How can you enhance Sijosae splitter precision without a trimmer?

On 12/15/20 8:24 AM, Phil Hobbs wrote:

Like this.

Cheers

Phil Hobbs

------------ Version 4 SHEET 1 1344 1808 WIRE 80 -80 -112 -80 WIRE 576 -80 80 -80 WIRE 880 -80 576 -80 WIRE 880 -48 880 -80 WIRE 800 -32 32 -32 WIRE 576 0 576 -80 WIRE -112 16 -112 -80 WIRE 512 48 464 48 WIRE 880 48 880 32 WIRE 80 96 80 -80 WIRE 32 112 32 -32 WIRE 48 112 32 112 WIRE 368 128 112 128 WIRE 464 128 464 48 WIRE 464 128 448 128 WIRE 480 128 464 128 WIRE 576 128 576 96 WIRE 576 128 560 128 WIRE -112 144 -112 96 WIRE 48 144 -112 144 WIRE 576 160 576 128 WIRE 592 160 576 160 WIRE 720 160 672 160 WIRE 800 160 800 -32 WIRE 800 160 720 160 WIRE 848 160 800 160 WIRE 864 160 848 160 WIRE -112 176 -112 144 WIRE 576 176 576 160 WIRE 864 176 864 160 WIRE 720 208 720 160 WIRE 464 224 464 128 WIRE 512 224 464 224 WIRE -112 288 -112 256 WIRE -48 288 -112 288 WIRE 80 288 80 160 WIRE 80 288 -48 288 WIRE 864 288 864 256 WIRE 576 304 576 272 WIRE 592 304 576 304 WIRE 976 304 960 304 WIRE 976 320 976 304 WIRE -112 352 -112 288 WIRE 720 368 720 352 WIRE 976 432 976 400 WIRE -112 448 -112 432 FLAG 880 48 0 FLAG -112 448 0 FLAG 720 368 0 FLAG 864 288 0 FLAG 848 160 split FLAG 976 432 0 FLAG 960 304 split FLAG 880 -80 V+ FLAG -48 288 V- FLAG 592 304 V- SYMBOL voltage 880 -64 R0 SYMATTR InstName V1 SYMATTR Value 6 SYMBOL res -128 0 R0 SYMATTR InstName R1 SYMATTR Value 5k SYMBOL res -128 160 R0 SYMATTR InstName R2 SYMATTR Value 5k SYMBOL npn 512 0 R0 WINDOW 0 67 20 Left 2 WINDOW 3 35 58 Left 2 SYMATTR InstName Q1 SYMATTR Value 2N4401 SYMBOL pnp 512 272 M180 WINDOW 0 39 41 Left 2 WINDOW 3 80 17 Left 2 SYMATTR InstName Q2 SYMATTR Value 2N4403 SYMBOL res 576 112 R90 WINDOW 0 0 56 VBottom 2 WINDOW 3 32 56 VTop 2 SYMATTR InstName R3 SYMATTR Value 100 SYMBOL cap 704 208 R0 SYMATTR InstName C1 SYMATTR Value 100u SYMBOL current 864 176 R0 WINDOW 123 0 0 Left 0 WINDOW 39 0 0 Left 0 SYMATTR InstName I1 SYMATTR Value PULSE(-50m 50m 1m 50n 50n 5u 10u) SYMBOL Opamps\\opamp2 80 64 R0 WINDOW 3 19 134 Left 2 SYMATTR InstName U1 SYMATTR Value TL084 SYMBOL voltage -112 336 R0 SYMATTR InstName V2 SYMATTR Value -6 SYMBOL res 704 256 R0 SYMATTR InstName R5 SYMATTR Value 47m SYMBOL res 464 112 R90 WINDOW 0 0 56 VBottom 2 WINDOW 3 32 56 VTop 2 SYMATTR InstName R6 SYMATTR Value 1m SYMBOL current 976 320 R0 WINDOW 3 -283 177 Left 2 WINDOW 123 0 0 Left 0 WINDOW 39 0 0 Left 0 SYMATTR InstName I2 SYMATTR Value PULSE(-50m 50m 10m 50n 50n 5m 10m) SYMBOL res 688 144 R90 WINDOW 0 -12 56 VBottom 2 WINDOW 3 38 55 VTop 2 SYMATTR InstName R8 SYMATTR Value 1m TEXT 520 464 Left 2 !.tran 30m TEXT -96 480 Left 2 !* TEXT -96 490 Left 2 !*$ TEXT -96 500 Left 2 !* TL084 OP AMP "MACROMODEL" SUBCIRCUIT TEXT -96 510 Left 2 !* CREATED with PARTS REL 4.01 ON 06/16/89 TEXT -96 520 Left 2 !* (REV N/A) SUPPLY VOLTAGE: +/-15V TEXT -96 530 Left 2 !* CONNECTIONS: NON-INVERTING INPUT TEXT -96 540 Left 2 !* | INVERTING INPUT TEXT -96 550 Left 2 !* | | POSITIVE POWER SUPPLY TEXT -96 560 Left 2 !* | | | NEGATIVE POWER SUPPLY TEXT -96 570 Left 2 !* | | | | OUTPUT TEXT -96 580 Left 2 !* | | | | | TEXT -96 590 Left 2 !.SUBCKT TL084 1 2 3 4 5 TEXT -96 600 Left 2 !* TEXT -96 610 Left 2 ! C1 11 12 3.498E-12 TEXT -96 620 Left 2 ! C2 6 7 15.00E-12 TEXT -96 630 Left 2 ! DC 5 53 DX TEXT -96 640 Left 2 ! DE 54 5 DX TEXT -96 650 Left 2 ! DLP 90 91 DX TEXT -96 660 Left 2 ! DLN 92 90 DX TEXT -96 670 Left 2 ! DP 4 3 DX TEXT -96 680 Left 2 ! EGND 99 0 POLY(2) (3,0) (4,0) 0 .5 .5 TEXT -96 690 Left 2 !FB 7 99 POLY(5) VB VC VE VLP VLN 0 4.715E6 -5E6 5E6 TEXT -96 700 Left 2 ! TEXT -96 710 Left 2 ! GA 6 0 11 12 282.8E-6 TEXT -96 720 Left 2 ! GCM 0 6 10 99 8.942E-9 TEXT -96 730 Left 2 ! ISS 3 10 DC 195.0E-6 TEXT -96 740 Left 2 ! HLIM 90 0 VLIM 1K TEXT -96 750 Left 2 ! J1 11 2 10 JX TEXT -96 760 Left 2 ! J2 12 1 10 JX TEXT -96 770 Left 2 ! R2 6 9 100.0E3 TEXT -96 780 Left 2 ! RD1 4 11 3.536E3 TEXT -96 790 Left 2 ! RD2 4 12 3.536E3 TEXT -96 800 Left 2 ! RO1 8 5 150 TEXT -96 810 Left 2 ! RO2 7 99 150 TEXT -96 820 Left 2 ! RP 3 4 2.143E3 TEXT -96 830 Left 2 ! RSS 10 99 1.026E6 TEXT -96 840 Left 2 ! VB 9 0 DC 0 TEXT -96 850 Left 2 ! VC 3 53 DC 2.200 TEXT -96 860 Left 2 ! VE 54 4 DC 2.200 TEXT -96 870 Left 2 ! VLIM 7 8 DC 0 TEXT -96 880 Left 2 ! VLP 91 0 DC 25 TEXT -96 890 Left 2 ! VLN 0 92 DC 25 TEXT -96 900 Left 2 !.MODEL DX D(IS=800.0E-18) TEXT -96 910 Left 2 !.MODEL JX PJF(IS=15.00E-12 BETA=270.1E-6 VTO=-1) TEXT -96 920 Left 2 !.ENDS

TCA0372 is stable as a follower with a big aluminum or polymer or tantalum to ground. A 1-ohm series resistor can be included for the skeptical.

I love that part. I use it as a voltage regulator, a fan controller, or a relay driver. Sometimes both.

You can verify that by scoping the supply current, of course.

Yeah, I'm fond of it too. I usually use it for floating grounds of one sort or another, e.g. biasing some fast front end made with a +-5V op amp up at a higher voltage. Supplies are a bit iffy, but the L272 is very similar except for crummy max supplies, 28V vs. 40V. Not pin-compatible, unfortunately.

Cheers

Phil Hobbs

Here are a few notes. There is some question as to whether the thermal shutdown is really protective at higher supply voltages.

I think ST or someone makes an equivalent part.

What's a BFC?

One of the worst LTspice circuits on the web. You should be an example to others instead of an example of poor design:

- no labels on nodes. The node number can change if you add or modify the crcuit. Any plots will point to a different node making it impossible to decipher the circuit.

- the 1m resistors serve no purpose.

- what is the point of adding pulsed square wave current loads at 100KHz and 100Hz. This does not allow you to evaluate the distortion in the follower.

- a continuous square wave load does not represent a transient. There is no steady-state DC output.

- the model statements overlap and are impossible to read.

There is still no explanation of BFC.

The datasheet only goes to 2nF. A big capacitor would imply no phase margin. See Figure 5. Phase Margin versus Output Load Capacitance

Thanks. Looks about like my experience with it.

Yeah, ST makes the L272, a lower-voltage near-equivalent.

Cheers

Phil Hobbs

What gives you that idea?

I have been using it since the late 1960's. It has no tendency to oscillate in pure analog or digital circuits. It is wideband, low distortion, and stable as a rock.

That's why I tested it.

One would want to identify the weaknesses and improve on them; the 'wall wart' might have transients, and balance of the input capacitor values is required to keep those transients off the output. But, the output impedance as shown is five or ten ohms (and changes slightly if there's a bias current at the GND pin, because one of the output transistors turns off). So, there's limited ability to handle a ground-pin transient currents without some ground bounce.

In terms of the overall (wallwart with two wire cord feeding split power at target) problem, one can find a 48V wallwart, and use a telecom-style converter at the target to get symmetric plus/minus power. It's a tad more cost, but easy with off-the-shelf gizmos, and I got a load of +/- 12V, +5V units at surplus prices some years ago... As a bonus, such converters are useful without the Class II inputs, you can ground any pin at will (so the same converter can be used for -24V or +17V...).

The issue only arises when the poles are too close together. By making the output pole into a lead/lag network (100 uF with ~0.05 ohms ESR) the overall feedback loop becomes stable. It's the same way you stabilize a switcher that doesn't like all-ceramic output caps.

Cheers

Phil Hobbs

That's the Jim Williams Stabilization Technique: add big caps until it stops oscillating.

Lots of opamps are stable with a big modest-ESR cap to ground. Many are stable with a lot of ceramics to ground too. But the only way to be sure is to test them.

Some opamps are pretty good comparators, ditto.

Oh, come on.

1. The circuit is trivial, so I really don't expect anybody to do much with my version. Its primary purpose is to show that your worries about distortion and oscillation are misplaced. Node labels don't add anything to that. I hereby release it under the Creative Commons license, so you can add your own node labels if you like. Open source hardware and so forth. ;)

1. The 1 milliohm resistors are there to make it easy to plot the currents. You can use zero-volt voltage sources for that, but they take up too much screen space.

2. The follower's distortion is completely irrelevant, which is one of the points I'm making. What sane people expect in a rail splitter is that it splits the rail without making any drama when you bang on it a bit. This one copes with unpleasant loads with no issues. If I were going to use it in real life, I'd spend a lot more time on protection circuitry, but the basic operation is fine.

1. You don't think that a load that switches between +-100 mA, in 100-mA steps with 50-ns edges and a 1000:1 frequency range counts as a transient? G'wan.

2. If you really want to read a boring TL084 macromodel from over 30 years ago, you can open the .asc file with a text editor. I did it that way so that people would know what it was, but without causing line wrap problems or taking up all the screen space for the model.

I picked the TL084, 2N4401, and 2N4403 because if Don doesn't have _those_ in his junkbox, it must be full of rusty bolts and springs instead of electronics.

1. BFC stands for "Big Fat Capacitor", or something like that. ;)

Cheers

Phil Hobbs

Jim W was pretty smart about a lot of things. ;) He did tend to use a crapload of parts sometimes.

Scoping the supply rails to make sure there's no low-amplitude oscillation. It takes a lot to make 100 uF of ceramic cap move very fast.

Just not the ones with the diodes across the inputs. ;)

Cheers

Phil Hobbs

Oh, he was great. I met him a couple times at the Foothill Flea Market (also sadly gone) and he was nice, interested, but very shy. His two books are wonderful. But I think he never got an EE education, so he ran more on instinct than theory.

Ouch. Bad memories.

My current gumdrop, OPA197, is stable with a big cap and is a decent RRIO comparator.

Use the power pins of an op-amp to drive the transistors now organised as common emitter

In this circuit the power to run the op-amp flows through the transistor bases turning them on to assist the op-amp

As the power pins are involved you want a single op-amp in this application, as if you use dual or quad the extra sub-parts won't be usable.

The above image is taken from this blog post.

Which has lots of other good looking circuits too.

That doesn't follow. I've never had an EE education, but you can get the theory out of books and papers. If you do, you are an autodidact. Instinctive reactions are inherited - good luck with finding any that help you with electronics

There will be holes in that education - you tend to learn what you need to - but there are plenty of holes in formal electrical engineering courses.

Electronics has changed a lot since 1965 - when I started learning about it - and people who stopped learning about it as soon as they got out of university wouldn't be much use.