YG Group
No. This article uses the explicitly scoped TI base device. A purchasing decision still needs the variant, package, temperature grade and carrier details of a complete order. The SOIC examples in Table 5 are separate comparison entities; none silently defines the package of the primary LM358 request.
No. The recommended common-mode ceiling is V+ − 2 V over the applicable temperature range. At a ground-referenced 5 V supply, that gives 3 V. Some 25°C electrical rows use a different headroom condition, but they do not support an unrestricted positive-rail input claim.
The supplied TI document recommends 3–30 V for original LM358 and LM358A, versus 3–36 V for LM358B and LM358BA. Do not transfer the newer variants' supply range to an original device. Also distinguish total supply voltage from the individual positive and negative rail values.
Author: YG Group | Primary part: LM358 | Manufacturer: Texas Instruments
LM358 is a Texas Instruments dual operational-amplifier family commonly used in power and control circuits. The original LM358 and newer LM358B differ in supply range, offset and bandwidth, but neither offers rail-to-rail input operation. This comparison separates those versions, then checks input common-mode limits, output loading and the package-specific evidence needed before an existing circuit is upgraded.
Not as an unconditional rail-to-rail output capability. Positive-rail headroom depends on the output current and electrical conditions; Table 3 includes maximum headroom of 1.42 V at 50 µA and 1.48 V at 1 mA under the B-section conditions. Calculate the required swing at the actual load before choosing the part.
The BA version has a tighter maximum input-offset specification at 25°C: 2 mV rather than 3 mV. Preserve the detailed test conditions and temperature limits when building an error budget. The lower offset does not imply different rail-to-rail input behavior or automatic compatibility with an unspecified package.
No. TI identifies an upgrade path, but an unspecified existing device leaves manufacturer, package, loading and operating conditions unresolved. Verify the exact candidate and circuit before making a replacement claim. A base-family comparison cannot authorize every cross-manufacturer or cross-package combination.
LM358B is a newer TI device with useful specification changes, but the suffix does not mean that every original circuit limitation disappears. Start by separating original, A, B and BA variants. Otherwise, a family datasheet can appear to promise 36 V operation or 1.2 MHz bandwidth for an original LM358 that does not carry those specifications.
This article uses TI's base-device scope because the requested primary designation is LM358, not a complete ordering code. No package or carrier is selected for that primary entity. The technical source is the supplied TI SLOS068AB datasheet, revised October 2024, including its July 15, 2026 package addendum. Exact SOIC orders later in the article are separate comparison examples, not a silent choice for the user's existing circuit.
| TI variant | Recommended total supply | Operating ambient range | Maximum input offset at 25°C | Typical gain-bandwidth product |
|---|---|---|---|---|
| LM358 | 3–30 V | 0°C to +70°C | 7 mV | 0.7 MHz |
| LM358A | 3–30 V | 0°C to +70°C | 3 mV | 0.7 MHz |
| LM358B | 3–36 V | −40°C to +85°C | 3 mV | 1.2 MHz |
| LM358BA | 3–36 V | −40°C to +85°C | 2 mV | 1.2 MHz |
Table 1: TI LM358 variant limits, with maximum offset and typical bandwidth kept separate | Source: TI SLOS068AB, family comparison and sections 5.3, 5.5 and 5.7 | Compiled by: YG Group
The offset values are not universal circuit-error limits. For LM358/LM358A, the detailed offset row uses a 5–30 V supply, zero common-mode input and 1.4 V output. The B/BA electrical section uses its own header: 5–36 V supply, 25°C, with common-mode and output at half the supply and the specified load arrangement unless a row says otherwise. Table 1 is a variant summary; a precision design must return to those detailed conditions.
The A and B letters also answer different questions. LM358A tightens the original device's offset specification without changing its typical bandwidth to the B value. LM358BA tightens the B-version maximum offset further. If the existing design is limited by high source impedance or positive-rail input range, selecting the smallest offset number may leave the dominant error untouched.
TI's LM358 product page, checked September 10, 2026, points to LM358B as an upgrade. That is useful manufacturer guidance, but it does not identify the package, manufacturer marking or circuit conditions of an unspecified board. Treat the manufacturer upgrade path as the starting point for exact-order and application checks, not as blanket approval for every part sold with LM358 in its name.
The supply figures in Table 1 are total voltage between the positive and negative supply pins. A ±15 V arrangement therefore has 30 V total supply, not 15 V. The signal's permissible position within those rails remains a separate calculation. Writing the two physical rail voltages on the schematic avoids confusing a bipolar supply notation with a component's total operating range.
Neither LM358 nor LM358B accepts an arbitrary signal all the way to the positive rail. TI's recommended operating conditions give a common-mode input range from the negative rail to 2 V below the positive rail over the applicable operating temperature range. The B upgrade extends the supply ceiling but retains this recommended positive-rail headroom rule.
For a ground-referenced supply, the upper recommended input level is simply V+ − 2 V. At 5 V, that is 3 V. A signal near 4.5 V is therefore not made valid by changing the original LM358 to LM358B while leaving the supply and circuit unchanged. This direct boundary is more useful than describing both devices broadly as single-supply amplifiers. See section 5.3, page 5.
| Positive supply, with V− = 0 V | Calculated recommended upper common-mode limit | Original LM358/LM358A | LM358B/LM358BA | Interpretation |
|---|---|---|---|---|
| 3 V | 1 V | Within recommended supply range | Within recommended supply range | Recommended range screen; detailed electrical guarantees remain condition-specific |
| 5 V | 3 V | Applicable | Applicable | A 3.4 V input exceeds this recommended ceiling |
| 12 V | 10 V | Applicable | Applicable | Supply headroom remains necessary |
| 24 V | 22 V | Applicable | Applicable | Not a high-side input beyond the supply |
| 30 V | 28 V | Applicable | Applicable | Original device is at its recommended supply ceiling |
| 36 V | 34 V | Outside original recommended supply range | Applicable | Do not transfer this point to the original versions |
Table 2: Calculated recommended input ceilings, not measured transfer curves | Source: TI SLOS068AB, page 5 recommended operating conditions | Compiled by: YG Group
Some 25°C electrical rows show a V+ − 1.5 V upper input boundary under their stated supplies. That does not supersede the more conservative full-temperature recommendation used in Table 2. It also does not imply that every parameter is characterized at every point of the recommended range. Keep the row conditions visible when the design depends on a small margin near a limit.
Consider a hypothetical 5 V-powered voltage follower intended to buffer a sensor that ranges from 0.4 V to 3.4 V. In normal linear operation, both amplifier inputs track approximately the sensor voltage. The low end is inside the recommended common-mode range, but the 3.4 V upper end exceeds its 3 V ceiling by 0.4 V. Lower input offset will not repair that range violation. A valid redesign might scale the signal, change the available supply with all other limits checked, or select a different input architecture.
Now include a stated supply tolerance rather than evaluating only the nominal value. If the hypothetical 5 V rail can fall to 4.75 V, the recommended upper common-mode limit becomes 2.75 V. A 2.9 V sensor output that seemed acceptable at nominal supply no longer passes that conservative check. The original arithmetic is simple: 4.75 − 2 = 2.75 V, leaving −0.15 V margin against the 2.9 V signal. The assumed supply tolerance is an example, not a TI specification.
The voltage-follower example is not a rule that the external signal must always equal the common-mode voltage. In an inverting amplifier, feedback may hold the input nodes near a reference while the external signal arrives through a resistor. In a noninverting amplifier, the input node follows the external signal. Inspect the actual voltages at both amplifier input pins during normal operation and abnormal states; a block-diagram gain label alone does not establish them.
The near-ground capability is useful in control circuits, but it must not be generalized into a high-side measurement capability. A current-sense stage may need to measure a small differential voltage riding on a common-mode voltage far above the amplifier's supply. That calls for a dedicated input architecture whose common-mode limits must be checked separately. Its behavior is not evidence for a general-purpose LM358 input.
Finally, an absolute input-stress limit and a linear common-mode limit answer different questions. Surviving a voltage does not mean amplifying it accurately. When a circuit spends time outside the recommended input range, document that state and evaluate recovery and system behavior separately instead of counting it as normal linear operation.
Valid input voltage does not guarantee the requested output voltage. LM358-family output swing depends on how much current the output sources or sinks, where the load is connected, and the applicable supply and temperature conditions. Neither original nor B-version behavior supports an unconditional rail-to-rail output claim.
The useful quantity is headroom: the distance between the requested output and the relevant supply rail. Compare that distance with the corresponding electrical row. A typical graph under a light load cannot prove a maximum swing under a heavier load, and a load to ground is not equivalent to the same resistance connected to mid-supply.
| Variant and rail | Stated supply/temperature context | Output load or current condition | Headroom value | Classification and source |
|---|---|---|---|---|
| LM358B/BA, positive rail | Section 5.5 header: 5–36 V, 25°C unless noted | IOUT = 50 µA | 1.35 V typical; 1.42 V maximum | Page 6; current-conditioned output row |
| LM358B/BA, positive rail | Same section header | IOUT = 1 mA | 1.40 V typical; 1.48 V maximum | Page 6; do not transfer the lighter-load value |
| LM358B/BA, negative rail | Same section header | IOUT = 50 µA | 100 mV typical; 150 mV maximum | Page 6; not the separate load-to-ground condition |
| LM358B/BA, near negative rail | VS = 5 V, −40°C to +85°C | RL ≤ 10 kΩ connected to V−, as printed | 5 mV typical; 20 mV maximum | Page 6; retain the exact stated load arrangement |
| Original LM358/LM358A, positive rail | VS = 30 V, 0°C to +70°C | RL = 2 kΩ | 4 V maximum | Page 10; different conditions from B current rows |
| Original LM358/LM358A, positive rail | VS = 30 V, 25°C | RL ≥ 10 kΩ | 2 V typical; 3 V maximum | Page 10; not a matched-condition upgrade comparison |
Table 3: Output headroom values with their distinct conditions preserved | Source: TI SLOS068AB, pages 6 and 10 electrical tables | Compiled by: YG Group
For a conditional 5 V, 25°C example using the B-version 50 µA sourcing row, the maximum specified 1.42 V positive-rail headroom corresponds to a 3.58 V upper output level. At the 1 mA row, the analogous calculation is 5 − 1.48 = 3.52 V. These are deductions from those specific rows, not a universal output range for every load or the full temperature range.
Return to a hypothetical gain-of-two stage whose input varies from 0.2 V to 2 V. Its intended output spans 0.4–4 V. The input can fit the 5 V common-mode recommendation while the requested 4 V output fails the positive-swing budget in the B-version examples above. This is an output limitation, not an input limitation. Replacing the device based only on a lower offset specification leaves the requested voltage span unresolved.
Load resistance also needs interpretation. A 10 kΩ load to ground draws 0.35 mA when the output is 3.5 V. A 10 kΩ load to a 2.5 V reference draws only 0.10 mA from an output at 3.5 V. The resistor value is identical, but the output-current demand is not. Use the actual load-reference voltage and add other paths, such as feedback resistance or an attached measurement input, before selecting the applicable swing row.
Near ground, do not assume that a small no-load output residual guarantees the same behavior while sinking current from another source. The separate rows in Table 3 make that distinction visible. If a load can pull current into the amplifier output, review the sinking condition explicitly. An output intended to drive an ADC, transistor base or long cable may also have dynamic demands that a static resistance calculation does not capture.
A practical bench check sweeps the requested output range while recording the actual load current and input common-mode voltage. Observe whether the output follows linearly before comparing endpoint numbers. Repeat at the intended supply extremes and temperatures. If an output cannot reach its target, isolate whether the cause is current demand, input range, saturation, oscillation or the surrounding feedback network; these mechanisms can produce superficially similar clipping.
An LM358B upgrade is most useful when its verified improvements address the circuit's dominant limitation. Offset and bandwidth are legitimate reasons to investigate it, but neither substitutes for headroom. A circuit already outside its input or output range cannot be rescued by treating an offset calculation as its complete error budget.
For an original low-frequency example, assume a noninverting stage with noise gain 11, valid input/output range and otherwise ideal components. Multiplying each 25°C maximum input offset by 11 gives an offset-only output-error bound of 77 mV for LM358, 33 mV for LM358A, 33 mV for LM358B and 22 mV for LM358BA. This is a calculated contribution, not a total system accuracy claim; it excludes bias-current effects, resistor errors, reference errors, drift and noise.
The result is useful because it connects the variant table to a real question: how much of the output error allocation can offset consume? If a hypothetical design allows 20 mV total error at that gain, even the 22 mV BA offset-only calculation does not satisfy the allocation at the stated maximum. A different amplifier, calibration strategy or gain distribution may be required. If the total allowance is 100 mV, other error terms still need room rather than assuming that a 77 mV contribution leaves the system qualified.
Temperature changes the offset budget. The detailed tables give larger full-range maxima than the 25°C headline for these variants: for example, 9 mV for original LM358 over 0–70°C and 4 mV for LM358B over −40–85°C. At the same assumed noise gain of 11, those contributions become 99 mV and 44 mV respectively. The temperature ranges differ, so these are each variant's own bounds, not an equal-temperature measurement of improvement. See the relevant offset rows.
Source impedance creates another error path. A hypothetical 50 nA input current through 100 kΩ produces 5 mV of voltage error by Ohm's law. That illustrative current is not assigned as the guaranteed value of every LM358 variant. The actual result depends on each input's source resistance, bias current, offset current and feedback arrangement. When buffering a high-impedance sensor, investigate this input-current mechanism before choosing an amplifier; a low-offset specification alone does not establish an appropriate replacement.
Bandwidth should be handled with comparable care. Dividing typical gain-bandwidth product by a noise gain of 11 gives first-order estimates of about 63.6 kHz for original LM358 and 109.1 kHz for LM358B. These estimates assume a simple dominant-pole closed-loop response and suitable loading. They are not guaranteed flat-response bandwidths or settling times. The gain-bandwidth values themselves are typical, and the real circuit can introduce additional poles.
Large-signal speed is separate from small-signal bandwidth. A sine wave of 1 V peak at 10 kHz requires a peak slope of 2π × 10000 × 1, or approximately 0.0628 V/µs. That calculation identifies a slew-rate demand, but does not prove low distortion or adequate settling. The original and B tables contain typical slew-rate values under their own conditions; a production design must preserve enough margin and validate the intended amplitude, frequency and load.
A direct-digital-synthesis analog output provides a relevant application context: a buffer's bandwidth, headroom and filtering role must fit the generated signal. More bandwidth is not valuable if the stage clips, adds unacceptable noise or becomes unstable with its actual load. Evaluate that output stage as a circuit, with the required signal amplitude, frequency and load stated explicitly.
Approve an upgrade against the existing circuit and a complete candidate order, not against a family name. TI's base LM358 designation leaves the primary package unspecified here. The correct next step is to identify the installed part and intended replacement, then document which electrical and mechanical conditions have actually been checked.
| Requirement | Evidence from existing circuit | Candidate evidence to collect | Pass condition | Unresolved input in this article |
|---|---|---|---|---|
| Manufacturer and variant | Marking and approved BOM | Exact manufacturer order and datasheet | Same intended identity or explicitly approved change | User's physical part not supplied |
| Package and pinout | Released footprint and symbol | Orderable package drawing and pin functions | Mechanical and net-by-net compatibility | Primary package unspecified |
| Supply and input range | Rail tolerance and input extrema | Recommended range and electrical conditions | Both inputs valid through intended states | Actual circuit voltages unknown |
| Output load | Resistive, current and capacitive loading | Source/sink swing and stability evidence | Required output range and settling achieved | Actual load unknown |
| Error budget | Allowed offset, gain and noise error | Offset, drift, bias and noise conditions | Sum of contributions meets the allocation | Accuracy target unknown |
| Dynamic operation | Signal amplitude, frequency and steps | Bandwidth, slew, settling and recovery | Required response with margin | Signal profile unknown |
| Startup and abnormal states | Sequencing, open inputs and overloads | Applicable limits and validation plan | No unreviewed stress or control failure | Test results not supplied |
| Lifecycle and procurement | Approved order and restrictions | Current official status and seller offer | Exact approved variant procured | Inventory and delivery unverified |
Table 4: Original LM358 upgrade qualification checklist with unknowns left explicit | Source: TI SLOS068AB recommended, electrical and ordering sections | Compiled by: YG Group
Use simulation to investigate a candidate, but first check which model and variant the simulator represents. A model may reproduce typical linear behavior without bounding every offset, temperature, recovery or protection condition. An apparently clean simulated waveform is not proof that an overrange input is supported. Keep datasheet limits in the review even when the model seems tolerant.
Package verification should include the land pattern, pin numbering, orientation and assembly requirements, not merely the number of pins. An eight-pin SOIC example does not establish interchangeability with an eight-pin package of a different type. Likewise, a wider operating temperature range on one amplifier does not qualify the entire board's passives, connectors or calibration over that range.
| Device or exact order | Scope and relationship | Relevant difference | Package boundary | Required interpretation |
|---|---|---|---|---|
| LM358 | Primary TI base-device baseline | Original 3–30 V, 0–70°C device scope | Not selected | Not a complete procurement order |
| LM358BIDR | Related family member | B version, −40–85°C | Eight-pin SOIC example | Check full electrical and footprint conditions |
| LM358BAIDR | Related family member | BA maximum offset 2 mV at 25°C | Eight-pin SOIC example | Tighter offset does not create rail-to-rail input |
| LM358ADR | Related family member | Original A version, 0–70°C | Eight-pin SOIC example | Offset option, not the B architecture |
| LM2904BIDR | Comparison candidate | Related B-family −40–125°C scope | Eight-pin SOIC example | No automotive qualification inferred from the designation |
Table 5: Related TI order examples, not a blanket replacement authorization | Source: TI SLOS068AB, electrical sections and package addendum pages 36, 38 and 39 | Compiled by: YG Group
An effective release note states what the upgrade is expected to fix and what remains unchanged. For example: lower the allocated offset contribution, retain a stated gain, and preserve the original input/output range restrictions. Then attach the measured acceptance results for those requirements when testing is performed. This prevents a future engineer from interpreting “upgraded to B” as permission to widen the signal range or substitute another manufacturer's part without review.
Check present availability only after the complete intended order has been selected. Use the TI LM358 datasheet as the technical starting point and include the approved variant, package, quantity and packing in any quotation inquiry. The original technical source check is dated September 10, 2026; neither a family-level active status nor this article establishes inventory, price or lead time for the eventual order.
LM358B can be a useful upgrade when its supply range, offset or bandwidth changes address the actual circuit constraint. It does not make the input reach the positive rail or remove load-dependent output headroom. Identify the installed TI variant, calculate both input and output margins, then qualify the exact package and error budget. That sequence turns a suffix comparison into a defensible engineering decision.
