulate it on LTSpice)
- Output protection circuit
- Thermal protection circuit
- Speaker DC offset protection circuit
- TMC or TPC compensation method (I don't know much about this compensation methods, I've started a thread entitled "Need help with TPC/TMC")
- An output tripple??? (I don't know if it's stable)
- A second output pair (I'm concerned about stabillity)
- Supply bypass capacitors
- Vbe multiplier bypass capacitor

I don't know how to calculate PSRR, and I don't know how to improve it.
It's a good idea to use a regulated power supply for IPS and VAS?
About the OPS it could be connected to the raw supply with only the bypass capacitors?

LTSpice circuit diagram attached

Best regards,
Daniel
 

Attachments

Last edited:
#12
Status
Not open for further replies.
Top Bottom