It depends on how much standing current you want to pass.
If you dont mind settling for just enough to get rid of crossover distortion on the scope then a simple Vbe multiplier works not on the heatsink.
If you want 100mA or more then you need the Vbe multiplier on the heatsink or glued on an output transistor.
I found using 240/9240 I had no problems.
However, it got a bit more difficult when using a quasi output stage.
I found I needed the Vbe multiplier on the heatsink.
I have built about 40 complimentary pcb's and 2 quasi's to get my data.
Most topologies used so called "true complementary" power stages, even by IRF MOSFETs. If such power stages runs in pure class A, no matter regarded the sound.
But if such stages runs in Class AB (mostly 50-150mA idle current/each pair) the sonic results always worse, because the term "true complementary" itself is an error (present only in therory and schematic).
Only inexperienced developers choose "true complementary" power stages topologies. But experienced developers like L. Ollson or diyaudio member "X-PRO" choose only N-Channel MOSFETs in the output with the related front end and get better results especially by not "pure ClassA" resp. hard running bias adjust. My favorite topology is the CSPP/PPP circuit (inside also by commercial brand SUMO).
Check out this threads:
http://www.diyaudio.com/forums/soli...-mk2-circuit-descr-arround-q12a-wanted-2.html
http://www.ant-audio.co.uk/Theory/N-channel D-MOSFET output stage with improved linearity.pdf
http://www.diyaudio.com/forums/soli...better-audio-non-complements-audio-power.html
http://www.diyaudio.com/forums/soli...e-ended-related-solid-state-output-stage.html
http://www.diyaudio.com/forums/solid-state/154388-its-cheap-its-n-its-dirty-its-circlomos.html
http://www.diyaudio.com/forums/solid-state/162310-how-identify-quasi-complementary-amplifier.html
http://www.diyaudio.com/forums/solid-state/23888-sumo-power-amp.html
But if such stages runs in Class AB (mostly 50-150mA idle current/each pair) the sonic results always worse, because the term "true complementary" itself is an error (present only in therory and schematic).
Only inexperienced developers choose "true complementary" power stages topologies. But experienced developers like L. Ollson or diyaudio member "X-PRO" choose only N-Channel MOSFETs in the output with the related front end and get better results especially by not "pure ClassA" resp. hard running bias adjust. My favorite topology is the CSPP/PPP circuit (inside also by commercial brand SUMO).
Check out this threads:
http://www.diyaudio.com/forums/soli...-mk2-circuit-descr-arround-q12a-wanted-2.html
http://www.ant-audio.co.uk/Theory/N-channel D-MOSFET output stage with improved linearity.pdf
http://www.diyaudio.com/forums/soli...better-audio-non-complements-audio-power.html
http://www.diyaudio.com/forums/soli...e-ended-related-solid-state-output-stage.html
http://www.diyaudio.com/forums/solid-state/154388-its-cheap-its-n-its-dirty-its-circlomos.html
http://www.diyaudio.com/forums/solid-state/162310-how-identify-quasi-complementary-amplifier.html
http://www.diyaudio.com/forums/solid-state/23888-sumo-power-amp.html
Last edited:
Thats is right. It has thermal problem (thermal runaway) when junction temperature start to reach 70C or above. Better to keep the temperature below 65C. DMOS working faster and stronger than LateralMOS, but lateral mosfet has exelent linearity and thermal stability.
Hi.. I'm resurrecting zombie thread
Is is possible to directly replace (with minor modification) those Hitachi mosfets with TO3 IRF230/9230??
Thanks
okky
The simple answer to that is no
Whether it's possible depends on the circuit... the bias generator would need to provide the required voltage for the IRF's and as others have eluded, also be able to provide thermal tracking.
The laterals are far superior for a class b (ab) design at low bias current.
Why are you looking to replace with IRF's ? have the originals failed ?
I have built several versions o a design that is extremely stable thermally, using a single pair of IRFP240/9140 without any source resistors at all. The bias generator is a simple Vgs multiplier - a MOSFET version of the classic Vbe multiplier - using an IRF610. This particular part was chosen simply because it was at hand and because it uses the exact same geometry on the silicon except fewer cells, so thermal tracking is almost perfect. The metal tab makes it very easy to track the temperature of the heatsink.
The exact factor of Vgs multiplication is usually a bit over 2 but depends on the ratio fo desired bias current in the outputs and current through the Vgs multiplier.
One particulairly interesting version of this output stage was biassed to 100mA, and it's driver also uses 100mA as the standing current, so this same current passes through the Vgs multiplier (the current sounds high but it's perfectly easily attainable since the driver is actually a MOSFET 'VAS' - higher current in it just makes it more linear and capable of driving the output stage and the only extra cost is the heat). As an experiment, an IRF640 was sued in the Vgs multiplier, since it is very close in specs as the outputs and it passes the same idle current. Predictably, the multiplication factor was near exactly 2, and the bias was actually very slightly overcompensated (falls from 100mA at 20C to about 97mA at 50C). This means the compensating MOSFET needs to work with a current closer to it's zero tempco than the outputs, in other words, it's better if it's a lower Idmax device compared to the outputs. Needless to say, in a situation where there are no source resistors, you want the most complementary devices you can choose.
The exact factor of Vgs multiplication is usually a bit over 2 but depends on the ratio fo desired bias current in the outputs and current through the Vgs multiplier.
One particulairly interesting version of this output stage was biassed to 100mA, and it's driver also uses 100mA as the standing current, so this same current passes through the Vgs multiplier (the current sounds high but it's perfectly easily attainable since the driver is actually a MOSFET 'VAS' - higher current in it just makes it more linear and capable of driving the output stage and the only extra cost is the heat). As an experiment, an IRF640 was sued in the Vgs multiplier, since it is very close in specs as the outputs and it passes the same idle current. Predictably, the multiplication factor was near exactly 2, and the bias was actually very slightly overcompensated (falls from 100mA at 20C to about 97mA at 50C). This means the compensating MOSFET needs to work with a current closer to it's zero tempco than the outputs, in other words, it's better if it's a lower Idmax device compared to the outputs. Needless to say, in a situation where there are no source resistors, you want the most complementary devices you can choose.
Hi
The amp I'm currently working on uses comp hexfets with HEC. I took a chance with the PCB and tried an experiment using SOT-23 RF transistors for the error amplifiers. With HEC, the error devices are part of the thermal transfer function so they must be within the thermal compensation loop. I mounted them on the PCB so they are directly under and in contact with the drain pin right next to the case of the respective output transistor. I was worried about how it would track the bias, but it turns out that it seems to be working. Right now, it has a slight negative coefficient, which is better than a positive. I left room so that the position could be adjusted, a few mm closer or further from the transistor, sort of a "thermal pot" so to speak. The idea is that since the drain is where the heat is generated, the pin temp would better represent the die temperature. The tracking appears to have little delay, which can sometimes happen with the thermal compensating device mounted on the heatsink. I find bias for these type of outputs needs to be at least 150mA, which by observing the error signal seems to be enough to keep Gm more constant wrt Id. Gfs = 22S for these devices.
The amp I'm currently working on uses comp hexfets with HEC. I took a chance with the PCB and tried an experiment using SOT-23 RF transistors for the error amplifiers. With HEC, the error devices are part of the thermal transfer function so they must be within the thermal compensation loop. I mounted them on the PCB so they are directly under and in contact with the drain pin right next to the case of the respective output transistor. I was worried about how it would track the bias, but it turns out that it seems to be working. Right now, it has a slight negative coefficient, which is better than a positive
. I left room so that the position could be adjusted, a few mm closer or further from the transistor, sort of a "thermal pot" so to speak. The idea is that since the drain is where the heat is generated, the pin temp would better represent the die temperature. The tracking appears to have little delay, which can sometimes happen with the thermal compensating device mounted on the heatsink. I find bias for these type of outputs needs to be at least 150mA, which by observing the error signal seems to be enough to keep Gm more constant wrt Id. Gfs = 22S for these devices.The simple answer to that is no
Whether it's possible depends on the circuit... the bias generator would need to provide the required voltage for the IRF's and as others have eluded, also be able to provide thermal tracking.
The laterals are far superior for a class b (ab) design at low bias current.
Why are you looking to replace with IRF's ? have the originals failed ?
Hi Mooly,
Thanks for the answer. Actually I am looking for good high power class AB MOSFET amplifier to drive my new magneplanar. I feel that my DIY amplifier (using a pair of Hitachi) doesn't deliver enough power. I prefer MOSFET's sound than BJT. Lateral MOSFETs are very expensive nowdays, but I don't want high power class A IRF's either, too much heat (and more expensive)..
. Lateral MOSFETs are very expensive nowdays, but I don't want high power class A IRF's either, too much heat (and more expensive)..
They doesn t cost too much if they live long enough...
I bought 6 pairs of Hitachi s back in 1988 , and they
still work flawlessly on the same amp..
IRFP240 and IRFP940.
This Mosfets are not expensive and used in some of the best amps in the world like Pass
or Audionet.
The Audionet Max2 is the least coloured amp i ever heard. It uses this Fets with enormous amouts of feedback but without sounding sterile,, so good ( say near perfect) sound with feedback is posible.
This Mosfets are not expensive and used in some of the best amps in the world like Pass
or Audionet.
The Audionet Max2 is the least coloured amp i ever heard. It uses this Fets with enormous amouts of feedback but without sounding sterile,, so good ( say near perfect) sound with feedback is posible.
I pay £1-50 for IRFP240 and £2-00 for IRFP9240.
Some of these FET's are meant for switching but work very well as linear devices.
Wise words wahab
This amp serves me well for 15 years. I think you're correct, those devices are not so expensive eventually.
The Rds(on) is only a secondary factor when operating the FETs in linear mode, as in the output stage of an audio amp. Of more concern is the gm - the Rds(on) can be an indicator of this, but its not a particularly good one. You're right that the two parts aren't particularly complimentary, but this is because the gms of the P and N channel devices aren't well matched.
The damping factor is also dependent on the gm, not the Rds(on) as the FETs in general won't be fully enhanced, its also strongly a function of the amount of feedback. The evils which aren't corrected by global feedback can be mitigated by local feedforward, as Bob Cordell showed in the Siliconix application note from decades ago.
The IRFP9240 is about 0.5 ohms when fully on.
Compared to a 4 ohm speaker this is pretty high and so gives a very poor damping factor. It also means there will be a few volts dropped across the IRFP9240 when driven hard. Add to this the gate voltage and the amp will be a long way from reaching a rail to rail output.
Damping factor and FET Rds(on)
The damping factor is not measured when the FET is fully on as its not a realistic condition - if its fully on your amp is clipping. To do a damping factor measurement with the FET hard on, the amp would have to be presented with a DC signal sufficient to drive it close enough to one or other of the supply rails. Most amps will have response roll off meaning that such a DC signal will potentially overload the input common mode voltage (assuming its servoed or has unity gain at DC), making the test invalid. If the amp is AC coupled, a DC input will fail to drive the output anywhere at all.
Yes, but this is another matter entirely. The lateral MOSFETs from Hitachi have a whopping great 1.7ohm Rds(on) but still get used in output stages.
The damping factor is not measured when the FET is fully on as its not a realistic condition - if its fully on your amp is clipping. To do a damping factor measurement with the FET hard on, the amp would have to be presented with a DC signal sufficient to drive it close enough to one or other of the supply rails. Most amps will have response roll off meaning that such a DC signal will potentially overload the input common mode voltage (assuming its servoed or has unity gain at DC), making the test invalid. If the amp is AC coupled, a DC input will fail to drive the output anywhere at all.
Yes, but this is another matter entirely. The lateral MOSFETs from Hitachi have a whopping great 1.7ohm Rds(on) but still get used in output stages.
Very rarely does the mosfet turn on fully saturated. But when it does, there has to be some limitation to the gate charge or else the fet will not turn off completely before the other turns on leading to destruction. Usually the driver stage will need to have a higher voltage in order to drive the fet to its full conductance with a resistive load, but it must be clamped at some level to prevent excess gate charge. The Rds on variable is really a worthless parameter, more important is the Gm vs Vgs relationship. Vgs for lateral fets is much higher wrt a relative Gm compared to vertical or hex type devices. But verticals tend to have a much greater Cfs. However the Ciss and Crss for verticals are much higher and this has to be taken into account when choosing the driver stage output Z. A speaker is not a resistor and is quite reactive. Therefore full output current does not necessarily happen at the same time as full output voltage. Damping has more to do than with just Gm, it also includes the fb loop that has to make up the change in Vgs wrt Gm. In addition the change in Vgs wrt Gm is far from linear and contains high order frequencies, which in turn require more current from the driver stage into the input capacitance. Not having enough bias to keep Gm constant with Id throughout the zero current crossover exacerbates the problem. Local feedback and/or feedforward can help take this burden away from the overall feedback loop, resulting in higher BW and better damping. The expense is more complexity.
Damping factor is the ratio of output impedance to load, which in turn is a function of the circuit topology/feedback etc... the internal resistance of the devices used as outputs is irrelevant.
Electrical damping of speakers is misunderstood too as the speaker is a mechanical device. An amp with a "high" damping factor is no better able to control a speaker than one of "low" damping within reason.
The only way to get reasonable linearity from the HEXFET's is to run them at high current... as in Nelsons class A designs... but to me while certainly effective and capable of very good results, it's a bit of a brute force approach.
Electrical damping of speakers is misunderstood too as the speaker is a mechanical device. An amp with a "high" damping factor is no better able to control a speaker than one of "low" damping within reason.
The only way to get reasonable linearity from the HEXFET's is to run them at high current... as in Nelsons class A designs... but to me while certainly effective and capable of very good results, it's a bit of a brute force approach.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.