Continue to Site

Welcome to our site!

Electro Tech is an online community (with over 170,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.

  • Welcome to our site! Electro Tech is an online community (with over 170,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.

Sawtooth generator (3 transistors) behaviour

Status
Not open for further replies.

Elerion

Member
I'm facing something I cannot fully understand even though I tried for hours.

**broken link removed**

Under simulation (LTSpice and Orcad), it sometimes, under some circusntances, stops working (Q1/Q2 keep always ON, and so capacitor stops charging again, thus oscillation ends). I found it to be dependant on R1 value and Q1 hFE.

Here's an example (LTSpice).
Works - > 1.5kohm, Q1 BC847A
Does not - > 1.8kohm

**broken link removed**
**broken link removed**

If Q1's hFE is below 200 it works for some R1 values that don't work for higher Q1's hFE.

Orcad simulation has its own R1 valid range, and the output frequency differs from LTSpice's.
Q1 as BC547B leads to:
- LTSpice works for R1 as low as 800 ohm, but Orcad stops working around 1.1 kohm.
- LTSpice works for R1 as high as 1,3k ohm, but Orcad stops working around 1.9 kohm.

It seems that this kind of circuit is not very well suited for simulations. Maybe it is not a neat desing?

Thanks for any idea.
 

Attachments

  • sawtooth3.asc
    1.5 KB · Views: 150
Hi Elerion,
Q1 and Q2 behave like a diac which switches to the conducting state when the voltage on the capacitor reaches a certain value. It remains in the conducting state until the current drops below the holding current. I think what is happening is that when it stops oscillating it is because the capacitor charging current is greater than the holding current.

Les.
 
Sorry, I did not see your schematic because I got distracted by the MONEY and PRETTY GIRLS at Postimg. You should attach your schematic HERE on this website instead of over there.
 
Sorry, I did not see your schematic because I got distracted by the MONEY and PRETTY GIRLS at Postimg. You should attach your schematic HERE on this website instead of over there.
Just use Adblock like everyone sane and you will never ever see an advertisement again.
I can´t install plugins at my work computer (where I started working just last month), and was quite horrified how many ads are on websites I visit frequently.
 
But I like to see MONEY and PRETTY GIRLS. My OS is Windows 10 that does not support plugins like Java so maybe it also will not block the pretty girls.
 
Listen to Les...

The latching occurs even if the current source consisting of R1, R2, R3, and Q3 is replaced with a simple resistor that charges C1.
In this circuit, the current source is replaced with just R1. This latches (and doesn't oscillate) for R1=100K, 200K, 300K, 400K and 500K, but resets and oscillates when R1=>600K. If R1 can source a current less than 7.5uA, then that current is not sufficient to keep the "scr" triggered.
See the gray trace; it is the first one that oscillates...

osc.gif

Here is a closer look at the conditions that cause the "scr" to trigger and reset:

osc1.gif

Note that it triggers when V(e) is ~2.8V and current into the emitter of Q2 is only 10nA!
It resets when V(e) is ~0.65V and the current into the emitter of Q2 is less than ~40uA. Note that in this simulation, I have the voltage changing much slower than in the oscillator, above.
 
Lots of circuits like this and multivibrators were designed in the past and found to be very unreliable.
Some needed special transistors, special component values and precise rail voltages, otherwise they failed to work.
This circuit could fall into that category and you need to use only circuits that work with almost any component values and a wide range of rail voltages.
Don't even think of using a dickey circuit.
 
Many years ago I used a similar PUT (2N6027 and 2N6028 programmable unijunction transistor) to make a sawtooth waveform or pulses and it worked perfectly. Of course they are not available anymore.
 

Attachments

  • PUT.png
    PUT.png
    31.7 KB · Views: 213
Thanks for every comment.

If I understud, capacitor charge current must be low enough for discharge transistors to cut-off. Right?

Is there any reason a high enough R1 could make the circuit fail?
As far as I know, it should only make signal oscillate at a slower pace.
 
Hi Elerion,

I would think that increasing the value of R1 (up to a certain point.) would INCREACE the charging current into the capacitor which would result in an INCREACE in frequency not a decreace.

Les.
 
I'm facing something I cannot fully understand even though I tried for hours...
Under simulation (LTSpice and Orcad), it sometimes, under some circusntances, stops working (Q1/Q2 keep always ON, and so capacitor stops charging again, thus oscillation ends). I found it to be dependant on R1 value and Q1 hFE..
If Q1's hFE is below 200 it works for some R1 values that don't work for higher Q1's hFE...
It seems that this kind of circuit is not very well suited for simulations. Maybe it is not a neat design? Thanks for any idea.


Hi Elerion,

I wrote this response before most of the other posts came in, so it covers some of the points that the other members have already made.

(1) General

Sorry to say this, but as you have found, this circuit has a few basic problems and to even work at all depends on component parameters- not a good thing. Having said that, though, it is quite clever and uses the minimum components. You are hoping for a multivibrator saw-tooth generator, but the circuit is basically a monostable and will stay in the low state indefinitely.

(2) Initial Points

(2.1) As has already been noted, the constant current (Ik) from the constant current generator Q2 is way too low at around 5uA. At a current that low, transistor hFE and ft drop-off badly. Worse still, the low current will not be able to charge up parasitic and virtual capacitances very quickly. The end result is that the circuit will tend to be unpredictable.

(2.2) Not only is Ik to low, but it is ill defined and will depend greatly on Q2 junction temperature, hFE to an extent, and individual transistor VBEs (bulk resistance effects).

(2.3) The saw-tooth high and low points are not well defined and will also depend on temperature, Ik and individual transistor characteristics.

(2.4) When Q2 is turned off, Q1 base is open circuit, so the node Q2C/Q1B is a high impedance. Consequently, that node will be badly affected by stay pick-up and leakage currents, especially at high temperatures. Also, when Q2 turns off there is no clear discharge path for the capacitances. The net result is that Q1 will turn on fast but will take an age to turn off. Finally, with a high impedance node like that there is more chance of parasitic oscillations.

(3) Solutions

The attached image shows your original circuit (CURCUIT #1) with a few voltages and currents indicated. CIRCUIT #2 shows the same circuit with some minor modifications, which should make the saw-tooth generator work- after a fashion. CIRCUIT 3 shows more extensive modifications which will, if I am right and there are no gross errors, work OK.

(3.1) CIRCUIT #1

There are two main problems with CIRCUIT #1:

(3.1.1) The very low and ill-defined Ik (note the very low voltage across R3).

(3.1.2) The mechanism for multivibrator operation relies on Ik not being sufficient to turn Q1 on. As Q1's base is not terminated this is impossible with an acceptable value of Ik, as discussed by Les and mikeA.

(3.1.3) Not a problem as such, but for development, it would be better to make C1 larger; increasing from 10nF to 100nF is recommended.

(3.2) CIRCUIT #2

(3.2.1) Ik has been increased to a much better 110uA, and is also reasonably well defined by changing the values of R1 and R3.

(3.2.2) A mechanism for multivibrator operation has been incorporated at the base of Q1. As a consequence Q1 base is also terminated. The principle is that when Ik flows through R6 the voltage drop across R6 will not be sufficient to turn Q1 on, so that the saw-tooth generator will not latch in the low state. But when Q2 conducts at the positive excursion of the saw-tooth, C1 discharge current plus Ik will be sufficient to turn Q1 on. Once C1 has discharge the current will drop back to Ik, so Q1 will turn off again. And so the cycle will continue- hopefully! You may need to adjust the value of R6 for optimum results.

(3.3) CIRCUIT #3

CIRCUIT #3 builds on CIRCUIT#2, buy adding a longer and better defined time for the saw-tooth low period, by virtue of C2 and associated resistors. I haven't bothered to work out the exact time, but you could experiment with different values of C2 to get the low period that you want.

(4) Summary

I guess that you are investigating this approach for the experience rather than to build a saw-tooth generator per se. In case you don't already know, there are much easier and better performing ways. Hope this is of some use and you haven't been bored to much reading it.

Chuck

(EOE)

ETO_sawtooth_generator_iss01.00_2015_12_02_Sh1~N.png


ERRATA
(1) '@= CHANGED COMPONENT' should read: '@= NEW COMPONENT'
(2) On CIRCUIT #1, label '700mV to 3V1' should attach to 'OUTPUT SAWTOOTH' not 'OUTPUT 0V'
(3) '4V16' in two locations should read '4V36'
(4) 'Ik=100uA' in two locations should read 'Ik=110uA'
(5) Q2-4= Q2-1, Q2-5= Q2-3, Q1-6= Q2-2
 
Last edited:
Les, I got mixed up, I meant to say decrease R1.

spec, thank you so much for your detailed answer. I got the idea. You're right, this is just for the experience.

The 555 circuit looks neat. I thought on using an Schmitt Trigger OpAmp to generate a square wave, then integrating, then adding a third OpAmp as voltage follower to drive something, like LEDs (though, their fordward bias barrier won't allow a full travel in intensity; I suppose I would need somekind of current source, from the integrator output).
 
No Probs Elerion.

I thought you were just investigating.

I spent years designing scope time bases, so this brings back memories. The integrating approach is another way of generating a saw tooth and a big advantage is the low output impedence, unlike the constant current approach, but overall the constant current method tends to be the most flexible and easier to deal with

When you say you would like to drive LEDs, what did you have in mind?
 
Last edited:
Yes, just investigating. Even though I'm an electronic engineer, what you learn in the university is just not enough for me.

I meant, make LED light intensity follow a smooth progression up, then a sudden fall (in other words, make the light intensity imitate a sawtooth function)
 
Yes, just investigating. Even though I'm an electronic engineer, what you learn in the university is just not enough for me.

Know what you mean. I bet many of us experienced the same- I certainly did. Have you seen item (1.1) here: https://www.electro-tech-online.com/articles/books-articles-data.758/ Many of the grads in the UK found it a help when transitioning fom achedemia to the practical side. Where are you from: US?

I meant, make LED light intensity follow a smooth progression up, then a sudden fall (in other words, make the light intensity imitate a sawtooth function)

What you would like to do sounds simple but it is quite difficult for two reasons:

(1) logarithmic response of the eye to white light- more complicated for single colour, I think

(2) Non linear Id/Luminance of LED.

If you forced a current through the LED proportional to the ramp voltage I think the LED would tend to come on in the first 20% say of the ramp and then not do much. Rather like replacing a log volume control pot on an audio amplifier with a lin version: the perceived volume increase is all over in the first few degrees rotation of the volume control.

Can I suggest that you consider an array of individual LEDs that progressively light as the ramp increases. You could implement that yourself or use one of the chips designed for the job.

You could make it really fancy by having the first two LEDs red, the next two green, the next two blue and the final two white. Or some combination- just thinking out loud
 
Last edited:
Thanks for the tips.
I knew that the light intensity modulation won't be smooth and clean, but I forgot to consider the points you just stated.
Again this is just pure experimentation,...

The last circuit's ID would really be (Vinput - Vfdiode) / RI ?
LEDs fordward voltage is quite high, so for voltages around 5V, it cannot be neglected.

By the way, I'm not from US, but from the same side of the Atlantic Ocean as you, but a little bit to the south; Spain.
 
Thanks for the tips.
I knew that the light intensity modulation won't be smooth and clean, but I forgot to consider the points you just stated.
Again this is just pure experimentation,...

The last circuit's ID would really be (Vinput - Vfdiode) / RI ?
LEDs fordward voltage is quite high, so for voltages around 5V, it cannot be neglected.

By the way, I'm not from US, but from the same side of the Atlantic Ocean as you, but a little bit to the south; Spain.

Hi Elerion,

I'm a newbee on ETO, but welcome all the same.

Ah! Espania- you are lucky: been a few times to mainland and Balierics. My son is getting married in Barcelona in September 2016.

(1) The last circuit's ID would really be (Vinput - Vfdiode) / RI?:
No, it is as stated on circuit. As you are investigating, I will leave you to work out the reasons (I am right by the way)

(2) LEDs fordward voltage is quite high, so for voltages around 5V, it cannot be neglected:
Yes, that is true, a point I had omitted, but it does not affect the fudamental priciple of the circuit. For your analysis best make V supply higher, say 12V. If you wanted to stay with a 5V supply, you would need to scale the input waveform amplitude and the resistor to suit. The only limitation on ID is the output current capability of the IC (about 60mA for the OPA192), its power dissipation, and the power dissipation of the LED itself. LED VF ranges from about 1V2 for red thru 4V for white/blue so the circuit should handle all colour LEDs. In practice a 5V supply would not be ideal: 9V to 12V would be much better.

Hope you have fun!

PS: I now know that you live in Spain, but are you Spanish?
 
Last edited:
Interesting and informative article about LEDs: **broken link removed**
 
Status
Not open for further replies.

Latest threads

New Articles From Microcontroller Tips

Back
Top