AoE x-Chapters, using power MOSFETs as linear transistors

On 7/25/19 1:10 PM, bitrex wrote:
On 7/24/19 6:20 PM, John Larkin wrote:
On Wed, 24 Jul 2019 18:01:51 -0400, bitrex <user@example.net> wrote:

On 7/24/19 10:59 AM, Winfield Hill wrote:
DRAFT copy of a useful section in the upcoming
AoE x-Chapters, about using power MOSFETs as
linear transistors.  Subthreshold region, etc.
Comments, errors found, etc.  Thanks!

https://www.dropbox.com/s/0ph15moamo0mlt4/3x.5_MOSFETs_Linear-Transistors_DRAFT.pdf?dl=1




Do you have anything in the X or AoE III (sorry, don't have a copy yet,
just my well-worn AoE II) on low-voltage current sources? JFETs in
linear region, depletion MOSFETs, etc?

This old chestnut works okay in some situations down to ~2 volts it kind
of sucks the lemon wrt tempco and output impedance, though:

https://tinyurl.com/yxn7mej4


For low currents, this works pretty well:

https://www.dropbox.com/s/d4ntmq7fdzah69a/LED_Isrc_data.JPG?raw=1

The tempcos track nicely, as long as the transistor doesn't self-heat
much.

There are a bunch of IC current limiters around now.





I have a client who wants low-voltage oscillator (~2 volts) with very
low Q "meander" inductors, printed on a piece of flexible plastic film
or something. The inductance varies as the film is compressed or
deformed and hence the oscillator frequency.

Not an insurmountable problem but I'm a bit stumped for a temperature
stable current source down there.

If I'm lucky maybe some TinyLogic schmitts are in the budget
 
On Thu, 25 Jul 2019 13:10:45 -0400, bitrex <user@example.net> wrote:

On 7/24/19 6:20 PM, John Larkin wrote:
On Wed, 24 Jul 2019 18:01:51 -0400, bitrex <user@example.net> wrote:

On 7/24/19 10:59 AM, Winfield Hill wrote:
DRAFT copy of a useful section in the upcoming
AoE x-Chapters, about using power MOSFETs as
linear transistors. Subthreshold region, etc.
Comments, errors found, etc. Thanks!

https://www.dropbox.com/s/0ph15moamo0mlt4/3x.5_MOSFETs_Linear-Transistors_DRAFT.pdf?dl=1



Do you have anything in the X or AoE III (sorry, don't have a copy yet,
just my well-worn AoE II) on low-voltage current sources? JFETs in
linear region, depletion MOSFETs, etc?

This old chestnut works okay in some situations down to ~2 volts it kind
of sucks the lemon wrt tempco and output impedance, though:

https://tinyurl.com/yxn7mej4


For low currents, this works pretty well:

https://www.dropbox.com/s/d4ntmq7fdzah69a/LED_Isrc_data.JPG?raw=1

The tempcos track nicely, as long as the transistor doesn't self-heat
much.

There are a bunch of IC current limiters around now.





I have a client who wants low-voltage oscillator (~2 volts) with very
low Q "meander" inductors, printed on a piece of flexible plastic film
or something. The inductance varies as the film is compressed or
deformed and hence the oscillator frequency.

Not an insurmountable problem but I'm a bit stumped for a temperature
stable current source down there.

Why does it need a current source?


--

John Larkin Highland Technology, Inc
picosecond timing precision measurement

jlarkin att highlandtechnology dott com
http://www.highlandtechnology.com
 
On 7/25/19 4:24 PM, bitrex wrote:
On 7/25/19 3:17 PM, John Larkin wrote:

I have a client who wants low-voltage oscillator (~2 volts) with very
low Q "meander" inductors, printed on a piece of flexible plastic film
or something. The inductance varies as the film is compressed or
deformed and hence the oscillator frequency.

Not an insurmountable problem but I'm a bit stumped for a temperature
stable current source down there.

Why does it need a current source?



They don't wanna pay for fast, low voltage op amps or logic and such.
The circuit is gonna be in a semi-disposable product so cost is not
no-object :(

The idea i'm messing with is bring the frequency down to something more
sensible from the self-resonant of a couple hundred MHz with some
low-loss inductance in series. Then use some negative resistance to
compensate for the bad Q down there. Can all be done with FETs and BJTs,
at low voltages.

I'm not under any kind of NDA so I'll post my result if I don't fail and
am sent to work camp in forest.
 
On 7/25/19 3:17 PM, John Larkin wrote:

I have a client who wants low-voltage oscillator (~2 volts) with very
low Q "meander" inductors, printed on a piece of flexible plastic film
or something. The inductance varies as the film is compressed or
deformed and hence the oscillator frequency.

Not an insurmountable problem but I'm a bit stumped for a temperature
stable current source down there.

Why does it need a current source?

They don't wanna pay for fast, low voltage op amps or logic and such.
The circuit is gonna be in a semi-disposable product so cost is not
no-object :(
 
On Wednesday, July 24, 2019 at 5:54:15 PM UTC-7, Winfield Hill wrote:
> No, that's not an easy subject. Source resistors...

Hey, if you guys don't cover it, who will? If it were easy, nobody
would look to you for hints. :)

Bob Cordell brings up a few points in his book on audio power
amps that you could repeat, or perhaps reference in the bibliography.
E.g., source resistors alone may not be enough to make up for poorly-
matched FETs, but FETs from the same manufacturing lot are usually
easy to match. Lateral MOS behaves (very) differently than VMOS
due to its tempco, but lateral parts are expensive and hard to come by.
Gate resistors are more important than usual when paralleling FETs due
to their tendency to oscillate, but there goes your phase response.

He suggests Zobel networks between the gate and drain terminals as an
alternative. He also references a note on stability of paralleled
FETs by Ed Oxner that I can't find.

-- john, KE5FX
 
John Larkin wrote:

Back when we were in the NMR gradient driver business, we ran
switcher-type mosfets as constant-current linear amps, with PPM levels
of settling and noise; one box had 20 KW peak output. We just wrapped
an opamp around each fet to hide all the sub-threshold complexities.

What we couldn't hide was blowing up the fets in the far corners of
the SOAR graph.

** However, Larkin failed to do the *bleeding obvious thing* and use MOSFETs characterised for liner operation - ie Lateral types as made by Hitachi and others.

Another example of his massive and insane prejudice against any and all things that even smack of "audio".

Wot a cretin.



..... Phil
 
On 7/25/19 7:55 PM, Phil Allison wrote:
John Larkin wrote:



Back when we were in the NMR gradient driver business, we ran
switcher-type mosfets as constant-current linear amps, with PPM levels
of settling and noise; one box had 20 KW peak output. We just wrapped
an opamp around each fet to hide all the sub-threshold complexities.

What we couldn't hide was blowing up the fets in the far corners of
the SOAR graph.


** However, Larkin failed to do the *bleeding obvious thing* and use MOSFETs characterised for liner operation - ie Lateral types as made by Hitachi and others.

Another example of his massive and insane prejudice against any and all things that even smack of "audio".

Wot a cretin.

Imagine anybody wanting to go faster than 20 kHz. The very _idea_!

Cheers

Phil Hobbs
 
bitrex wrote...
The idea i'm messing with is bring the frequency down to something
more sensible from the self-resonant of a couple hundred MHz with
some low-loss inductance in series. Then use some negative resistance
to compensate for the bad Q down there. Can all be done with FETs
and BJTs, at low voltages.

You might want to experiment with piglet (Eric Wagner)'s idea
of adapting a foldback current-limit circuit into a negative-
resistance current source. We wrote it up for section 2x.12,
in the AoE x-Chapters book, see draft copy on DropBox.

https://www.dropbox.com/s/2urywtwwlpt6sjb/2x.12_negative-resistance_WH.pdf?dl=1

I played with SPICE and added a 100uA 1.5-volt low-voltage
version. It might be fast enough, if you scale the resistors
to run it at 2mA or even 10mA. You might do a pnp version.


--
Thanks,
- Win
 
bitrex wrote...
Winfield Hill wrote:
bitrex wrote...

The idea i'm messing with is bring the frequency down to something
more sensible from the self-resonant of a couple hundred MHz with
some low-loss inductance in series. Then use some negative
resistance to compensate for the bad Q down there. Can all be
done with FETs and BJTs, at low voltages.

You might want to experiment with piglet's idea ...

Hi Win, thank you, very nice!

If I'm allowed a more luxurious ~2.5 volts I came up with this
modified Lambda diode circuit.

What oscillation frequency do you envision? From 200 MHz down
to low MHz? Will your circuit work up that high?, I see 10k
resistors, etc.?

BTW, your added inductor esr doesn't show on the schematic
(it's better to add parts to show these things explicitly).


--
Thanks,
- Win
 
On Friday, July 26, 2019 at 11:30:05 AM UTC-4, Winfield Hill wrote:
bitrex wrote...

The idea i'm messing with is bring the frequency down to something
more sensible from the self-resonant of a couple hundred MHz with
some low-loss inductance in series. Then use some negative resistance
to compensate for the bad Q down there. Can all be done with FETs
and BJTs, at low voltages.

You might want to experiment with piglet (Eric Wagner)'s idea
of adapting a foldback current-limit circuit into a negative-
resistance current source. We wrote it up for section 2x.12,
in the AoE x-Chapters book, see draft copy on DropBox.

https://www.dropbox.com/s/2urywtwwlpt6sjb/2x.12_negative-resistance_WH.pdf?dl=1

Instead of "With a little prodding" you might substitute,
"For a few haycorns". :^)

GH
I played with SPICE and added a 100uA 1.5-volt low-voltage
version. It might be fast enough, if you scale the resistors
to run it at 2mA or even 10mA. You might do a pnp version.


--
Thanks,
- Win
 
On 7/26/19 11:29 AM, Winfield Hill wrote:
bitrex wrote...

The idea i'm messing with is bring the frequency down to something
more sensible from the self-resonant of a couple hundred MHz with
some low-loss inductance in series. Then use some negative resistance
to compensate for the bad Q down there. Can all be done with FETs
and BJTs, at low voltages.

You might want to experiment with piglet (Eric Wagner)'s idea
of adapting a foldback current-limit circuit into a negative-
resistance current source. We wrote it up for section 2x.12,
in the AoE x-Chapters book, see draft copy on DropBox.

https://www.dropbox.com/s/2urywtwwlpt6sjb/2x.12_negative-resistance_WH.pdf?dl=1

I played with SPICE and added a 100uA 1.5-volt low-voltage
version. It might be fast enough, if you scale the resistors
to run it at 2mA or even 10mA. You might do a pnp version.

Hi Win, thank you, very nice!

If I'm allowed a more luxurious ~2.5 volts I came up with this modified
Lambda diode circuit. It uses a JFET and pnp for the negative
resistance, plus current mirror for biasing up the diode (it won't start
into a lossy inductor on its own) and some Colpitts-like feedback.

The 10u has 35 ohms of series resistance to represent a somewhat crappy L.

I am concerned about Vgs spread on real JFETs, though, and whether a
circuit of this topology will actually oscillate in the real world
consistently. It can be fiddly to start in LTSpice. An AC 1 current
source into the node shows a current gain of about 20dB into the inductor


Version 4
SHEET 1 2432 1620
WIRE 432 -320 432 -432
WIRE 432 -320 224 -320
WIRE 640 -320 432 -320
WIRE 224 -256 224 -320
WIRE 176 -192 -144 -192
WIRE 640 -160 640 -320
WIRE 224 32 224 -160
WIRE 848 32 224 32
WIRE 1280 32 912 32
WIRE 224 144 224 32
WIRE 640 192 640 -80
WIRE 640 192 288 192
WIRE 640 288 640 192
WIRE -144 416 -144 -192
WIRE 224 416 224 240
WIRE 224 416 -144 416
WIRE 640 416 640 368
WIRE 640 416 224 416
WIRE 848 416 640 416
WIRE 1024 416 848 416
WIRE 1024 464 1024 416
WIRE 640 480 640 416
WIRE -352 528 -352 448
WIRE -144 544 -144 416
WIRE 848 560 848 416
WIRE 1024 592 1024 528
WIRE 1280 592 1280 32
WIRE 1280 592 1024 592
WIRE 640 640 640 560
WIRE 640 640 432 640
WIRE 1024 656 1024 592
WIRE -352 704 -352 608
WIRE -144 704 -144 624
WIRE 224 736 224 416
WIRE 640 736 640 640
WIRE 432 784 432 640
WIRE 432 784 288 784
WIRE 576 784 432 784
WIRE 848 800 848 640
WIRE 1024 800 1024 720
WIRE 640 896 640 832
WIRE 224 912 224 832
WIRE 224 1056 224 992
WIRE 640 1056 640 976
FLAG -352 704 0
FLAG -352 448 Vcc
FLAG 432 -432 Vcc
FLAG 224 1056 0
FLAG 640 1056 0
FLAG 1024 800 0
FLAG 848 800 0
FLAG -144 704 0
SYMBOL pnp 288 240 R180
SYMATTR InstName Q1
SYMATTR Value BC557B
SYMBOL res 624 -176 R0
SYMATTR InstName R1
SYMATTR Value 10k
SYMBOL res 624 272 R0
SYMATTR InstName R2
SYMATTR Value 2.2k
SYMBOL voltage -352 512 R0
WINDOW 3 24 44 Left 2
WINDOW 123 0 0 Left 0
WINDOW 39 0 0 Left 0
SYMATTR Value 2.5
SYMATTR InstName V1
SYMBOL njf 176 -256 R0
SYMATTR InstName J1
SYMATTR Value J113
SYMBOL npn 288 736 M0
SYMATTR InstName Q3
SYMATTR Value BC546B
SYMBOL res 208 896 R0
SYMATTR InstName R7
SYMATTR Value 50
SYMBOL npn 576 736 R0
WINDOW 3 66 66 Left 2
SYMATTR InstName Q4
SYMATTR Value BC546B
SYMBOL res 624 464 R0
SYMATTR InstName R8
SYMATTR Value 10k
SYMBOL ind 832 544 R0
SYMATTR InstName L2
SYMATTR Value 10Âľ
SYMATTR SpiceLine Rser=35
SYMBOL cap 1008 464 R0
SYMATTR InstName C1
SYMATTR Value 470p
SYMBOL cap 1008 656 R0
SYMATTR InstName C2
SYMATTR Value 470p
SYMBOL cap 912 16 R90
WINDOW 0 0 32 VBottom 2
WINDOW 3 32 32 VTop 2
SYMATTR InstName C3
SYMATTR Value 47p
SYMBOL res 624 880 R0
SYMATTR InstName R6
SYMATTR Value 10k
SYMBOL current -144 544 R0
WINDOW 123 24 108 Left 2
WINDOW 39 0 0 Left 0
SYMATTR InstName I1
SYMATTR Value ""
SYMATTR Value2 AC 1
TEXT -472 768 Left 2 !.ac oct 10 100k 10Meg
 
On Thu, 25 Jul 2019 16:55:14 -0700 (PDT), Phil Allison
<pallison49@gmail.com> wrote:

John Larkin wrote:



Back when we were in the NMR gradient driver business, we ran
switcher-type mosfets as constant-current linear amps, with PPM levels
of settling and noise; one box had 20 KW peak output. We just wrapped
an opamp around each fet to hide all the sub-threshold complexities.

What we couldn't hide was blowing up the fets in the far corners of
the SOAR graph.


** However, Larkin failed to do the *bleeding obvious thing* and use MOSFETs characterised for liner operation - ie Lateral types as made by Hitachi and others.

Another example of his massive and insane prejudice against any and all things that even smack of "audio".

Wot a cretin.

This cretin designs and sells a lot of electronics, and has fun doing
it. I guess I'm dumb and happy.

Can you suggest some part numbers for high power n and p channel
lateral fets? Something in the 200 volt, 400 watt ballpark?

I'm not fundamentally prejudiced against audio, but most audio design
is really bad and mostly subjective. And boring.

We're using a very nice TI class D "audio" amp in a couple of designs
now.




--

John Larkin Highland Technology, Inc
picosecond timing precision measurement

jlarkin att highlandtechnology dott com
http://www.highlandtechnology.com
 
In article <lnemjeh2pckf2thlfsjcv419rdnp88mfhn@4ax.com>,
John Larkin <jjlarkin@highland_snip_technology.com> wrote:

Can you suggest some part numbers for high power n and p channel
lateral fets? Something in the 200 volt, 400 watt ballpark?

Exicon has the ECW20N20/ECW20P20, which are 200-volt 250-watt parts in
TO-264. ECF20N20/ECF20P20 seem to be the same dice, in TO-3. These
types look as if they're two 125-watt dice in a single package.

Their ProFusion arm sells 'em directly.

As far as I can tell, Exicon is the only company making the
lateral-MOSFET dice these days; there may be multiple companies
packaging them.
 
On 7/26/19 3:30 PM, bitrex wrote:
On 7/26/19 2:32 PM, Winfield Hill wrote:
bitrex wrote...

There was an error with that .asc here is the corrected version:

Version 4
SHEET 1 2432 1748
WIRE 432 -320 432 -432
WIRE 432 -320 224 -320
WIRE 640 -320 432 -320
WIRE 224 -256 224 -320
WIRE -288 -192 -288 -272
WIRE 176 -192 16 -192
WIRE 640 -160 640 -320
WIRE -288 -16 -288 -112
WIRE 224 0 224 -160
WIRE 1184 0 224 0
WIRE 1664 0 1248 0
WIRE 224 112 224 0
WIRE 640 160 640 -80
WIRE 640 160 288 160
WIRE 640 288 640 160
WIRE -176 416 -288 416
WIRE 16 416 16 -192
WIRE 16 416 -96 416
WIRE 224 416 224 208
WIRE 224 416 16 416
WIRE 640 416 640 368
WIRE 640 416 224 416
WIRE 960 416 640 416
WIRE 1200 416 960 416
WIRE 1408 416 1200 416
WIRE 1472 416 1408 416
WIRE 1408 464 1408 416
WIRE -288 528 -288 416
WIRE 640 560 640 416
WIRE 960 560 960 416
WIRE 1408 592 1408 528
WIRE 1664 592 1664 0
WIRE 1664 592 1408 592
WIRE 1408 656 1408 592
WIRE 1200 672 1200 416
WIRE 640 688 640 640
WIRE 640 688 432 688
WIRE -288 704 -288 608
WIRE 960 720 960 640
WIRE 960 720 816 720
WIRE 1136 720 960 720
WIRE 224 752 224 416
WIRE 640 752 640 688
WIRE 432 800 432 688
WIRE 432 800 288 800
WIRE 576 800 432 800
WIRE 1408 800 1408 720
WIRE 816 832 816 720
WIRE 960 960 960 720
WIRE 1024 960 960 960
WIRE 1200 960 1200 768
WIRE 1200 960 1104 960
WIRE 640 992 640 848
WIRE 224 1008 224 848
WIRE 1200 1024 1200 960
WIRE 224 1152 224 1088
WIRE 640 1152 640 1072
WIRE 816 1200 816 896
WIRE 1200 1200 1200 1104
WIRE 1200 1200 816 1200
WIRE 1200 1328 1200 1200
WIRE 1344 1328 1200 1328
WIRE 1488 1328 1344 1328
WIRE 1200 1392 1200 1328
WIRE 1344 1456 1344 1328
WIRE 1488 1456 1488 1328
WIRE 1200 1536 1200 1472
WIRE 1200 1648 1200 1616
WIRE 1344 1648 1344 1536
WIRE 1344 1648 1200 1648
WIRE 1488 1648 1488 1520
WIRE 1488 1648 1344 1648
WIRE 1200 1728 1200 1648
FLAG -288 -16 0
FLAG -288 -272 Vcc
FLAG 432 -432 Vcc
FLAG 224 1152 0
FLAG 640 1152 0
FLAG 1408 800 0
FLAG 1200 1728 0
FLAG -288 704 0
FLAG 1472 416 Out
IOPIN 1472 416 Out
SYMBOL pnp 288 208 R180
SYMATTR InstName Q1
SYMATTR Value BC557B
SYMBOL res 624 -176 R0
SYMATTR InstName R1
SYMATTR Value 10k
SYMBOL res 624 272 R0
SYMATTR InstName R2
SYMATTR Value 2.2k
SYMBOL voltage -288 -208 R0
WINDOW 3 24 44 Left 2
WINDOW 123 0 0 Left 0
WINDOW 39 0 0 Left 0
SYMATTR Value 2.5
SYMATTR InstName V1
SYMBOL njf 176 -256 R0
SYMATTR InstName J1
SYMATTR Value J113
SYMBOL npn 288 752 M0
SYMATTR InstName Q3
SYMATTR Value BC546B
SYMBOL res 208 992 R0
SYMATTR InstName R7
SYMATTR Value 50
SYMBOL npn 576 752 R0
WINDOW 3 66 66 Left 2
SYMATTR Value BC546B
SYMATTR InstName Q4
SYMBOL cap 1392 464 R0
SYMATTR InstName C1
SYMATTR Value 470p
SYMBOL cap 1392 656 R0
SYMATTR InstName C2
SYMATTR Value 470p
SYMBOL cap 1248 -16 R90
WINDOW 0 0 32 VBottom 2
WINDOW 3 58 36 VTop 2
SYMATTR InstName C3
SYMATTR Value 47p
SYMBOL res 624 976 R0
SYMATTR InstName R6
SYMATTR Value 10k
SYMBOL npn 1136 672 R0
WINDOW 3 66 66 Left 2
SYMATTR Value BC546B
SYMATTR InstName Q2
SYMBOL res 1184 1008 R0
SYMATTR InstName R3
SYMATTR Value 47
SYMBOL res 944 544 R0
SYMATTR InstName R4
SYMATTR Value 2.2k
SYMBOL res 1120 944 R90
WINDOW 0 0 56 VBottom 2
WINDOW 3 32 56 VTop 2
SYMATTR InstName R5
SYMATTR Value 68k
SYMBOL cap 800 832 R0
SYMATTR InstName C4
SYMATTR Value 470p
SYMBOL ind 1184 1376 R0
SYMATTR InstName L1
SYMATTR Value 4Âľ
SYMATTR SpiceLine Rser=0 Cpar=0
SYMBOL res 624 544 R0
SYMATTR InstName R8
SYMATTR Value 10k
SYMBOL voltage -288 512 R0
WINDOW 123 0 0 Left 0
WINDOW 39 0 0 Left 0
SYMATTR InstName V2
SYMATTR Value SINE(0 1 1Meg 0 0 0 1000)
SYMBOL res -80 400 R90
WINDOW 0 0 56 VBottom 2
WINDOW 3 32 56 VTop 2
SYMATTR InstName R9
SYMATTR Value 10k
SYMBOL res 1184 1520 R0
SYMATTR InstName R10
SYMATTR Value 50
SYMBOL res 1328 1440 R0
SYMATTR InstName R11
SYMATTR Value 100k
SYMBOL cap 1472 1456 R0
SYMATTR InstName C5
SYMATTR Value 10p
TEXT -352 48 Left 2 !.tran 100m
TEXT 1592 1488 Left 5 ;Lossy L
TEXT -632 296 Left 5 ;Kickstarter (sim only)
 
On 7/26/19 2:32 PM, Winfield Hill wrote:
bitrex wrote...

Winfield Hill wrote:
bitrex wrote...

The idea i'm messing with is bring the frequency down to something
more sensible from the self-resonant of a couple hundred MHz with
some low-loss inductance in series. Then use some negative
resistance to compensate for the bad Q down there. Can all be
done with FETs and BJTs, at low voltages.

You might want to experiment with piglet's idea ...

Hi Win, thank you, very nice!

If I'm allowed a more luxurious ~2.5 volts I came up with this
modified Lambda diode circuit.

What oscillation frequency do you envision? From 200 MHz down
to low MHz? Will your circuit work up that high?, I see 10k
resistors, etc.?

BTW, your added inductor esr doesn't show on the schematic
(it's better to add parts to show these things explicitly).

Here is a time domain of the fleshed-out idea oscillating OK in the sim.
A few hundred cycles of the resonant frequency injected into the tank
gets it moving, in the simulation.

There's a gyrator in series with the lossy-L to bring the "self-resonant
frequency" down.

The desired result is that when the inductance of the lossy-L is changed
(by compressing or stretching the physical coil) the frequency deviates
a bit.

That's about it.

Anyway it looks good here but I'm concerned if it will be annoying to
get operating reliably on a breadboard in particular the part-to-part
tolerance on the JFET, any suggestions would be appreciated on that front

Version 4
SHEET 1 2432 1748
WIRE 432 -320 432 -432
WIRE 432 -320 224 -320
WIRE 640 -320 432 -320
WIRE 224 -256 224 -320
WIRE -304 -208 -304 -288
WIRE 176 -192 16 -192
WIRE 640 -160 640 -320
WIRE -304 -32 -304 -128
WIRE 224 0 224 -160
WIRE 1184 0 224 0
WIRE 1664 0 1248 0
WIRE 224 112 224 0
WIRE 640 160 640 -80
WIRE 640 160 288 160
WIRE 640 288 640 160
WIRE -176 416 -288 416
WIRE 16 416 16 -192
WIRE 16 416 -96 416
WIRE 224 416 224 208
WIRE 224 416 16 416
WIRE 640 416 640 368
WIRE 640 416 224 416
WIRE 960 416 640 416
WIRE 1200 416 960 416
WIRE 1408 416 1200 416
WIRE 1472 416 1408 416
WIRE 1408 464 1408 416
WIRE -288 528 -288 416
WIRE 640 560 640 416
WIRE 960 560 960 416
WIRE 1408 592 1408 528
WIRE 1664 592 1664 0
WIRE 1664 592 1408 592
WIRE 1408 656 1408 592
WIRE 1200 672 1200 416
WIRE -288 704 -288 608
WIRE 960 720 960 640
WIRE 960 720 880 720
WIRE 1136 720 960 720
WIRE 224 752 224 416
WIRE 640 752 640 640
WIRE 576 800 288 800
WIRE 1408 800 1408 720
WIRE 880 832 880 720
WIRE 960 960 960 720
WIRE 1024 960 960 960
WIRE 1200 960 1200 768
WIRE 1200 960 1104 960
WIRE 640 992 640 848
WIRE 224 1008 224 848
WIRE 1200 1024 1200 960
WIRE 224 1152 224 1088
WIRE 640 1152 640 1072
WIRE 880 1200 880 896
WIRE 1200 1200 1200 1104
WIRE 1200 1200 880 1200
WIRE 1200 1328 1200 1200
WIRE 1344 1328 1200 1328
WIRE 1488 1328 1344 1328
WIRE 1200 1392 1200 1328
WIRE 1344 1456 1344 1328
WIRE 1488 1456 1488 1328
WIRE 1200 1536 1200 1472
WIRE 1200 1648 1200 1616
WIRE 1344 1648 1344 1536
WIRE 1344 1648 1200 1648
WIRE 1488 1648 1488 1520
WIRE 1488 1648 1344 1648
WIRE 1200 1728 1200 1648
FLAG -304 -32 0
FLAG -304 -288 Vcc
FLAG 432 -432 Vcc
FLAG 224 1152 0
FLAG 640 1152 0
FLAG 1408 800 0
FLAG 1200 1728 0
FLAG -288 704 0
FLAG 1472 416 Out
IOPIN 1472 416 Out
SYMBOL pnp 288 208 R180
SYMATTR InstName Q1
SYMATTR Value BC557B
SYMBOL res 624 -176 R0
SYMATTR InstName R1
SYMATTR Value 10k
SYMBOL res 624 272 R0
SYMATTR InstName R2
SYMATTR Value 2.2k
SYMBOL voltage -304 -224 R0
WINDOW 3 24 44 Left 2
WINDOW 123 0 0 Left 0
WINDOW 39 0 0 Left 0
SYMATTR Value 2.5
SYMATTR InstName V1
SYMBOL njf 176 -256 R0
SYMATTR InstName J1
SYMATTR Value J113
SYMBOL npn 288 752 M0
SYMATTR InstName Q3
SYMATTR Value BC546B
SYMBOL res 208 992 R0
SYMATTR InstName R7
SYMATTR Value 50
SYMBOL npn 576 752 R0
WINDOW 3 66 66 Left 2
SYMATTR InstName Q4
SYMATTR Value BC546B
SYMBOL cap 1392 464 R0
SYMATTR InstName C1
SYMATTR Value 470p
SYMBOL cap 1392 656 R0
SYMATTR InstName C2
SYMATTR Value 470p
SYMBOL cap 1248 -16 R90
WINDOW 0 0 32 VBottom 2
WINDOW 3 58 36 VTop 2
SYMATTR InstName C3
SYMATTR Value 47p
SYMBOL res 624 976 R0
SYMATTR InstName R6
SYMATTR Value 10k
SYMBOL npn 1136 672 R0
WINDOW 3 66 66 Left 2
SYMATTR InstName Q2
SYMATTR Value BC546B
SYMBOL res 1184 1008 R0
SYMATTR InstName R3
SYMATTR Value 47
SYMBOL res 944 544 R0
SYMATTR InstName R4
SYMATTR Value 2.2k
SYMBOL res 1120 944 R90
WINDOW 0 0 56 VBottom 2
WINDOW 3 32 56 VTop 2
SYMATTR InstName R5
SYMATTR Value 68k
SYMBOL cap 864 832 R0
SYMATTR InstName C4
SYMATTR Value 470p
SYMBOL ind 1184 1376 R0
SYMATTR InstName L1
SYMATTR Value 4Âľ
SYMATTR SpiceLine Rser=0 Cpar=0
SYMBOL res 624 544 R0
SYMATTR InstName R8
SYMATTR Value 10k
SYMBOL voltage -288 512 R0
WINDOW 123 0 0 Left 0
WINDOW 39 0 0 Left 0
SYMATTR InstName V2
SYMATTR Value SINE(0 1 1Meg 0 0 0 1000)
SYMBOL res -80 400 R90
WINDOW 0 0 56 VBottom 2
WINDOW 3 32 56 VTop 2
SYMATTR InstName R9
SYMATTR Value 10k
SYMBOL res 1184 1520 R0
SYMATTR InstName R10
SYMATTR Value 50
SYMBOL res 1328 1440 R0
SYMATTR InstName R11
SYMATTR Value 100k
SYMBOL cap 1472 1456 R0
SYMATTR InstName C5
SYMATTR Value 10p
TEXT -352 48 Left 2 !.tran 100m
TEXT 1592 1488 Left 5 ;Lossy L
TEXT -632 296 Left 5 ;Kickstarter (sim only)
 
"Winfield Hill" wrote in message news:qhav1702std@drn.newsguy.com...

Kevin Aylward wrote...

"Winfield Hill" wrote in message news:qh9roj0fdi@drn.newsguy.com...

DRAFT copy of a useful section in the upcoming
AoE x-Chapters, about using power MOSFETs as
linear transistors. Subthreshold region, etc.
Comments, errors found, etc. Thanks!

https://www.dropbox.com/s/0ph15moamo0mlt4/3x.5_MOSFETs_Linear-Transistors_DRAFT.pdf?dl=1

Well, Win.... I would be interested in how well your mosfet
data could be fitted to the DMOS extension that models
subthreshold shown here

http://www.anasoft.co.uk/MOS1Model.htm

Thanks, Kevin, I'll have to study your material.

Also, the fit to the capacitances of the above model ...

Although we show one company's attempt at SPICE
modelling capacitance, it's an area I've stayed away
from. I've been happy to model HV amplifiers in the
region well away from the supply rails, where I can
simply use a fixed value appropriate to the region.

So... I would say though, that I am not particular enamoured of the Sandler
technique of adding subthreshold modelling. In principle, its a little bit
precarious. It has potential for convergence issues. Its correcting for a
zero current discontinuity by adding a feedback loop.

It would be preferable to simply add a controlled current using a B source
in parallel to the mosfet. One would then use an exp() function directly
rather than using a diode. One could then wrap that with a tanh() to
terminate its effect at larger currents.

A case in point, is... at my day job... we had an updated kit from X-Fab for
the process that were using just this last couple of weeks. All my existing
spice tests failed due to convergence issues. I traced it down to some
modelling problems in the natural mosfets. They had newly introduced B
Sources into the subckts, for reasons unknown. They were flagging up
convergence issues all over the place. A bare mosfet worked ok, but once one
put them into a real circuit, it fell apart.

So, if a mainstream fab can release a model kit that fails...

It might be worth mentioning in that section that LTSpice does model the
non-linear gd capacitance and has a subthreshold extension as well. Being
that millions of people use LTSpice, its getting to be a tad moot that
Spice3 don't support certain features. Additionally, there are other spices
that also do, with even other enhancements possible worth a mention... :)


-- Kevin Aylward
http://www.anasoft.co.uk - SuperSpice
http://www.kevinaylward.co.uk/ee/index.html
 
On 7/26/19 2:32 PM, Winfield Hill wrote:
bitrex wrote...

Winfield Hill wrote:
bitrex wrote...

The idea i'm messing with is bring the frequency down to something
more sensible from the self-resonant of a couple hundred MHz with
some low-loss inductance in series. Then use some negative
resistance to compensate for the bad Q down there. Can all be
done with FETs and BJTs, at low voltages.

You might want to experiment with piglet's idea ...

Hi Win, thank you, very nice!

If I'm allowed a more luxurious ~2.5 volts I came up with this
modified Lambda diode circuit.

What oscillation frequency do you envision? From 200 MHz down
to low MHz? Will your circuit work up that high?, I see 10k
resistors, etc.?

1-2MHz is fine. The actual lossy inductor will be about 0.5uH to 4 uH, I
intend to "swamp" it with a somewhat larger low-loss inductor so it
starts reliably despite the variation. And then the idea is that the
oscillation frequency will deviate a bit in response to the smaller one
being stretched/compressed. a few 10s of khz deviation is fine

BTW, your added inductor esr doesn't show on the schematic
(it's better to add parts to show these things explicitly).

Got it
 
On 7/26/19 3:30 PM, bitrex wrote:
On 7/26/19 2:32 PM, Winfield Hill wrote:
bitrex wrote...

Winfield Hill wrote:
bitrex wrote...

The idea i'm messing with is bring the frequency down to something
more sensible from the self-resonant of a couple hundred MHz with
some low-loss inductance in series.  Then use some negative
resistance to compensate for the bad Q down there.  Can all be
done with FETs and BJTs, at low voltages.

You might want to experiment with piglet's idea ...

Hi Win, thank you, very nice!

If I'm allowed a more luxurious ~2.5 volts I came up with this
modified Lambda diode circuit.

  What oscillation frequency do you envision?  From 200 MHz down
  to low MHz?   Will your circuit work up that high?, I see 10k
  resistors, etc.?

  BTW, your added inductor esr doesn't show on the schematic
  (it's better to add parts to show these things explicitly).



Here is a time domain of the fleshed-out idea oscillating OK in the sim.
A few hundred cycles of the resonant frequency injected into the tank
gets it moving, in the simulation.

<snip>

No, it's not going to work right. That bias network is wrong.
 
On Fri, 26 Jul 2019 11:40:53 -0700, dplatt@coop.radagast.org (Dave
Platt) wrote:

In article <lnemjeh2pckf2thlfsjcv419rdnp88mfhn@4ax.com>,
John Larkin <jjlarkin@highland_snip_technology.com> wrote:

Can you suggest some part numbers for high power n and p channel
lateral fets? Something in the 200 volt, 400 watt ballpark?

Exicon has the ECW20N20/ECW20P20, which are 200-volt 250-watt parts in
TO-264. ECF20N20/ECF20P20 seem to be the same dice, in TO-3. These
types look as if they're two 125-watt dice in a single package.

Their ProFusion arm sells 'em directly.

As far as I can tell, Exicon is the only company making the
lateral-MOSFET dice these days; there may be multiple companies
packaging them.

https://www.profusionplc.com/parts/ecw20n20

"They offer superb sonic characteristics"

What can that mean?

Does Hitachi still make laterals? I couldn't find them.


--

John Larkin Highland Technology, Inc
picosecond timing precision measurement

jlarkin att highlandtechnology dott com
http://www.highlandtechnology.com
 
On 26/07/2019 7:23 pm, George Herold wrote:
Instead of "With a little prodding" you might substitute,
"For a few haycorns". :^)

GH

:) I get a mention in the great book - Yay!

piglet
 

Welcome to EDABoard.com

Sponsor

Back
Top