EEVblog Electronics Community Forum
General => General Technical Chat => Topic started by: mrnuke on October 05, 2017, 06:52:44 pm
-
I was playing around with a low-side current sense circuit, and noticed that even with a single 5V supply, the output of an LM358 can swing very close to GND. The datasheet value for low-level output voltage is a max of 20mV. And I whack it on a breadboard, and yes, it easily swings very very close to GND.
How can this be? The functional block diagram shows the output stage to be a complimentary emitter follower, which I would expect to swing to be around 1V above ground -- V_BE drop of output PNP transistor and V_CE saturation voltage of NPN driver transistor.
(https://www.eevblog.com/forum/chat/how-does-the-lm358-achieve-a-low-level-output-voltage-of-20mv-with-single-supply/?action=dlattach;attach=357636;image)
-
The PNP transistor at the output should be able to swing down to its VCE(sat) saturation value, which is less than the VBE DC value.
-
The PNP transistor at the output should be able to swing down to its VCE(sat) saturation value, which is less than the VBE DC value.
Wouldn't that require the base of the PNP to swing below ground in order to properly bias the BE junction and bring the transistor into conduction?
-
I have been caught by this. It can only sink a very low current, I guess around 50 uA via the current regulator.
-
I have been caught by this. It can only sink a very low current, I guess around 50 uA via the current regulator.
Yes, the output characteristics in the data sheet span many orders of magnitude. Once the output is loaded with anything more than a trivial load the saturation voltage jumps right up.
-
The PNP transistor at the output should be able to swing down to its VCE(sat) saturation value, which is less than the VBE DC value.
Nope, the PNP is an emitter follower, so it can't saturate.
The secret is that 50uA current sink... but I don't think the schematic has been published.
I have been caught by this. It can only sink a very low current, I guess around 50 uA via the current regulator.
It's common to add a pulldown resistor, if you need a little more sink current.
-
The secret is that 50uA current sink... but I don't think the schematic has been published.
It's likely just a current mirror, nothing special. If you want to reverse engineer it, here are die photos:
http://vintageteardown.com/die-photo/lm358-dual-op-amp/ (http://vintageteardown.com/die-photo/lm358-dual-op-amp/)
The current sinking issue is spelled out clearly in the datasheet. Refer to "Output Current" specification.
Also, the plot of current sinking characteristics:
(https://www.eevblog.com/forum/chat/how-does-the-lm358-achieve-a-low-level-output-voltage-of-20mv-with-single-supply/?action=dlattach;attach=357687)
-
The 50uA sink plays nicely with conventional "single supply" circuits, where you're using loads referenced to ground anyway. For such loads, the NPN follower is almost always active. The only time it's not, is during relatively fast changes, especially with reactive loads: then the PNP will have to pull the output down, with concomitant crossover distortion when it turns on and off (class C output stage).
Real RRIO amps (not just "single supply"), like TLV2372, have excellent saturation characteristics, to both rails, and don't suffer from crossover distortion. They may not be as cheap, but they're also not as... cheap. ;)
Tim
-
Yes, it's as simple as a current mirror connected to the output. A similar trick can be used with any op-amp. I've done it with the TL072 but don't expect it to work in SPICE, using the standard models, as the output stage is simplified.
https://www.eevblog.com/forum/projects/better-ltspice-tl072-model/msg999066/#msg999066 (https://www.eevblog.com/forum/projects/better-ltspice-tl072-model/msg999066/#msg999066)
-
The 50 microamp pull down is just the output from a current mirror as shown in the ON Semiconductor schematic below as Q10. This schematic also shows the transconductance reduction in the input stage which allows for a much smaller compensation capacitor, 5 instead of 30 picofarads, yielding a smaller die for lower cost.
The only time it's not, is during relatively fast changes, especially with reactive loads: then the PNP will have to pull the output down, with concomitant crossover distortion when it turns on and off (class C output stage).
I do not know why the 324/358 ended up with a class-B output stage. Maybe it was for cost reasons or to avoid a patent. When they came out, the Fairchild versions (single/dual/quad uA799/uA798/uA3503) included diodes or a single transistor Vbe multiplier to remove the dead zone and were advertised as having a class-AB output stage. The Fairchild schematics do not show the 50 microamp current sink on the output either but I assume it is there because they are advertised as sinking current to ground and the specifications support that; I get the feeling that they were hiding how they managed it.
Real RRIO amps (not just "single supply"), like TLV2372, have excellent saturation characteristics, to both rails, and don't suffer from crossover distortion. They may not be as cheap, but they're also not as... cheap. ;)
Not just rail-to-rail output operational amplifiers achieve this. The LT1013/LT1014 from Linear Technology are improved replacements for the 358/324 which use an active NPN pull down for much better negative saturation characteristics. They also have a class-AB output stage. At one point I was using quite a few of them because they were the least expensive dual precision operational amplifier available in a standard pinout.
-
I didn't realize a 50 uA drain could bring the output so close to zero. I did try putting a pull-up on the output, and sure enough we went up in voltage a bit. However, for my case, which is a low-side current sense amp, the LM358 works excellent, so I'll stick with it.
Thank you everyone for the insights!
-
Keep in mind that that 50uA current sink is always present and constant and is not modulated by the previous stages like the class-B output stage is. It just serves as a continuous 50uA pull-down and it adds a constant 50uA to the quiescent current. It is exactly as if an external 50uA pull-down based on an NPN current mirror was used.
The improved LM358 that Linear Technology designed, the LT1013, lacks the 50uA current sink pull-down and instead uses a class-AB all NPN quasi-complementary output stage where the lower NPN transistor is actively driven so it can sink considerably more current at low output voltages without affecting the quiescent current and 50uA down to about 50mV which is 10 times better than the LM358. The frequency compensation of just that part of the design has more capacitance than the entire rest of the operational amplifier so it is a large expensive chip although one of the cheapest "precision" amplifiers available.