This is because the variation in thickness and doping profile of the base region. When the voltage is high enough the depletion region in the base extend far enough to make electron tunneling possible. This circuit also known as Esaki oscillator.
Back in the day in 'Everyday Electronics' they published a project called "The Dream Machine" which was a transistor in this configuration and a simple amplifier to drive a small speaker and it would generate 'Pink Noise' this can aid sleep in some folks by mimicking the noise of light rain or the gentle babble of a stream. It worked exactly as described and in the theory bit it was stated that 'actual Zener breakdown' ended around the 4v area and the avalanche kicked in after that. There was also talk of experimenting with different transistors as some 'Won't work' as you clearly demonstrated !!......cheers.
Transistors are designed asymmetrically on purpose, but collector and emitter can always be interchanged. The asymmetry affects the gain: lots of gain in the normal orientation, very little gain (or even gain less than 1) when reversed. But we get something back for this gain degradation: the saturation voltage is lower. So, back in the days when bipolar transistors were the power devices in power supplies, they could be used as switches in low-voltage switching supplies that way.
Nice demo! Those who are unaware of the result of the following circuit should also try this: Turn off the +12 V source. Put a voltmeter in place of the LED. Ground the base of the transistor. You can remove or leave the capacitor at your option. Predict the voltage to be read by the voltmeter. Now turn on the +12 V source and note the collector voltage. The result can be surprising. (This is an example suggested by the late great Bob Pease to show that the spice circuit analysis program does not always give the right answer.)
Thanks for the time to put this relaxation oscillator circuit on your bench. It's been years since I've touched this circuit. Talking about knocking the rust off. Already got my breadboard out...
Back in my "Starving Student" days - I used the reverse-biased E-B junction of a 2N3904 as a ~5V Zener reference in a lab project. Worked perfectly, with no apparent damage the part !
Another good one is the Unijunction Transistor. Makes good relaxation oscillators. I don't think UJTs are made anymore but you can still find them around. Old school for sure!
From a past search, I believe they are. In the “worst case”, one can use its younger brother = PUT, programmable unijunction transistor, operating quite similar, but NOT interchangeable.
There are special avalanche transistors that use the same principle for generating pulses with very fast edges, in the picosecond range, useful for testing equipment and transmission lines. In the same league there is the "step recovery diode" another obscure microwave tech device, also for generating pulses with lots of harmonics.
This is the design fault made in many power supplies that start discharging your laptop battery if plugged into the latop but pulled out of the wall outlet during charging.
@@stefflus08Good thinking. Well, nothing limits the current. Some “cheapened” linear supplies just blow up when you back-drive them when off. Back when I used to go to T&M tradeshows, I’d do this back-drive test. Managed to blow a couple supplies up that way. “Folks, this needs some work before it hits the market”. I felt sorry for the folks in the booth, but the few potential customers who avoided a snafu were worth it.
Interesting, I did a lot of bipolar transistor characterization in the 1980 and 1990's but, never witnessed the base-emitter breakdown unto a negative resistance region. Most small signal transistors would breakdown at about 6 volts and appear as a 6 volt zener. Most bipolar transistors do not spec the base-emitter (B-E) breakdown. The transistors that were better spec'ed will show a minimum breakdown voltage. Sustained operation in the breakdown region causes an early failure. I witnessed many of the early solid state TVs improperly applying inputs to the base that would cause B-E breakdown and even some Heathkit designs that would use a reverse biased B-E in place of a small signal zener.
As Absurdengineering pointed out, you can get very low saturation voltage, when you swap the collector and emitter on certain transistors. That was used (before the MOSFET produced nearly zero saturation voltages) for turning a low DC measurement voltage to AC for purpose of greater amplification in AC stage and then again rectifying it to DC. I recall benefit of more than a decade, say from 200 mV to 2 or 5 mV saturation. I also recall there were some special transistors sold exactly for this application. Finally, I think the “best” special transistor may have been PNP, rather than NPN types. Maybe you like to try and demonstrate this feature?
You could make this much more reliable by using 2 transistors connected to be the equivalent of an SCR. Just make sure the charging current isn’t enough to keep it latched!
Heat up the transistor and the LED stops flashing )) Just will light steady. I use old Ge transistors as Heat and light sensors, attaching them into the scheme this way.
You can use a similar circuit to generate pulses with really fast rise time. I remeber using it as a hamonic generator for a frequency multiplier. Like a step recovery diode for poor people!
Only works if base is open circuit. Connect base to collector and it will act like a regular zener, and base to emitter it will be a regular diode. Long term it will degrade the base emitter junction, and the voltage will creep up, till it stops running at all. You got special transistors that were doped heavily in both junctions, so they would operate as a transistor either way, the early Ge point contact transistors, and the alloyed junction ones, would operate like this, generally having somewhat symmetrical breakdown voltages, though better processing, and making the emitter blob bigger, so it diffused to be larger, did change that. that is why you had to be careful soldering them, heat made the alloyed junctions migrate through the slice, and as they got closer the gain went up, to the point the 2 alloy diffusions met in the middle, and the gain went infinite.
My father told me that in some Sony and Tektronix equipment, they deliberately used a very specific bjt "upside down" with the emmiter and collector switched. It still functioned as a bjt, but with different characteristics. They did this because they wanted the unusual characteristics this provided. I have no experience with this, just something he told me many many years ago. (the '70s)
I have already used this circuit, as a curiosity. But your comments supported by the curve tracer are great! Another point is to know how-stable this would be in the long run. Would this current burst enabling LED blink, damage the BJT? Would this damage the transistor? Would it vary per temperature? These were some digressions about the usefulness of this kind of circuit.
One can build a really fast rising edge (few 100s of ps) pulse generator with a circuit very similar to this one by using special avalanche transistors (alas, very rare and expensive, mostly new old stock) instead of a jellybean… I've found that some 3904s also work, but have to be hand picked.
@@IMSAIGuy Oh well, didn't know about this video :). Anyway, I was referring to the well known AN72/Appendix B, it's the same thing :). Btw, there's another (cheap) way to accomplish the same - a stock SFP / SFP+ / XSFP module and its TX driver...
What about a one zener oscillator, would that have the same result? (I guess there would need to be the avalanche breakdown - I'm just too lazy to get my components to try).
@@jamesmoffat9754Avalanche breakdown is only modeled in special models made to capture that behavior. Typical SPICE models just don’t do it. For all such circuits, breadboard is king, and simulation will blatantly lie to you.
i wondered if its a thing to float/bias a capacitor like the following problem example. you got a 1000uf cap, on a 45v supply and when your load draws current pulses that the cap is there to filter/catch those, but as it has esr and finite capacitance, it drops to lets say 40v in the dips. so a 5v ripple. you just use the energy thats in the 5v range, the other stored energy that make up the 40v is unused. so what if you bias the cap to 40v on the negative lead or slightly below, then the whole capacity should be accessible to the 40...45v range isnt it? or at least you could use a cap with a lower rating (10v for example) thats either much smaller or has much more capacitance instead. i think i saw one of those arrangements in a yamaha power amplifier it had 40v and 80v rails and just two 50v caps in series so basically it had its negative lead floating at 40v. instead of using a 50 and 100v cap
No, that is just using the one rail to bias the capacitor above ground, and does result in the ripple current in the upper capacitor being also impressed on the lower rail. the energy stored in a capacitor is proportional to voltage squared, so the amount of energy stored from 40 to 45V is a lot more than the energy stored from 0V to 5V. Biasing it to 40V will just mean the other power rail has to supply the ripple current that flows between the plates. The yamaha amplifier with 2 rails and 2 capacitors is just to save cost, or, look on it as a symmetrical power supply, with a centre ground, and a +40V rail and a -40V rail. load current from + rail to - rail has ripple flowing through both capacitors, but any current to the centre is only in that rail capacitor, and on average it is no more than having 2 separate capacitors, but now you need one 50V one and one 100v one, which will be a larger can, and a separate part number. Cheaper to use 2, get economy of scale when building them by the thousands, and save on the assembly needing to stock an extra part, and on the larger board area you need, or a higher cabinet to handle the height of it.
Do any of the home brew or cheap curve tracers coupled with an oscilloscope allow us to see any of the behaviours you have been showing with your fancy Tektronix curve tracer lately?
Absolutely. The curve tracer is in many cases just a handy combination of multiple other tools. Most notably, the CC/CV PSU with multiple preset points and pairing it with the display scaling.
A basic 12V AC transformer supply output is enough to show this. Use a current sense resistor in series with the reverse-biased junction. Ground clip goes to the resistor-transistor connection. CH1 goes to the other end of the transistor and is the voltage (“x”) channel. CH2 goes to the other end of the current sense resistor and is the current readout (“y”) channel. CH2 has to be inverted to get a right-side-up plot. The ground clip must be in-between the device being characterized and the current sense resistor, not at the bottom of the sense resistor - otherwise the voltage drop on the sense resistor introduces an error in the plot. No problem of course - you paid for that “CH2 INV” button, might as well get your money’s worth :)
Does this damage the transistor over time? I put this in a circuit to flash an LED to remind me of those times I connected 12 volts. It seems to work fine but I didn't know if the transistor would eventually fail. Also, like you said it needs a high enough voltage. When I was first playing around with this on a breadboard, I only had a 9v battery handy. I thought the LED would just be dim, but it didn't flash at all.
I remember reading about using the reverse breakdown of a transistor junction to generate white noise. The author of the article mentioned to never try to use this transistor in an ordinary manner after doing this, the junction might be damaged. I read this in some electronics magazine in the '70s or '80s. I have no experience with this.
The collector-emitter breakdown voltage is 40 volts, but another way around it isn't. Emitter-collector breakdown voltage is lower. Datasheet does not necessarily include that value, but the emitter-base breakdown is about 6.5 volts.
The VBEx (base-emitter breakdown voltage) is on the order of 6-7V in many transistors; some however have it even as low as 4-5V. This is something to remember even when designing regular circuits like amplifiers, because too high a negative signal on the amplifier transistor base will lead to this sort of behavior.
This is because the variation in thickness and doping profile of the base region. When the voltage is high enough the depletion region in the base extend far enough to make electron tunneling possible.
This circuit also known as Esaki oscillator.
Back in the day in 'Everyday Electronics' they published a project called "The Dream Machine" which was a transistor in this configuration and a simple amplifier to drive a small speaker and it would generate 'Pink Noise' this can aid sleep in some folks by mimicking the noise of light rain or the gentle babble of a stream. It worked exactly as described and in the theory bit it was stated that 'actual Zener breakdown' ended around the 4v area and the avalanche kicked in after that. There was also talk of experimenting with different transistors as some 'Won't work' as you clearly demonstrated !!......cheers.
Transistors are designed asymmetrically on purpose, but collector and emitter can always be interchanged. The asymmetry affects the gain: lots of gain in the normal orientation, very little gain (or even gain less than 1) when reversed. But we get something back for this gain degradation: the saturation voltage is lower. So, back in the days when bipolar transistors were the power devices in power supplies, they could be used as switches in low-voltage switching supplies that way.
Nice demo! Those who are unaware of the result of the following circuit should also try this: Turn off the +12 V source. Put a voltmeter in place of the LED. Ground the base of the transistor. You can remove or leave the capacitor at your option. Predict the voltage to be read by the voltmeter. Now turn on the +12 V source and note the collector voltage. The result can be surprising. (This is an example suggested by the late great Bob Pease to show that the spice circuit analysis program does not always give the right answer.)
Thanks for the time to put this relaxation oscillator circuit on your bench. It's been years since I've touched this circuit. Talking about knocking the rust off. Already got my breadboard out...
Back in my "Starving Student" days - I used the reverse-biased E-B junction of a 2N3904 as a ~5V Zener reference in a lab project. Worked perfectly, with no apparent damage the part !
Another good one is the Unijunction Transistor. Makes good relaxation oscillators. I don't think UJTs are made anymore but you can still find them around. Old school for sure!
You might be able to use jfet instead of UJT.
From a past search, I believe they are. In the “worst case”, one can use its younger brother = PUT, programmable unijunction transistor, operating quite similar, but NOT interchangeable.
And it's possible to emulate a PUT with a PNP and an NPN transistor. All of those old unijunction circuits can live on if you want them to.
There are special avalanche transistors that use the same principle for generating pulses with very fast edges, in the picosecond range, useful for testing equipment and transmission lines.
In the same league there is the "step recovery diode" another obscure microwave tech device, also for generating pulses with lots of harmonics.
pulse: ua-cam.com/video/QCq1REgpBCA/v-deo.htmlsi=j8J7cGE-j6iGvZZT
step: ua-cam.com/video/pQVQIufVmJk/v-deo.htmlsi=pUVRRmtrWCz012fp
This is the design fault made in many power supplies that start discharging your laptop battery if plugged into the latop but pulled out of the wall outlet during charging.
With such a fault, what limits the current?
@@stefflus08Good thinking. Well, nothing limits the current. Some “cheapened” linear supplies just blow up when you back-drive them when off. Back when I used to go to T&M tradeshows, I’d do this back-drive test. Managed to blow a couple supplies up that way. “Folks, this needs some work before it hits the market”. I felt sorry for the folks in the booth, but the few potential customers who avoided a snafu were worth it.
Yes. The negative resistance is necessary for this oscillator to function. 8 V Zener diodes are more reliable in the base of a chopper transistor.
Interesting, I did a lot of bipolar transistor characterization in the 1980 and 1990's but, never witnessed the base-emitter breakdown unto a negative resistance region. Most small signal transistors would breakdown at about 6 volts and appear as a 6 volt zener.
Most bipolar transistors do not spec the base-emitter (B-E) breakdown. The transistors that were better spec'ed will show a minimum breakdown voltage.
Sustained operation in the breakdown region causes an early failure.
I witnessed many of the early solid state TVs improperly applying inputs to the base that would cause B-E breakdown and even some Heathkit designs that would use a reverse biased B-E in place of a small signal zener.
I just tried this one a week ago Messing up a transistor is so much fun. I used a bc547 but needed above 12 volts to let the led flash.
All against transistors abuse.
Didn’t Look Mum No Computer make 1000 audio oscillators like this for his museum?
Instead of using a diode, I guess we could also use a resistor?
That way you can make an oscillator with a more even duty cycle.
As Absurdengineering pointed out, you can get very low saturation voltage, when you swap the collector and emitter on certain transistors. That was used (before the MOSFET produced nearly zero saturation voltages) for turning a low DC measurement voltage to AC for purpose of greater amplification in AC stage and then again rectifying it to DC. I recall benefit of more than a decade, say from 200 mV to 2 or 5 mV saturation. I also recall there were some special transistors sold exactly for this application. Finally, I think the “best” special transistor may have been PNP, rather than NPN types. Maybe you like to try and demonstrate this feature?
You could make this much more reliable by using 2 transistors connected to be the equivalent of an SCR. Just make sure the charging current isn’t enough to keep it latched!
That is a really cool effect. I like your breadboard power rail PCB!!
www.pcbway.com/project/shareproject/Protoboard_Power_Connector_55084a5d.html
You're awesome, thanks!!@@IMSAIGuy
Heat up the transistor and the LED stops flashing )) Just will light steady. I use old Ge transistors as Heat and light sensors, attaching them into the scheme this way.
You can use a similar circuit to generate pulses with really fast rise time. I remeber using it as a hamonic generator for a frequency multiplier. Like a step recovery diode for poor people!
Connection wise it is dissimilar: The NPN emitter is the low side output and the base is connected to ground.
Any reason why this wouldn't work with diodes(s) only?
Only works if base is open circuit. Connect base to collector and it will act like a regular zener, and base to emitter it will be a regular diode. Long term it will degrade the base emitter junction, and the voltage will creep up, till it stops running at all.
You got special transistors that were doped heavily in both junctions, so they would operate as a transistor either way, the early Ge point contact transistors, and the alloyed junction ones, would operate like this, generally having somewhat symmetrical breakdown voltages, though better processing, and making the emitter blob bigger, so it diffused to be larger, did change that. that is why you had to be careful soldering them, heat made the alloyed junctions migrate through the slice, and as they got closer the gain went up, to the point the 2 alloy diffusions met in the middle, and the gain went infinite.
Base connection can be used to control the breakdown voltage.
My father told me that in some Sony and Tektronix equipment, they deliberately used a very specific bjt "upside down" with the emmiter and collector switched. It still functioned as a bjt, but with different characteristics. They did this because they wanted the unusual characteristics this provided. I have no experience with this, just something he told me many many years ago. (the '70s)
@@jbuchanaThat’s right. They used it to get a lower saturation voltage - it was used as a switch I bet.
Interesting!
There is no such thing in the data sheet as to how I abuse a transistor.
Neat circiut.
I have already used this circuit, as a curiosity. But your comments supported by the curve tracer are great!
Another point is to know how-stable this would be in the long run. Would this current burst enabling LED blink, damage the BJT? Would this damage the transistor? Would it vary per temperature? These were some digressions about the usefulness of this kind of circuit.
I have heard but not confirmed this sort of use/breakdown will lower the beta a lot.
How does this relate to the breakdown at much higher voltage? >100V Where the Emitter is the output ?
One can build a really fast rising edge (few 100s of ps) pulse generator with a circuit very similar to this one by using special avalanche transistors (alas, very rare and expensive, mostly new old stock) instead of a jellybean… I've found that some 3904s also work, but have to be hand picked.
ua-cam.com/video/QCq1REgpBCA/v-deo.htmlsi=Sqne0Gmw0ojUP0bw
@@IMSAIGuy Oh well, didn't know about this video :). Anyway, I was referring to the well known AN72/Appendix B, it's the same thing :). Btw, there's another (cheap) way to accomplish the same - a stock SFP / SFP+ / XSFP module and its TX driver...
my friend designs those things and uses this: ua-cam.com/video/edHw3qz3jMQ/v-deo.htmlsi=t344NZG4zrnYsHd_
What about a one zener oscillator, would that have the same result? (I guess there would need to be the avalanche breakdown - I'm just too lazy to get my components to try).
I'm going to try it on LT spice
@@jamesmoffat9754 Wonder if breakdown voltages are in the program? Maybe try a 2222 first.
@@jamesmoffat9754Avalanche breakdown is only modeled in special models made to capture that behavior. Typical SPICE models just don’t do it. For all such circuits, breadboard is king, and simulation will blatantly lie to you.
Most Zeners don’t have the negative resistance region needed for oscillation.
i wondered if its a thing to float/bias a capacitor
like the following problem example.
you got a 1000uf cap, on a 45v supply and when your load draws current pulses that the cap is there to filter/catch those, but as it has esr and finite capacitance, it drops to lets say 40v in the dips. so a 5v ripple.
you just use the energy thats in the 5v range, the other stored energy that make up the 40v is unused.
so what if you bias the cap to 40v on the negative lead or slightly below, then the whole capacity should be accessible to the 40...45v range isnt it?
or at least you could use a cap with a lower rating (10v for example) thats either much smaller or has much more capacitance instead.
i think i saw one of those arrangements in a yamaha power amplifier
it had 40v and 80v rails and just two 50v caps in series so basically it had its negative lead floating at 40v. instead of using a 50 and 100v cap
No, that is just using the one rail to bias the capacitor above ground, and does result in the ripple current in the upper capacitor being also impressed on the lower rail. the energy stored in a capacitor is proportional to voltage squared, so the amount of energy stored from 40 to 45V is a lot more than the energy stored from 0V to 5V. Biasing it to 40V will just mean the other power rail has to supply the ripple current that flows between the plates. The yamaha amplifier with 2 rails and 2 capacitors is just to save cost, or, look on it as a symmetrical power supply, with a centre ground, and a +40V rail and a -40V rail. load current from + rail to - rail has ripple flowing through both capacitors, but any current to the centre is only in that rail capacitor, and on average it is no more than having 2 separate capacitors, but now you need one 50V one and one 100v one, which will be a larger can, and a separate part number. Cheaper to use 2, get economy of scale when building them by the thousands, and save on the assembly needing to stock an extra part, and on the larger board area you need, or a higher cabinet to handle the height of it.
Negative resistance region.
Do any of the home brew or cheap curve tracers coupled with an oscilloscope allow us to see any of the behaviours you have been showing with your fancy Tektronix curve tracer lately?
Absolutely. The curve tracer is in many cases just a handy combination of multiple other tools. Most notably, the CC/CV PSU with multiple preset points and pairing it with the display scaling.
A basic 12V AC transformer supply output is enough to show this. Use a current sense resistor in series with the reverse-biased junction. Ground clip goes to the resistor-transistor connection. CH1 goes to the other end of the transistor and is the voltage (“x”) channel. CH2 goes to the other end of the current sense resistor and is the current readout (“y”) channel. CH2 has to be inverted to get a right-side-up plot. The ground clip must be in-between the device being characterized and the current sense resistor, not at the bottom of the sense resistor - otherwise the voltage drop on the sense resistor introduces an error in the plot. No problem of course - you paid for that “CH2 INV” button, might as well get your money’s worth :)
Does this damage the transistor over time?
I put this in a circuit to flash an LED to remind me of those times I connected 12 volts. It seems to work fine but I didn't know if the transistor would eventually fail.
Also, like you said it needs a high enough voltage. When I was first playing around with this on a breadboard, I only had a 9v battery handy. I thought the LED would just be dim, but it didn't flash at all.
I do think it theoretically does damage the transistor. Perhaps you're unlikely to see it happen in these circuits.
I remember reading about using the reverse breakdown of a transistor junction to generate white noise. The author of the article mentioned to never try to use this transistor in an ordinary manner after doing this, the junction might be damaged. I read this in some electronics magazine in the '70s or '80s. I have no experience with this.
Incredible !
Do you think we could use this avalanche characteristics to build a HF oscillator, using a resonant circuit ?
Yes, but does it work with a 2N4401? :)
How can the voltage go down during the curve tracer test? I thought that the curve tracer controlled the voltage and the current was measured.
look up negative resistance
Thank you! That makes sense!@@IMSAIGuy
But how 2n2222 breakdown voltage is triggered at 12v if the transistor is rated for 40v?
The collector-emitter breakdown voltage is 40 volts, but another way around it isn't. Emitter-collector breakdown voltage is lower. Datasheet does not necessarily include that value, but the emitter-base breakdown is about 6.5 volts.
it's in backwards
The VBEx (base-emitter breakdown voltage) is on the order of 6-7V in many transistors; some however have it even as low as 4-5V. This is something to remember even when designing regular circuits like amplifiers, because too high a negative signal on the amplifier transistor base will lead to this sort of behavior.
A bit Off topic but is 7:41 Not the best time for a Classic opamp Clip ;)