LM6171: How to keep the op amp off its saturation region?

Hi Yatao,

I don't think that this will work, because this would mean to short circuit the output to the rails.

One method to prevent saturation is to clamp within the feedback loop. In your case a lot of diodes would be needed, which does not look very realistic:

yatao1.TSC

Another way is to use low capacitance LEDs:

yatao2.TSC

Kai

Hi Kai,
Somehow I didn't receive the notification email, so I thought you haven't replied me. Sorry for the delay.
As to your second suggestion, different from a normal integrator whose '-' terminal voltage stabilizes to 0 Volt, the '-' terminal voltage is very large when stable, bringing some trouble to my control goal.
I wonder if you are sure my proposal won't work. In my opinion, if LM6171's output stage is comprised of push-pull BJTs, then when I use the clamping diodes, their BJTs's Vce are about 2V or 3V, causing them to work in amplification region. This link http://www.analogzoo.com/2014/12/op-amp-recovery-and-slew-rates/, gives an example of clamping op amp's output with a diode. If this is true, then I think this is worth a try.

I come up with another circuit like the one below. 15V and 10V TVS are used and each of them are in series with a diode to avoid the TVS cause short circuit.

If you're sure that these won't work, are there other convenient alternatives?(Is my application requirement very unique?)

Regars
Yatao

Hi Sean,

Thanks to you and Kai, and I decide to use 2 Zener diodes to prevent op amp’s saturation.

Just to confirm, I plan my configuration as the schematic below, with one diode’s Zener voltage 12V and the other 8V. Meanwhile, the input voltage Vin has magnitude of +3V and -2V. Then, as the schematic shows, this integrator’s output voltage will be within -10V and +15V and the op amp will operate away from the saturation region. Am I right?

PS: I will choose low-capacitance zener diode. But I wonder if TVS diode is more suitable, considering TVS’s switching time is less than 1 picosecond?

Regards

Yatao

Hi Yateo,

clamping the output of LM6171 to signal ground by the help of zener diodes or to a bias voltage by the help of small signal diodes causes an output current of more than 80mA! Driving the output into the short circuit current limit of OPAmp also causes a sort of saturation and I don't think that doing this is a good idea.

Using zener diodes in the feedback path sounds to be a good idea, but unfortunately the junction capacitance of a standard small current zener diode is rather high, which will affect the integration capacitance. Also, the junction capacitance is not linear with the applied voltage. Because of this I suggested the series circuit of many BAV99. The effective capacitance of such a series circuit is much smaller than your integration capacitance and doesn't play any role. Even a series circuit of LEDs can be useful. So, if you try zener diodes look for variants with very low junction capacitance.

Using a TVS might not be a good idea. They usually have a very short turn-on time, but suffer from a long turn-off time. Look for a TVS with little junction capacitance and short reverse recovery time.

Kai

Hi Yatao,

I have made a simulation with your proposed low capacitance zener diodes:

yatao3.TSC

Kai

Hi Kai
I'm very thankful for your kindness to help me.
In fact, I haven't decided to use your two proposals, just because that my application requires +15V and -10V, and errors of several hundreds of mV is permitted. However, your proposal gives -9V and 13.38V, which is just too far away.
The two diodes I proposed are limitted by 100mW(maximum power rate) and can not be used here. I tried hard and find Panasonic's(Japan) low capacitance zener diode industrial.panasonic.com/.../DD3X062J0L_E.pdf, which can be connected to form the right zener voltage.
I really don't want to bother you. But can you please tell me how to use regular diode(BAV99 and so on) to create +15V and -10V. I'm very careful with the design because if the PCB I make fails to work, then I'll need to modify it and make a new version PCB, which takes too long time.

Regards
Yatao

Hi Yatao,

I would give this scheme a try:

I think the BZX84B11 and BZX84B7V5 can be used here. Please note that I have taken the THS4631 again.

Kai

Hi Kai,

You are so nice.

Just now, I was about to use several LEDs in series to clamp the output. Then you give new hope to achieve my goal with fewer diodes.

I’ve tried TINA simulation with BZX84B11 and BZX84B7V5 with lower integration resistance, 200ohms. The result is as Fig1. I’d give some analysis here.

Fig1

As you can see, at the input step, the output rises sharply due to the input difference and the inverting node rises at the same speed since BZX84B7V5’s clamping. Then at cursor a, input difference reaches zero. After cursor a, the integration resistor’s current begins to discharge BZX84B7V5 and the process is the short interval between cursor a and b, during which VM1 and Vout rises with same low speed. At cursor b, BZX84B7V5 no longer serves as a 8V dc supply but a voltage-dependent capacitor. Then this integrator comes back to the behavior of a normal one, with no zener diodes. If the plateau process between cursors is seen as the desaturation of BZX84B7V5, then we are actually using BZX84B7V5’s saturation to replace op amp’s saturation. Luckily, the former one features a much smaller recovery time, approximately 9ns. Beside, when input drops to -2V, the plateau for the decreasing process lasts 20ns.

You emphasize the use of THS4631, which leads me to try LM6171. The simulation result with 200ohms integration resistance is as Fig2.Its performance is slightly worse than THS4631. For example, inverting node’s voltage suffers from an obvious overshoot.

By the way, I’ve always doubted that, to what extent can we trust TINA’s result. Because as you see, the integrator’s desaturation delay, which is obvious in my experiment, never appears in the simulation results.

Best regards

Yatao

Hi Yatao,

your analysis is correct. Zener diodes are not optimal for fast switchings. But when using a BAV99 instead of the forward biased zener diode, you can improve the performance a bit:

And when additionally decreasing the integrating resistance, the performance can be furtherly optimized:

yatao5.TSC

Kai

Hi Kai，
You already have given me a lot of choices and I'll try to pick one from them.
I've learned a lot from you guys, what can I do to express my thanks to you?

Regards
Yatao