1

Here is a nice and straight-forward sawtooth oscillator that I learned from this video.

enter image description here

It is simple and nice. When we come to higher frequencies though, I am dealing with very finicky adjustments that end up breaking down. I can reduce R3 and C1 to get the slope up, but that only works until a few-fold increase, and then I'm at the end. The R1 can be tuned a little, but needs to be 1k for higher speeds. And at some point the difference between 1.5k and 1.8k makes the difference between life (oscillation) vs. death.

Like, even setting C1 to 1nF, with that R1 of 2k, nothing works. Reducing R1 to 1k gets the oscillations back. But we're not at much higher frequency, actually, we're at lower frequency.

We can do C1 = 100pF and R1 = 1kOhm and get about 2.5kHz. But now we're beginning to have non-linear starts and ends of the sawtooth:

enter image description here

and from there it gets only worse:

enter image description here

is it the switching speed of the BJTs that's causing this non-linearitiy? I am surprised that it is happening at frequencies that are still pretty low. Here we are only at 20 kHz, i.e., even if I wanted to use this as an sawtooth for an audio use case (sound synthesizer) I am already stretched to get that right.

Is this circuit known for this limitation or is there some better way to scale it to higher frequencies that I just don't know?

UPDATE: I took the suggestion of using the two diodes instead of R1, and unfortunately it's flat-lining and I can't find a proper C1, R2 and R3 that would make it oscillate.

enter image description here

Gunther Schadow
  • 1,441
  • 7
  • 22
  • @jonk I am looking for the simplest approach (as usual) where the principle can be seen the best. If I trigger externally, say from a 555, then ultimately that depends on some transistor like Q3/4 to open and drain that cap. 555 would just hide the real electrical action in a black box. If I have any other 74LSxx TTL or 74HCxx CMOS, I have that current limit problem. How about a MOSFET to drain C1? I will soon do a MOSFET to short an LC to ground at high frequency to make a DC to DC boost converter 5V to 12V (and then you will wonder about my Rube Goldberg machine again, haha) – Gunther Schadow Aug 25 '20 at 02:45
  • Simple in approach is not necessarily simple in layout. Simple in layout is quite often not at all simple in approach. For example, can you explain the principles behind your first schematic? There are even simpler ones to build that would depend on the base-emitter avalanche, which isn't even in the datasheet. Not simple in concept. But simple in layout. The joule thief is simple in layout, but understanding it? There are whole debates I've read where people get pretty caught up in the debate. So just saying "simplest" isn't that helpful. Just FYI. (A 555 is pretty simple in concept, too.) – jonk Aug 25 '20 at 03:00
  • You can also look [at this](https://electronics.stackexchange.com/a/481317/38098) on the topic of relatively predictable current sink/source design. – jonk Aug 25 '20 at 03:46
  • The comments under the answer are quite interesting. It is being mused that the thing needs instability to oscillate. Can I explain how it works? Well, to me it is intuitive: the transistor Q1 is a constant current source. Charging C1. When the potential over C1 rises above the threshold of Q3 the pair Q3+Q2 enters a positive feedback loop and only shuts off when the current from the capacitor is exhausted. For that, the constant current must not be enough to keep the pair going, that might be a bit of a mystery here. – Gunther Schadow Aug 25 '20 at 03:46
  • Okay. Well, it's not worth getting into debates about intuition. I consider human intuition to be consistently unreliable as a source of anything useful. But I've spent too long working with teams to solve problems no one else has solved before. So I've learned a healthy disrespect for the idea, that you've yet to find. That said, have a look at the link about making a decent current source/sink. It may help with part of what's ahead. Also, [here](https://i.stack.imgur.com/wTfc2.png) is a Spice run of my improvement on your circuit. It uses four BJTs total but includes positive FB. – jonk Aug 25 '20 at 03:55
  • @jonk I don't know either what the discussion is really about. I've been doing fine in software for 25 years and intuition has always been my guide. Anyway, what's your schematics? – Gunther Schadow Aug 25 '20 at 05:00
  • This question here is about an oscillator. My external triggered one is the other. And I just helped myself with a Darlington pair to effect the fast discharge. This should be easy for me to build with what I have. https://electronics.stackexchange.com/questions/518223/ltspice-and-ttl-logic-gate-parameters-my-edge-triggered-ramp-generator-is-curre – Gunther Schadow Aug 25 '20 at 05:20
  • It's [this](https://i.stack.imgur.com/nX0o7.png). – jonk Aug 25 '20 at 06:02
  • Should we guess what V(n00x) screenshots are? Ultimately (short term) what are you trying to achieve? – Andy aka Aug 25 '20 at 07:50
  • @Andyaka, the flat line V(n00x) should be the output of the oscillator between C1 and collector of Q1. Why does it matter what I am trying to achieve? I want to learn! I built a Wien bridge oscillator just to see it working. A rectangle oscillator is boring and done it many times, 555, crystal, etc. Integrating a rectangle oscillator to a triangle is easy and an RC network to go from rectangle to triangle and sine is under my belt already. Sawtooth was missing and is nice because it can be used for ramp/sweep generators but also amplified to sound like trumpets. – Gunther Schadow Aug 25 '20 at 16:46

2 Answers2

3
  1. This design depends on constant-current from Q1, and its GBW and \$beta\$.
  2. If this is real hardware you seriously need a bypass capacitor of 100 nF and 100 uF across 5 volts to ground. This is what helps make HF designs stable.
  3. Replace R1 with 2 1N4148 diodes in series to lock the voltage of R3 at ~650 millivolts. Reduce R2 so bias current for diodes and Q1 is ample. R2 can be as low as 2.2 K without heating the diodes or Q1.
  4. Still, R3 will have a low limit based on the gain of Q1. A super-beta transistor might give you more range for high and low current for the ramp.
  5. I assume that the 5 volt supply is very stable. If not R2 should be a 2 mA current sink.
  6. I understand your pushing boundaries here, so you will always find upper and lower limits, however the right type for Q1 plus the diodes might get you to 1 MHZ.
  7. Scour the datasheets to find a Q1 with a GBW of 100 MHZ, as full power frequency is a tiny fraction of this value.
  8. @jonk is offering better designs of more-or-less the same circuit. If you want something better (but not laboratory grade), listen to what he says.

EDIT: A 10 uH tiny inductor in series with the diodes would create a lot of instability, or replace 1 diode with a 620 ohm resistor. Out of many sawtooth generator schematics, you picked the most minimal design, so it will be the most finicky. Due to lack of self-oscillating parts and feedback, it seems to work by accident, depending on abstract bjt properties.

  • I followed your instructions, the pair of diodes needed to be done, but I couldn't find the right values to bring it back to oscillating. Updated in the OP. – Gunther Schadow Aug 25 '20 at 02:54
  • 1
    It is possible that it needs instability to make it oscillate. The diodes would have stabilized the current, yet dynamics in the current help make it oscillate. I am siding with jonk in that you could use a better design. –  Aug 25 '20 at 03:15
  • I added a EDIT section with some observations. –  Aug 25 '20 at 03:26
  • I find your observation that instability is needed quite deep and important, and possibly a drawback of all simulators, as they might not add any noise which would trigger an oscillation. In this case is seems to obvious that the circuit should oscillate, and it does if I use a timed source to trigger the discharge. But somehow if I use this SCR thingy or even my Darlington configuration and a simple BJT based comparator, the simulator enters into some weird steady state, in-between, and probably in real life it would oscillate fine. – Gunther Schadow Aug 25 '20 at 16:37
  • The SCR Q2-Q3 seems to work over a limited range. It maybe getting stuck ON and not going to an OFF state to allow C1 to charge. C1 cannot be too low a value as it looses its ability to toggle the simple SCR between charge and discharge. –  Aug 25 '20 at 17:34
  • 1
    Possibly a 100 K resistor across C1 could help the SCR reset, else adjust the R6/R4 ratio to create a lower trip point. A 2 uH inductor in series with Q1 collector would provide a slight back-EMF to shut OFF the SCR, but EMF peaks could damage Q1. Personally in the past I built sweep generators using a TLC555 and a constant-current charger for the capacitor, and a op-amp buffer. –  Aug 25 '20 at 21:56
2

16kHz oscillator with 2 semiconductors: ujt

csabahu
  • 2,156
  • 4
  • 5
  • I like your simplistic answer, so +1. However if you Google "Sawtooth generator schematic" you will find thousands of examples. Remember the OP may want an easy-to-adjust frequency. –  Aug 26 '20 at 00:25
  • I love it. What is that D1 E102 component? A constant current diode of 0.8 - 1.1 mA. So that means we could adjust the frequency by using a variable capacitor or by adjusting the charge level / amplitude. and you use the J-FET to reset. Very interesting. – Gunther Schadow Aug 26 '20 at 02:27
  • Yes, the frequency can be adjusted with the capacitor or to a lesser extent with the voltage of UJT B2. Alternatively, the diode can be replaced with a low current JFET and then the current can be controlled by the resistor SG of the JFET. – csabahu Aug 26 '20 at 07:38