Discrete transistor circuit to generate variable PPM signal?

[Refrigerator]

I've been searching high and low and couldn't for the life of me find a discrete PPM circuit, neither could i come up with one that would generate a simple two channel R/C PPM.

I want to go discrete and as little components as possible just as a challenge and an exercise in learning.

What I've tried to do so far was to extend the simple two transistor multi-vibrator but it seems like after three transistors it stops working as intended, i then tried thinking of SCSs and different kinds of oscillators but still no cigar.

So here i am asking for a little help in solving this puzzle.

Thanks.
 

[uncle_bob]

Hi

Just to check before this spirals off in the wrong direction.

PPM = pulse position modulation
PWM = pulse width modulation

Are you sure you want pulse position? PWM is a lot more common these days.

Bob

[Refrigerator]

Hi

Just to check before this spirals off in the wrong direction.

PPM = pulse position modulation
PWM = pulse width modulation

Are you sure you want pulse position? PWM is a lot more common these days.

Bob

As i mentioned i need two channel PPM, a pair two short 0.5 to 1.5 mS variable pulses with a longer pulse to seperate each pair.
Seems easy in theory but has proven to be reasonably difficult in practice.
Although i have seen a  circuit that takes a PWM signal and turns it into PPM but there's nothing to learn from replicating a circuit seen on the internet.

[Mechatrommer]

simple two transistor multi-vibrator but it seems like after three transistors it stops working as intended..

with high probability that you wont be able to make it in 2 or 3 components... with exception.. one word one component.... pic10f206... or... atTiny13a.. soic package is fun...

[uncle_bob]

Hi

The standard approach is to generate PWM first and then pull the PPM pulses off the edges of the PWM. A transistor for each edge should do the trick once you have the PWM.

Bob

[rfeecs]

Again you can go back to the old 1970's RCs. 
A schematic of a 5 channel Heathkit transmitter is attached.
It looks like the free running multivibrator drives a sort of shift register that generates periods of different adjustable lengths.
The edges of those pulses drive another monostable that outputs a fixed length pulse on each edge.

[Muxr]

For a discrete design you'd probably want to use op-amps: http://www.engineersgarage.com/tutorials/diy-circuit-design-pulse-position-modulation
http://www.engineersgarage.com/tutorials/diy-circuit-design-pulse-position-demodulation

[Refrigerator]

For a discrete design you'd probably want to use op-amps: http://www.engineersgarage.com/tutorials/diy-circuit-design-pulse-position-modulation
http://www.engineersgarage.com/tutorials/diy-circuit-design-pulse-position-demodulation

Thanks, but although functional the circuits you linked were quite over-complicated in my opinion, since while playing around in LTspice i managed to make a circuit with the same functionality and using only two transistors, it's also variable from 0.5 to 1.5 ms with a 10k pot, i also need to make a more than one channel PPM which is where most of the difficulty comes from.

Again you can go back to the old 1970's RCs. 
A schematic of a 5 channel Heathkit transmitter is attached.
It looks like the free running multivibrator drives a sort of shift register that generates periods of different adjustable lengths.
The edges of those pulses drive another monostable that outputs a fixed length pulse on each edge.

I think it's funny how it's called digital proportional although it's totally analog but it's really cool to see just how few components it has.

Somebody should turn these circuits into kits, much like the walkie-talkies and discrete 555s, and silkscreen every waveform at each test point on the PCB, it would also be compatible with standard RC stuff.

Also what i've been thinking about were the SCS and i thought of cascading them in such a way that one SCS would set another SCS while each SCS would reset itself through an RC circuit that would determine my ON time while there would be another RC after that so it would go to the set pin on the following SCS, just to set the delay between the pulses.
Sounds like a dog's breakfast but i hope you get the idea.

[uncle_bob]

Hi

Much cheaper / easier to do with a low pin count MCU if you are building this as a "for sale" product. Your connectors and packaging each will cost more than the only chip on the board....One board to stuff and stock. Turn them into whatever you need when the order comes in.

Bob

[Refrigerator]

Hi

Much cheaper / easier to do with a low pin count MCU if you are building this as a "for sale" product. Your connectors and packaging each will cost more than the only chip on the board....One board to stuff and stock. Turn them into whatever you need when the order comes in.

Bob

Not a product, just a project. Although a MCU would be an easier solution it's not what i want since i like a little challenge and it's cool to see just with how few transistors can i pull this off with. Seems like everything today is pretty much done with MCUs.

[uncle_bob]

Hi

Well, be sure to count proper output buffering in your totals .... It's only a solution once it drops into the system.

Bob

[Zero999]

For a discrete design you'd probably want to use op-amps: http://www.engineersgarage.com/tutorials/diy-circuit-design-pulse-position-modulation
http://www.engineersgarage.com/tutorials/diy-circuit-design-pulse-position-demodulation

Thanks, but although functional the circuits you linked were quite over-complicated in my opinion, since while playing around in LTspice i managed to make a circuit with the same functionality and using only two transistors, it's also variable from 0.5 to 1.5 ms with a 10k pot, i also need to make a more than one channel PPM which is where most of the difficulty comes from.

Please post a schematic.

[Refrigerator]

For a discrete design you'd probably want to use op-amps: http://www.engineersgarage.com/tutorials/diy-circuit-design-pulse-position-modulation
http://www.engineersgarage.com/tutorials/diy-circuit-design-pulse-position-demodulation

Thanks, but although functional the circuits you linked were quite over-complicated in my opinion, since while playing around in LTspice i managed to make a circuit with the same functionality and using only two transistors, it's also variable from 0.5 to 1.5 ms with a 10k pot, i also need to make a more than one channel PPM which is where most of the difficulty comes from.

Please post a schematic.

Here's a screenshot of the setup on LTspice:
https://gyazo.com/32f5b2ae0eee7b2c8abb10b3236ac8de
The measurement was taken at the collector of Q2 and is controlled with resistor R3, which is a 5k resistor in series with the 10k pot. With this circuit the on pulse can be changed from around 0.5 to just over 1.5 mS, verified by changing R3 value from 5k to 15k (full travel of the pot).
Also the signal can be squared up by adding another transistor as a buffer on the collector of Q1.
Edit: like this (collector of Q3): https://gyazo.com/ca61cadbd90d8d1ddfa642509976b3f2

[alsetalokin4017]

But your 2- and 3- transistor circuits are PWM circuits, not PPM circuits, aren't they? They don't have the same functionality as the op-amp based PPM circuit in Muxr's link.

And adjusting R3 changes the frequency as well as the pulse width, yes? Because the pulse HI width changes but the pulse LO width remains (nearly) constant as R3 is adjusted.

So now to achieve PPM, you need a stage that will generate a constant width pulse at the trailing edge of every PWM pulse produced by your 3-transistor circuit, right? Something like a monostable pulse generator triggered by the edges of the PWM pulses your circuit makes. Varying R3 would then vary the position of this constant width pulse in the pulse train.

Or am I missing something......  ?

[uncle_bob]

Hi

Indeed, what you have described there is not quite true PPM. If that meets your needs, then it's a fine way to go. Be sure to add the buffering or your load will mess things up big time.

Bob

[Refrigerator]

Hi

Indeed, what you have described there is not quite true PPM. If that meets your needs, then it's a fine way to go. Be sure to add the buffering or your load will mess things up big time.

Bob

The Q3 should be plenty for the relatively high impedances it'll be driving.

But your 2- and 3- transistor circuits are PWM circuits, not PPM circuits, aren't they? They don't have the same functionality as the op-amp based PPM circuit in Muxr's link.

And adjusting R3 changes the frequency as well as the pulse width, yes? Because the pulse HI width changes but the pulse LO width remains (nearly) constant as R3 is adjusted.

So now to achieve PPM, you need a stage that will generate a constant width pulse at the trailing edge of every PWM pulse produced by your 3-transistor circuit, right? Something like a monostable pulse generator triggered by the edges of the PWM pulses your circuit makes. Varying R3 would then vary the position of this constant width pulse in the pulse train.

Or am I missing something......  ?

From what i understand what matters is the ON time and the off time is not critical, it's pretty much the same as in the circuit rfeecs attached to his reply.
The signal in that circuit seems to have very short off time, which for me seemed a bit unusual but the engineers who made it probably know more that i do so who am i to judge.

Edit: i just tried the circuit out on a breadboard and it seems to work very well, although i replaced the 5k resistor to a 10k since the ON time was a bit too short. I also noticed while turning the power supply up and down that the signal stays very constant.

This little test gave me alot of new ideas for the circuit, i think i could modify this circuit to give me a 2ch PPM.

[orolo]

Hm, how about a modified LED chaser? Pass through a schmitt trigger to get nicer edges. Trim the base resistors to modulate pulse lenght. If I'm not mistaken, it is essentially what you were aiming for: taking the bistable multivibrator one further step.

[Refrigerator]

Hm, how about a modified LED chaser? Pass through a schmitt trigger to get nicer edges. Trim the base resistors to modulate pulse lenght. If I'm not mistaken, it is essentially what you were aiming for: taking the bistable multivibrator one further step.

I have tried this exact circuit and as i mentioned it stops working if more than 3 transistors are added, i've even used this circuit to drive three phase BLDC motors before.
Although what you have there looks almost like 2CH PPM signal, just needs some delay between the ON signals.

Now i'm thinking to replicate the circuit i mentioned in my previous reply and change the long delay in between the ON pulses and mix the signal, i could also make them oscillate in sync (trigger at the same time i guess ?).

A bit hard to explain the rush of ideas i'm having, might have to make a little doodle in paint to help get the idea across.

Edit: just doodled it : https://gyazo.com/7d4d5ad4ddac7c9b275eeb5d1830c7e9
The first transistor would give the initial pulse, each delay would cause the long low signal but would be slightly different thus separating the ON signals (i guess).

[rfeecs]

I thought it might be useful to point out that what we are talking about is PPM for a radio control.  PPM has been used for radio control for at least 40 years.  It has recently been replaced by newer systems but is still used for very simple cheap radio controls.

The PPM timing is standardized.  This page explains it pretty nicely:
http://skymixer.net/electronics/84-rc-receivers/78-rc-ppm-signal

The nice thing about PPM is it is very easy to decode.  Just use a serial-in parallel-out shift register.  Clear the shift register if no pulse is received for the blanking/sync time.  Then clock the shift register.  The servo PWM signals appear at the parallel outputs of the shift register.

For the transmitter, you have the example from the Heathkit I posted earlier in the thread.  Your approach of a ring oscillator would probably work as well, but you need to pick off say the rising edge of each pulse and wire-or them together (you can see this on the Heathkit schematic where they capacitively couple the pulses through diodes and tie them all together).  You might want to then use that to trigger a monostable to give you a nice pulse on each edge.

[orolo]

Hm, how about a modified LED chaser? Pass through a schmitt trigger to get nicer edges. Trim the base resistors to modulate pulse lenght. If I'm not mistaken, it is essentially what you were aiming for: taking the bistable multivibrator one further step.

I have tried this exact circuit and as i mentioned it stops working if more than 3 transistors are added, i've even used this circuit to drive three phase BLDC motors before.
Although what you have there looks almost like 2CH PPM signal, just needs some delay between the ON signals.

It works for odd numbers of transistors, not only three. For example, in page 75 of http://www.talkingelectronics.com/projects/200TrCcts/1-100TransistorCircuits.pdf  you have a 5 transistor example. Altough I have not implemented it myself, I don't see why it souldn't work. The FET variation in page 76 is interesting too. Once you have 5 or 7 stages, it seems only a matter of combining whatever emitter outputs you need via a DTL OR gate. The discarded outputs should give a time delay.

Another idea that comes to mind is using PUTs (or UJTs if you can get them) to do the switching work. You could use a npn-pnp pair instead of a PUT, but in my experience that configuration is rather flaky.

[Refrigerator]

Hm, how about a modified LED chaser? Pass through a schmitt trigger to get nicer edges. Trim the base resistors to modulate pulse lenght. If I'm not mistaken, it is essentially what you were aiming for: taking the bistable multivibrator one further step.

I have tried this exact circuit and as i mentioned it stops working if more than 3 transistors are added, i've even used this circuit to drive three phase BLDC motors before.
Although what you have there looks almost like 2CH PPM signal, just needs some delay between the ON signals.

It works for odd numbers of transistors, not only three. For example, in page 75 of http://www.talkingelectronics.com/projects/200TrCcts/1-100TransistorCircuits.pdf  you have a 5 transistor example. Altough I have not implemented it myself, I don't see why it souldn't work. The FET variation in page 76 is interesting too. Once you have 5 or 7 stages, it seems only a matter of combining whatever emitter outputs you need via a DTL OR gate. The discarded outputs should give a time delay.

Another idea that comes to mind is using PUTs (or UJTs if you can get them) to do the switching work. You could use a npn-pnp pair instead of a PUT, but in my experience that configuration is rather flaky.

I noticed that too when i was playing around with LTspice yesterday.
Edit : after a little playing around in LTspice i got something that looks promising !
https://gyazo.com/dfa33df1e59d8ff6bfbb01e9503ddac9

[orolo]

Well, how about this: a square wave clock triggers four cascaded monostables. The 500n caps at the emitters generate slow ramps which, when passed to the OR gate, cause delays between pulses. Besides, the 500n stabilize the cascade, which otherwise would respond to both edges of the clock. This is quick and dirty, but uses rather few transistors. There is some trimming to be done with this circuit for it to work to spec, but I guess it should work. By the way, reading rfeecs' reference on PPM signals, for two channels would you need three short pulses?

[uncle_bob]

Hi

The more parts that get tossed into this, the more complex it is getting. The alternative solution is $0.58 at Mouser. The PIC with essentially zero added parts will blow any of the multi transistor approaches away.

Yes, I know, it's not as much of a challenge. When the challenge solution gets more complex / higher power / lower accuracy / higher cost and less repeatability ..... I question the validity of the solution.

Bob

[Refrigerator]

Well, how about this: a square wave clock triggers four cascaded monostables. The 500n caps at the emitters generate slow ramps which, when passed to the OR gate, cause delays between pulses. Besides, the 500n stabilize the cascade, which otherwise would respond to both edges of the clock. This is quick and dirty, but uses rather few transistors. There is some trimming to be done with this circuit for it to work to spec, but I guess it should work. By the way, reading rfeecs' reference on PPM signals, for two channels would you need three short pulses?

Thanks, i guess i went a bit to the wrong direction, your approach seems much better, the signal there looks like my PPM decoder ( also discrete transistors ) could handle it, although the decoder i have now is only 2CH and the signal there looks like 3CH so i'll have to make another one, but that's no problem.
If all this fails i'll give the Heathkit transmitter circuit a try since it's guaranteed to work.

Hi

The more parts that get tossed into this, the more complex it is getting. The alternative solution is $0.58 at Mouser. The PIC with essentially zero added parts will blow any of the multi transistor approaches away.

Yes, I know, it's not as much of a challenge. When the challenge solution gets more complex / higher power / lower accuracy / higher cost and less repeatability ..... I question the validity of the solution.

Bob

I have a couple Attiny 45s and i could just program it to do what i need but i won't feel satisfied afterwards.
I want to learn and doing things the hard way increases the chance of failure thus increasing the rate of learning.
 

[orolo]

Thanks, i guess i went a bit to the wrong direction, your approach seems much better, the signal there looks like my PPM decoder ( also discrete transistors ) could handle it, although the decoder i have now is only 2CH and the signal there looks like 3CH so i'll have to make another one, but that's no problem.
If all this fails i'll give the Heathkit transmitter circuit a try since it's guaranteed to work.

I built it with three pulses because I though two channels = three pulses. With only two pulses you need one less transistor  . Since I was in a hurry I did the or gate with mosfets; a bipolar one would be much more old-style. Anyway, thanks for the interesting discussion! If you get a working prototype let me know.