FAQ
TL;DR: LM2900 dates to 1972; “Check the LM139 LM239 LM339 quad comparators.” Build the 0–1 V and 4–5 V detector with two standard comparators, not Norton amps. [Elektroda, Anonymous, post #21684581]
Why it matters: This FAQ helps designers quickly choose reliable, low-cost parts and avoid LM2900 input pitfalls for dual-range window detection.
Quick-Facts
- Two op-amps/comparators can implement the 0–1 V and 4–5 V windows with minimal parts. [Elektroda, Giovanni Di Maria, #21684568]
- LM2900/LM3900 use Norton current-differencing inputs; avoid direct voltage-drive. [Elektroda, Anonymous, post #21684579]
- Sub‑5 V Zeners have soft turn‑on at low current; thresholds drift. [Elektroda, Elizabeth Simon, #21684573]
- LTspice accepts vendor SPICE models; download from component makers. [Elektroda, Anonymous, post #21684575]
- LM339 is a quad comparator that suits voltage window tasks well. [Elektroda, Anonymous, post #21684581]
Quick Facts
- Two op-amps/comparators can implement the 0–1 V and 4–5 V windows with minimal parts. [Elektroda, Giovanni Di Maria, #21684568]
- LM2900/LM3900 use Norton current-differencing inputs; avoid direct voltage-drive. [Elektroda, Anonymous, post #21684579]
- Sub‑5 V Zeners have soft turn‑on at low current; thresholds drift. [Elektroda, Elizabeth Simon, #21684573]
- LTspice accepts vendor SPICE models; download from component makers. [Elektroda, Anonymous, post #21684575]
- LM339 is a quad comparator that suits voltage window tasks well. [Elektroda, Anonymous, post #21684581]
What’s the simplest way to detect 0–1 V and 4–5 V windows?
Use two comparators set as window detectors. One triggers between 0–1 V, the other between 4–5 V. This approach is stable and easy to scale. It beats transistor-only tricks for repeatable thresholds. “Much easier to use two op-amps.” [Elektroda, Anonymous, post #21684571]
Can I do it with just a couple of transistors or MOSFETs?
Yes, but thresholds vary with device parameters, temperature, and Vth spread. That makes calibration painful across many units. Designers in the thread recommended moving to comparator-based windows for consistency and fewer surprises. [Elektroda, Anonymous, post #21684571]
Why did my LM2900/LM3900 window comparator misbehave?
LM2900 is a Norton amplifier with current inputs. It sets input nodes near a diode drop above ground. Driving it from a low-impedance voltage source causes errors and may damage inputs. Use resistor networks or choose a voltage comparator. [Elektroda, Anonymous, post #21684579]
What does “Norton amplifier” actually mean here?
It follows Norton’s Theorem: current source behavior with resistive networks. The LM2900/LM3900 compare input currents rather than voltages. Designers suggested LM339-family comparators for voltage windows instead. [Elektroda, Anonymous, post #21684581]
Which comparator should I pick for dual windows?
Pick a quad comparator like LM339/LM239/LM139. You get four channels per package, so two handle your two windows, leaving two spares. “Check the LM139 LM239 LM339 quad comparators.” [Elektroda, Anonymous, post #21684581]
How do I build the two windows with standard comparators?
- Set comparator A thresholds near 0 V and 1 V using a divider and reference.
- Set comparator B thresholds near 4 V and 5 V similarly.
- OR each window’s in/out logic to two outputs as needed.
Two-op-amp examples were shared and simulated successfully. [Elektroda, Giovanni Di Maria, #21684568]
Can I just program an MCU instead of adding comparators?
Yes. Sample the input and assert two digital outputs for the windows. The posted pseudo‑code shows the logic structure. Hardware may not reduce, but flexibility improves. [Elektroda, Giovanni Di Maria, #21684567]
Are Zeners a good way to set 1.2 V and 3.9 V thresholds?
Not for precision at low current. Sub‑5 V Zeners and diode stacks have soft knees. Their breakdown varies with current and temperature. Expect drift and fuzzy switching without careful biasing. [Elektroda, Elizabeth Simon, #21684573]
My LTspice result looked great, but the hardware is mushy—why?
A model set may lack accurate low‑voltage Zeners, and idealized diodes hide softness. Real parts show rounded knees and higher tolerance. The OP saw this and switched to quad op‑amps. [Elektroda, Anonymous, post #21684574]
How do I add real vendor parts to LTspice fast?
Download SPICE models from manufacturer pages and include them in LTspice. Many op‑amps, diodes, and transistors provide ready models. A quick “part + model” search helps. [Elektroda, Anonymous, post #21684575]
Any small tweak to stabilize a BJT-as-diode threshold?
Add a resistor from base to emitter. It raises diode current and sharpens the effective knee, improving repeatability slightly. This still trails a comparator for accuracy. [Elektroda, Anonymous, post #21684577]
Is there a risk if I connect a voltage source straight to LM2900 inputs?
Yes. The inputs prefer current drive through resistors. Direct voltage drive violates its input structure and can damage inputs. Route through resistive networks or choose LM339 instead. [Elektroda, Anonymous, post #21684579]
What did the original designer finally choose?
They moved away from Zeners and Norton amps, and ordered quad op‑amps instead. Simulations and experience led to the simpler, more reliable comparator path. [Elektroda, Anonymous, post #21684574]
Does this approach scale for many channels on a board?
Yes. A quad comparator gives four channels per IC, reducing cost and space. That’s 4 inputs per package, useful when duplicating windows across sensors. [Elektroda, Anonymous, post #21684581]
What if my signal source is a Hall sensor like in the thread?
Buffer or condition the Hall output, then feed the comparators. The OP’s node toggled ~0.2–4.8 V with magnet polarity, suiting window detection well. [Elektroda, Anonymous, post #21684578]
