Re: There is no such thing as a free lunch
Hello Fred,
No, the vertical devices have roughly twice as much Cgd (the parameter that matters in a follower configuration) per unit of transconductance as a lateral device. Furthermore, the variation in Cgd as Vds changes is several times worse in the vertical devices.
This was discussed in another thread:
http://www.diyaudio.com/forums/showthread.php?postid=336264#post336264
Best regards,
Charles Hansen
Fred Dieckmann said:
Hello Fred,
No, the vertical devices have roughly twice as much Cgd (the parameter that matters in a follower configuration) per unit of transconductance as a lateral device. Furthermore, the variation in Cgd as Vds changes is several times worse in the vertical devices.
This was discussed in another thread:
http://www.diyaudio.com/forums/showthread.php?postid=336264#post336264
Best regards,
Charles Hansen
I thought as much
Thanks Mr. Hanson...... I missed that one. It confirmed my suspicions. It would seem the Hexfet (or whatever the other parties in litigation with IR call it) would be a more efficient geometry.
Now if my nose would just stop hurting without another Vicodine........Inka Dinka Doo.
Thanks Mr. Hanson...... I missed that one. It confirmed my suspicions. It would seem the Hexfet (or whatever the other parties in litigation with IR call it) would be a more efficient geometry.
Now if my nose would just stop hurting without another Vicodine........Inka Dinka Doo.
OK, but my point was the opposite. The vertical structure used for nearly all modern power MOSFETs is less efficient in those terms. The lateral devices are unmatched in that regard.
There are many variants of the vertical structure, such as HEXFET, TrenchFET, et cetera. ("HEXFET" is a trademark of International Rectifier.) I believe that the idea behind all of these variants is to produce the lowest "on" resistance (best for switching applications, which is 99.99% of the MOSFET market) with the lowest die size (cost).
Jon Marsh is an expert in these areas, and perhaps he could elaborate on this and/or correct me.
There are many variants of the vertical structure, such as HEXFET, TrenchFET, et cetera. ("HEXFET" is a trademark of International Rectifier.) I believe that the idea behind all of these variants is to produce the lowest "on" resistance (best for switching applications, which is 99.99% of the MOSFET market) with the lowest die size (cost).
Jon Marsh is an expert in these areas, and perhaps he could elaborate on this and/or correct me.
Oh my other left!........Thanks Charles
I see your point now. I am recovering from sinus surgery and am not very sharp right now...... I knew the transconductance was much higher for vertical mosfets than lateral and made the jump on the capacitance ratios without studying the curves. I wonder when we will see the first DIY aleph with 4 or 5 lateral mosfets for every Hexfet used in the original designs. I wonder who that is grinding their teeth I hear?
I see your point now. I am recovering from sinus surgery and am not very sharp right now...... I knew the transconductance was much higher for vertical mosfets than lateral and made the jump on the capacitance ratios without studying the curves. I wonder when we will see the first DIY aleph with 4 or 5 lateral mosfets for every Hexfet used in the original designs. I wonder who that is grinding their teeth I hear?
It seems we've argued this out before. I quote myself:
Let's take an example of a 2SK2220, a modern lateral part, and
it's vertical counterpart, and IRF230.
Both are rated about the same voltage, current, and wattage,
but the lateral's transconductance is about 1 and the vertical's
is about 5.
Ciss capacitance of both parts is the same at 600 pF, but the
Cdss of the lateral is about 3 times higher. The Crss of the
lateral is 1/10 the vertical (score a point there).
The turn-on turn-off times are quite different, with the vertical
being 30 and 50 ns and the lateral being 240 and 90 ns.
So depending on the application, paralleling several lateral
Mosfets to get the transconductance figure of a comparable
vertical device may give you trouble with the capacitance
numbers that pile up.
:Endquote
The only place where you can point to a better capacitance
figure is the Crss, but in the example above, when you
parallel 5 of the 2SK2220, your Ciss is about 5 times higher
and and your Cdss is about 15 times higher.
Let's take an example of a 2SK2220, a modern lateral part, and
it's vertical counterpart, and IRF230.
Both are rated about the same voltage, current, and wattage,
but the lateral's transconductance is about 1 and the vertical's
is about 5.
Ciss capacitance of both parts is the same at 600 pF, but the
Cdss of the lateral is about 3 times higher. The Crss of the
lateral is 1/10 the vertical (score a point there).
The turn-on turn-off times are quite different, with the vertical
being 30 and 50 ns and the lateral being 240 and 90 ns.
So depending on the application, paralleling several lateral
Mosfets to get the transconductance figure of a comparable
vertical device may give you trouble with the capacitance
numbers that pile up.
:Endquote
The only place where you can point to a better capacitance
figure is the Crss, but in the example above, when you
parallel 5 of the 2SK2220, your Ciss is about 5 times higher
and and your Cdss is about 15 times higher.
Nelson is right to qualify my numbers. I was referring to Cgd, which applies in a follower application. I've never looked at a vertical versus lateral comparison in a common source connection as was proposed in this thread.
One important point to keep in mind is that the capacitances quoted in the data sheets are generally under conditions that don't apply in an audio circuit. For example, the Hitachi lateral devices are spec'ed at 10 Vds (not so far off), but at 5 Vgs, which means that they are turned on virtually all the way. On the other hand, the vertical parts are spec'ed at 0 Vgs, which is simply useless for audio work. This has no relationship whatsoever to how the part will be used in an audio amplifier.
The problem is to get accurate information that is representative of the conditions under which the parts will actually be used. I, for one, don't have this type of equipment at hand. However, I do have a few rare reference books that discuss the actual behavior of both lateral and vertical parts under conditions that more closely match those found in an audio amp. This is what led me to try the lateral devices. I found the results with lateral devices to be superior (when used in a follower configuration). Your mileage may vary.
One important point to keep in mind is that the capacitances quoted in the data sheets are generally under conditions that don't apply in an audio circuit. For example, the Hitachi lateral devices are spec'ed at 10 Vds (not so far off), but at 5 Vgs, which means that they are turned on virtually all the way. On the other hand, the vertical parts are spec'ed at 0 Vgs, which is simply useless for audio work. This has no relationship whatsoever to how the part will be used in an audio amplifier.
The problem is to get accurate information that is representative of the conditions under which the parts will actually be used. I, for one, don't have this type of equipment at hand. However, I do have a few rare reference books that discuss the actual behavior of both lateral and vertical parts under conditions that more closely match those found in an audio amp. This is what led me to try the lateral devices. I found the results with lateral devices to be superior (when used in a follower configuration). Your mileage may vary.
Fred Dieckmann said:
This will require the presumable quite rare unit Siemens per
Farad, so maybe we should coin a new name for that unit.
Maybe it should be called a Pass or a Hansen, since they brought
it up first in the other thread, but I think Dieckman is a much
better name for a unit, so maybe we should define:
1 Dieckmann = 1 S/F
I am to lazy to calculate typical values now, but I guess we
would typically end up in the range of kiloDieckmanns to
gigaDieckmanns.
Fred, sorry to dissapoint you, but be prepared your nose will
hurt for quite a while. I was on strong painkillers for at least
two weeks after my nose surgery. I thought I'd be able to
start working in a few days or a week, but my doctor said,
let's start with a two week sick-leave and then we'll see. He
knew what he was talking about.
I vote for Dieckmanns. This would be expressed in:
Amps per Farad per Volt. I / C / V
And as I recall, a Farad is 1 Volt per second per amp.
Since a Farad is 1V per Second per Amp V/S/I
Then a Dieckmann might logically be I^2 Sec / V^2
In any case the 2SK2220 would be 1 Siemen / 8 pF,
which would be 1.25 * 10 ^11 Dieckmanns.
Personally, I think that's an astronomical number of
Dieckmanns for that much Siemen
Amps per Farad per Volt. I / C / V
And as I recall, a Farad is 1 Volt per second per amp.
Since a Farad is 1V per Second per Amp V/S/I
Then a Dieckmann might logically be I^2 Sec / V^2
In any case the 2SK2220 would be 1 Siemen / 8 pF,
which would be 1.25 * 10 ^11 Dieckmanns.
Personally, I think that's an astronomical number of
Dieckmanns for that much Siemen
Jam:
We definately want to keep Fred on his toes and off his nose!
This thread (like a few others lately), has been great! I wish all this was around years ago when I had more brain cells to soak it all up.
I've had surgery too (though nothing at all as severe as others): We're moving, so ALL my "fun" stuff (notes, papers, databooks, parts, scopes, THE GEAR!, CDs and albums, etc.) has been in storage for months. If it wasn't for you guys, I'd be climbing the walls!
I guess I'll go ahead and admit my fondness for the fully complementary - symmetric designs. I also like jfets at the front. I was always annoyed that I'd build them and measure the gains of the top and bottom jfet diff amps and find they're not the same. I'm sure that does wonderful things to the distortion spectra . Too bad DIY doesn't include doing your own semiconductors - the ultimate in matched devices. Heh, heh ...
That being said, you can probably imagine now that I like the circuit in your post #85. A simple platform for some experiments... like adding small resistors to the RHS drains of the jfets and tying the sources of the output mosfets to the respective drains of the jfets instead of the rails and see what sounds different/better/etc.
Sometimes, in the intensisty of all the heated discussions, we forget this is supposed to be fun.
mlloyd1
(speedy recovery, fred!)
We definately want to keep Fred on his toes and off his nose!
This thread (like a few others lately), has been great! I wish all this was around years ago when I had more brain cells to soak it all up.
I've had surgery too (though nothing at all as severe as others): We're moving, so ALL my "fun" stuff (notes, papers, databooks, parts, scopes, THE GEAR!, CDs and albums, etc.) has been in storage for months. If it wasn't for you guys, I'd be climbing the walls!
I guess I'll go ahead and admit my fondness for the fully complementary - symmetric designs. I also like jfets at the front. I was always annoyed that I'd build them and measure the gains of the top and bottom jfet diff amps and find they're not the same
. I'm sure that does wonderful things to the distortion spectra . Too bad DIY doesn't include doing your own semiconductors - the ultimate in matched devices. Heh, heh ...That being said, you can probably imagine now that I like the circuit in your post #85. A simple platform for some experiments... like adding small resistors to the RHS drains of the jfets and tying the sources of the output mosfets to the respective drains of the jfets instead of the rails and see what sounds different/better/etc.
Sometimes, in the intensisty of all the heated discussions, we forget this is supposed to be fun.
mlloyd1
(speedy recovery, fred!)
jam said:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.