The Kuartlotron - keantoken's simple error-correction superbuffer

[keantoken]

Looks good so far even with an IRF510, which makes it far faster. I was using the wrong lytic in my simulation. The 1MHz resonance also seems to be from the SMPS.

[Salas]

Originally Posted by keantoken

I just threw one together on the breadboard. I just happened to have a single IRF630. Testing with an SMPS at ~75mA I see resonances at 20MHz (seems to be from the SMPS) and 1MHz.

IRF9630 it could use, not IRF630, thats for the negative polarity. Termination lytic should not be low ESR.

[Salas]

Originally Posted by keantoken

Looks good so far even with an IRF510, which makes it far faster. I was using the wrong lytic in my simulation. The 1MHz resonance also seems to be from the SMPS.

No problem pretty soon then and its just makeshift. You will test again in your prototype PCB layout. My sim gives 133 deg phase margin at gain=0dB open loop analysis with the original TO-247s and 0.3R ESR 5nH ESL assumed terminal cap parasitics. It never resonated on its dedicated board. Try use no wimpy MOSFET if you will leave it open to user hot-rod. You will sort it out, no worries.

[wintermute]

Project16's "all in one" posts moved here http://www.diyaudio.com/forums/analo...onception.html

[Project16]

Originally Posted by wintermute

Project16's "all in one" posts moved here http://www.diyaudio.com/forums/analo...onception.html

Many thank's wintermute!

[keantoken]

The IRF510 can handle at least 20W. The main thermal hazard is not FET power, but inadequate sinking, which will be equally destructive for IRF510 and IRF240. Considering this I don't see a point in using TO-247 devices. They are slower and probably not even more stable than the IRF510. With good sinking the IRF510 in this design can handle more than 1A of current.

[jean-paul]

The Lego piece you can use has been tested and proven to work OK.You have permission of the author to use his tested and working design. Why loose time questioning the size of the original parts ? The MOSFETs are not expensive and with TO247 you won't need heatsinks with low current. You could click this Lego piece on your Lego piece and test and measure. Maybe it is because I am a builder and have never simulated anything but things seem to delay without necessity. Sorry but, as said before, IMO focus should be on completing this design to a complete working device that can be used in 99% of cases in everyday life. For this reason I will focus on Project16's design. Good luck with the project, the topology is promising.

[Salas]

Originally Posted by keantoken

The IRF510 can handle at least 20W. The main thermal hazard is not FET power, but inadequate sinking, which will be equally destructive for IRF510 and IRF240. Considering this I don't see a point in using TO-247 devices. They are slower and probably not even more stable than the IRF510. With good sinking the IRF510 in this design can handle more than 1A of current.

As I wrote, the topology lends itself to various devices substitution. If you will decide a fixed constant current then you can pre-package the thermal policy too. If left open to user hot-rod it must be taken into account that the bigger MOSFET has much better RθJC, RθJA, RθCS. Constant current power dissipation is not giving cooling breaks, its not lending itself to the pulsed rating spec of a device. I.e. the power spec gets derated and then again for long life reliability. There is no rule you should leave it open to user hot-rod. You may as well optimize a level. Your choice.

[keantoken]

If there is any significant pulsing in a shunt regulator, then it's already broken. It is straightforward to set operating points according to thermal limits. Am I missing something? A drain resistor should be enough to prevent FET destruction in the prototype.

[Salas]

The SOA curves for power MOSFETS are usually given in pulsed current durations since they are made for switching applications primarily in mind. Thus their constant power abilities should not be based on their 25C case temperature pulsed current spec in the data sheet. That's what I am saying.