Power amp under development

Andrew,
I have got one board running fine now. The other still has a problem in that if you replace the test 10 ohm resistors in place of the fuses with fuses then the bias starts to run away. Now this board did work before but the fets were not matched and I tried to do it but something went wrong. I have read somewhere where Quasi have said that the circuit were developed to work with a Vgs of the fets being around 4V.This is what I have found.
Vgs of fets in working board.
Pos rail: 3.894V, 3.819V, 3.819V
Neg rail: 3.768V, 3.711V, 3.743V

Faulty board Vgs.
Pos rail: 3.602V, 3.608V, 3.616V
Neg rail: 3.547V, 3.570V, 3.574V
Can it be that the Vgs of the fets of the problematic board is too low, that it needs to be in excess of 3.7 V ? Maybe I need to change the drivers to MJE15030 /MJE 15031.
I have changed all semi's and all caps and checked the resistors. This is the last thing that I can think of, otherwise I will make a new pcb.😕
 
The other still has a problem in that if you replace the test 10 ohm resistors in place of the fuses with fuses then the bias starts to run away.
I have advised many times that this method of setting output bias is wrong.

Most amplifiers change their bias current and heating when the supply voltage is changed.
Usually the bias increases when the rail resistor is removed. Now that the quiescent checking resistor is removed you don't see the increase nor do you see whether it settles at a new higher value or keeps on increasing until the amp blows up.

The resistor can be used to check front end quiescent current during the early stages of testing, but never during output bias setting.
 
Andrew, you are right, but the resistors are just for checking and safety and I do back the bias pot down to zero before I replace the resistors with fuses, so do you think it might be the lower Vgs fets which causes this calamity ? No way for me to get the bias lower at the lowest setting.
 
Andrew,
I have replaced R13(470ohm) with a 390ohm resistor. I can set the bias perfectly now without any safety resistors. The current through R4 and R8 is ok at 1.3mA. The current through R12 is a little higher. Is 14mA should be 12.5mA but that should not be a problem. The current through R20 should be 18mA but is 15ma. Does this sound Ok to you. I have not had any of the two boards on a scope. Will do tomorrow if all is good. 🙂
 
The current through R4 and R8 are not needed. What you should be checking are the currents through R5 and R10. These should be identical. Did you match R5 and R10 before you fitted them? Did you fit VR1? It should not be needed.

The LTP currents are determined by the current through R7.

The VAS current is determined by the current through R15.

The current through R20 is determined by the Vbe multiplier output.

It seems from your questions, you don't understand the quiescent settings given by the schematic.
 
The bias current/bias voltage of a mosFET output is not critical to ClassAB operation.
This accurate bias voltage applies to ClassAB EF BJT output stage.
Here bias voltage should be ~26mVre, or a bit less with very small Re values.

For a mosFET output stage the crossover distortion gets less as the bias is increased, all the way up to the bias to obtain ClassA.
Borbely tells us to aim for at least 500mA of output bias irrespective of how many output pairs are used. Many others tell us that a mosFET output stage is very suited to ClassA operation.

500mA for a three pair output stage would be 167mA through each 0r33, developing Vrs = 55mV
The less you use, the more the crossover distortion, the harder the feedback has to work to try to correct for the error.
 
Last edited:
Thanks Andrew, in this case of this specific amp Quasi recommends 30mA across each 0.33 ohm source resistor for an output pair, there are 3 pairs so that is 90 mA. It means 10mV across each source resistor. I am a little scared of flames if I take it too high. Do you think it is safe to up the bias to 55mV over each gate resistor, output devices are IRFP460.
 
You would need to do a power dissipation assessment and thereafter determine the remaining power rating available to handle first resistive test loads and finally reactive speaker loads.

You can run a ClassA amp very hot and use up all the Power rating just getting rid of the heat. But that leaves nothing to cope with the loads you intend to pass current through.

Have you ever downloaded Bensen's spreadsheet?
 
No , I haven't but I will check it out just now thanks. Just a little update. I had one module running at 10mV over each .33ohm source resistor. Quasi instructed to use 100ohm as testing resistors in the place of the fuse. Got 9 V over the resistors. Then replace with 10 ohm resistors. Set bias to .9 V. He then suggests to insert the fuses and bias would be good as final adjustment. I did that and the vd over the source resistors went up to 55mV and the came down rapidly until the fets of the negative rail blew. Luckily I have plenty. I have read in the thread that many people had this problem. I suppose there is stability problems with this specific design. I think that I will be looking rather at the quasi mosfet amp of Apex. It might even be fake components causing this. First I'll check out Bensen's paper. Do you maybe have a link ?
 
it's in here.
But also download David Eather's article showing the long hand method. It too is in here.

Or you can send me an Email requesting my modified versions covering FETs, or BJTs or even BJT Drivers.
There are a few alternative devices in each excel.
Others can be added, but doing the SOA from the datasheet is a bit complicated. I did it by trial and error until each device matched it's datasheet.

I also added in a few versions of IV protection. It is capable of showing the "bad" version adopted by Leach and how the improved version in Jens' PCB is very much better.
 
Andrew, I have fixed the board again, only one fet that went faulty. I have adjusted the bias over the 0.33ohm source resistors to 10mV which is 30mA, it has been running for more than an hour with bias varying between 8.9-10.2 mV. It should be okay now. Played some music and it sound good. Plse remember to mail me the promised info. 🙂
 
Well, rail voltage is still 50V and you should simply add more output transistors, like nmos 350 version.
I also made a layout with 2 pairs of FET, see below
 

Attachments

  • nmos 200 2P.jpg
    nmos 200 2P.jpg
    123.1 KB · Views: 451
Using mosFETs for the output stage makes it very easy to decide how many are required to give reliable operation.
Maximum output power equals the Total dissipation of all the output devices divided by four (4).

If you have a +-50Vdc supply, then expect around 100W into 8ohms. That would require 400W of output devices.

If one decides to build for 4ohms duty, then the expected output would be just short of 200W. That would require 800W of output devices.
But to realise that target 4ohms power you will need to consider seriously where the current will come from and how you can deliver that current through the amplifier to the load. That is not easy. Many fail in this "current capability" and instead of getting close to 200W they end up with only 150W to 160W. In my book they have failed: 0 out of 10 !
 
Last edited: