25W Mosfet Amp troubleshooting

Hi,
as one of my first 'complex' projects I decided to build the 25W amp from REDCircuits. ( 25 Watt MosFet Audio Amplifier - RED - Page2 )

I designed the PCB using Eagle:

[IMGDEAD]http://imgx.gxzone.com/images/9/6/968266f6a40.png[/IMGDEAD]

I built the circuit as it was presented on the web. ( +-33V supply from a 120VA 220/25+25 TR with 4700uf caps)

The first test was made using a +-28V regulated lab supply and it worked flawlessly. Even tough there was some oscillations at output when using as an input a 1kHz sine-wave generator; my first guess is the oscilations where on a non-audible frequency. (Music did sound pretty clear)

I did forgot to leave some sort of pins for measuring the quiescent current from the fets so I had to do it by hearing. Is there any other way of measuring it?

[IMGDEAD]http://imgx.gxzone.com/images/1/a/1a79feb7667.jpg[/IMGDEAD]

So I built the power source according to schematic's and run another test.
On the scope it started to show the amplified 1kHz wave but the 10Ohm R16 resistor started smoking. I shut everything off and checked the whole circuit again. The power source seems pretty strait-forward to built..
On the next run the same resistor now started to glow like a 45Watt lightbulb so I had to turn it off again.

[IMGDEAD]http://imgx.gxzone.com/images/9/8/98f3baae3ad.jpg[/IMGDEAD]

I did the simulations on Spice and with the R11 trimmer on about 250Ohm, R16 is supposed to draw only about 20mA. I checked the C5/C6 caps that link R16 to ground and they are alive and working. I asked for help on checking the mosfets to one of my university teachers and he told me they were on good shape.

I have to admit I'm pretty lost. The only fehacible answer is that R16 is somehow connected to ground in order for such a large current to be flowing through it. I re-checked the PCB and could not find such a path.

[IMGDEAD]http://imgx.gxzone.com/images/0/3/03aeb695f6c.jpg[/IMGDEAD]

Any thoughts?

Thank you in advance,
Agustin
 
Parasitic oscillations will smoke an amplifier.

If they are audible, that is very unstable.
If they are only visible on a scope then it is still unstable and needs to be stabilized.

Layout might have an effect on that.
The schematic looks ok. Nothing glaring at first look.

Make sure you have the right values in the right holes. It is hard to see your own errors, it is for everyone.

Inserting a 0.1ohm or approximately that value +200% on the source side of a Mosfet is fine will let you measure the bias current.

R16, check the polarity of the two caps on the supply rail on the board...


_-_-
 
I thought that these oscillations were product of the LC tank made by the C11 output cap and the coil from the speaker. Is there a way to stop that kind of oscillations or if they are produced by pcb layout, is there a basic thumb rule to how to design a new pcb?

I already took the time to re-check every polarity and resistor value, couldn't find any misplaced one. My last resource is to get a new pair of mosfet, but since shops are closed for the whole week in my city, I wanted to study every other cause for this problem.

The 0.1ohm resistor should go between Q8 drain and +33?
 
I checked every path for any shortcircuit with surrounding/earth and all seem to be fine.
My biggest concern is why it did worked on the first try, but when supplied (two days later) with the schematic's power source it burned R16.

Every electrolitic cap seems to be ok. Even though I was told Mosfet's are ok, how could a failure in Q9 affect the current flowing through R16?
 
The circuit is designed for lateral mosfets. This is obvious because there is no thermal feedback, or Vgs multiplier. Vertical fets require thermal compensation at class AB bias.(class B???:whazzat: ) There are no source ballast resistors. There is no gate to source Zener protection. The PCB has no de-coupling caps for the output stage. Proper grounding is very important. There is no local compensation for the mosfets. Parasitic oscillations from the mosfet will be in the 10's of MHz range and will likely be dependent on current conduction due to the dependence on transconductance (Gm) which is not the same at low currents vs high currents. It is very likely that the bias is too low and cross-over distortion is present. Mosfets suffer from reduced Gm at low channel conduction. This means that at low conduction around crossover they require a greater change in Vgs vs Id. A typical bias should be >100mA. There is no way this will be steady without source ballast resistors and a Vgs multiplier.
 
Last edited:
I think as a first project using fets, this circuit could be practical if you used lateral type fets. You would need to add source resistors and lower the value of R11. Lateral fets ususally have Zener protection built in and also a much lower threshold voltage, Vth. Local decoupling caps with a robust trace to the star gnd point would help alot. You should try to keep signal traces as short as possible, especially for high impedance nodes like the base of the input transistor. Also you must seperate the gnd paths from the signal gnd and the input stage/feedback gnd references from the power supply and high current return paths. To put it another way, there should be seperate roads back home to the star gnd point.
 
Hincha de Peñarol!!!!! ;) ;) ;)
That's your mistake: you must connect C5 **positive** leg to ground !!!
By the way, it's shown in the original schematic ;)

That said, "RED circuits" are terrible. :(
This one does not have thermal compensation and is unstable :(

Nacional Nacional..
I actually made the mistake when writing, it has it's positive leg to ground. :p

As my first 'complex' circuit I really wanted to finish it before starting a new one.
I have been reading a lot about class D amps and I believe they are the way to go on the next project. Even though I found a lot of nice 15-25W amps on the class D forum, most of them needed some relative new IC's (like the TDA8920, TDA7492, etc) which mostly cannot be found in my country. I also found a few with a more common triangle wave generator + comparator using op amps, but they mostly use an IC to drive the mosfets (like the one from ejtagle that uses an IR 2110).
Since I really want to stick to small amps to keep studying and practicing before getting into a bigger/cleaner-sounding amp, could you advice me about a class D low-power amp that fits that description?

Thank you a lot for all of your advices!
Agustin
 
A bit late now, but anyway....

There's something else wrong in that schematic, aside from what others have mentioned. It doesn't seem to have any sort of compensation scheme for high frequency stability.

At first glance, it looks like C9 is supposed to do the job but it's connected in series with a 150K resistor (R8), so it won't have any effect at all. Maybe R8 is supposed to be 150 Ohms, or maybe it should just be left out.

As it is, R8 and C9 don't do anything.
 
Agree.
It also lacks a Zobel network.
Red circuits are homemade designs very crude and unfinished, they will probably simulate well, even work (sort of) marginally stable on a bench, driving resistive loads a few inches away, and for a few minutes, but in the real world, driving real speakers (very complex impedance) through long wires (lots of parasite inductance and capacitance) is something else.
 
I think as a first project using fets, this circuit could be practical if you used lateral type fets. You would need to add source resistors and lower the value of R11. Lateral fets ususally have Zener protection built in and also a much lower threshold voltage, Vth. Local decoupling caps with a robust trace to the star gnd point would help alot. You should try to keep signal traces as short as possible, especially for high impedance nodes like the base of the input transistor. Also you must seperate the gnd paths from the signal gnd and the input stage/feedback gnd references from the power supply and high current return paths. To put it another way, there should be seperate roads back home to the star gnd point.

Like CBS is saying, even with lateral fets this circuit isn't that good. Cloning the Hafler DH120 would be better which has all the improvements CBS has listed.
 
Like CBS is saying, even with lateral fets this circuit isn't that good. Cloning the Hafler DH120 would be better which has all the improvements CBS has listed.

The Hafler looks like a very nice option for a higher complexity project. Thank you!

As of now I'm trying to figure out a little bit more about the modifications that jerluwoo proposed in another topic:
http://www.diyaudio.com/forums/solid-state/234253-advice-choosing-new-20w-project.html#post3457285
I draw the new circuit in LtSpice with some modifications:

+-VCC at 33V (as of now I was trying to figure out how to model my power source in order to get the exact voltage on load, but according to the builders info it should be around 33-34)

- The same decoupling cap as RedCircuits schematic: 0.47uF

- Same output filter as RedCircuits schematic


Do you have any other advice for such modificated circuit?

Thanks in advance!