Positive-Negative ±5V current-boosted references

Thread Starter

cmartinez

Joined Jan 17, 2007
8,218
I need to build a ±5V voltage reference with a current output of at least 100ma (actually, the full working load in my design will be 55mA) based in the following diagram

Capture.JPG
Never mind the MAX865 (it's a single supply ±10V generator), I am going to replace it with simple 7812 and 7912 regulators connected to a switching ±15V power source. This way there will be far more power available than the meager 15mA the MAX865 is able to deliver.


I intend to boost the positive current output of the MAX6350 using the following technique.

x.jpg

The thing is that I just found out that the MAX400 has been discontinued, and I'm trying to find a good replacement for it. For that purpose, I'm considering using the LM7171. Although that chip has a much larger voltage offset than the MAX400, I'm under the impression that the 1K pot in the diagram is being used to adjust negative voltage output anyway.

Would the LM7171 be a viable replacement? Or would you suggest something else?


Thanks in advance for reading this post.
 

OBW0549

Joined Mar 2, 2015
3,566
Would the LM7171 be a viable replacement? Or would you suggest something else?
While the LM7171 would probably do the job, I think you will find that part VERY difficult to use because of its extremely high speed-- unless you take the utmost care in circuit layout and decoupling, you're apt to end up with a high-power RF oscillator rather than an amplifier. High-speed opamps are very tricky, and the LM7171 is among the trickiest of the tricky.

My preferred approach would be to use a precision, moderate-speed (i.e., "tame") opamp followed by a high-current buffer such as the LT1010, as in Linear Technology appnote AN-18 (http://cds.linear.com/docs/en/application-note/an18f.pdf). Much less likely to destroy your sanity that way...

The Linear Tech appnotes (see http://www.linear.com/designtools/doc_list.php?dt=2) are a treasure-trove of good design ideas, BTW.
 

Thread Starter

cmartinez

Joined Jan 17, 2007
8,218
While the LM7171 would probably do the job, I think you will find that part VERY difficult to use because of its extremely high speed-- unless you take the utmost care in circuit layout and decoupling, you're apt to end up with a high-power RF oscillator rather than an amplifier. High-speed opamps are very tricky, and the LM7171 is among the trickiest of the tricky.

My preferred approach would be to use a precision, moderate-speed (i.e., "tame") opamp followed by a high-current buffer such as the LT1010, as in Linear Technology appnote AN-18 (http://cds.linear.com/docs/en/application-note/an18f.pdf). Much less likely to destroy your sanity that way...

The Linear Tech appnotes (see http://www.linear.com/designtools/doc_list.php?dt=2) are a treasure-trove of good design ideas, BTW.
Excellent reference, thanks! You've probably also saved me a few more bucks that I would have had to spend on Advil pills... (and spare my liver a little bit suffering too)
I'm not sure if the LT1010 is capable of handling negative output voltages, though....
 

Thread Starter

cmartinez

Joined Jan 17, 2007
8,218
BTW, after the op-amp stage, why can't I just use a circuit similar to this one:

x.jpg

but using an NPN transistor instead of the LT1010 to handle the negative output?
 

Thread Starter

cmartinez

Joined Jan 17, 2007
8,218
Sure it can:

http://cds.linear.com/docs/en/datasheet/1010fe.pdf



Too funky. Depending on the opamp characteristics, it could oscillate. Use the LT1010.
Silly me... the spreadsheet's title says it all..
Anyway, how about if I use a simple OPA244 (which I'm already familiar with, BTW) and connect its output to the LT1010? What should I do with the LT1010 bias pin? connect it to V+ to improve stability?
 

OBW0549

Joined Mar 2, 2015
3,566
From what I can see in the LT1010 datasheet, it looks like you should just be able to leave the bias pin unconnected. In any case, don't connect it directly to V+, but rather through a resistor. As the text explains, the bias pin apparently aids negative-going slew rate: "...the rate of rise of sink current is noticeably less than for source current. This can be mitigated by connecting a resistor between the bias terminal and V+, raising quiescent current." Sounds to me like for your application, this is probably a "don't-care."

As for the OPA244, are its input offset voltage and offset voltage tempco OK for your application? If so, I should think the OPA244 would be acceptable.
 

Thread Starter

cmartinez

Joined Jan 17, 2007
8,218
From what I can see in the LT1010 datasheet, it looks like you should just be able to leave the bias pin unconnected. In any case, don't connect it directly to V+, but rather through a resistor. As the text explains, the bias pin apparently aids negative-going slew rate: "...the rate of rise of sink current is noticeably less than for source current. This can be mitigated by connecting a resistor between the bias terminal and V+, raising quiescent current." Sounds to me like for your application, this is probably a "don't-care."

As for the OPA244, are its input offset voltage and offset voltage tempco OK for your application? If so, I should think the OPA244 would be acceptable.
I've just looked at the OP07DP, and it looks like a better candidate than the OPA244, it's got a very low Vtempco (2.5µV/°C), 0.3 V/µs slew rate, and it's also offset-adjustable. Although that last feature is already taken care of by the the 1K trimpot shown in my first post.
Also, it's even cheaper than the OPA244! ($1.15 vs $1.70)
 

OBW0549

Joined Mar 2, 2015
3,566
I've just looked at the OP07DP, and it looks like a better candidate than the OPA244, it's got a very low Vtempco (2.5µV/°C), 0.3 V/µs slew rate, and it's also offset-adjustable. Although that last feature is already taken care of by the the 1K trimpot shown in my first post.
Also, it's even cheaper than the OPA244! ($1.15 vs $1.70)
The OPA277 (http://www.ti.com/product/opa277) would be even better: +/-0.1µV/°C tempco and 10µV initial offset.

The OP07 is a true "oldie but goodie"; did you take a look at the copyright dates on the datasheet?
 

OBW0549

Joined Mar 2, 2015
3,566
Nice!... and wow... 1983 uh?... I was a only a bratty teenager back then!
The OP-07 is actually older than that, even: it was introduced by Precision Monolithics, Inc. back in 1975. Forty years ago! I remember when they came out, as a major improvement over the earlier OP-05 which, when it appeared, was practically the last word in precision opamps.

Here's an interesting historical reference:

http://www.ee.bgu.ac.il/~angcirc/History/Solutions_2003_2004_B/SomeStuff/History18opamp.pdf
 

Thread Starter

cmartinez

Joined Jan 17, 2007
8,218
Would you recommend I use the LT1010 as a current booster instead of the transistor for the positive reference as well?
 

OBW0549

Joined Mar 2, 2015
3,566
Would you recommend I use the LT1010 as a current booster instead of the transistor for the positive reference as well?
Not unless you used it in conjunction with a precision opamp, like with the negative reference.

Note that in either case, positive or negative, the LT1010 needs to be inside the opamp feedback loop, so that its own offset voltage doesn't ruin your accuracy; merely using the LT1010 as a buffer amp after the opamp (i.e., outside the opamp feedback loop) would completely negate the benefits of having a precision voltage reference since it has an output offset voltage of 200 millivolts or more.
 

Thread Starter

cmartinez

Joined Jan 17, 2007
8,218
So then how exactly should I connect the LT1010 after the OPA277? Because I think I almost did that same mistake that you just mentioned.
Is there a figure in its datasheet that clearly shows how to place it inside the opamp feedback loop? Is that what the Bias pin is supposed to do?
 

OBW0549

Joined Mar 2, 2015
3,566
So then how exactly should I connect the LT1010 after the OPA277? Because I think I almost did that same mistake that you just mentioned.
Is there a figure in its datasheet that clearly shows how to place it inside the opamp feedback loop? Is that what the Bias pin is supposed to do?
No, forget the bias pin, it has nothing to do with this.

Take a look at the figure at the bottom of page 11 of the LT1010 data sheet. Notice A1, an LT118, and the LT1010, A2. A1 and A2 are being used as a current-boosted unity-gain buffer here. See how A1's feedback resistor, R2, is coming from the output of A2? That's what I mean when I say the LT1010 is "inside the feedback loop" of the opamp that's providing all the gain and the good input offset specs.

The two circuits at the bottom of page 9 also show this arrangement, one in an inverting configuration and the other in a non-inverting configuration.
 

Thread Starter

cmartinez

Joined Jan 17, 2007
8,218
No, forget the bias pin, it has nothing to do with this.

Take a look at the figure at the bottom of page 11 of the LT1010 data sheet. Notice A1, an LT118, and the LT1010, A2. A1 and A2 are being used as a current-boosted unity-gain buffer here. See how A1's feedback resistor, R2, is coming from the output of A2? That's what I mean when I say the LT1010 is "inside the feedback loop" of the opamp that's providing all the gain and the good input offset specs.

The two circuits at the bottom of page 9 also show this arrangement, one in an inverting configuration and the other in a non-inverting configuration.
Ha!... that's exactly what I did prior to getting your post! I placed a 2k2 resistor from the LT1010 output back into the opamp's negative input, and a 1nf cap. Exactly as shown in page 9 bottom right figure. That is, I used a follower configuration... See? I can (accidentally) get the right answer every once in a while... :D
 

Thread Starter

cmartinez

Joined Jan 17, 2007
8,218
Alright, here's where things are standing so far...

The reason why I wanted to make a boosted-current ±5V reference is that I want to use it to excite a load cell with it, by generating a 400Hz AC square wave for that purpose. I was going to use a mosfet H-bridge for that purpose. But then I found the following circuit:

135-9498.png

I simmed it in LTspice and it worked like a charm. Also, I chose to use the OPA277, since we've already decided it's the best OpAmp for this application. I'm using the OP07 in the sim because that's what I have available in LTspice's library.

The thing that got me stuck for a moment, is that approx 1V is lost in the square wave generator. But then I remembered that I can excite the load cell with up to ±7V if there's need, so I'm sure ±6V won't damage it, and maybe resolution will also improve a little bit in the end. That made me decide to use an LT1021-7 reference, which according to its datasheet, it's also the best of the line, with long term stability and all.

upload_2015-6-5_14-47-47.png

What I liked about this design, is that both positive and negative current is boosted with a single LT1010, and I did away with the 2N2905 transistor arrangement that I intended to use for the positive stage.

The only thing that I'm a little concerned about is that I'm not sure if I connected the LT1010's output properly, configuring it as a follower. The original figure says that it should be connected through a resistor to the negative input of the OpAmp that's feeding it, but that input is connected to ground already! I guess the load cell's resistance is already taking care of that feedback loop?
 

Attachments

OBW0549

Joined Mar 2, 2015
3,566
Uh... no.

If you're going to use that opamp oscillator circuit to generate the square wave drive for the strain gauge, what's the point of having a precision voltage reference???? HINT: look at the OPA277's "output voltage swing" spec on the datasheet. What is it telling you? 2ND HINT: Never, never, never, **NEVER** rely on an opamp's output saturation characteristics to give you precise voltage levels.

Bad juju. VERY bad juju.
 

Thread Starter

cmartinez

Joined Jan 17, 2007
8,218
Uh... no.

If you're going to use that opamp oscillator circuit to generate the square wave drive for the strain gauge, what's the point of having a precision voltage reference???? HINT: look at the OPA277's "output voltage swing" spec on the datasheet. What is it telling you? 2ND HINT: Never, never, never, **NEVER** rely on an opamp's output saturation characteristics to give you precise voltage levels.

Bad juju. VERY bad juju.
OUCH! :confused::(
Let me first pick the pieces and sweep the ashes of my destroyed pride off the floor, and I'll be right back....
 

Thread Starter

cmartinez

Joined Jan 17, 2007
8,218
Alright... first, I want you to know that I'm very grateful for all you've helped me with so far, and also I thank you for pointing out my huge mistake in my last design. You've probably saved me a lot of headaches already.

Now, here's what I plan to do:
I'm going to use a crystal oscillator to switch a mosfet half h-bridge. I intend to run that oscillator at 420Hz, which is a multiple of 60, so that maybe that will minimize noise picked up from mains. What do you think?
For the h-bridge, I'll use a MIC4428ZN chip to directly drive the load cell, since it seems perfectly capable of supplying the current needed, and it also says here, in page 41, that that's not only possible, but also convenient. Probably an important detail is to guarantee dead time between mosfet switching to minimize current, but that's something that can easily be arranged.

Is that better?
 
Top