micro power square wave oscillator

[James Arthur]

<snip LTSpice LC oscillator file>

Cute. You could make a bundle selling those ideal 0.5H inductors. :O)

That was a nice comparator--betcha could redo this with
a crystal and kick tail.

Cheers,
James Arthur

 

[Joerg]

I already spec'd a 2v regulated supply--don't make me make it
1.8v and save even more power!


The 2n3904 only shows about 4pF in LTSpice, so there's
really not much to save here with fancy-pants transistors.
I deem it more than good enough for our long-departed OP,
who never really needed any of this anyhow.

4pf at 100kHz and a couple of volts or so matters a whole lot. Unless
you'd file a petition to revoke Ampere's, Coulomb's and Kirchhoff's laws
or have them declared unconstitutional ;-)

 

[Joerg]

Has anybody suggestesa 4093 Schmitt trigger NAND yet?
Although that only goes down to 3 V, the input cap i s7.5pF,
and with a single feedback resistor think the charge current will be
low.
Dunno how symmetrical it will be... or how stable.
Maybe there are better suited Schmitt trigger CMOS chips too.

The cross currents of the input stage will kill ya.

 

[JMini]

I'm looking for a schematic for a square wave oscillator which draws
around 10uA, powered by 2 to 5V and oscillates at around 100kHz with a
50% duty cycle. I tried out some schematics around a 4007 but none can
meet all the requirements.

any suggestions here?

regards,
nukeymusic

Try the Micrel. MIC1557. SOT-23-5 package.

 

[Joerg]

Try the Micrel. MIC1557. SOT-23-5 package.

Guzzles 200uA at 3V.

 

[JMini]

Guzzles 200uA at 3V.

Oops. By bad. I live mostly in the mA world.

 

[Joerg]

Oops. By bad. I live mostly in the mA world.

I used to live in the kilowatt world (RF, ham radio). Then I married ...

 

[Jan Panteltje]

The cross currents of the input stage will kill ya.

Probably, well it is not an easy thing 2V, 10uA, 100kHz.
In fact I have been unable to come up with something that I know works 100%.

 

[Joerg]

Probably, well it is not an easy thing 2V, 10uA, 100kHz.
In fact I have been unable to come up with something that I know works 100%.

Probably either a multivibrator with RF transistors or really small
geometry regular BJT (hard to find nowadays) or one of those battery
monitor chips used as an oscillator. But I can't try that out right now,
the missus sez it's clean-up day ...

 

[MooseFET]

On Jul 4, 3:03 am, Jim Thompson <To-Email-Use-The-Envelope-I...@My-Web-
Site.com> wrote:
[...]

Not exactly a realistic circuit. Where is the OP going to get 500k
resistors? It isn't an E12 value, or even E24.

C'mon, that's just an entry from the sims. Use 470K, 499K, whatever.

And a 0,265V voltage swing isn't all that useful. The classic two-
transistor emitter-coupled multivibrator has a free collector output,
usually used to drive a third - complementary - transistor as a
saturating switch to provide a rail-to-rail output.

And why use the 2N3904? The BFR92 has much lower internal capacitance
(and a much higher bandwidth) and could probably run with less than a
microamp in either leg. The BFT92 could serve as the saturating
switch.

Well, then go all out and use a BFP620

0.22pF C-E. But be careful, that thing really rocks, ft is 65GHz.

The aren't much good at very low currents and there's only one in the
package. The UPA861TD looks better on these points.

[...]

 

[Joerg]

On Jul 4, 3:03 am, Jim Thompson <To-Email-Use-The-Envelope-I...@My-Web-
Site.com> wrote: [...]

http://analog-innovations.com/SED/EmitterCoupledOscillator.pdf
Not exactly a realistic circuit. Where is the OP going to get 500k
resistors? It isn't an E12 value, or even E24.

C'mon, that's just an entry from the sims. Use 470K, 499K, whatever.

And a 0,265V voltage swing isn't all that useful. The classic two-
transistor emitter-coupled multivibrator has a free collector output,
usually used to drive a third - complementary - transistor as a
saturating switch to provide a rail-to-rail output.
And why use the 2N3904? The BFR92 has much lower internal capacitance
(and a much higher bandwidth) and could probably run with less than a
microamp in either leg. The BFT92 could serve as the saturating
switch.

Well, then go all out and use a BFP620

0.22pF C-E. But be careful, that thing really rocks, ft is 65GHz.

The aren't much good at very low currents and there's only one in the
package. The UPA861TD looks better on these points.

Interesting that one is 0.35pF and the other 0.25pF, probably because of
the not quite symmetrical pinout:

http://www.cel.com/pdf/datasheets/upa861td.pdf

Both package variants non-stock at Digikey though :-(

 

[MooseFET]

[email protected] wrote:
On Jul 4, 3:03 am, Jim Thompson <To-Email-Use-The-Envelope-I...@My-Web-
Site.com> wrote:
[...]
http://analog-innovations.com/SED/EmitterCoupledOscillator.pdf
Not exactly a realistic circuit. Where is the OP going to get 500k
resistors? It isn't an E12 value, or even E24.
C'mon, that's just an entry from the sims. Use 470K, 499K, whatever.
And a 0,265V voltage swing isn't all that useful. The classic two-
transistor emitter-coupled multivibrator has a free collector output,
usually used to drive a third - complementary - transistor as a
saturating switch to provide a rail-to-rail output.
And why use the 2N3904? The BFR92 has much lower internal capacitance
(and a much higher bandwidth) and could probably run with less than a
microamp in either leg. The BFT92 could serve as the saturating
switch.
Well, then go all out and use a BFP620
0.22pF C-E. But be careful, that thing really rocks, ft is 65GHz.

The aren't much good at very low currents and there's only one in the
package.  The UPA861TD looks better on these points.

Interesting that one is 0.35pF and the other 0.25pF, probably because of
the not quite symmetrical pinout:

http://www.cel.com/pdf/datasheets/upa861td.pdf

Both package variants non-stock at Digikey though :-(

Try mouser if you want some.

I assume you would resist designing them in because you can't get them
from digikey. I try to make it a rule never to spec a part that can't
be bought from them with no lead time.

 

[James Arthur]

4pf at 100kHz and a couple of volts or so matters a whole lot. Unless
you'd file a petition to revoke Ampere's, Coulomb's and Kirchhoff's laws
or have them declared unconstitutional ;-)

Coulomb's law is easily skirted: the collector swing is
limited to about 280mV, or roughly 1/8th the supply in
both oscillators, Jim's and mine. By design.

The dissipation arising from driving the 33pF timing cap is
far more important at this hefty power level. (20uW is
not low power.)

Ct sees about the same swing, but has a much higher capacitance.
This burns power, but still comes in way under budget. And
that was deliberate. I kept Ct high to minimize component
sensitivities and the influence of strays.

If you wanted to save power you could cut Ct and raise
impedances. Passive loads dissipate ~40% of this
circuits' input power. Active loads would help, but add
comlexity.

If we wanted to go to _really_ low powers then we'd want to
recycle the energy pumped into the timing scheme elements
--like with an LC or crystal--rather than dumping it each time.
That could save 80% of the power or more. Then better
transistors would be useful, but not needed here.

So, again, 2n3904s are good enough.

But then there's the question of what exactly are you going
to drive with this? A zero pF load? What accuaracy and
stability are needed? Rise and fall times?

All silliness.

Cheers,
James Arthur

 

[Joerg]

On Jul 3, 5:13 pm, Joerg <[email protected]>
wrote:
[email protected] wrote:
On Jul 4, 3:03 am, Jim Thompson <To-Email-Use-The-Envelope-I...@My-Web-
Site.com> wrote:
[...]
http://analog-innovations.com/SED/EmitterCoupledOscillator.pdf
Not exactly a realistic circuit. Where is the OP going to get 500k
resistors? It isn't an E12 value, or even E24.
C'mon, that's just an entry from the sims. Use 470K, 499K, whatever.
And a 0,265V voltage swing isn't all that useful. The classic two-
transistor emitter-coupled multivibrator has a free collector output,
usually used to drive a third - complementary - transistor as a
saturating switch to provide a rail-to-rail output.
And why use the 2N3904? The BFR92 has much lower internal capacitance
(and a much higher bandwidth) and could probably run with less than a
microamp in either leg. The BFT92 could serve as the saturating
switch.
Well, then go all out and use a BFP620
0.22pF C-E. But be careful, that thing really rocks, ft is 65GHz.
The aren't much good at very low currents and there's only one in the
package. The UPA861TD looks better on these points.

Interesting that one is 0.35pF and the other 0.25pF, probably because of
the not quite symmetrical pinout:

http://www.cel.com/pdf/datasheets/upa861td.pdf

Both package variants non-stock at Digikey though :-(

Try mouser if you want some.

They've got 37 parts. Not exactly a lot ...

I assume you would resist designing them in because you can't get them
from digikey. I try to make it a rule never to spec a part that can't
be bought from them with no lead time.

Yep, same here. Some manufacturers understand that, some don't. Many of
the ones that don't are relying on key account business and consequently
are posting lots of quarterly losses. No surprise to me.

 

[Joerg]

Coulomb's law is easily skirted: the collector swing is
limited to about 280mV, or roughly 1/8th the supply in
both oscillators, Jim's and mine. By design.

The dissipation arising from driving the 33pF timing cap is
far more important at this hefty power level. (20uW is
not low power.)

That 33pF is going to spoil the broth when you need useful logic level
amplitudes.

Ct sees about the same swing, but has a much higher capacitance.
This burns power, but still comes in way under budget. And
that was deliberate. I kept Ct high to minimize component
sensitivities and the influence of strays.

If you wanted to save power you could cut Ct and raise
impedances. Passive loads dissipate ~40% of this
circuits' input power. Active loads would help, but add
comlexity.

If we wanted to go to _really_ low powers then we'd want to
recycle the energy pumped into the timing scheme elements
--like with an LC or crystal--rather than dumping it each time.
That could save 80% of the power or more. Then better
transistors would be useful, but not needed here.

Yep, but adds a lot of complexity.

So, again, 2n3904s are good enough.

If you don't have to drive anything and just win a bet that you can make
a 100kHz oscillator under 10uA, maybe.

But then there's the question of what exactly are you going
to drive with this? A zero pF load? What accuaracy and
stability are needed? Rise and fall times?

That's why RF transistors may be needed. You may have to leave some
headroom for the load capacitance. It ain't going to be zero.

All silliness.

Hard to say without knowing the purpose of this oscillator. Watch
manufacturers certainly would not call their designs silly, the
difference between a two-year battery lifetime and five years can be a
strong sales argument ;-)

 

[James Arthur]

That 33pF is going to spoil the broth when you need useful logic level
amplitudes.






Yep, but adds a lot of complexity.


If you don't have to drive anything and just win a bet that you can make
a 100kHz oscillator under 10uA, maybe.


That's why RF transistors may be needed. You may have to leave some
headroom for the load capacitance. It ain't going to be zero.


Hard to say without knowing the purpose of this oscillator. Watch
manufacturers certainly would not call their designs silly, the
difference between a two-year battery lifetime and five years can be a
strong sales argument ;-)

I missed the part where we were designing a watch oscillator
with a 5-year life, 100kHz +/- 5ppm 0-40ºC. (If we wanted to
do that the lower frequency makes 2n3904s sing even sweeter

I thought we--Jim, John Fields and I--were just having some
fun with a nonsense requirement. Yeah, I could
do any and all of what you want, but that's not the
assignment.

Wanna play? Post your favorite circuit.

Cheers,
James Arthur

 

[krw]

On Jul 3, 5:13 pm, Joerg <[email protected]>
wrote:
[email protected] wrote:
On Jul 4, 3:03 am, Jim Thompson <To-Email-Use-The-Envelope-I...@My-Web-
Site.com> wrote:
[...]

http://analog-innovations.com/SED/EmitterCoupledOscillator.pdf
Not exactly a realistic circuit. Where is the OP going to get 500k
resistors? It isn't an E12 value, or even E24.
C'mon, that's just an entry from the sims. Use 470K, 499K, whatever.

And a 0,265V voltage swing isn't all that useful. The classic two-
transistor emitter-coupled multivibrator has a free collector output,
usually used to drive a third - complementary - transistor as a
saturating switch to provide a rail-to-rail output.
And why use the 2N3904? The BFR92 has much lower internal capacitance
(and a much higher bandwidth) and could probably run with less thana
microamp in either leg. The BFT92 could serve as the saturating
switch.
Well, then go all out and use a BFP620

0.22pF C-E. But be careful, that thing really rocks, ft is 65GHz.

The aren't much good at very low currents and there's only one in the
package.  The UPA861TD looks better on these points.

Interesting that one is 0.35pF and the other 0.25pF, probably because of
the not quite symmetrical pinout:

http://www.cel.com/pdf/datasheets/upa861td.pdf

Both package variants non-stock at Digikey though :-(

Try mouser if you want some.

I assume you would resist designing them in because you can't get them
from digikey. I try to make it a rule never to spec a part that can't
be bought from them with no lead time.

I always have at least enough parts for the prototypes in my desk
before the schematic is finalized. However, I'm not the one
controlling the BOM this time (the boards are now six months late).
:-(

 

[Joerg]

I missed the part where we were designing a watch oscillator
with a 5-year life, 100kHz +/- 5ppm 0-40ºC. (If we wanted to
do that the lower frequency makes 2n3904s sing even sweeter

The temp range won't do in watch apps. Just wanted to point out that
such requirements are not always nonsense.

I thought we--Jim, John Fields and I--were just having some
fun with a nonsense requirement. Yeah, I could
do any and all of what you want, but that's not the
assignment.

Wanna play? Post your favorite circuit.

Maybe but first I'll have to crawl underneath the house some more. I
finally fixed out little valve problem just to be replaced by a new
problem: Water hammer. Darn!

 

[[email protected]]

<snip LTSpice LC oscillator file>

Cute.  You could make a bundle selling those ideal 0.5H inductors. :O)

With more nearly real components, plugged into the standard LTspice
example astable you can get 1.5uA

Version 4
SHEET 1 2848 1240
WIRE -576 576 -736 576
WIRE -112 576 -576 576
WIRE -576 608 -576 576
WIRE -112 608 -112 576
WIRE -576 704 -576 688
WIRE -464 704 -576 704
WIRE -384 704 -464 704
WIRE -240 704 -320 704
WIRE -112 704 -112 688
WIRE -112 704 -240 704
WIRE -736 720 -736 576
WIRE -464 736 -464 704
WIRE -240 736 -240 704
WIRE -576 832 -576 704
WIRE -112 832 -112 704
WIRE -496 880 -512 880
WIRE -464 880 -464 816
WIRE -464 880 -496 880
WIRE -448 880 -464 880
WIRE -384 880 -320 704
WIRE -320 880 -384 704
WIRE -240 880 -240 816
WIRE -240 880 -256 880
WIRE -208 880 -240 880
WIRE -176 880 -208 880
WIRE -496 928 -496 880
WIRE -208 928 -208 880
WIRE -576 1008 -576 928
WIRE -496 1008 -496 992
WIRE -496 1008 -576 1008
WIRE -208 1008 -208 992
WIRE -112 1008 -112 928
WIRE -112 1008 -208 1008
WIRE -736 1040 -736 800
WIRE -576 1040 -576 1008
WIRE -112 1040 -112 1008
FLAG -112 1040 0
FLAG -576 1040 0
FLAG -736 1040 0
SYMBOL RES -592 592 R0
SYMATTR InstName R1
SYMATTR Value 5100K
SYMBOL res -128 592 R0
SYMATTR InstName R2
SYMATTR Value 5100K
SYMBOL cap -384 864 R90
WINDOW 0 0 32 VBottom 0
WINDOW 3 32 32 VTop 0
SYMATTR InstName C1
SYMATTR Value 3.3p
SYMBOL cap -256 864 R90
WINDOW 0 0 32 VBottom 0
WINDOW 3 32 32 VTop 0
SYMATTR InstName C2
SYMATTR Value 3.3p
SYMBOL VOLTAGE -736 704 R0
SYMATTR InstName V1
SYMATTR Value 5
SYMBOL npn -176 832 R0
SYMATTR InstName Q1
SYMATTR Value BFR92A
SYMBOL npn -512 832 M0
SYMATTR InstName Q2
SYMATTR Value BFR92A
SYMBOL res -256 720 R0
SYMATTR InstName R3
SYMATTR Value 15000K
SYMBOL res -480 720 R0
SYMATTR InstName R4
SYMATTR Value 15000.1K
SYMBOL cap -512 928 R0
SYMATTR InstName C3
SYMATTR Value 3.3p
SYMBOL cap -224 928 R0
SYMATTR InstName C4
SYMATTR Value 3.3p
TEXT -768 1176 Left 0 !.tran 0 1.5m 0.5m startup
TEXT -776 1224 Left 0 ;This example schematic is supplied for
informational/educational purposes only.
TEXT -480 1176 Left 0 !.model BFR92A NPN(IS=0.1213E-15 VAF=30 BF=94.73
IKF=0.46227 XTB=0 BR=10.729 CJC=946.47E-15 CJE=10.416E-15
TR=1.2744E-9 TF=26.796E-12 ITF=0.0044601 VTF=0.32861 XTF=0.3817
RB=14.998 RC=0.13793 RE=0.29088 Vceo=15 Icrating=4m mfg=Infineon)

 

[Joerg]

James Arthur wrote:

[...]

I missed the part where we were designing a watch oscillator
with a 5-year life, 100kHz +/- 5ppm 0-40ºC. (If we wanted to
do that the lower frequency makes 2n3904s sing even sweeter

I thought we--Jim, John Fields and I--were just having some
fun with a nonsense requirement. Yeah, I could
do any and all of what you want, but that's not the
assignment.

Wanna play? Post your favorite circuit.

Ok, example:


3.3V 3.3V
| |
\ \
/ 1M /
\ 1M \
/ /
| |
| |
o-------o-------. '-------o-------o
| | | 2.2p| | |
| \ --- --- \ |
| / --- --- / |
| \ 2M |2.2p | \ 2M |
| / | | / |
| | | | | |
\| --------)---------' | |/
|-----' | -o-----|
<| '---------------' |> BFP620
| BFP620 |


GND GND
(created by AACircuit v1.28.6 beta 04/19/05 www.tech-chat.de)


Forgive the lousy schematic, I am not used to these ASCII editors.
Anyhow, 3.3V is beyond the BFP620's 2.3V limit but there are similar
oens that reach there. Uses around 5uA in SPICE. Thing is, with a 2N3904
this won't work at all. It's capacitances snuff it out. Of course this
is rather hokey in terms of stray capacitance and would need the usual
"anti-RF" precautions such as ferrite beads or at least small resistors
in front of the bases.