BJT behaviour at ridiculously low current levels...

[Piotr Wyderski]

Hi,

I am experimenting with ultra-low power ideas and I was able to simulate
a 30Hz relaxation oscillator running at 45nA/1V. I am waiting for the
delivery of high-value resistors, but the same oscillator with resistors
scaled down 10x works on the bench and consumes ~380nA/3V/360Hz which is
still pretty awesome.

This real device has been made of BSS123/BSS84 operating in the
subthreshold realm -- the highest AC p-p voltage I can observe is
~500mV. There is a lot of interesting papers on subthreshold mode, so it
is all good.

But, out of curiosity, I replaced the FETs with BC847C/BC857C and it
appears to work (in LTSpice!) even better, with V_BE sort of 350mV.
I admit, no prototype so far. I didn\'t expect this -- could you please
you recommend me some good book/papers on BJTs operated at nA levels?
I would like to learn the theory in a proper way, as inferring it from a
SPICE behavioural model might not be the best way to go. For example, it
looks like the beta is extremely high there, even exponential. Does this
BJT mode have a googlable name?

Best regards, Piotr

 

[whit3rd]

On Wednesday, May 11, 2022 at 11:39:29 AM UTC-7, Piotr Wyderski wrote:

Hi,

I am experimenting with ultra-low power ideas and I was able to simulate
a 30Hz relaxation oscillator running at 45nA/1V.
....
recommend me some good book/papers on BJTs operated at nA levels?
I would like to learn the theory in a proper way, as inferring it from a
SPICE behavioural model might not be the best way to go. For example, it
looks like the beta is extremely high there, even exponential. Does this
BJT mode have a googlable name?

Alas, it\'s a region that has high dependence on collector leakage currents
(and cosmic rays, etc). Some of the space-rated research does get into
this, because radiation damage moves the effects into higher-current parts
of the characteristic. Beta is \'high\' because the collector leakage is
adding to the external base wire\'s contribution. It takes negative base bias
to get an \'off\' state, through Rbb resistance (base spreading resistance).

 

[John Larkin]

On Wed, 11 May 2022 20:39:08 +0200, Piotr Wyderski
<bombald@protonmail.com> wrote:

Hi,

I am experimenting with ultra-low power ideas and I was able to simulate
a 30Hz relaxation oscillator running at 45nA/1V. I am waiting for the
delivery of high-value resistors, but the same oscillator with resistors
scaled down 10x works on the bench and consumes ~380nA/3V/360Hz which is
still pretty awesome.

This real device has been made of BSS123/BSS84 operating in the
subthreshold realm -- the highest AC p-p voltage I can observe is
~500mV. There is a lot of interesting papers on subthreshold mode, so it
is all good.

But, out of curiosity, I replaced the FETs with BC847C/BC857C and it
appears to work (in LTSpice!) even better, with V_BE sort of 350mV.
I admit, no prototype so far. I didn\'t expect this -- could you please
you recommend me some good book/papers on BJTs operated at nA levels?
I would like to learn the theory in a proper way, as inferring it from a
SPICE behavioural model might not be the best way to go. For example, it
looks like the beta is extremely high there, even exponential. Does this
BJT mode have a googlable name?

Best regards, Piotr

There is a niche of building oscillators that run at very low supply
voltages, ballpark 10 mV. I think jfets do pretty good.

A depletion PHEMT might be good too.

--

If a man will begin with certainties, he shall end with doubts,
but if he will be content to begin with doubts he shall end in certainties.
Francis Bacon

 

[Piotr Wyderski]

John Larkin wrote:

There is a niche of building oscillators that run at very low supply
voltages, ballpark 10 mV. I think jfets do pretty good.

Yes, there are several interesting examples. The lowest voltage I know
of is 20mV. But in this off-time research application related to the
recent CeraCharge purchase I am trying to minimize current consumption,
not voltage. It can stop oscillating at 1.2V, no problem with that. But
if the current could be as low as 20nA, that would be something.

Best regards, Piotr

 

[John Larkin]

On Wed, 11 May 2022 21:23:17 +0200, Piotr Wyderski
<bombald@protonmail.com> wrote:

John Larkin wrote:

There is a niche of building oscillators that run at very low supply
voltages, ballpark 10 mV. I think jfets do pretty good.

Yes, there are several interesting examples. The lowest voltage I know
of is 20mV. But in this off-time research application related to the
recent CeraCharge purchase I am trying to minimize current consumption,
not voltage. It can stop oscillating at 1.2V, no problem with that. But
if the current could be as low as 20nA, that would be something.

Best regards, Piotr

The AOE X-chapter book has some data on using BJTs at pA levels.
Section 2x.9

2N4401 holds up pretty well at Ic of 100 pA. I\'d expect some very
small RF transistors to be pretty good.



--

If a man will begin with certainties, he shall end with doubts,
but if he will be content to begin with doubts he shall end in certainties.
Francis Bacon

 

[Phil Hobbs]

Piotr Wyderski wrote:

Hi,

I am experimenting with ultra-low power ideas and I was able to simulate
a 30Hz relaxation oscillator running at 45nA/1V. I am waiting for the
delivery of high-value resistors, but the same oscillator with resistors
scaled down 10x works on the bench and consumes ~380nA/3V/360Hz which is
still pretty awesome.

This real device has been made of BSS123/BSS84 operating in the
subthreshold realm -- the highest AC p-p voltage I can observe is
~500mV. There is a lot of interesting papers on subthreshold mode, so it
is all good.

But, out of curiosity, I replaced the FETs with BC847C/BC857C and it
appears to work (in LTSpice!) even better, with V_BE sort of 350mV.
I admit, no prototype so far. I didn\'t expect this -- could you please
you recommend me some good book/papers on BJTs operated at nA levels?
I would like to learn the theory in a proper way, as inferring it from a
SPICE behavioural model might not be the best way to go. For example, it
looks like the beta is extremely high there, even exponential. Does this
BJT mode have a googlable name?

    Best regards, Piotr

Lots of small BJTs work fine in the low nanoamps. I\'ve used BFT25As as
transdiodes down to 5 nA or so. They crapped out about 1 nA.

The transconductance is a lot better than FETs down there!

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]

John Larkin wrote:

On Wed, 11 May 2022 21:23:17 +0200, Piotr Wyderski
bombald@protonmail.com> wrote:

John Larkin wrote:

There is a niche of building oscillators that run at very low supply
voltages, ballpark 10 mV. I think jfets do pretty good.

Yes, there are several interesting examples. The lowest voltage I know
of is 20mV. But in this off-time research application related to the
recent CeraCharge purchase I am trying to minimize current consumption,
not voltage. It can stop oscillating at 1.2V, no problem with that. But
if the current could be as low as 20nA, that would be something.

Best regards, Piotr

The AOE X-chapter book has some data on using BJTs at pA levels.
Section 2x.9

2N4401 holds up pretty well at Ic of 100 pA. I\'d expect some very
small RF transistors to be pretty good.

The f_T goes in the tank--at 1 nA, even a BFT25A is a < 1-MHz transistor.

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

 

[Piotr Wyderski]

Phil Hobbs wrote:

Lots of small BJTs work fine in the low nanoamps.  I\'ve used BFT25As as
transdiodes down to 5 nA or so.  They crapped out about 1 nA.

This is getting really interesting. I didn\'t know that BJTs have useful
subthreshold region too.

> The transconductance is a lot better than FETs down there!

Indeed, Spice appears to say the same. I need to build a JFET probe in
order not to disturb the single electron microcosmos down there with a
dingy default 1Meg probe. And grab some radial resistors in the range of
47Meg+, as I don\'t fancy the idea of connecting hordes of 4.7Meg ones to
get there.

Best regards, Piotr

 

[John Larkin]

On Wed, 11 May 2022 17:14:08 -0400, Phil Hobbs
<pcdhSpamMeSenseless@electrooptical.net> wrote:

John Larkin wrote:
On Wed, 11 May 2022 21:23:17 +0200, Piotr Wyderski
bombald@protonmail.com> wrote:

John Larkin wrote:

There is a niche of building oscillators that run at very low supply
voltages, ballpark 10 mV. I think jfets do pretty good.

Yes, there are several interesting examples. The lowest voltage I know
of is 20mV. But in this off-time research application related to the
recent CeraCharge purchase I am trying to minimize current consumption,
not voltage. It can stop oscillating at 1.2V, no problem with that. But
if the current could be as low as 20nA, that would be something.

Best regards, Piotr

The AOE X-chapter book has some data on using BJTs at pA levels.
Section 2x.9

2N4401 holds up pretty well at Ic of 100 pA. I\'d expect some very
small RF transistors to be pretty good.




The f_T goes in the tank--at 1 nA, even a BFT25A is a < 1-MHz transistor.

Cheers

Phil Hobbs

Well, 1 pA into 1 pF gets you 1 volt per second. Speed takes power.

Big savings on heat sinks!




--

If a man will begin with certainties, he shall end with doubts,
but if he will be content to begin with doubts he shall end in certainties.
Francis Bacon

 

[John Larkin]

On Wed, 11 May 2022 23:38:24 +0200, Piotr Wyderski
<bombald@protonmail.com> wrote:

Phil Hobbs wrote:

Lots of small BJTs work fine in the low nanoamps.  I\'ve used BFT25As as
transdiodes down to 5 nA or so.  They crapped out about 1 nA.

This is getting really interesting. I didn\'t know that BJTs have useful
subthreshold region too.

I was theorizing that an LED, at low current, wouldn\'t have enough
voltage across it to make visible photons. A typical green LED runs
about 80 mV per decade current.

I tried some green LEDs at 1 nA and they made visible light.

I conjecture (ie guess) that base current makes collector current down
at the single-electron level. 1 pA is just 6 milion electrons per
second.

--

If a man will begin with certainties, he shall end with doubts,
but if he will be content to begin with doubts he shall end in certainties.
Francis Bacon

 

[John Larkin]

On Wed, 11 May 2022 23:38:24 +0200, Piotr Wyderski
<bombald@protonmail.com> wrote:

Phil Hobbs wrote:

Lots of small BJTs work fine in the low nanoamps.  I\'ve used BFT25As as
transdiodes down to 5 nA or so.  They crapped out about 1 nA.

This is getting really interesting. I didn\'t know that BJTs have useful
subthreshold region too.

The transconductance is a lot better than FETs down there!

Indeed, Spice appears to say the same. I need to build a JFET probe in
order not to disturb the single electron microcosmos down there with a
dingy default 1Meg probe. And grab some radial resistors in the range of
47Meg+, as I don\'t fancy the idea of connecting hordes of 4.7Meg ones to
get there.

I have some 1T ohm 0805 resistors. High values are available.

--

If a man will begin with certainties, he shall end with doubts,
but if he will be content to begin with doubts he shall end in certainties.
Francis Bacon

 

[bitrex]

On 5/11/2022 6:21 PM, John Larkin wrote:

On Wed, 11 May 2022 17:14:08 -0400, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

John Larkin wrote:
On Wed, 11 May 2022 21:23:17 +0200, Piotr Wyderski
bombald@protonmail.com> wrote:

John Larkin wrote:

There is a niche of building oscillators that run at very low supply
voltages, ballpark 10 mV. I think jfets do pretty good.

Yes, there are several interesting examples. The lowest voltage I know
of is 20mV. But in this off-time research application related to the
recent CeraCharge purchase I am trying to minimize current consumption,
not voltage. It can stop oscillating at 1.2V, no problem with that. But
if the current could be as low as 20nA, that would be something.

Best regards, Piotr

The AOE X-chapter book has some data on using BJTs at pA levels.
Section 2x.9

2N4401 holds up pretty well at Ic of 100 pA. I\'d expect some very
small RF transistors to be pretty good.




The f_T goes in the tank--at 1 nA, even a BFT25A is a < 1-MHz transistor.

Cheers

Phil Hobbs

Well, 1 pA into 1 pF gets you 1 volt per second. Speed takes power.

Big savings on heat sinks!

I remember seeing a sci-fi illustration when I was a kid of an alien
species discovering the Voyager golden record many millions of years in
the future - their little scout-ships the size of what would be 1:1000
scale toys of a Space Shuttle touching down on its surface to investigate.

Those guys really got it good with respect to the energy requirements of
\"manned\" space travel.

I also recall an alternative interpretation of general relativity (shape
dynamics) says that time isn\'t relative and instead the sizes of objects
in the Universe are what\'s relative.

 

[Piotr Wyderski]

John Larkin wrote:

> I have some 1T ohm 0805 resistors. High values are available.

I have 1G in 1206. I have quite a lot of 10M-33M in 0603 too, but they
make prototyping difficult. Wires help, hence this purchase of their THT
equivalents. Should arrive today, so no problem here.

Best regards, Piotr

 

[Arie de Muijnck]

On 2022-05-12 00:30, John Larkin wrote:

I conjecture (ie guess) that base current makes collector current down
at the single-electron level. 1 pA is just 6 milion electrons per
second.

AFAIK, it\'s not the base current but base voltage that controls the collector current. The base current is just an unfortunate side effect. The relation between them happens to be rather constant, hence the idea that hfe is the main factor. Ideally, the base \'leakage\' could be so low that base current electrons can be counted. In that case, would the much higher collector current come in packets?

Arie

 

[Martin Brown]

On 11/05/2022 23:30, John Larkin wrote:

On Wed, 11 May 2022 23:38:24 +0200, Piotr Wyderski
bombald@protonmail.com> wrote:

Phil Hobbs wrote:

Lots of small BJTs work fine in the low nanoamps.  I\'ve used BFT25As as
transdiodes down to 5 nA or so.  They crapped out about 1 nA.

This is getting really interesting. I didn\'t know that BJTs have useful
subthreshold region too.

I was theorizing that an LED, at low current, wouldn\'t have enough
voltage across it to make visible photons. A typical green LED runs
about 80 mV per decade current.

I tried some green LEDs at 1 nA and they made visible light.

Certainly it a dark room you will see the die light up. I\'d be surprised
if it did it with less voltage across the junction than energy of the
the photons being emitted (give or take a bit of slop for thermal phonon
interactions boosting the odd one).

These days at 5uA current drive you can see the white LED dies lit up
glowing a faint blue white under normal lab conditions.

I conjecture (ie guess) that base current makes collector current down
at the single-electron level. 1 pA is just 6 milion electrons per
second.

I suspect that there was a good reason why we used 10^11 ohm precision
resistors on Faraday amplifiers in mass specs but I never thought to
ask. They would sometimes put a 10^10 or 10^9 resistor on one channel
that was expected to have a higher beam current for some applications.

\"Resistors\" in that very high value regime don\'t behave at all ideally
either (maybe this has improved - I don\'t know). Considerable effort
went into correcting their non-ideal behaviour in post processing.

When the guy making them retired the first few batches with just his
apprentice at the helm unsupervised were unusable. There was an art to
cooking them just right to have the best combination of properties.

It isn\'t just the active components that behave oddly at ultra low
currents so does the pcb and the passive components. High value
resistors tend to behave like a distributed capacitance at >10^10.

--
Regards,
Martin Brown

 

[server]

On Thu, 12 May 2022 07:47:39 +0200, Piotr Wyderski
<bombald@protonmail.com> wrote:

John Larkin wrote:

I have some 1T ohm 0805 resistors. High values are available.

I have 1G in 1206. I have quite a lot of 10M-33M in 0603 too, but they
make prototyping difficult. Wires help, hence this purchase of their THT
equivalents. Should arrive today, so no problem here.

Best regards, Piotr

I did one photodiode amp that was all surface-mount but I couldn\'t
find suitable high-value thinfilm resistors, so I had to fly-over a
series pair of axial thinfilms.

The high value surfmounts are all thickfilm, at least the ones I could
get easily.

I don\'t like very low current design. Measurements take ages to
settle. Picosecond stuff happens right now.

Ditto thermal measurements. Tedious.



--

Anybody can count to one.

- Robert Widlar

 

[Jan Panteltje]

On a sunny day (Wed, 11 May 2022 12:17:51 -0700) it happened John Larkin
<jlarkin@highland_atwork_technology.com> wrote in
<0n2o7h9e69qdlfnd2j843un2b2ffldj3t4@4ax.com>:

On Wed, 11 May 2022 20:39:08 +0200, Piotr Wyderski
bombald@protonmail.com> wrote:

Hi,

I am experimenting with ultra-low power ideas and I was able to simulate
a 30Hz relaxation oscillator running at 45nA/1V. I am waiting for the
delivery of high-value resistors, but the same oscillator with resistors
scaled down 10x works on the bench and consumes ~380nA/3V/360Hz which is
still pretty awesome.

This real device has been made of BSS123/BSS84 operating in the
subthreshold realm -- the highest AC p-p voltage I can observe is
~500mV. There is a lot of interesting papers on subthreshold mode, so it
is all good.

But, out of curiosity, I replaced the FETs with BC847C/BC857C and it
appears to work (in LTSpice!) even better, with V_BE sort of 350mV.
I admit, no prototype so far. I didn\'t expect this -- could you please
you recommend me some good book/papers on BJTs operated at nA levels?
I would like to learn the theory in a proper way, as inferring it from a
SPICE behavioural model might not be the best way to go. For example, it
looks like the beta is extremely high there, even exponential. Does this
BJT mode have a googlable name?

Best regards, Piotr



There is a niche of building oscillators that run at very low supply
voltages, ballpark 10 mV. I think jfets do pretty good.

Have a JFET LC running from a thermocouple:
http://panteltje.com/pub/40_mV_oscillator_IMG_3597.GIF
http://panteltje.com/pub/40_mV_oscillator_waveform_on_gate_2Vpp_IMG_3598.GIF

Circuit diagram:
http://panteltje.com/pub/lighting_a_LED_with_a_candle_IMG_3604.GIF
But did not design for very low current.


Some PIC microprocessors take very little current
The 18LF1XK22 uses 34 nA in sleep mode, and 650 nA @32 kHz clock (says dataheet)
and can do a lot more than just oscillate...

>A depletion PHEMT might be good too.

Temperature range would be interesting.

 

[Phil Hobbs]

Arie de Muijnck wrote:

On 2022-05-12 00:30, John Larkin wrote:

I conjecture (ie guess) that base current makes collector current down
at the single-electron level. 1 pA is just 6 milion electrons per
second.

AFAIK, it\'s not the base current but base voltage that controls the
collector current. The base current is just an unfortunate side effect.
The relation between them happens to be rather constant, hence the idea
that hfe is the main factor. Ideally, the base \'leakage\' could be so low
that base current electrons can be counted. In that case, would the much
higher collector current come in packets?

Arie

Yup, a BJT is a transconductance device with a loss mechanism
(recombination in the base region) that produces leakage current. IIRC
Mikko Kiviranta or somebody said that the beta of a BFP650 goes up to
above 10000 at low temperature.

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

 

[Don]

Arie de Muijnck wrote:

John Larkin wrote:

I conjecture (ie guess) that base current makes collector current down
at the single-electron level. 1 pA is just 6 milion electrons per
second.

AFAIK, it\'s not the base current but base voltage that controls the
collector current. The base current is just an unfortunate side effect.
The relation between them happens to be rather constant, hence the idea
that hfe is the main factor. Ideally, the base \'leakage\' could be so low
that base current electrons can be counted. In that case, would the much
higher collector current come in packets?

Good point. You must be within the active region to use the hybrid model
to ignore transconductance.

Danke,

--
Don, KB7RPU, https://www.qsl.net/kb7rpu
There was a young lady named Bright Whose speed was far faster than light;
She set out one day In a relative way And returned on the previous night.

 

[Phil Hobbs]

Don wrote:

Arie de Muijnck wrote:
John Larkin wrote:

I conjecture (ie guess) that base current makes collector current down
at the single-electron level. 1 pA is just 6 milion electrons per
second.

AFAIK, it\'s not the base current but base voltage that controls the
collector current. The base current is just an unfortunate side effect.
The relation between them happens to be rather constant, hence the idea
that hfe is the main factor. Ideally, the base \'leakage\' could be so low
that base current electrons can be counted. In that case, would the much
higher collector current come in packets?

Good point. You must be within the active region to use the hybrid model
to ignore transconductance.

Danke,

Don\'t ever ignore transconductance.

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