I meant that;
IRFP240 has 280W power dissipation capability, with 6v2 zener we can provide ~2,1A bias that more than enough...
And without feedback dc bloackage capacitor, we would have a DC offset problem but with 100K tirmpot between offset comp pins of NE5534 we can fix the output offset..
I've designed that about one year ago and one of my friend has tested it!
What do you say?
An externally hosted image should be here but it was not working when we last tested it.
IRFP240 has 280W power dissipation capability, with 6v2 zener we can provide ~2,1A bias that more than enough...
And without feedback dc bloackage capacitor, we would have a DC offset problem but with 100K tirmpot between offset comp pins of NE5534 we can fix the output offset..
I've designed that about one year ago and one of my friend has tested it!
What do you say?
The opamp theoretically runs in class A in this design as it sources current to the base of the output transistor but it does not sink current from it. The problem is that if the output transistor is slow the opamp will probably try to suck carriers from the base and enter the class B. The mosfet version will be more class-B (unless you add a constant current source/sink to the opamp because it will sink and suck carriers from the gate capacitance. This can involve several ma's at high slew rates).
I have just ordered a couple of MJL4281A and MJE15034 to try arrange them in either EF or Sziklai.
The Sziklai arrangement would be that:
http://sound.westhost.com/project3a.htm
I have no idea if this will work here
I have tryed to drop-down a mosfet with a gate resistor and it did not work (no idea why), but once i got an LM4562 to drive mosfets and there where problems with clipping voltage (you should "lift" the gate voltage with a ccs-loaded vbe multiplier unless you want to listen at really low volumes.
Another problem with mosfets is that if you consider them in conjunction with the gate stopper resistor they aren't as fast as people belive (-3dB point at few MHz ignoring Cgs and taking only in account Cgd*) apart from distorting more open-loop because they have lower transconductance. I would not do it but this does not mean it should not work/not sound very good. If you have experience in building amps and don't mind if you break one then just try.
* Cgs is bootstrapped by the load to some 5/10% of its value in the source follower arrangement.
About the opamp i would say "the better the phase margins the better it will work". I have no experiences with this particular opamp but i suggest to place it on a socket so you can try many (there are even discrete opamps and discrete opamps with tube input stages out there).
The current sinks take 0.65/R A each one at room temperature, where R is the 0.5 Ohms resistor (R2-R8). This means they will sink some 1.3A each. You can make them sink a bit less to get the desired 2A or place a single one (which will get really hot). I don't know if a mosfet ccs would do better. But remember that loudspeakers actually dip much lower than its rated impedance. I belive a BW CM1 rated 8Ohms dips to 3.1 at some frequency, so a monitor audio that does not dip below 5Ohms would be a better choice.
If you have a scope feed a sine wave to it (it would be better to have the loudspeakers in another room). If it looks clipped as usual at both sides then the current is enough. If it has a dip at one side then it lacks current. Try it over all the frequency spectrum to see that clipping does not happen anywhere. If the amp is ok but does not have enough current somewhere they you may notice that it sounds good always but in some particular passages of some program material.
If you have built an amplifier that you like you can always take its output stage and place it in replacement of the upper tip-142 (The two transistors and the one or two resistors). Emmiter/source degeneration makes no sense there because bias current does not depend on the gate voltage, and adding one will only reduce feedback and increase distortion if it does anything at all.
The 100pF transistors is for the TIP142 output stage to make the opamp a tad slower and not going crazy if the output does not respond as fast as it wants it to do. If you have a scope you should choose the least value that makes the trace not look fuzzy at any output voltage.
Another idea i bear in mind is running the output stage open-loop, as the sziklai approach goes to 0.05/0.01 % THD at full power in 8Ohms, but this requeres offsetting the signal with a diode thermally coupled to the output stage and i don't know how to do this without passing the signal to a large cap (or worse, an unbypassed diode).
I have just ordered a couple of MJL4281A and MJE15034 to try arrange them in either EF or Sziklai.
The Sziklai arrangement would be that:
http://sound.westhost.com/project3a.htm
I have no idea if this will work hereI have tryed to drop-down a mosfet with a gate resistor and it did not work (no idea why), but once i got an LM4562 to drive mosfets and there where problems with clipping voltage (you should "lift" the gate voltage with a ccs-loaded vbe multiplier unless you want to listen at really low volumes.
Another problem with mosfets is that if you consider them in conjunction with the gate stopper resistor they aren't as fast as people belive (-3dB point at few MHz ignoring Cgs and taking only in account Cgd*) apart from distorting more open-loop because they have lower transconductance. I would not do it but this does not mean it should not work/not sound very good. If you have experience in building amps and don't mind if you break one then just try.
* Cgs is bootstrapped by the load to some 5/10% of its value in the source follower arrangement.
About the opamp i would say "the better the phase margins the better it will work". I have no experiences with this particular opamp but i suggest to place it on a socket so you can try many (there are even discrete opamps and discrete opamps with tube input stages out there).
The current sinks take 0.65/R A each one at room temperature, where R is the 0.5 Ohms resistor (R2-R8). This means they will sink some 1.3A each. You can make them sink a bit less to get the desired 2A or place a single one (which will get really hot). I don't know if a mosfet ccs would do better. But remember that loudspeakers actually dip much lower than its rated impedance. I belive a BW CM1 rated 8Ohms dips to 3.1 at some frequency, so a monitor audio that does not dip below 5Ohms would be a better choice.
If you have a scope feed a sine wave to it (it would be better to have the loudspeakers in another room). If it looks clipped as usual at both sides then the current is enough. If it has a dip at one side then it lacks current. Try it over all the frequency spectrum to see that clipping does not happen anywhere. If the amp is ok but does not have enough current somewhere they you may notice that it sounds good always but in some particular passages of some program material.
If you have built an amplifier that you like you can always take its output stage and place it in replacement of the upper tip-142 (The two transistors and the one or two resistors). Emmiter/source degeneration makes no sense there because bias current does not depend on the gate voltage, and adding one will only reduce feedback and increase distortion if it does anything at all.
The 100pF transistors is for the TIP142 output stage to make the opamp a tad slower and not going crazy if the output does not respond as fast as it wants it to do. If you have a scope you should choose the least value that makes the trace not look fuzzy at any output voltage.
Another idea i bear in mind is running the output stage open-loop, as the sziklai approach goes to 0.05/0.01 % THD at full power in 8Ohms, but this requeres offsetting the signal with a diode thermally coupled to the output stage and i don't know how to do this without passing the signal to a large cap (or worse, an unbypassed diode).
I have just seen your shematic, but remember, if the opamp can go, say 2V from the rails, then the output mosfet will go maybe 8V from the rails (assuming you are using high-vgs mosfets like the IR's) This voltage will not be (much) affected by the drain voltage.
If you can live with extra dissipation instead of placing a mosfet i would cascode. It's strange but i belive that mine sounds much better as it heats up, this is reasonable as some plots about beta-vs collector current get flatter at higher temperatures. this would remove some "thermal memory effect" as i belive that the die can get some 10K hotter with current peaks. This affects more the ccs transistor than the active one, but it might cause some compression in the upper one as it colds down at the extremes.
This may also give better FR in the Sziklai arrangement but i'm not sure.
Anyways the TIP-142 has a politeness that i belived exclusive to tubes! Maybe the not outstanding frequency response of the TIPS does a bit of smoothing (this would be terrible in a class B amp as the feedback cancellation of crossover artifacts depends on the feedback at high frequencies and slow transistors would probably sound harsher).
If you can live with extra dissipation instead of placing a mosfet i would cascode. It's strange but i belive that mine sounds much better as it heats up, this is reasonable as some plots about beta-vs collector current get flatter at higher temperatures. this would remove some "thermal memory effect" as i belive that the die can get some 10K hotter with current peaks. This affects more the ccs transistor than the active one, but it might cause some compression in the upper one as it colds down at the extremes.
This may also give better FR in the Sziklai arrangement but i'm not sure.
Anyways the TIP-142 has a politeness that i belived exclusive to tubes! Maybe the not outstanding frequency response of the TIPS does a bit of smoothing (this would be terrible in a class B amp as the feedback cancellation of crossover artifacts depends on the feedback at high frequencies and slow transistors would probably sound harsher).
ionomolo said:
Right, I was not considering that Vgs matter. In fact my original design was based on IRF510 and working with 100mA bias as a headphone amplifier. And its still working with very nice sounding.
This version has been built (but designed by me) by one of my friend. I guess He's happy with it because until now I heard no compaint from him. May be He's using it with very sensitive speakers (may be a pair of Klipsch) then he had no need that much power until now.
Anyway,
May be MOSFETs would replace with 2SK1058 type lower Vgs ones and useable as parallelled.
And about the sound quality of it, I agree with you, yours will have a better sound. I dont like MOSFET sound too much also. But its simpler than yours.
May be I can add a gate driver BJT to improve sound for upper MOSFET.. I heard that some people, may be it works..
I don't see the bjt version to be more complicated, the mosfet ccs depends on parameters that are a bit harder to control (Vgs) and the upper part of my current prototype is a single transistor, so it has one part less than the bjt one (the gate stopper).
BJT designs may get harder in class B because they need thermal compensation, but that's not the case here!
BJT designs may get harder in class B because they need thermal compensation, but that's not the case here!
Ok, I was talking about the current version of mine. I am using one MOSFET as CCS and yours two!
But if I should modify it (so it looks I must) then it will be more complex than yours.
However what about the opamps output current in your design?
Youre talking about 10mA... In this case the opamp cannot work in class A... Or am I mistaken?
But if I should modify it (so it looks I must) then it will be more complex than yours.
However what about the opamps output current in your design?
Youre talking about 10mA... In this case the opamp cannot work in class A... Or am I mistaken?
I have to confess that the current prototype i'm listening now has a single ccs because my transformer can't supply more than 2A.
Class A is not about current. Class B means that there are two devices sinking and sourcing current and that one works for one half of the cycle and the other works during the other half.
If you place a constant current source at the output of the opamp then the whole thing will be able to source-sink 10 mA's of current without the opamp changing between "sourcing mode" and "sinking mode" so there will be no crossover. That's because the opamp will source 10 mA's for zero output, 20 mA for maximum output and 0 mA's for negative output (or the opposite if you are using a source instead of a sink.) But it will be always sourcing so the only operating pair of the chipamp will be the upper one and there will be no crossover. If you place a load that is so light that its under the ompamp output stage bias current you will also have class A operation, but i belive that with a mosfet load the impedance of the gate at high frequencies will eat more than this and the opamp will enter class B.
The LM4562 can give a bit more than 20 mA's of current, that does mean that 10 ma's will be the choice that gives the most swing in the class A region, but not the best-sounding. If you need less than 10 mA's putting less stress on the opamp will probably sound better.
Class A is not about current. Class B means that there are two devices sinking and sourcing current and that one works for one half of the cycle and the other works during the other half.
If you place a constant current source at the output of the opamp then the whole thing will be able to source-sink 10 mA's of current without the opamp changing between "sourcing mode" and "sinking mode" so there will be no crossover. That's because the opamp will source 10 mA's for zero output, 20 mA for maximum output and 0 mA's for negative output (or the opposite if you are using a source instead of a sink.) But it will be always sourcing so the only operating pair of the chipamp will be the upper one and there will be no crossover. If you place a load that is so light that its under the ompamp output stage bias current you will also have class A operation, but i belive that with a mosfet load the impedance of the gate at high frequencies will eat more than this and the opamp will enter class B.
The LM4562 can give a bit more than 20 mA's of current, that does mean that 10 ma's will be the choice that gives the most swing in the class A region, but not the best-sounding. If you need less than 10 mA's putting less stress on the opamp will probably sound better.
Dxvideo said:
Yes, i was talking about the capacitance. It looks like a resistance + a phase shift, and the resistance is 1/(2*pi*C*f). This looks like some 3 KOhms at 100KHz, but assuming the cutoff frequency of the opamp to be about 5.5 MHZ (55 MHz gain-bandiwidth product at a gain of 10) the worst case would be 57 Ohms. This is not high. Also, the current requiered for 20V/usec slew rate (maximum of the LM4562) will be 10mA. If you use the OPA637 (not sure if this would be a good idea) the current drive to keep its 135V/usec slew rate would be 68.5 mA. This is much!
Not now!
I am currently working on a buffered and regulated gainclone project and mostly finished.
And the next one will be a LM4702+SAP15 project. I have already purchased the components and sent the PCB files to my PCB maker..
By the way, do you have any experience with that SAP darlingtons?
I am currently working on a buffered and regulated gainclone project and mostly finished.
And the next one will be a LM4702+SAP15 project. I have already purchased the components and sent the PCB files to my PCB maker..
By the way, do you have any experience with that SAP darlingtons?
Yes I'd a chance to examine some papers about the LME49810. Especially its output current capability and slew rate seems impressive.
I've requested two samples from National, however not arrived yet.
May be I can built a dual mono darlington (but which darlingtons?) with them as the next of next..
I've requested two samples from National, however not arrived yet.
May be I can built a dual mono darlington (but which darlingtons?) with them as the next of next..
The green ones are to filter the signal that enters the board. I belive they are 1000uF or so but i'm not sure. They serve the same purpose as the ones in a LM3886 amp.
The REV3 schematic includes the black one at one leg of the "active" transistor. It should be > 10uF because there are largue signals here. i suggest you to try anything between 47 uF and 220 uF. The larguer will add some spikes when the caps load at the rectifier pulses, so it's better to have a low one while keeping on the safe side, what about 100uF?
I strongly suggest to build it with only one ccs to see if you like it, but the most important, if it gets too hot. If everything is ok and not burning then add the second. (It will be hot anyways but you should be able to keep your hand on it for a while).
The REV3 schematic includes the black one at one leg of the "active" transistor. It should be > 10uF because there are largue signals here. i suggest you to try anything between 47 uF and 220 uF. The larguer will add some spikes when the caps load at the rectifier pulses, so it's better to have a low one while keeping on the safe side, what about 100uF?
I strongly suggest to build it with only one ccs to see if you like it, but the most important, if it gets too hot. If everything is ok and not burning then add the second. (It will be hot anyways but you should be able to keep your hand on it for a while).
I have built an amp like this before. It seems like a very cool design and with good quality components, it should sound very good.
I had oscillation problems like you had. I put a Zobel Network on the output to get rid of the oscillation.
Have you thought of modulating the constant current source? It keeps the system in Class A and gives better efficiency all together.
I had oscillation problems like you had. I put a Zobel Network on the output to get rid of the oscillation.
Have you thought of modulating the constant current source? It keeps the system in Class A and gives better efficiency all together.
davidallancole said:
I have just ordered a couple of beta-enhanced devices to try if this gives better sound. They also have an Ft of 60Mhz at 2A instead of the 1MHz of the TIP142.
I will try to arrange them as Sziklai and pray it to be stable.
The idea of modulating the ccs is not bad at all, but i reserve it for mk II.
Thanks for the interest and happy that you get godd results.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.