Hexfets and lateral MOSFETs are two rather different animals. How do hexfets compare to regular vertical MOS?
Tom
Several amps of drain current. The rate of gate charge accumulation has a weird plateau versus applied vgs, making a nonlinear capacitance. Turn it “good and on” and that capacitance starts to linearize. Hence, the required high bias current to be well behaved as well as it’s better behavior when driven from low Z as opposed to high Z.
I think I mentioned in the original post that the pot on the LTP was a mistake. It doesn't do anything because of the current mirrors.
The 330ohm gate stoppers will be changed to 47ohm in the next rev. The slew rate before the output coil is about 10-15V/us. Not super fast, but still good enough for anything in the audible range and way beyond.
The "upside down" LTP just kind of happened as I was playing around with various ideas. That's the best explanation I have. 🙂 I tried reversing it in SPICE, but it didn't make much of a difference. I might revisit it, but the slew rate is at least OK the way it sits right now.
I've been lamenting the lack of LEDs too. That will be the first order of business!
Hexfets are a form of vertical MOS. All of them are better at being switches than amplifiers, although older types were not bad and they lent themselves very well to class A amps. Laterals were originally intended as linear amplifiers.
OK. That makes sense. But how does that relate to gate current and gate drivers? You will ONLY need >50mA (and I'm adding a bunch of margin due to non linearities), when you're driving dV/dt > 10^7v/s. Rod (or maybe it's Mitch) goes on about needing 350mA of gate current. I don't see the need for that unless you want a frequency response into the RF range.
Again, I might be getting this wrong, but there's a major disconnect between wanting to drive the Ids pretty hard (which you do using bias) and needing lots of gate current.
Body diodes are parasitic and you can’t get away from them. You are stuck with whatever it’s performance is. As free wheel diodes in a linear amp that only come on if you go into current limit nothing special is required. The one inside is good enough. Newer types have the diode “improved” so that adding a parallel fast switching type is not required in SMPS and class D. The parallel diode added externally might have a higher forward drop than th body diode - and that would involve extra complexity to “correct”. In situations where a 1N5406 would do (like this one) just use the body diode.
It’s not so much just running out of current, but making sure the load on the VAS doesn’t go nonlinear through the critical crossover region. That results in high frequency crossover distortion that your global NFB will fight to correct. The higher you go in frequency, the harder it’s job.
Driving the gate with low impedance pushes the pole to a very high frequency. It is not the current that matters but the impedance.
Ed
Or, it's like this joke: What happens when you put 10 economists in a room? You'll get 11 opinions.
Ok. That makes sense, especially for transients. But now we’re talking about non-linearities in the BJT-based parts, not the MOSFET. But maybe I’m splitting hairs.
That whole article is a bit of a mess, since they mix drain current and gate current freely and out of context. You helped clear that up!
I'm aware. As far as I can remember the hex structure was a way to avoid current crowding in the drain, but I could be wrong on that. I was just curious about the difference between the hexfets and plain vertical MOS (assuming such things exist) when it comes to input capacitance, that'l all.
Tom
geometry of the current flow.
well known laterals have lower gate capacitance compared to vertical mosfets
Hexfet lowered the typical capacitance of vertical but generalized are optimized for high power.
So no need to beat the dead horse of why the high power lateral market is pretty much non existent.
well known laterals have lower gate capacitance compared to vertical mosfets
Hexfet lowered the typical capacitance of vertical but generalized are optimized for high power.
So no need to beat the dead horse of why the high power lateral market is pretty much non existent.
The power MOSFETs of your circuit are obsolete.2. Interesting. I was told by someone that I should INCREASE the value of the source resistors. I’ll certainly experiment with that. Thanks!
Not for mosfets. One pair of mosfets does not need source resistors. Search this forum, this was discussed several times (E.g. by Ian Hegglun).
3. Dreadful how? It measures fine as far as I can tell and sounds good too. What aspect of the sound goes bad? THD? Frequency response? The only thing I may have noticed that the string section in large orchestras sound a little “metallic”, but that could be the room or the decent but not great Dayton towers I have it hooked up to at the moment.
I can't measure it or sim it. All I know, that when I tried to build hexfet amps without drivers, they sounded terrible.
If drivers were "optional", I'm sure commercial companies/designers would be skipping them to save money, but they don't.
You'll rarely see hexfet amp without drivers.
4. With the VBE multiplier in close thermal contact with the MOSFETs I have a negative temperature coefficient. If I listen at a high volume for a minute, the bias current is lower than before I cranked up the volume by about 25%.
I'm sure you have negative coefficient. The question is 'how accurate' ? Cordell has a whole chapter on Vbe multipliers in his book, including cases for hexfet outputs. Thermal tracking for BJTs and mosfets is not the same, they are different.
There is several ways to implement Vbe multiplier for them, and have as good as possible thermal tracking.
Single transistor will work, but it's not perfect. The easiest way to improve it, is to add a diode (some people use red LED) in the emitter of the tracking transistor.
Again - you can find plenty of hexfet amps showing this. e.g. here:
Amp from the post https://www.diyaudio.com/community/threads/unusual-amp-from-1987.357369/post-7068145
has been built (1 channel so far). Works like a champ. the only correction was to change C19 from 5pF to 9pF.
One channel built and tested so far. Will build 2nd channel, and then test with real music.
Idle current 40-50mA per output fet, output DC voltage (with C20 installed): 0.4 mV. Without out - 160mV.
All of the vertical structures have to deal with that nasty CV curve (plateau in the gate charge). Laterals less so, but it’s still there, and the capacitance is lower to begin with. The only thing that doesn’t are the glass FETs.
Todays “hexfets” are moving away from the hex structure, which helped minimize that Spirito effect. Now all that is back with a vengeance. Great. Just f*cking great.
Thanks for starting this thread. Your design goals for using basic parts is worthy and often exotics are not needed for great sound. It should be noted that there are many amp designs by member Apex Audio also follows this general theme of no fancy parts and good enough sound for most applications or uses. One in particular looks similar. One could replace the drivers with more readily available TTA004/TTC004 alternates. Here is FX14 design.
Somebody somewhere will be making IRF(P)240/9240. You might not like who is making them these days, but…..
27N25 N channel is still in production. No longer available as larger T03P package it is current production as T0-220
FQA36P15 P channel is in high demand its status has been reactivated.
To meet the volume demands manufacturing location has been relocated
High power T03P package match for P channel 36P15 is N channel 28N15
FQA28N15
FQA36P15
currently both in production, p channel lead time is 47 weeks from 9/10/24
is likely n channel production to be moved to p channel location.
Until then there is current stock of both in US states
Avnet Holds 60 units, MicroChip USA holds 9,975 units
If Digikey or Mouser does not hold stock, likely your order will process from there.
or directly from OnSemi also a option
Last edited:
Isn't that just the Miller effect? There is usually a gate-source voltage versus charge charge plot with a resistor from the drain to the supply. At first, the charge largely goes into the gate-source capacitance, so the gate-source voltage increases. Then the MOSFET starts to turn on, the drain voltage drops and the charge almost completely goes into the gate-drain capacitance, at almost constant gate-source voltage. Once the drain voltage is almost zero, there is no Miller effect anymore and the gate-source voltage increases again.
The drain-gate capacitance is non-linear, sometimes very non-linear, but you would also get a Miller plateau if it were linear.
You need to use fairly old-fashioned power MOSFETs to avoid Spirito instability/thermal second breakdown, but I think the MOSFETs in the schematic of post #1 are old-fashioned enough.