# Low Power Relaxation Oscillator Design Problem

#### Sensacell

Joined Jun 19, 2012
3,098
I want to create an oscillator circuit that functions in a manner similar to the classic neon relaxation oscillator- but with a strike voltage around 10-30 volts.
This is for an energy harvesting concept, the off-state leakage of the trigger circuit is critical.

The neon circuit has some characteristics that are optimal for my project:

1) Super low leakage current - almost no energy is lost to switch leakage- the neon is almost an open circuit until the strike voltage is reached.
2) The neon stays conductive until a relatively low voltage level - most of the capacitor energy is used.

I could use an SCR to switch my main load, so the trigger circuit could be a relatively high impedance when it triggers.
The current sourced to charge the capacitor must / will be below the SCR's minimum hold current so it will turn off when the capacitor discharges.

Perhaps a DIAC would do it? I have to investigate the leakage and voltage ratings of these devices.

Any ideas?

#### Attachments

• 2.5 KB Views: 14
• 601 KB Views: 17

#### Ian0

Joined Aug 7, 2020
6,696
Yes, a diac would do it. You could use a 2N4991 or BS08D if you want it to trigger at a lower and more accurate voltage, or a 2N6027, if you wanted to set the exact trigger voltage.

#### Sensacell

Joined Jun 19, 2012
3,098
Yes, a diac would do it. You could use a 2N4991 or BS08D if you want it to trigger at a lower and more accurate voltage, or a 2N6027, if you wanted to set the exact trigger voltage.
Thanks for the suggestion.

I looked up the BS08D device, the threshold voltage is lower than I would like, but the leakage is good, 1-10 uA.
These are surprisingly expensive- around \$1.00 in small quantities, I would love to find a 20 cent solution for 1K volume.

#### Ian0

Joined Aug 7, 2020
6,696
I don’t know why they have got so pricey, we used to use them to trigger a crowbar on the output of an audio amplifier, and only paid pence for them.

How about a thyristor with a zener between anode and gate? A 24V zener leaks less than 100nA at 18V, and the thyristor <1uA for a Thomson 0.8A thyristor with 200uA trigger current.

#### BobTPH

Joined Jun 5, 2013
6,076
Have you looked at energy harvesting chips? I suspect that they will do better than a small simple circuit made from discrete components.

Bob

#### AnalogKid

Joined Aug 1, 2013
10,174
UJT or PUT?

Or, given that a PUT is a four-layer device, basically two transistors, a discrete circuit equivalent so you can adjust the threshold voltage - ?

ak

#### Audioguru again

Joined Oct 21, 2019
5,413
I used a 2N6027 PUT almost 50 years ago. Today Digikey does not list it, maybe it is obsolete.
Newark says, "No longer available".

#### Ian0

Joined Aug 7, 2020
6,696
OnSemi has only its SPICE model - sort of like the ghost of the real thing.
Just make your own from a pair of bipolar transistors, but remember that half the total current goes through the base, so the limiting factor is the maximum base current.

#### neonstrobe

Joined May 15, 2009
181
If you can tolerate 36V then a DB3 diode can be used (<50uA before breakdown).
This is a common diode used in fluorescent tube starter circuits.
A complementary bipolar pair of devices can latch on if you don't provide a leakage path which causes the voltage to fall (eventually) to less than the holding voltage. That means one of the voltage divider chains (or the chain) should pass a leakage current. Base currents can be limited by additional resistors in series, which may help to reduce the local gain too (helping to turn off).

#### Alec_t

Joined Sep 17, 2013
13,228

U1-3 (better labelled U1a, U1b, U1c) are all in one IC. Rload is merely an example load for extracting energy from the cap.
U1/U2/R1/R2 together form an extended hysteresis Schmitt trigger.
The cap voltage can swing close to both rails with careful selection of R1 or R2. These resistors could have higher values to minimise energy wastage.

#### Attachments

• 1.9 KB Views: 1

#### Ian0

Joined Aug 7, 2020
6,696
Don't underestimate how much current a 40106 takes when it is biassed into a linear mode!

#### Alec_t

Joined Sep 17, 2013
13,228
I wouldn't dream of doing so!
None of the gates in my circuit is in the linear mode for long.
LTspice says for U1 the current draw is about 2uA for 1uS, but only during the change of state between high and low or vice versa. For U2 and U3 the current draw is about 80uA for 80nS during the state change.

#### AnalogKid

Joined Aug 1, 2013
10,174
None of the gates in my circuit is in the linear mode for long.
Actually, none of the gates are in the "linear mode", ever. Ripping through the transition region at CMOS edge rates doesn't count. Neither does just about anything with a Schmitt trigger input stage, or in a hysteretic circuit.

The bad news is that none of the energy built up in the timing capacitor is dumped into the load, a common thing in the harvesting world. Depending on what the load actually is, it might be possible to replace D1 with it.

ak

#### Ian0

Joined Aug 7, 2020
6,696
Even though the output stages of a Schmitt trigger aren't in a linear mode, the input stages are.
1mA is a lot of current for a CMOS gate.

#### Bordodynov

Joined May 20, 2015
3,025
It is a bad idea to use a Schmidt trigger in this case. The circuit is not functional with a 3 V supply voltage. This is because the input of your compound Schmidt trigger (two logic elements + two resistors) does not get the necessary voltages for switching.

#### Sensacell

Joined Jun 19, 2012
3,098
Here is my simple circuit so far- it works for the first cycle, but of course, the transistor pair SCR holding current is minute - so it does not shut off, therefore it will not cycle.

The concept is to get a tiny DC motor to move from a very high impedance source of ~ 25-35 Volts.
The current sourced by the supply is way too low to overcome the starting friction of the mechanics, but this circuit 'bumps' the motor perfectly.
The leakage is really low so it functions with a very low charging current.

Trying to think of a way to get the SCR to shut off when the current drops below some value, or maybe a simple timeout. All the ideas I have tried are too linear, lacking the necessary hysteresis to make it oscillate.

Simple and cheap is the order of the day - this will become a super cheap kid's toy.

Any ideas on how to reset the SCR?

Last edited:

#### iimagine

Joined Dec 20, 2010
506
Can try this:

#### Attachments

• 974 bytes Views: 4

#### Sensacell

Joined Jun 19, 2012
3,098
This is a very inspired circuit!

If I understand this correctly, the voltage drop across R2 keeps the BE junction of Q2 reverse biased until the charging current drops?
When it fires, the SCR pair then has no choice but to commutate off when the capacitor discharges- Q2 BE junction is reverse biased again.

I changed the diode to a 1N4148 and it would not oscillate, adding a 1 meg resistor from Q2 Base to GND fixed it, at the expense of the leakage.

Love it! fiendishly clever!