very good mosfet design failure

Hi nothumans, i designed this amp and it works :p in Multisim with 0.000% THD :D,
+-15.82V output voltage.

I am using 5532 opamp and +-37V input voltage feeding the opamp and the 3.5V bias with 20V zener with 1k resistor :))

I built it and it sounded only a few seconds when mosftets failed instantly :mad:@ :mad: i tried 20 times with different modifications and it burns the fets without asking me and without heat them at all, they were cold. i dont have any fets now and i dont want to lose :RIP: :RIP: other fets, they are priceless to me.

I think would it possible the problem to be the bias because in the datasheet threshold voltage is between 2 and 4 volt ? and 3.5 to cause many amps flowing throught both fets ?

I am very confused, PLease help me, Any ideas ? :confused:
 

Attachments

  • problem1.PNG
    problem1.PNG
    84.3 KB · Views: 682
  • problem2.PNG
    problem2.PNG
    91.2 KB · Views: 689

Mooly

Administrator
Paid Member
2007-09-15 8:14 am
The circuit looks a bit erm ;) bizarre.

The biasing for the FET's is totally wrong and uncontrollable... but hey, lets not lose sight of the fact you designed this... and nothing beats practical experience.

Tip... use a bulb tester in all your experiments. That way the bulb lights, and the FET's live to fight another day.

It might be possible to make this work but I haven't time at the moment to look closely at it. The secret will be in rigging up a proper bias network between the FET gates because as you say, even a slight increase in bias voltage will cause the FET's to conduct very heavily.

Also (a bit hard to make out) but the feedback return (R6) needs to connect to the FET's and not the output end of the speaker coupling cap.

It needs a lot of work but it might just be possible to get some music from it.
 
hey initially feedback was before the cap but changing it improved the THD performance and I added 2k to ground before the cap to allow fet sources "touch ground","clearly".

Look at this attachement, initially i built this & actually i listened to it 2 days.. It was fantastic magical experience :D, sounded very good but i made those modifications and everything went to hell :eek:

modifications include adding zener diodes (before i used another supply for opamp and bias was from the high voltage 37) and output stage's input is changed from "resistored" to "capacitored" (in multisim that improved THD and decided to use it with two capacitors)

why u think bias is not stable ? zener diode isn't stable ? is 3.8v way too high ?
 

Attachments

  • jhkljk.PNG
    jhkljk.PNG
    152.9 KB · Views: 635

jlny

Member
2014-01-24 9:08 pm
Mosfet54,
I agree with Mooly's suggestion to look into a more controlled biasing method. I recently got some help here designing an amp that had a pretty similar problem and almost the same MOSFET pairing, so it would definitely be applicable in your case:

http://www.diyaudio.com/forums/solid-state/286540-ab-amplifier-oscillating-seeking-critique.html

I used discrete transistors in place of the Op-amp for mine, but the basic idea is quite similar. :)

One additional suggestion for future experiments would be to put a fuse in line with your power cables. For many years I used to work with a power supply with no internal current limiting, so what I did was I spliced an in-line fuse holder onto my power leads (you can usually find them in auto parts stores or online). Maybe that trick might work for you.
 
Welcome to the real world, i.e. the perils of designing with a simulator without actually understanding the parts and circuits.
For some pointers see if you can figure out the data sheet diagrams for dependance of Vgs versus Id curves on temperature, and also what it means when the treshold voltage specification says 'typical'.\
Also, did you ever think why you can only get output voltage to be around the same as OP-amp power supply voltage, and where the 'rest' of it goes?
 

jlny

Member
2014-01-24 9:08 pm
why u think bias is not stable ? zener diode isn't stable ? is 3.8v way too high ?

The issue is the problem of "thermal runaway." When the MOSFETs get hot, the bias current increases, and this increase in current causes the MOSFETs to get even hotter, which causes the bias current to increase even more, and so on until the entire system overheats and fails.

Alternatively, if the system is unstable and oscillating at high frequency, that might also be causing excessive current spikes that could cause the system to fail. you would probably need to look at the signal using an oscilloscope to determine if this is the case.
 
thank u jlny i readed your thread and learned about the dynamic bias thermo coupled. :):):)

however, im wondering why this happen because fets are cold and this happen only a few seconds after play, i do not think thermal coupling would do the job.

anyway im gonna add thermal coupling, gate stoppers, low pass filter, a bulb for protection and set bias voltage to 2.5v-3v
 
ilimzn, i think output voltage is limited by the limited output of the op amp, i am thinking of long tailed pair but firstly need to learn it :))

another questions: is it possible to run tl072 +-20v when max ratings are +-18 :DD
if it possible to parallel tl072 and ne5532 for synergie?
is it acceptable to implement this without opamp, feedback and with some crossover distortion ?

tommorow will test again and post completed circuit :))
it could be oscillation, gonna test with the scope. Where to see ? At the gates or at the doors ? :D

Even i implemented some discrete op amp with bc 549 and 559 but it is from ready-to-use schematic and i dont want to use it i like when i understand it :D
Im gonna learn long tailed pair and build myself one. With this ready **** there is a lot of distortion in Multisim.
 

jlny

Member
2014-01-24 9:08 pm
Mosfet54,

I would look for oscillation at the output of the power stage first. if it is oscilliating, it will likely be visible there.

I would not use any components past their maximum ratings. Connecting multiple op-amps in parallel should not be necessary.

I think that the main thrust of ilimzn's comment about the output swing is that, if your output swing is limited to +/- 16V by your op-amp, and your supply voltage for you MOSFETs is +/- 37V, most of that excess voltage drop is just being dissipated as heat through the MOSFETs. This is not very efficient, and it would be better to either reduce the main supply voltage to the FETS, reducing the heat being produced, or redesign the rest of your system to run at +/- 37V and make use of the full voltage swing of your system.

It should be possible to replace your existing op-amp with a long-tailed pair similar to the configuration I used in my amplifier. the 2N3904 small signal transistors that I use would not be rated to a high enough voltage for this application, but something similar should be possible here if you want to run from the full supply voltage. That said, that would make for a much higher power amplifier-- probably a lot more than you need. you might be better off going the other direction and reducing your power supply voltage.
 
Last edited:
Why use +-37V while testing a circuit which cannot give you more than +-16V or so. i asked where you think the rest goes, and until you figure that out you will not understand how MOSFETs can fail so quickly. And you are lucky - if it was BJT it would be 1/10s of seconds.
Also - you do NOT fix a parameter which varies from MOSFET to MOSFET. If you set bias voltage to 2.5V it will likely not conduct at all. And if you set 3V, on one MOSFET it may be 100mA, on the next it might be 500mA, on the third 10mA. Mosfet treshold voltage has high tolerances between the same type MOSFET at the same temperature. You cannot rely on this parameter. The proper way is to make the bia system adjustable. Then you start with the lowest setting while checking bias CURRENT through MOSFETs and slowly adjust upwards until you get the desired current. If you have done your job right with the thermo-coupled bias, it will remain more or less constant.
Also - do not even try running the amp with no heatsink on the MOSFETs. Think about it - the silicon crystal inside the MOSFET is tiny and it can heat up far over the destruction level before the outside case even becomes warm.
I am sorry but none of this you will learn from a simulator. Read some basics on electronics circuits and elements.
 
since i do not have another bipolar supply, i use this and i plan to use all of the voltage using long-tailed pair :))

here are some pictures of my build and atmosphere.

i am wondering if transistors resistance go low with temperature highs or not.
i will examine some thermobiases from other circuits and find out :headshot:

i wanted to test today but it will be tommorow :gnasher:

attached pictures include TL494 class D i invented myself and later see its on the net too but mine uses not the error amp of the tl but feedback and dtc pins :D both them work and sound is amazing :D i apply some dc offset to signal with a pot and rock and roll :D

only problem - no deadtime when using single-ended grounded output ctrl for more duty cycle, but i think gonna solve this using totem pole or not ? am i right for this ? i hate the short circuit in high frequency.

another question: how many watts approx. is this transformator ?

---------------------
there was error upploading pictures:

Your submission could not be processed because a security token was missing.

If this occurred unexpectedly, please inform the administrator and describe the action you performed before you received this error.
---

when i tried to upload with URL the problem was they are too big and
they are there -> Index of /diyaudio/
 
Look at the attachment, Is this succes ? :D

Im posting again photos link because its not very visible at the end of my last post :)

Index of /diyaudio/ <-- here

Is long-tailed pair beat the performance of op-amps ?
Or is it good without feedback ?

Heey, this works without R14 when R3 and R4 are 10k
Is it right without R14 ?
 

Attachments

  • heheheh.PNG
    heheheh.PNG
    99 KB · Views: 210
Last edited:

kct

Member
Paid Member
2006-12-24 2:45 am
KW, Ontario
Hi mosfet54,

seeing the pictures, I also suspect transient issue / oscillation that may kill your transistors.
For test purpose, you may want to install current limiting resistors in the supply line to the transistors, so at worst you have a cheap resistor burn up.
This way it is safer to investigate a possible instability and oscillation.
Also, you may want to consider having the wires that come of the board twisted.