pHEMT model tweaked for 1/f noise

[Phil Hobbs]

Hi, all,

In one post on the 'purpose of precision' thread, I noted that the
ATF38143 model I posted awhile back predicted way, way too much low
frequency noise. The real pHEMTs tend to have a pretty accurately 1/f
PSD with corner frequencies between 10 and 50 MHz.

Their 1/f characteristic makes them pretty good for bootstraps, because
the f**2 PSD caused by eN*C noise squashes the flicker noise pretty well.

Here's the old one:

..MODEL ATF38143_chip NMF( vto=-0.75, Beta=0.3, Lambda=0.07, Alpha=4,
+ B=0.8, Pb=0.7, Cgs=0.997E-12,
+ Cgd=0.176E-12, Rd=0.084, Rs=0.054, Kf=1e6, Af=1)

The relevant parameters are AF, which is the reciprocal of the noise
exponent, and Kf, which is the noise amplitude.

This produces a low-frequency noise PSD that goes as 1/f**2, and at 10
MHz is 340 *millivolts*/sqrt(Hz).

Here's the new one, with AF = 2, which gives a 1/f PSD, and KF = 5E-11,
which gives a 1/f corner of 22 MHz.

..MODEL SAV551_chip NMF( vto=0.08, Beta=0.3,
+ Lambda=0.07, Alpha=4 B=0.8, Pb=0.7,
+ Cgs=0.997E-12, Cgd=0.176E-12, Rd=0.084,
+ Rs=0.054, Kf=5e-11, Af=2)

I've attached the .asc file to this blog post at
<https://electrooptical.net/www/sed/SAV551noise.asc>.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

 

[Phil Hobbs]

On 2020-02-23 09:14, Phil Hobbs wrote:

Hi, all,

In one post on the 'purpose of precision' thread, I noted that the
ATF38143 model I posted awhile back predicted way, way too much low
frequency noise.  The real pHEMTs tend to have a pretty accurately 1/f
PSD with corner frequencies between 10 and 50 MHz.

Their 1/f characteristic makes them pretty good for bootstraps, because
the f**2 PSD caused by eN*C noise squashes the flicker noise pretty well.

I posted a slightly expanded version of this post at
<https://electrooptical.net/News/low-frequency-noise-in-ingaas-heterojunction-fets/>,
with a screen shot and the .asc file link.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

 

On Sun, 23 Feb 2020 09:14:49 -0500, Phil Hobbs
<pcdhSpamMeSenseless@electrooptical.net> wrote:

Hi, all,

In one post on the 'purpose of precision' thread, I noted that the
ATF38143 model I posted awhile back predicted way, way too much low
frequency noise. The real pHEMTs tend to have a pretty accurately 1/f
PSD with corner frequencies between 10 and 50 MHz.

Their 1/f characteristic makes them pretty good for bootstraps, because
the f**2 PSD caused by eN*C noise squashes the flicker noise pretty well.

Here's the old one:

.MODEL ATF38143_chip NMF( vto=-0.75, Beta=0.3, Lambda=0.07, Alpha=4,
+ B=0.8, Pb=0.7, Cgs=0.997E-12,
+ Cgd=0.176E-12, Rd=0.084, Rs=0.054, Kf=1e6, Af=1)

The relevant parameters are AF, which is the reciprocal of the noise
exponent, and Kf, which is the noise amplitude.

This produces a low-frequency noise PSD that goes as 1/f**2, and at 10
MHz is 340 *millivolts*/sqrt(Hz).

Here's the new one, with AF = 2, which gives a 1/f PSD, and KF = 5E-11,
which gives a 1/f corner of 22 MHz.

.MODEL SAV551_chip NMF( vto=0.08, Beta=0.3,
+ Lambda=0.07, Alpha=4 B=0.8, Pb=0.7,
+ Cgs=0.997E-12, Cgd=0.176E-12, Rd=0.084,
+ Rs=0.054, Kf=5e-11, Af=2)

I've attached the .asc file to this blog post at
https://electrooptical.net/www/sed/SAV551noise.asc>.

Cheers

Phil Hobbs

That's wonderful. I have several active uses for SAV551, both linear
and switching.

The thing on the left is a 551.

https://www.dropbox.com/s/i5o7ie33xe8ibpt/Z474_Partial.jpg?raw=1

I hope to finish that and fire it up tomorrow, if enough people will
leave me alone.

I also have a new coaxial ceramic resonator oscillator that uses a
551, and I want minimal phase noise.







--

John Larkin Highland Technology, Inc

The cork popped merrily, and Lord Peter rose to his feet.
"Bunter", he said, "I give you a toast. The triumph of Instinct over Reason"

 

On Sun, 23 Feb 2020 09:14:49 -0500, Phil Hobbs
<pcdhSpamMeSenseless@electrooptical.net> wrote:

Hi, all,

In one post on the 'purpose of precision' thread, I noted that the
ATF38143 model I posted awhile back predicted way, way too much low
frequency noise. The real pHEMTs tend to have a pretty accurately 1/f
PSD with corner frequencies between 10 and 50 MHz.

Their 1/f characteristic makes them pretty good for bootstraps, because
the f**2 PSD caused by eN*C noise squashes the flicker noise pretty well.

Here's the old one:

.MODEL ATF38143_chip NMF( vto=-0.75, Beta=0.3, Lambda=0.07, Alpha=4,
+ B=0.8, Pb=0.7, Cgs=0.997E-12,
+ Cgd=0.176E-12, Rd=0.084, Rs=0.054, Kf=1e6, Af=1)

The relevant parameters are AF, which is the reciprocal of the noise
exponent, and Kf, which is the noise amplitude.

This produces a low-frequency noise PSD that goes as 1/f**2, and at 10
MHz is 340 *millivolts*/sqrt(Hz).

Here's the new one, with AF = 2, which gives a 1/f PSD, and KF = 5E-11,
which gives a 1/f corner of 22 MHz.

.MODEL SAV551_chip NMF( vto=0.08, Beta=0.3,
+ Lambda=0.07, Alpha=4 B=0.8, Pb=0.7,
+ Cgs=0.997E-12, Cgd=0.176E-12, Rd=0.084,
+ Rs=0.054, Kf=5e-11, Af=2)

I've attached the .asc file to this blog post at
https://electrooptical.net/www/sed/SAV551noise.asc>.

Cheers

Phil Hobbs

Your model is a bit off my DC measurements, which of course doesn't
matter in most amplifier applications.

At +600 mV dc on the gate, I measured Ig = 2 uA and Rds-on = 2.2 ohms.
Your model has more gate current and higher Rds-on, 230 uA and 5 ohms.

I'm driving some from 4 volt CMOS logic, high or low, through 5K
resistors. Sort of like it was a bipolar transistor. The gate behaves
a lot like the gate of an NPN. Like most such parts, the gate is
delicate, rated 2 mA max for this one.

I might tweak your model a tad to make it a little better for
switching.

Mini-Circuits does *not* have a Spice model for any of their phemts.

OK, time for my donut.



--

John Larkin Highland Technology, Inc

The cork popped merrily, and Lord Peter rose to his feet.
"Bunter", he said, "I give you a toast. The triumph of Instinct over Reason"

 

On Sun, 23 Feb 2020 17:04:57 -0500, Phil Hobbs
<pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2020-02-23 11:14, jlarkin@highlandsniptechnology.com wrote:
On Sun, 23 Feb 2020 09:14:49 -0500, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

Hi, all,

In one post on the 'purpose of precision' thread, I noted that the
ATF38143 model I posted awhile back predicted way, way too much low
frequency noise. The real pHEMTs tend to have a pretty accurately 1/f
PSD with corner frequencies between 10 and 50 MHz.

Their 1/f characteristic makes them pretty good for bootstraps, because
the f**2 PSD caused by eN*C noise squashes the flicker noise pretty well.

Here's the old one:

.MODEL ATF38143_chip NMF( vto=-0.75, Beta=0.3, Lambda=0.07, Alpha=4,
+ B=0.8, Pb=0.7, Cgs=0.997E-12,
+ Cgd=0.176E-12, Rd=0.084, Rs=0.054, Kf=1e6, Af=1)

The relevant parameters are AF, which is the reciprocal of the noise
exponent, and Kf, which is the noise amplitude.

This produces a low-frequency noise PSD that goes as 1/f**2, and at 10
MHz is 340 *millivolts*/sqrt(Hz).

Here's the new one, with AF = 2, which gives a 1/f PSD, and KF = 5E-11,
which gives a 1/f corner of 22 MHz.

.MODEL SAV551_chip NMF( vto=0.08, Beta=0.3,
+ Lambda=0.07, Alpha=4 B=0.8, Pb=0.7,
+ Cgs=0.997E-12, Cgd=0.176E-12, Rd=0.084,
+ Rs=0.054, Kf=5e-11, Af=2)

I've attached the .asc file to this blog post at
https://electrooptical.net/www/sed/SAV551noise.asc>.

Cheers

Phil Hobbs

Your model is a bit off my DC measurements, which of course doesn't
matter in most amplifier applications.

At +600 mV dc on the gate, I measured Ig = 2 uA and Rds-on = 2.2 ohms.
Your model has more gate current and higher Rds-on, 230 uA and 5 ohms.

My procedure was pretty simplistic--I just stuck 30 mA through it,
measured V_GS, and hacked VTO to make the two agree. I ignored base
current entirely.


I'm driving some from 4 volt CMOS logic, high or low, through 5K
resistors. Sort of like it was a bipolar transistor. The gate behaves
a lot like the gate of an NPN. Like most such parts, the gate is
delicate, rated 2 mA max for this one.

I might tweak your model a tad to make it a little better for
switching.

Mini-Circuits does *not* have a Spice model for any of their phemts.

OK, time for my donut.

Cake, bread, brioche, cruller, ...?

I cordially dislike cake doughnuts, but a good maple- or chocolate-iced
bread doughnut is a beautiful thing.

Cheers

Phil Hobbs

(Pecan sticky buns rule, as long as they're not the toffee ones);

Our nearby Safeway makes a super Boston Creme, white goo inside and
cholocate icing. They make great bagels too. All their other baked
stuff is terrible.

Sticky buns are wonderful, but hard to find here. We have "morning
buns" that are awfully good.

The Brat is working on her Guinness Cake. She's bringing the latest
revision over tonight for a design review.



--

John Larkin Highland Technology, Inc

The cork popped merrily, and Lord Peter rose to his feet.
"Bunter", he said, "I give you a toast. The triumph of Instinct over Reason"

 

[Phil Hobbs]

On 2020-02-23 10:53, jlarkin@highlandsniptechnology.com wrote:

On Sun, 23 Feb 2020 09:14:49 -0500, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

Hi, all,

In one post on the 'purpose of precision' thread, I noted that the
ATF38143 model I posted awhile back predicted way, way too much low
frequency noise. The real pHEMTs tend to have a pretty accurately 1/f
PSD with corner frequencies between 10 and 50 MHz.

Their 1/f characteristic makes them pretty good for bootstraps, because
the f**2 PSD caused by eN*C noise squashes the flicker noise pretty well.

Here's the old one:

.MODEL ATF38143_chip NMF( vto=-0.75, Beta=0.3, Lambda=0.07, Alpha=4,
+ B=0.8, Pb=0.7, Cgs=0.997E-12,
+ Cgd=0.176E-12, Rd=0.084, Rs=0.054, Kf=1e6, Af=1)

The relevant parameters are AF, which is the reciprocal of the noise
exponent, and Kf, which is the noise amplitude.

This produces a low-frequency noise PSD that goes as 1/f**2, and at 10
MHz is 340 *millivolts*/sqrt(Hz).

Here's the new one, with AF = 2, which gives a 1/f PSD, and KF = 5E-11,
which gives a 1/f corner of 22 MHz.

.MODEL SAV551_chip NMF( vto=0.08, Beta=0.3,
+ Lambda=0.07, Alpha=4 B=0.8, Pb=0.7,
+ Cgs=0.997E-12, Cgd=0.176E-12, Rd=0.084,
+ Rs=0.054, Kf=5e-11, Af=2)

I've attached the .asc file to this blog post at
https://electrooptical.net/www/sed/SAV551noise.asc>.

Cheers

Phil Hobbs

That's wonderful. I have several active uses for SAV551, both linear
and switching.

The thing on the left is a 551.

https://www.dropbox.com/s/i5o7ie33xe8ibpt/Z474_Partial.jpg?raw=1

I hope to finish that and fire it up tomorrow, if enough people will
leave me alone.

Fun. Keep us posted.

I also have a new coaxial ceramic resonator oscillator that uses a
551, and I want minimal phase noise.

Dunno if a device with a 20-MHz 1/f corner is what you want for that,
but maybe.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

 

On Sun, 23 Feb 2020 17:01:09 -0500, Phil Hobbs
<pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2020-02-23 10:53, jlarkin@highlandsniptechnology.com wrote:
On Sun, 23 Feb 2020 09:14:49 -0500, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

Hi, all,

In one post on the 'purpose of precision' thread, I noted that the
ATF38143 model I posted awhile back predicted way, way too much low
frequency noise. The real pHEMTs tend to have a pretty accurately 1/f
PSD with corner frequencies between 10 and 50 MHz.

Their 1/f characteristic makes them pretty good for bootstraps, because
the f**2 PSD caused by eN*C noise squashes the flicker noise pretty well.

Here's the old one:

.MODEL ATF38143_chip NMF( vto=-0.75, Beta=0.3, Lambda=0.07, Alpha=4,
+ B=0.8, Pb=0.7, Cgs=0.997E-12,
+ Cgd=0.176E-12, Rd=0.084, Rs=0.054, Kf=1e6, Af=1)

The relevant parameters are AF, which is the reciprocal of the noise
exponent, and Kf, which is the noise amplitude.

This produces a low-frequency noise PSD that goes as 1/f**2, and at 10
MHz is 340 *millivolts*/sqrt(Hz).

Here's the new one, with AF = 2, which gives a 1/f PSD, and KF = 5E-11,
which gives a 1/f corner of 22 MHz.

.MODEL SAV551_chip NMF( vto=0.08, Beta=0.3,
+ Lambda=0.07, Alpha=4 B=0.8, Pb=0.7,
+ Cgs=0.997E-12, Cgd=0.176E-12, Rd=0.084,
+ Rs=0.054, Kf=5e-11, Af=2)

I've attached the .asc file to this blog post at
https://electrooptical.net/www/sed/SAV551noise.asc>.

Cheers

Phil Hobbs

That's wonderful. I have several active uses for SAV551, both linear
and switching.

The thing on the left is a 551.

https://www.dropbox.com/s/i5o7ie33xe8ibpt/Z474_Partial.jpg?raw=1

I hope to finish that and fire it up tomorrow, if enough people will
leave me alone.

Fun. Keep us posted.


I also have a new coaxial ceramic resonator oscillator that uses a
551, and I want minimal phase noise.

Dunno if a device with a 20-MHz 1/f corner is what you want for that,
but maybe.

Probably not. I should probably use a fast NPN.

If you have any spare time, maybe you could help with this. Some
thinking, if nothing else. I think I have invented a digital PLL that
I don't understand!

(Want a new DDG?)

https://www.dropbox.com/s/j3fycoyhpus0vpc/a4.jpg?raw=1




--

John Larkin Highland Technology, Inc

The cork popped merrily, and Lord Peter rose to his feet.
"Bunter", he said, "I give you a toast. The triumph of Instinct over Reason"

 

[Phil Hobbs]

On 2020-02-23 11:14, jlarkin@highlandsniptechnology.com wrote:

On Sun, 23 Feb 2020 09:14:49 -0500, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

Hi, all,

In one post on the 'purpose of precision' thread, I noted that the
ATF38143 model I posted awhile back predicted way, way too much low
frequency noise. The real pHEMTs tend to have a pretty accurately 1/f
PSD with corner frequencies between 10 and 50 MHz.

Their 1/f characteristic makes them pretty good for bootstraps, because
the f**2 PSD caused by eN*C noise squashes the flicker noise pretty well.

Here's the old one:

.MODEL ATF38143_chip NMF( vto=-0.75, Beta=0.3, Lambda=0.07, Alpha=4,
+ B=0.8, Pb=0.7, Cgs=0.997E-12,
+ Cgd=0.176E-12, Rd=0.084, Rs=0.054, Kf=1e6, Af=1)

The relevant parameters are AF, which is the reciprocal of the noise
exponent, and Kf, which is the noise amplitude.

This produces a low-frequency noise PSD that goes as 1/f**2, and at 10
MHz is 340 *millivolts*/sqrt(Hz).

Here's the new one, with AF = 2, which gives a 1/f PSD, and KF = 5E-11,
which gives a 1/f corner of 22 MHz.

.MODEL SAV551_chip NMF( vto=0.08, Beta=0.3,
+ Lambda=0.07, Alpha=4 B=0.8, Pb=0.7,
+ Cgs=0.997E-12, Cgd=0.176E-12, Rd=0.084,
+ Rs=0.054, Kf=5e-11, Af=2)

I've attached the .asc file to this blog post at
https://electrooptical.net/www/sed/SAV551noise.asc>.

Cheers

Phil Hobbs

Your model is a bit off my DC measurements, which of course doesn't
matter in most amplifier applications.

At +600 mV dc on the gate, I measured Ig = 2 uA and Rds-on = 2.2 ohms.
Your model has more gate current and higher Rds-on, 230 uA and 5 ohms.

My procedure was pretty simplistic--I just stuck 30 mA through it,
measured V_GS, and hacked VTO to make the two agree. I ignored base
current entirely.

I'm driving some from 4 volt CMOS logic, high or low, through 5K
resistors. Sort of like it was a bipolar transistor. The gate behaves
a lot like the gate of an NPN. Like most such parts, the gate is
delicate, rated 2 mA max for this one.

I might tweak your model a tad to make it a little better for
switching.

Mini-Circuits does *not* have a Spice model for any of their phemts.

OK, time for my donut.

Cake, bread, brioche, cruller, ...?

I cordially dislike cake doughnuts, but a good maple- or chocolate-iced
bread doughnut is a beautiful thing.

Cheers

Phil Hobbs

(Pecan sticky buns rule, as long as they're not the toffee ones);


--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

 

[David Nadlinger]

On 23.02.20 2:14 pm, Phil Hobbs wrote:

Here's the new one, with AF = 2, which gives a 1/f PSD, and KF = 5E-11,
which gives a 1/f corner of 22 MHz.

.MODEL SAV551_chip  NMF( vto=0.08, Beta=0.3,
+ Lambda=0.07,  Alpha=4 B=0.8, Pb=0.7,
+ Cgs=0.997E-12,  Cgd=0.176E-12, Rd=0.084,
+ Rs=0.054, Kf=5e-11, Af=2)

Have you tried basing it on the ADS-style quadratic Curtice model for
the ATF-55134?

Something like (with your noise parameters)

..MODEL SAV551 NMF(Vto=0.3, Beta=0.444, B=0.8, Alpha=13, Lambda=0.072,
Rd=2.025, Rs=0.675, Cgs=0.6193E-12, Cgd=0.1435E-12, Pb=0.7, Fc=0.65,
Kf=5e-11, Af=2)

although that Vto looks a bit odd; definitely needs fiddling to make it
match measured parameters.

— David

 

[Phil Hobbs]

On 2020-02-23 18:30, David Nadlinger wrote:

On 23.02.20 2:14 pm, Phil Hobbs wrote:
Here's the new one, with AF = 2, which gives a 1/f PSD, and KF =
5E-11, which gives a 1/f corner of 22 MHz.

.MODEL SAV551_chip  NMF( vto=0.08, Beta=0.3,
+ Lambda=0.07,  Alpha=4 B=0.8, Pb=0.7,
+ Cgs=0.997E-12,  Cgd=0.176E-12, Rd=0.084,
+ Rs=0.054, Kf=5e-11, Af=2)

Have you tried basing it on the ADS-style quadratic Curtice model for
the ATF-55134?

Something like (with your noise parameters)

..MODEL SAV551 NMF(Vto=0.3, Beta=0.444, B=0.8, Alpha=13, Lambda=0.072,
Rd=2.025, Rs=0.675, Cgs=0.6193E-12, Cgd=0.1435E-12, Pb=0.7, Fc=0.65,
Kf=5e-11, Af=2)

although that Vto looks a bit odd; definitely needs fiddling to make it
match measured parameters.

 — David

I'd have to actually figure out what all those parameters do, though.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

 

[Phil Hobbs]

On 2020-02-23 17:25, jlarkin@highlandsniptechnology.com wrote:

On Sun, 23 Feb 2020 17:04:57 -0500, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2020-02-23 11:14, jlarkin@highlandsniptechnology.com wrote:
On Sun, 23 Feb 2020 09:14:49 -0500, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

Hi, all,

In one post on the 'purpose of precision' thread, I noted that the
ATF38143 model I posted awhile back predicted way, way too much low
frequency noise. The real pHEMTs tend to have a pretty accurately 1/f
PSD with corner frequencies between 10 and 50 MHz.

Their 1/f characteristic makes them pretty good for bootstraps, because
the f**2 PSD caused by eN*C noise squashes the flicker noise pretty well.

Here's the old one:

.MODEL ATF38143_chip NMF( vto=-0.75, Beta=0.3, Lambda=0.07, Alpha=4,
+ B=0.8, Pb=0.7, Cgs=0.997E-12,
+ Cgd=0.176E-12, Rd=0.084, Rs=0.054, Kf=1e6, Af=1)

The relevant parameters are AF, which is the reciprocal of the noise
exponent, and Kf, which is the noise amplitude.

This produces a low-frequency noise PSD that goes as 1/f**2, and at 10
MHz is 340 *millivolts*/sqrt(Hz).

Here's the new one, with AF = 2, which gives a 1/f PSD, and KF = 5E-11,
which gives a 1/f corner of 22 MHz.

.MODEL SAV551_chip NMF( vto=0.08, Beta=0.3,
+ Lambda=0.07, Alpha=4 B=0.8, Pb=0.7,
+ Cgs=0.997E-12, Cgd=0.176E-12, Rd=0.084,
+ Rs=0.054, Kf=5e-11, Af=2)

I've attached the .asc file to this blog post at
https://electrooptical.net/www/sed/SAV551noise.asc>.

Cheers

Phil Hobbs

Your model is a bit off my DC measurements, which of course doesn't
matter in most amplifier applications.

At +600 mV dc on the gate, I measured Ig = 2 uA and Rds-on = 2.2 ohms.
Your model has more gate current and higher Rds-on, 230 uA and 5 ohms.

My procedure was pretty simplistic--I just stuck 30 mA through it,
measured V_GS, and hacked VTO to make the two agree. I ignored base
current entirely.


I'm driving some from 4 volt CMOS logic, high or low, through 5K
resistors. Sort of like it was a bipolar transistor. The gate behaves
a lot like the gate of an NPN. Like most such parts, the gate is
delicate, rated 2 mA max for this one.

I might tweak your model a tad to make it a little better for
switching.

Mini-Circuits does *not* have a Spice model for any of their phemts.

OK, time for my donut.

Cake, bread, brioche, cruller, ...?

I cordially dislike cake doughnuts, but a good maple- or chocolate-iced
bread doughnut is a beautiful thing.

Cheers

Phil Hobbs

(Pecan sticky buns rule, as long as they're not the toffee ones);

Our nearby Safeway makes a super Boston Creme, white goo inside and
cholocate icing. They make great bagels too. All their other baked
stuff is terrible.

Sticky buns are wonderful, but hard to find here. We have "morning
buns" that are awfully good.

That place down the hill from you, across from the used record store,
had very good sticky buns.

The Brat is working on her Guinness Cake. She's bringing the latest
revision over tonight for a design review.

Could make a mess of your git repos.

Cheers

Phil Hobbs



--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

 

[Phil Hobbs]

On 2020-02-23 17:11, jlarkin@highlandsniptechnology.com wrote:

On Sun, 23 Feb 2020 17:01:09 -0500, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2020-02-23 10:53, jlarkin@highlandsniptechnology.com wrote:
On Sun, 23 Feb 2020 09:14:49 -0500, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

Hi, all,

In one post on the 'purpose of precision' thread, I noted that the
ATF38143 model I posted awhile back predicted way, way too much low
frequency noise. The real pHEMTs tend to have a pretty accurately 1/f
PSD with corner frequencies between 10 and 50 MHz.

Their 1/f characteristic makes them pretty good for bootstraps, because
the f**2 PSD caused by eN*C noise squashes the flicker noise pretty well.

Here's the old one:

.MODEL ATF38143_chip NMF( vto=-0.75, Beta=0.3, Lambda=0.07, Alpha=4,
+ B=0.8, Pb=0.7, Cgs=0.997E-12,
+ Cgd=0.176E-12, Rd=0.084, Rs=0.054, Kf=1e6, Af=1)

The relevant parameters are AF, which is the reciprocal of the noise
exponent, and Kf, which is the noise amplitude.

This produces a low-frequency noise PSD that goes as 1/f**2, and at 10
MHz is 340 *millivolts*/sqrt(Hz).

Here's the new one, with AF = 2, which gives a 1/f PSD, and KF = 5E-11,
which gives a 1/f corner of 22 MHz.

.MODEL SAV551_chip NMF( vto=0.08, Beta=0.3,
+ Lambda=0.07, Alpha=4 B=0.8, Pb=0.7,
+ Cgs=0.997E-12, Cgd=0.176E-12, Rd=0.084,
+ Rs=0.054, Kf=5e-11, Af=2)

I've attached the .asc file to this blog post at
https://electrooptical.net/www/sed/SAV551noise.asc>.

Cheers

Phil Hobbs

That's wonderful. I have several active uses for SAV551, both linear
and switching.

The thing on the left is a 551.

https://www.dropbox.com/s/i5o7ie33xe8ibpt/Z474_Partial.jpg?raw=1

I hope to finish that and fire it up tomorrow, if enough people will
leave me alone.

Fun. Keep us posted.


I also have a new coaxial ceramic resonator oscillator that uses a
551, and I want minimal phase noise.

Dunno if a device with a 20-MHz 1/f corner is what you want for that,
but maybe.

Probably not. I should probably use a fast NPN.

If you have any spare time, maybe you could help with this. Some
thinking, if nothing else. I think I have invented a digital PLL that
I don't understand!

Sounds like fun. Send me an email.

(Want a new DDG?)

https://www.dropbox.com/s/j3fycoyhpus0vpc/a4.jpg?raw=1

Very nice. More modern looking than VFDs, though not as pretty a colour.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

 

On Mon, 24 Feb 2020 11:01:34 -0500, Phil Hobbs
<pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2020-02-23 18:30, David Nadlinger wrote:
On 23.02.20 2:14 pm, Phil Hobbs wrote:
Here's the new one, with AF = 2, which gives a 1/f PSD, and KF =
5E-11, which gives a 1/f corner of 22 MHz.

.MODEL SAV551_chip  NMF( vto=0.08, Beta=0.3,
+ Lambda=0.07,  Alpha=4 B=0.8, Pb=0.7,
+ Cgs=0.997E-12,  Cgd=0.176E-12, Rd=0.084,
+ Rs=0.054, Kf=5e-11, Af=2)

Have you tried basing it on the ADS-style quadratic Curtice model for
the ATF-55134?

Something like (with your noise parameters)

..MODEL SAV551 NMF(Vto=0.3, Beta=0.444, B=0.8, Alpha=13, Lambda=0.072,
Rd=2.025, Rs=0.675, Cgs=0.6193E-12, Cgd=0.1435E-12, Pb=0.7, Fc=0.65,
Kf=5e-11, Af=2)

although that Vto looks a bit odd; definitely needs fiddling to make it
match measured parameters.

 — David

I'd have to actually figure out what all those parameters do, though.

Cheers

Phil Hobbs

Oh. I thought you knew.



--

John Larkin Highland Technology, Inc

The cork popped merrily, and Lord Peter rose to his feet.
"Bunter", he said, "I give you a toast. The triumph of Instinct over Reason"

 

On Mon, 24 Feb 2020 11:00:38 -0500, Phil Hobbs
<pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2020-02-23 17:25, jlarkin@highlandsniptechnology.com wrote:
On Sun, 23 Feb 2020 17:04:57 -0500, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2020-02-23 11:14, jlarkin@highlandsniptechnology.com wrote:
On Sun, 23 Feb 2020 09:14:49 -0500, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

Hi, all,

In one post on the 'purpose of precision' thread, I noted that the
ATF38143 model I posted awhile back predicted way, way too much low
frequency noise. The real pHEMTs tend to have a pretty accurately 1/f
PSD with corner frequencies between 10 and 50 MHz.

Their 1/f characteristic makes them pretty good for bootstraps, because
the f**2 PSD caused by eN*C noise squashes the flicker noise pretty well.

Here's the old one:

.MODEL ATF38143_chip NMF( vto=-0.75, Beta=0.3, Lambda=0.07, Alpha=4,
+ B=0.8, Pb=0.7, Cgs=0.997E-12,
+ Cgd=0.176E-12, Rd=0.084, Rs=0.054, Kf=1e6, Af=1)

The relevant parameters are AF, which is the reciprocal of the noise
exponent, and Kf, which is the noise amplitude.

This produces a low-frequency noise PSD that goes as 1/f**2, and at 10
MHz is 340 *millivolts*/sqrt(Hz).

Here's the new one, with AF = 2, which gives a 1/f PSD, and KF = 5E-11,
which gives a 1/f corner of 22 MHz.

.MODEL SAV551_chip NMF( vto=0.08, Beta=0.3,
+ Lambda=0.07, Alpha=4 B=0.8, Pb=0.7,
+ Cgs=0.997E-12, Cgd=0.176E-12, Rd=0.084,
+ Rs=0.054, Kf=5e-11, Af=2)

I've attached the .asc file to this blog post at
https://electrooptical.net/www/sed/SAV551noise.asc>.

Cheers

Phil Hobbs

Your model is a bit off my DC measurements, which of course doesn't
matter in most amplifier applications.

At +600 mV dc on the gate, I measured Ig = 2 uA and Rds-on = 2.2 ohms.
Your model has more gate current and higher Rds-on, 230 uA and 5 ohms.

My procedure was pretty simplistic--I just stuck 30 mA through it,
measured V_GS, and hacked VTO to make the two agree. I ignored base
current entirely.


I'm driving some from 4 volt CMOS logic, high or low, through 5K
resistors. Sort of like it was a bipolar transistor. The gate behaves
a lot like the gate of an NPN. Like most such parts, the gate is
delicate, rated 2 mA max for this one.

I might tweak your model a tad to make it a little better for
switching.

Mini-Circuits does *not* have a Spice model for any of their phemts.

OK, time for my donut.

Cake, bread, brioche, cruller, ...?

I cordially dislike cake doughnuts, but a good maple- or chocolate-iced
bread doughnut is a beautiful thing.

Cheers

Phil Hobbs

(Pecan sticky buns rule, as long as they're not the toffee ones);

Our nearby Safeway makes a super Boston Creme, white goo inside and
cholocate icing. They make great bagels too. All their other baked
stuff is terrible.

Sticky buns are wonderful, but hard to find here. We have "morning
buns" that are awfully good.

That place down the hill from you, across from the used record store,
had very good sticky buns.


The Brat is working on her Guinness Cake. She's bringing the latest
revision over tonight for a design review.

Could make a mess of your git repos.

Cheers

Phil Hobbs

https://www.simplyrecipes.com/wp-content/uploads/2012/03/chocolate-guinness-cake-square-600-wide-600x671.jpg\

She invented the multilayer Guinness Cake. It doesn't relly look like
a draft Guinness any more, but the icing/cake ratio is advantageous.



--

John Larkin Highland Technology, Inc

The cork popped merrily, and Lord Peter rose to his feet.
"Bunter", he said, "I give you a toast. The triumph of Instinct over Reason"

 

[Phil Hobbs]

On 2020-02-24 12:28, jlarkin@highlandsniptechnology.com wrote:

On Mon, 24 Feb 2020 11:01:34 -0500, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2020-02-23 18:30, David Nadlinger wrote:
On 23.02.20 2:14 pm, Phil Hobbs wrote:
Here's the new one, with AF = 2, which gives a 1/f PSD, and KF =
5E-11, which gives a 1/f corner of 22 MHz.

.MODEL SAV551_chip  NMF( vto=0.08, Beta=0.3,
+ Lambda=0.07,  Alpha=4 B=0.8, Pb=0.7,
+ Cgs=0.997E-12,  Cgd=0.176E-12, Rd=0.084,
+ Rs=0.054, Kf=5e-11, Af=2)

Have you tried basing it on the ADS-style quadratic Curtice model for
the ATF-55134?

Something like (with your noise parameters)

..MODEL SAV551 NMF(Vto=0.3, Beta=0.444, B=0.8, Alpha=13, Lambda=0.072,
Rd=2.025, Rs=0.675, Cgs=0.6193E-12, Cgd=0.1435E-12, Pb=0.7, Fc=0.65,
Kf=5e-11, Af=2)

although that Vto looks a bit odd; definitely needs fiddling to make it
match measured parameters.

 — David

I'd have to actually figure out what all those parameters do, though.

Cheers

Phil Hobbs

Oh. I thought you knew.

I did read up on the noise and threshold parameters for maybe 10
minutes. I might hook one up to the HP 4145B when I'm in the lab next
week--we're using a SAV-551+ bootstrap for a new
photomultiplier-replacement gizmo. In certain frequency ranges it's a
super good bootstrap, so it might be a good idea to try to generate a
better model.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com