1.5Mhz Linear Triangle Wave Using 2n3906/2n3904

dcbingaman

Joined Jun 30, 2021
498
I will have to study that circuit for a while just to understand what is going on. I will agree though, it is a nice triangle wave generator. The overall circuit has a lot of the techniques used in op-amps. In the real world, to reduce parts count, I would just use one or two op amps with some caps and resistors. But this is interesting from an intellectual perspective. Nice circuit.
 

dcbingaman

Joined Jun 30, 2021
498
Your sim originally used ideal PNP and NPN BJTs. I changed them to 2N3904 and 2N3906. The circuit still worked fine with no changes in output with that change.
 

dcbingaman

Joined Jun 30, 2021
498
After closer analysis I might be getting an idea how this circuit is working.

- Upon power on, C1 is not charged, this turns on Q3 and turns off Q2. It also turns off Q6 which turns on Q1 and Q8
- Q8 and Q7 form a current mirror with a constant current around (5V-1V)/110ohms or 36mA, R11 steals some of that current making the actual current around 24mA
- C1 is charging from current mirror with constant 24mA
- Ramp rate is then just i/C= 24mA/2.2nF or 11V per microsecond. Since we reach about 4V prior to circuit switching that is around 0.36 microsecond ramp up time.
- When the capacitor charge reaches about 4V, Q3 no longer has sufficient base current to remain on, Q3 switches off
- Q2 switches on, Q6 switches on, Q1 switches off turning off the current mirror for Q7
- Q2 being on turns on Q4/Q5 current mirror
- Q5 discharges C1 at same rate as the other current mirror
- Q3 would like to turn on during this time, but is prohibited by the lower emitter voltage caused by Q2 being on
- Only when C1 is practically completely discharged will Q5 current mirror stop working and turn back off
- The cycle repeats

Somehow D1, D2 and D3 improve performance and make a proper linear triangle wave. Holding the collector at no more than 1 diode drop (0.3V) above the base has some sort of benefit.

- By connecting Q3 collector to Q6 emitter we improve turn off speed of Q6. I experimented with connecting Q3 collector to ground, the circuit still works but with distortion in the waveform.

- Removing all three diodes, the circuit still oscillates but is far from generating a triangle wave.

Maybe the TS would like to provide some enlightenment on how the circuit was developed.
 
Last edited:

dcbingaman

Joined Jun 30, 2021
498
Studied that little diode trick. Here the base to collector voltage never reaches close to more than 0.3V when transistor turns off.

1634504387341.png
Input Red output green:
1634504468668.png

By adding the diode the base to collector max voltage cannot exceed 0.3V when the transistor is switched off, which speeds up the transistor:

1634505395698.png

Green trace in, blue trace out.

That is an interesting little trick for improved performance.
 

DickCappels

Joined Aug 21, 2008
7,984
Studied that little diode trick. Here the base to collector voltage never reaches close to more than 0.3V when transistor turns off.

View attachment 250516
(some of the post was deleted for clarity)

That is an interesting little trick for improved performance.
This is the Baker Clamp -it keeps the inverting transistor (Q9 in this case) from being saturated and/or over-driven. Great circuit.
 

dcbingaman

Joined Jun 30, 2021
498
Not sure the purpose of R11 and R12. The circuit works just fine without them at slightly higher frequency. I did have to adjust R6 from 10 to 8 after removing them, but you can regain the symmetry without them.
 

crutschow

Joined Mar 14, 2008
28,204
Certainly an interesting design for learning about discrete transistor circuits, but I see no advantage for that over a simpler and lower power circuit using high speed op amps.
 

atferrari

Joined Jan 6, 2004
4,473
This design reminded me the use of two constant current sources on the same cap; one of them feeding current 2I permanently and the other switched on/off periodically draining current I. Linear slopes both.
 

AnalogKid

Joined Aug 1, 2013
9,494
This design reminded me the use of two constant current sources on the same cap; one of them feeding current 2I permanently and the other switched on/off periodically draining current I. Linear slopes both.
I've seen that in commercial function generator schematics; it does a nice job of eliminating or reducing non-linearity at the transition points. I almost used it once, but the company didn't get the project.

ak
 

atferrari

Joined Jan 6, 2004
4,473
I've seen that in commercial function generator schematics; it does a nice job of eliminating or reducing non-linearity at the transition points. I almost used it once, but the company didn't get the project.

ak
IIRC, it was the ICL8038 where
I learnt of that topology in a simplified schematic in its datasheet.

Still trying to retrieve several LTS simulations I did some time ago.
 

dcbingaman

Joined Jun 30, 2021
498
Certainly an interesting design for learning about discrete transistor circuits, but I see no advantage for that over a simpler and lower power circuit using high speed op amps.
Granted. I thought the circuit was extremely interesting. I enjoyed reverse engineering it and learned a few things going through it. I am not sure how the TS came up with that design but it was extremely creative. Hat's off to the TS.
 

Thread Starter

iimagine

Joined Dec 20, 2010
498
I am not sure how the TS came up with that design
I have done quite a few of these triangle wave designs over the course of many years, whenever i get bored enough. Many were posted on here for comments and suggestion. The baker clamp diodes were suggested by the more knowledgeable, nice folks on here. The concept were solely based on one of my non conventional square wave generator, from there on, i constantly try to find a way to improve it. This is so far, my favoret version
 
Top