Simple lab amps...

[Phil Hobbs]

For some client work involving entropy sources for cryptography (*), I
needed a couple of 1-nV-class instrument amps (**) with about 20 MHz
bandwidth.

Weirdly it\'s hard to buy something like that as a module. I was going
to dead-bug a couple, but Simon offered to lay out a board and make it a
product. We have several reels of BF862s in stock, so I used a couple
of those as a diff pair with a gain of ~20 as the input stage of a
composite amp, with a THS4631 output amp. The input is AC-coupled, and
the feedback produces a DC gain of 1 and an AC gain of 100, with a 3-dB
bandwidth of 100 Hz - 30 MHz.

The BF862s came from a pretty good reel, apparently, because the total
input-referred noise is 1.07 nV/sqrt(Hz) from ~10 kHz to its full
bandwidth, and it\'s super well behaved.

No giant technological advance, obviously, but something that will make
my life easier, and may help other folks too.

Fun.

Cheers

Phil Hobbs


(*) \'One man\'s noise is another man\'s data\' and all that.

(**) not an \'instrumentation amp\'--these ones are AC coupled and
single-ended externally.


--
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

 

[John Larkin]

On Mon, 10 Oct 2022 20:49:22 -0400, Phil Hobbs
<pcdhSpamMeSenseless@electrooptical.net> wrote:

For some client work involving entropy sources for cryptography (*), I
needed a couple of 1-nV-class instrument amps (**) with about 20 MHz
bandwidth.

Weirdly it\'s hard to buy something like that as a module. I was going
to dead-bug a couple, but Simon offered to lay out a board and make it a
product. We have several reels of BF862s in stock, so I used a couple
of those as a diff pair with a gain of ~20 as the input stage of a
composite amp, with a THS4631 output amp. The input is AC-coupled, and
the feedback produces a DC gain of 1 and an AC gain of 100, with a 3-dB
bandwidth of 100 Hz - 30 MHz.

The BF862s came from a pretty good reel, apparently, because the total
input-referred noise is 1.07 nV/sqrt(Hz) from ~10 kHz to its full
bandwidth, and it\'s super well behaved.

No giant technological advance, obviously, but something that will make
my life easier, and may help other folks too.

Fun.

Cheers

Phil Hobbs


(*) \'One man\'s noise is another man\'s data\' and all that.

(**) not an \'instrumentation amp\'--these ones are AC coupled and
single-ended externally.

The amp sounds like a cool product (send me one!), but why do you need
a low noise amp to make noise?

 

[Anthony William Sloman]

On Tuesday, October 11, 2022 at 2:23:50 PM UTC+11, John Larkin wrote:

On Mon, 10 Oct 2022 20:49:22 -0400, Phil Hobbs
pcdhSpamM...@electrooptical.net> wrote:

For some client work involving entropy sources for cryptography (*), I
needed a couple of 1-nV-class instrument amps (**) with about 20 MHz
bandwidth.

Weirdly it\'s hard to buy something like that as a module. I was going
to dead-bug a couple, but Simon offered to lay out a board and make it a
product. We have several reels of BF862s in stock, so I used a couple
of those as a diff pair with a gain of ~20 as the input stage of a
composite amp, with a THS4631 output amp. The input is AC-coupled, and
the feedback produces a DC gain of 1 and an AC gain of 100, with a 3-dB
bandwidth of 100 Hz - 30 MHz.

The BF862s came from a pretty good reel, apparently, because the total
input-referred noise is 1.07 nV/sqrt(Hz) from ~10 kHz to its full
bandwidth, and it\'s super well behaved.

No giant technological advance, obviously, but something that will make
my life easier, and may help other folks too.

(*) \'One man\'s noise is another man\'s data\' and all that.

(**) not an \'instrumentation amp\'--these ones are AC coupled and
single-ended externally.
The amp sounds like a cool product (send me one!), but why do you need
a low noise amp to make noise?

Real parts always seem to add in their own 1/f noise component which cryptographers might not like. The cryptographic noise source presumably doesn\'t, and the low noise amplifier presumably doesn\'t add in enough 1/f noise to worry them. The FET noise corner is at a higher frequency than the bipolar one.

The BF862 puts it at about 100Hz, and if the customer doesn\'t care about the frequency content below 100Hz, that wouldn\'t worry them

https://www.mvaudiolabs.com/diy/modern-jfet-noise-measurements/

--
Bill Sloman, Sydney

 

[Jan Panteltje]

On a sunny day (Mon, 10 Oct 2022 20:49:22 -0400) it happened Phil Hobbs
<pcdhSpamMeSenseless@electrooptical.net> wrote in
<50a487d7-f25d-3215-7d41-a4de02a17cdd@electrooptical.net>:

For some client work involving entropy sources for cryptography (*), I

(*) \'One man\'s noise is another man\'s data\' and all that.

Note that if you go for some noise at low levels
an adversary gets the advantage to make their own not-so-random noise
by sending all sorts of RF signals to your location.
Apart from your 60 Hz wall-warts with their harmonics etc.
and local radio stations with the latest hits and jive.

 

[Gerhard Hoffmann]

Am 11.10.22 um 02:49 schrieb Phil Hobbs:

For some client work involving entropy sources for cryptography (*),  I
needed a couple of 1-nV-class instrument amps (**) with about 20 MHz
bandwidth.

Weirdly it\'s hard to buy something like that as a module.  I was going
to dead-bug a couple, but Simon offered to lay out a board and make it a
product.  We have several reels of BF862s in stock, so I used a couple
of those as a diff pair with a gain of ~20 as the input stage of a
composite amp, with a THS4631 output amp.  The input is AC-coupled, and
the feedback produces a DC gain of 1 and an AC gain of 100, with a 3-dB
bandwidth of 100 Hz - 30 MHz.

The BF862s came from a pretty good reel, apparently, because the total
input-referred noise is 1.07 nV/sqrt(Hz) from ~10 kHz to its full
bandwidth, and it\'s super well behaved.

No giant technological advance, obviously, but something that will make
my life easier, and may help other folks too.

Don\'t forget to simulate / measure the real part of the input impedance.
I found it impossible to combine 250pV/rtHz, > 1MHz,
feedback through opamp to input FETs and stability.
Yes, it won\' necessarilly oscillate, but -100 real Ohm input
impedance is sth. I don\'t want in am amplifier to work with.
And a 150 Ohm gate stopper destroys the low noise properties.
Cascode helps somewhat.

That can be measured:
<
https://www.flickr.com/photos/137684711@N07/52419417337/in/dateposted-public/
>
Where the green line leaves the Smith diagram\'s outer circle,
there comes more energy back from the port than we put in.
add an inductor, and we have an oscillator.


Oh, and Win\'s amplifier in AOE3 had that also.
<
https://www.flickr.com/photos/137684711@N07/52420209074/in/dateposted-public/

Yellow line between 100 KHz and 20 MHz.

Talking about Win, I hope he\'s well? Last time he disappeared
was during preparation of AOE3. Do we get a new book?

The only opamp that kinda worked was one from TI with 3 or 4
GHz BW. Its 1/f noise was so bad that it dwarfed the noise
of the FET stage even after the first stage gain.

I ended up without FB around the FETs. 16 CPH3910 at 45 mA total
seem to be a sweet spot where the pos and the neg TK of the FETs
cancel, at least in simulation. Then gain is constant over temp,
at least. For this new, overpriced TI FET that seems to happen
near Idss, no way.

<
https://www.flickr.com/photos/137684711@N07/52420371063/in/dateposted-public/
>

The third of the board to the right is switchable gain to 86dB,
BW limiting and such. 340 pV/rt Hz.

If someone wants to play with it, some empty boards are left.

Cheers,
Gerhard

 

[Phil Hobbs]

Gerhard Hoffmann wrote:

Am 11.10.22 um 02:49 schrieb Phil Hobbs:
For some client work involving entropy sources for cryptography (*),
I needed a couple of 1-nV-class instrument amps (**) with about 20 MHz
bandwidth.

Weirdly it\'s hard to buy something like that as a module.  I was going
to dead-bug a couple, but Simon offered to lay out a board and make it
a product.  We have several reels of BF862s in stock, so I used a
couple of those as a diff pair with a gain of ~20 as the input stage
of a composite amp, with a THS4631 output amp.  The input is
AC-coupled, and the feedback produces a DC gain of 1 and an AC gain of
100, with a 3-dB bandwidth of 100 Hz - 30 MHz.

The BF862s came from a pretty good reel, apparently, because the total
input-referred noise is 1.07 nV/sqrt(Hz) from ~10 kHz to its full
bandwidth, and it\'s super well behaved.

No giant technological advance, obviously, but something that will
make my life easier, and may help other folks too.

Don\'t forget to simulate / measure the real part of the input impedance.
I found it impossible to combine 250pV/rtHz, > 1MHz,
feedback through opamp to input FETs and stability.
Yes, it won\' necessarilly oscillate, but -100 real Ohm input
impedance is sth. I don\'t want in am amplifier to work with.
And a 150 Ohm gate stopper destroys the low noise properties.
Cascode helps somewhat.

That can be measured:

https://www.flickr.com/photos/137684711@N07/52419417337/in/dateposted-public/
 
Where the green line leaves the Smith diagram\'s outer circle,
there comes more energy back from the port than we put in.
add an inductor, and we have an oscillator.


Oh, and Win\'s amplifier in AOE3 had that also.

https://www.flickr.com/photos/137684711@N07/52420209074/in/dateposted-public/
  
Yellow line between 100 KHz and 20 MHz.


Talking about Win, I hope he\'s well? Last time he disappeared
was during preparation of AOE3. Do we get a new book?

The only opamp that kinda worked was one from TI with 3 or 4
GHz BW. Its 1/f noise was so bad that it dwarfed the noise
of the FET stage even after the first stage gain.

I ended up without FB around the FETs. 16 CPH3910 at 45 mA total
seem to be a sweet spot where the pos and the neg TK of the FETs
cancel, at least in simulation. Then gain is constant over temp,
at least. For this new, overpriced TI FET that seems to happen
near Idss, no way.


https://www.flickr.com/photos/137684711@N07/52420371063/in/dateposted-public/
  

The third of the board to the right is switchable gain to 86dB,
BW limiting and such. 340 pV/rt Hz.

If someone wants to play with it, some empty boards are left.

Cheers,
Gerhard



Hi, Gerhard,

Thanks. This one is okay even with an open-circuited cable on the
input, thanks to a BLM18BB100SN1D ferrite bead in series with the gate
of the input FET and 100-ohm base stoppers on their cascodes. (The
other one has a 990/10 ohm feedback divider.)

It\'s pretty simple, just a cascoded BF862 pair driving a THS4631
differentially, plus an RLC in the feedback network to flatten out the
frequency response.

Something slower and quieter is probably next--those SiGe BJTs have
super low Rbb\' and Ree\', at least according to the Spice model.

It\'ll also be fun trying out a TMUX1511 as a chopper amp--2 ohms, 5 pF.

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

 

[John Larkin]

On Tue, 11 Oct 2022 12:11:47 +0200, Gerhard Hoffmann <dk4xp@arcor.de>
wrote:

Am 11.10.22 um 02:49 schrieb Phil Hobbs:
For some client work involving entropy sources for cryptography (*),  I
needed a couple of 1-nV-class instrument amps (**) with about 20 MHz
bandwidth.

Weirdly it\'s hard to buy something like that as a module.  I was going
to dead-bug a couple, but Simon offered to lay out a board and make it a
product.  We have several reels of BF862s in stock, so I used a couple
of those as a diff pair with a gain of ~20 as the input stage of a
composite amp, with a THS4631 output amp.  The input is AC-coupled, and
the feedback produces a DC gain of 1 and an AC gain of 100, with a 3-dB
bandwidth of 100 Hz - 30 MHz.

The BF862s came from a pretty good reel, apparently, because the total
input-referred noise is 1.07 nV/sqrt(Hz) from ~10 kHz to its full
bandwidth, and it\'s super well behaved.

No giant technological advance, obviously, but something that will make
my life easier, and may help other folks too.

Don\'t forget to simulate / measure the real part of the input impedance.
I found it impossible to combine 250pV/rtHz, > 1MHz,
feedback through opamp to input FETs and stability.
Yes, it won\' necessarilly oscillate, but -100 real Ohm input
impedance is sth. I don\'t want in am amplifier to work with.
And a 150 Ohm gate stopper destroys the low noise properties.
Cascode helps somewhat.

That can be measured:

https://www.flickr.com/photos/137684711@N07/52419417337/in/dateposted-public/

Where the green line leaves the Smith diagram\'s outer circle,
there comes more energy back from the port than we put in.
add an inductor, and we have an oscillator.


Oh, and Win\'s amplifier in AOE3 had that also.

https://www.flickr.com/photos/137684711@N07/52420209074/in/dateposted-public/

Yellow line between 100 KHz and 20 MHz.


Talking about Win, I hope he\'s well? Last time he disappeared
was during preparation of AOE3. Do we get a new book?

I had cause to email Horowitz a few months ago and asked about Win.
He\'s reported to be well but won\'t do another edition.

 

[Phil Hobbs]

Phil Hobbs wrote:

Gerhard Hoffmann wrote:
Am 11.10.22 um 02:49 schrieb Phil Hobbs:
For some client work involving entropy sources for cryptography (*),
I needed a couple of 1-nV-class instrument amps (**) with about 20
MHz bandwidth.

Weirdly it\'s hard to buy something like that as a module.  I was
going to dead-bug a couple, but Simon offered to lay out a board and
make it a product.  We have several reels of BF862s in stock, so I
used a couple of those as a diff pair with a gain of ~20 as the input
stage of a composite amp, with a THS4631 output amp.  The input is
AC-coupled, and the feedback produces a DC gain of 1 and an AC gain
of 100, with a 3-dB bandwidth of 100 Hz - 30 MHz.

The BF862s came from a pretty good reel, apparently, because the
total input-referred noise is 1.07 nV/sqrt(Hz) from ~10 kHz to its
full bandwidth, and it\'s super well behaved.

No giant technological advance, obviously, but something that will
make my life easier, and may help other folks too.

Don\'t forget to simulate / measure the real part of the input
impedance. I found it impossible to combine 250pV/rtHz, > 1MHz,
feedback through opamp to input FETs and stability.
Yes, it won\' necessarilly oscillate, but -100 real Ohm input
impedance is sth. I don\'t want in am amplifier to work with.
And a 150 Ohm gate stopper destroys the low noise properties.
Cascode helps somewhat.

That can be measured:

https://www.flickr.com/photos/137684711@N07/52419417337/in/dateposted-public/
  
Where the green line leaves the Smith diagram\'s outer circle,
there comes more energy back from the port than we put in.
add an inductor, and we have an oscillator.


Oh, and Win\'s amplifier in AOE3 had that also.

https://www.flickr.com/photos/137684711@N07/52420209074/in/dateposted-public/
   
Yellow line between 100 KHz and 20 MHz.


Talking about Win, I hope he\'s well? Last time he disappeared
was during preparation of AOE3. Do we get a new book?

The only opamp that kinda worked was one from TI with 3 or 4
GHz BW. Its 1/f noise was so bad that it dwarfed the noise
of the FET stage even after the first stage gain.

I ended up without FB around the FETs. 16 CPH3910 at 45 mA total
seem to be a sweet spot where the pos and the neg TK of the FETs
cancel, at least in simulation. Then gain is constant over temp,
at least. For this new, overpriced TI FET that seems to happen
near Idss, no way.


https://www.flickr.com/photos/137684711@N07/52420371063/in/dateposted-public/
   

The third of the board to the right is switchable gain to 86dB,
BW limiting and such. 340 pV/rt Hz.

If someone wants to play with it, some empty boards are left.

Cheers,
Gerhard



Hi, Gerhard,

Thanks.  This one is okay even with an open-circuited cable on the
input, thanks to a BLM18BB100SN1D ferrite bead in series with the gate
of the input FET and 100-ohm base stoppers on their cascodes.  (The
other one has a 990/10 ohm feedback divider.)

It\'s pretty simple, just a cascoded BF862 pair driving a THS4631
differentially, plus an RLC in the feedback network to flatten out the
frequency response.

Something slower and quieter is probably next--those SiGe BJTs have
super low Rbb\' and Ree\', at least according to the Spice model.

It\'ll also be fun trying out a TMUX1511 as a chopper amp--2 ohms, 5 pF.

We ordered 50 boards, stuffed with most of the components excluding the
op amps and some of the transistors. Two hundred bucks total from
JLCPCB, parts included.

Not too shabby.

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