Emitter follower problems

[Alex_Baker]

I have been trying to construct an emitter follower power supply, but I have run into some unexpected problems. I attached a simplified schematic and scope capture.

The purpose of the circuit is to copy a control voltage, which is a 0-15V saw tooth waveform. The circuit works ok except for the 15 to zero edge. In the scope capture the red is my control voltage, and the blue is the output. As you can see the output follows the control pretty well except for the falling edge, it is quite lazy coming down from the peak. This capture is with a 220 ohm load, with less load the falling edge slope is way worse. At first I though it could be transistor leakage but there is no way it is that much leakage. I can post another scope capture later, but the base emitter junction is biased negative while falling like that.

I am not sure what I am missing here. I ran some LTspice sims, but they don't indicate any behavior like this. If somebody could school me on what I did wrong here that would be great 

[TimFox]

When the 358 tries to pull the output down, if the BE junction of the transistor is reverse-biased then there is a very high impedance in series with the 358’s output.
A cheap partial improvement:  add a reversed diode in parallel with the BE junction, although the negative-going current will still be limited by the 358 output stage.

[benseno]

Low output voltage will suffer from bjt BE induced limitation as mentioned above.
Proper way would be adding negative supply (-3.3V) and connect opamp supply and resistor to it.
Also, LM358 has limitation on maximum high output voltage which can impact this circuit too (if it expected to operate on such levels too).

[Benta]

You may also have run into the oscillation problem with emitter followers (there was a thread about it here a couple of months ago).
It can be hard to see, as it is often in the 10s MHz range.
Try adding a 1k resistor in series with the base.

EDIT: one month ago:
https://www.eevblog.com/forum/projects/1-transistor-reset-oscillates-at-100s-mhz-why-(source-bob-pease)/

[jwet]

I attached a schematic of the 358 from the data sheet.  These parts have a strange output especially on sinking current.  The output transistor is a PNP follower driven by a NPN collector.  If the NPN is saturated and pulling down hard on the base, the output can go no lower than .7V.  Note that there is a 120 uA current source shown, this to help but this isn't much current and it can't sink all the way down to ground- it has some headroom.  Often, putting an R load from the output to ground will help but fundamentally its not a great amp for this app (what do you expect for 3 cents!).  Look for an amp that has a diamond type output with a saturated NPN on the low side of the output or better yet something with a CMOS output that can sink.  The TI TLC272 is a medium voltage pin compatible jelly bean with a CMOS output- sorry its cost a few more cents.

You can also look at Fig 5.47 in the typ ops- the outputs sort of falls aparts at a about a volt.  They don't hide this but the EC table is written carefully to not advertise this.  They carefully choose conditions where this doesn't show up.


Also emitter followers themselves can't sink any more that their own pull down once the transistor is off.  The sourcing impedance is low but varies- its little re, which is 26/Ic in mA to give ohms but the sinking current at the bottom is 220 ohm.  If you have any speed or C, that's not a very stiff driver.

Others comments still apply but the root cause is the 358  output structure.

I couldn't embed the PDF piece in the document, its takes a click- its just one page.

Good luck

[Benta]

@jwet, I agree. The LM358 is not a good choice for this application, and the datasheet is tricky/misleading on this point.

[Zero999]

I am not sure what I am missing here. I ran some LTspice sims, but they don't indicate any behavior like this. If somebody could school me on what I did wrong here that would be great 

What model did you use for the LM358?

Many of the models are don't show the behaviour of the output stage very well. I've found TI's to be very good and seems to mirror your bench test quite well.



Note that this file won't simulate on LTSpice 24.1 due to a known bug. Use 24.0 or older.

Refer to the following thread for more information.
https://www.eevblog.com/forum/projects/ltspice24-external-models-syntax-error/

[magic]

Isn't it charge storage in Q1? I wonder if you could have better luck with a small MOSFET?

[Benta]

Isn't it charge storage in Q1? I wonder if you could have better luck with a small MOSFET?

No. Charge storage is phenomenon on saturated BJTs, which is not the case here.
The problem is the inability of the LM358's output to go low enough. A pull down resistor at the output might solve the problem (as indicated indirectly in the data sheet).

[Alex_Baker]

The op-amp supply is +26V and -7V, pushing the standard 30V total differential a bit but I will run with it for know.

Below is a scope capture showing a few things. Blue is emitter voltage to ground, red is base voltage to ground, and green is a math channel showing Vbe. From this it seems that the op-amp has no trouble pulling the base down.

[jwet]

This is a really different circuit, start a new thread.  You're exceeding the Veb max of a standard bipolar- you're looking at avalanche type operation very soon this might be what you're seeing.  You're a bit past the abs max of the 358.  What's the point- you've lost me.

Good luck.

[Zero999]

Please don't start another thread. There is already enough of a problem here with too many similar threads.

Add a diode in reverse parallel with the base-emitter to protect against reverse voltage.

I think LM358 is simply too slow and the slew rate is limiting the fall time. Does it work at a lower frequency?

[magic]

No need for a new thread and the problem is the transistor, not LM358. Charge storage is always a thing, not just in saturation.

I think a MOSFET may perform better, but frankly I have never investigated this case myself so it's just a guess.

A diode between base and emitter should also help, by allowing LM358 to actively pull the emitter down. This may or may not require negative supply for best results, try it and see it.

edit
I guess you could also use a PNP emitter follower instead of a diode. Base to Q1 base, emitter to Q1 emitter, collector to ground or some negative rail.

[PGPG]

No need for a new thread and the problem is the transistor, not LM358.

I'd add to it that only reason for output voltage to go down is R1.
Starting at 1V6 it 'takes' from emitter 16mA, but near slope end voltage (of your control signal) is about 0.05V and R1 'takes' from emitter only 0.5mA. You should not expect slope being the same when 16mA is drawn and when 0.5mA is drawn.

[benseno]

The op-amp supply is +26V and -7V, pushing the standard 30V total differential a bit but I will run with it for know.

Below is a scope capture showing a few things. Blue is emitter voltage to ground, red is base voltage to ground, and green is a math channel showing Vbe. From this it seems that the op-amp has no trouble pulling the base down.

This is quite interesting. Opamp negative is -7V but base voltage goes lower than -8V. Like a capacitor in charge pump circuit.
And opamp with 1MHz unity gain performs faster than 200Mhz current gain transistor.

update: oh my error - it is a red line. so i was wrong in -8V assumption.

[m k]

Maybe LCR meter can see something over R1 of non powered circuit.

[David Hess]

I will assume the Picotech DSO is performing properly since the input waveform is sharp.

How much capacitance is your probe?  Are you using x1 probes or coaxial connections?

I would probe the base signal.  Maybe it is not a 324 or it is damaged.

If the probe capacitance is the problem, then I would confirm it by adding a small capacitor to deliberately double it and see what that does.  You might need an active pull-down to get good performance at low voltage.  This would be a good application for a negative supply.

No need for a new thread and the problem is the transistor, not LM358. Charge storage is always a thing, not just in saturation.

Active pull-down from the 324, even 50 microamps, pulls excess charge out of the base.  This is not a charge storage problem.

I think a MOSFET may perform better, but frankly I have never investigated this case myself so it's just a guess.

Long ago I had the same problem with a 2N3904 emitter follower.  A 2N7000 made it worse because capacitance was higher.

[Alex_Baker]

My probes are switchable x1 x10, either mode makes no difference on the scope.

I have +- 26V available to this circuit, so I can try some voltage followers circuit with active pulldown.

What I am not understanding is why there is still current flowing out of the emitter of the transistor even when the base is reverse biased by several volts.

[Benta]

Re-read reply #5. Did you add the base resistor?

[David Hess]

What I am not understanding is why there is still current flowing out of the emitter of the transistor even when the base is reverse biased by several volts.

If you are switching the probe and there is no effect, then the base is not reversed.  Check the output of the operational amplifier.

[benseno]

What I am not understanding is why there is still current flowing out of the emitter of the transistor even when the base is reverse biased by several volts.

Yet, according to the latest graphs Emitter Base voltage enters breakdown region (<-5V). The breakdown voltage is measured with collector open.
Is it the cause of current through emitter doesn't go down to 0 instantly?

[Alex_Baker]

Re-read reply #5. Did you add the base resistor?

I tried adding a resistor from base to ground, it has very little effect. Adding any less than 2k or so just saturates the op-amp.

If you are switching the probe and there is no effect, then the base is not reversed.  Check the output of the operational amplifier.

I don't know what you mean. By no difference I meant possible capacitive or resistive loading from the probe in x1 or x10 mode.

Yet, according to the latest graphs Emitter Base voltage enters breakdown region (<-5V). The breakdown voltage is measured with collector open.
Is it the cause of current through emitter doesn't go down to 0 instantly?

I added a reverse parallel diode to limit the negative voltage swing, no noticeable change.

[sharow]

What is Q1 part number?

I can simulate your circuit behavior by BBSpice, here LM358(ST), R1=200ohm, Q1=2N2222, VCC=6V single supply.
But, I needed add more than 1000pF parasitic capacitance to do this.
I feel bit unrealistic.


no C


small C


large C


large C

So again, what is Q1?

PS: reversed diode between B-E might be help a bit.

[JonMorton]

@Alex_Baker
Did you check how the impulse looks on base of the Q1 transistor?
The LM358 has only 0.3V/uS slew rate, it may be cause of your problem. Try an op amp with >=10V/uS, for instance the inexpensive TL3472 (<$1.00).

[magic]

Yet, according to the latest graphs Emitter Base voltage enters breakdown region (<-5V). The breakdown voltage is measured with collector open.
Is it the cause of current through emitter doesn't go down to 0 instantly?

I added a reverse parallel diode to limit the negative voltage swing, no noticeable change.

Where specifically did you put a diode and how did it make no difference?

If you add a diode from Q1 emitter to Q1 base it has to make a difference. At the very least, the base should stop going a few volts negative.