Simple MOSFET amp

[BC]

...

Originally posted by ilimzn





STOP SUGGESTING UNSUITABLE REPLACEMENTS!!! ...


This has been mentioned countless times here, and IMHO whoever suggests replacing lateral MOSFETs with verticals next, shuld be banned. For the umpteenth time: TJHIS WILL NONT WORK, and worse, THE REPLACEMENT MOSFETS WILL GO INTO THERMAL RUNAWAy AND DESTROY THEMSELVES ...

IRF parts do not replace 2SJ/2SK laterals, unless considerable other modifications are made, period. In this particular schematic they will not be able to be biassed correctly, and will fail with a gate puncthrough the first time the amp clips.

Thanks ilimzn.


(To all readers:
I edited the quote. I do apologize to ilimzn in advance and anyone else who may not consider that correct procedure.)

[unclejed613]

that's NOT what i was suggesting...... just using them as TO-3P case dimension examples..... it's what i happened to have at hand for comparison...... didn't mean to open a big can of worms.... please note i used a couple of bipolars as examples as well..... i was only discussing dimensional and mounting considerations, NOT electrical characteristics......

[Nordic]

Hi, I am also busy with a small single mosfet headphone design and have a little quetsion...

Now I know its not a perfect CSS, but I have used it before on a heater grid in a tube based pproject..

I was wondering which is the better of these 2 evils...

useing a large wattage low impendance biasing resistor (8R2) which will see 5V dropped over it... so 610mA or 3W

Or useing an LM317 as a constant current source (this is what I was talking about in first paragraph), useing a 2R2 resistor from adjust to ouput tab should give me about 570mA...( a little less about 6%)

[ilimzn]

If you are asking about the output stage DC load, I would go with the current source, though not necesairly an LM317. The reason being, a current source will attempt to source the set current into any load no matter what the voltage is across the current source - ideally. In practice, there are limits of course. With a resistor, the maximum current sourced depends on the resistor AND the load, as there is a voltage divider thing going on. In general, for a given idle current, a current source will limit your output voltage swing less than a resistor.

In reality, however, things may be different. A LM317, for instance, needs a minimum voltage across it to work correctly, a few V. So, adding that to the 1.25V required on the current set resistor gives you some 4V as the minimum across this current source, for it to still act as a current source - so there you have an output voltage swing limiting mechanism. Fortunately, LM317 is not the only way to make a current source. It would help if we could see the rest of the schematic, but then we are going off topic with this, aren't we?

[Nordic]

Yes, I would hate to interrupt a productive thread, but I think my matter can be resolved with one or 2 small posts, and it relates directly to biasing a mosfet...

All the other Css methods I am familiar with tend to be pretty low wattage...

we need to manage about 3W over the CSS, as far as I see... which I think lies in the LM317's SOA

Ignoreing the pots etc before the fet, the circuit is basicaly 12v rail at the top, 8R2 resistor to irf610 0R82 resistor from irf610 to Gnd.

P.S. how do I calculate the voltage over the small resistor to ground?

[Nordic]

Lol, noone is useing this thread, so I might as well continue...

Got confused, between a few schematics, so ended up with 2r2 resistors instead of 0r82, when i came back from the shop.

Anyhow, I decided to parallel two of those for 1R1 ....

Well it runs pretty nicely... and I made a few test points for analysis, my last one was a beast to faulttrace...

I suppose the change in the resistor is responsible for the operateing point to be diffirent... about 8.85V is the lowest I can get the point marked 7V in the schematic... below this... a humlike distortion appears...I will swap in another 2R2, rarallel to the 1r1 I have currently, but wanted to know; does this affect the gain ?

Also, I have 47k pot on input... it hums slightly when turned to 0... but is dead quiet at 100% with source puased...

Kindly share any insights, advice, optimisations or usefull formulas that would help me conceptualise better...
I get the impression the bias current, input voltage and this resistor has an intimate relationship...

Power supply is 2x12VAC transformer with secondaries in parallel, followed by bridge made from 8A diodes, diodes bypassed with 100nf paper and film caps... followed by 2200uf, lm7812 and 1uf poly... was expecting to hear alot of PSU... but it is quiet...

Next I will construct the second channel, but this time with LM317 + 2r2 current source, in place of the power resistor... should make AB comparison easier... still have to swap the earpiece from ear to ear, due to diffirent frequency sensitivity of my ears... one seems to exlude low frequencies, and the other excludes very high frequencies (frequent grommet receiver of the year here)...

All of the technical things asside, this baby once again convince me that I would be a very easy Mosfet convert...

I love the solid powerfull bass, and slightly rolled of high frequency sensation, and lack of audiable artifacts, would make a very forgiveing headphone amp for bad sources as is... No sibbilance, which tends to be my main gripe with 70% of the headphone designs I try... something in the way high frequencies and rustles dont fade away properly...irritates me quickly...

If anything the sound reminds me of an e188cc cathode follower buffer...

The plan is to liberate a set of Logitech x-530 PC speakers of their crappy amp... they are quite efficient @ 94db/W, and not awefull sounding at all...this I want to use to drive front right and left channels at my PC.

[Lars Clausen]

No don't do that. This design is not meant for those vertical MOSFET types and it will go into thermal runaway.

I have worked quite a bit with IRF's in amplifier applications, and they will work fine. No thermal runaway problems to deal with, however the bias will fluctuate a bit, if the heat sink allows the temperature to swing.
I suggest adding a simple VBE multiplier of 3V, in series with a 3.9V zener. It will work dead stable.

I have always been a fan of simple designs. Not least because they are easy to make, and fun to start up.
However when trying to design an ultrasimple amplifier, you always run into big compromises, that favours throwing in a few more parts, to overcome the flaws of the components.

In case of the IRF's as output devices, they have a very high impedance at low currents. At higher currents the impedance drops off, so this gives problems with distortion. Around 0 output current the MOSFET only draws idle current, so the impedance is high, but as the signal goes up or down, the current rises.

The result is x-over distortion of 2-3%. Unless you go Class A, in which case you can easily get a good result.

Next problem is the high rail loss, that will make this design not only unltra simple, but also ultra inefficient. You can probably get twice the power out with the same heat sink, power supply and MOSFET's by optimizing rail loss.

Next again the poor bandwidth is a direct consequence of the over-simplified design.

[Nordic]

Down with efficiency!!!!!!!!

Why does killing the plannet have to sound this good?

[Nordic]

OK, so I dropped in the 0r82 resistors as per the schematic....

... and it is not behaveing as expected..

I can get it to work quietly unless I turn the trimpot till the 8r2 resistors drop less than 2V over themselves...

It worked alot better with the 2r2 and 1r1 to ground that I tried over the weekend...

I just dont unerstand how the designer claimed to get a 5V drop over the power resistors...

Could the load affect this point?

Changeing the pot on the input from 47k to 10k was an improvement

[ilimzn]

Originally posted by Lars Clausen

I have worked quite a bit with IRF's in amplifier applications, and they will work fine. No thermal runaway problems to deal with, however the bias will fluctuate a bit, if the heat sink allows the temperature to swing.
I suggest adding a simple VBE multiplier of 3V, in series with a 3.9V zener. It will work dead stable.

And did your IRF devices have source resistors? (note the MOSFETs in the simple schematic do not...). My last two amps have IRFs and no source resistors and they DO indeed go into thermal runaway without a bias servo, even though the heatsinks are huge. of course, the runaway may not be destructive, but the bias fluctuation can get way too high and the adjustment completely unreliable - not to mention the bias current being very impractical. So, when you say the bias fluctuates a bit, how much is 'a bit'?

In case of the IRF's as output devices, they have a very high impedance at low currents. At higher currents the impedance drops off, so this gives problems with distortion. Around 0 output current the MOSFET only draws idle current, so the impedance is high, but as the signal goes up or down, the current rises.

Hold on, I think something got lost in the translation here. What do you mean, impedance? Output impedance? If so, it is completely normal and not at all limited to IRFs. It has to do with gm rising with output current, which is one of the reasons higher bias currents are preferred with MOSFETs. With IRF devices this tends to be more obvious because the gm is so high. Higehr bias and some source resistance can aleviate the problem, but efficiency is down. lateral MOSFETs are, of course, not at all immune to the same effect.

Next problem is the high rail loss, that will make this design not only unltra simple, but also ultra inefficient. You can probably get twice the power out with the same heat sink, power supply and MOSFET's by optimizing rail loss.

Yes, but see my simple proposal for a modification (bootstrap 'current source' for VAS). The main reason for the rail loss is the low gm of the laterals, they need about 8V or so Vgs at mximum current, which means this is the rail loss one can expect for more difficult loads. A separate negative supply for the driver stage (positive is not needed with a slight modification of the VAS bootstrap arrangement) would take care of this but at the cost of higher distortion setting in at high power due to nonlinear Cgd effects with Vgd becoming negative. Assuming the amp is modified for IRFs with the inclusion of a Vbe or Vgs multiplier, the rail loss would be somewhat lower due to higher gm of the IRFs.

Next again the poor bandwidth is a direct consequence of the over-simplified design. [/B]

Yes, dependant on the stage currents. There is room to play around with, however. Doubling the currents would not be out of the question (low impedance sources would be preferred!), but even as it is it's usable. not optimal, but as you say, it is a very simple design.