Chris Davies
New Member
- Joined
- Jul 25, 2026
- Messages
- 3
- Likes
- 0
I have been thinking seriously about op-amp rolling, prompted by two recent experiences which appear contradictory.
In my study system, I replaced the NE5532 op-amps in the XLR and left/right signal-path positions of two Fosi ZA3 amplifiers, run as monoblocks, with OPA1622s. The change was clear, immediate, and easy to describe. It was not merely a vague preference. Individual sounds changed in character: metallic percussion became more convincingly metallic, reverb developed more clearly, transients became more compelling, and bass decays and tails became easier to follow.
The magnitude was comparable to a substantial improvement in speaker orientation. Not a “new amplifier” transformation, but far beyond something I had to strain to detect.
I then performed another experiment in a different system. At my store, I replaced Audience SX52Bs in a Douk D1 with the stock NE5532s. This system is reasonably well optimised, and I know its character very well. It currently drives mass-damped and high-passed JVC EX-A5 wood cone speakers, which I think are remarkable, and a single properly integrated mass-damped passive subwoofer.
The result was almost nothing.
I retained a very slight preference for the SX52Bs, but I could not describe the difference reliably, and I am confident I would not identify them consistently in a blind test.
These two observations led me away from the usual question:
- Do op-amps sound different?
I now think that question is too crude to be useful. The more useful question may be:
- Under what circuit and system conditions do op-amp substitutions produce detectable changes, and under what conditions do those changes disappear?
Op-amps are system components
An op-amp does not operate in isolation. It operates inside a circuit which determines:
After that comes the rest of the playback chain:
- Op-amp behaviour is a property of the op-amp–circuit–system interaction, not of the op-amp alone.
This seems obvious once stated, but much discussion treats op-amps as if they possess fixed sonic personalities which should follow them from product to product.
I now doubt that this is a useful model.
The failed-state observation
There is another piece of evidence which I think deserves much more attention.
The Audience SX52B produces no usable audio when installed in the Fosi ZA3 and V3 Stereo. Yet the same SX52B works normally in other products.
The failure is non-destructive. The op-amp is not damaged. The amplifier is not damaged. Reinstalling a compatible op-amp restores normal operation.
That means silence is not merely an absence of evidence. It is an actual operational outcome of the pairing.
The system enters a failed state without permanent damage.
This gives us a range of possible results from an op-amp substitution:
The same op-amp can function normally in one circuit and fail to pass audio in another.
This does not prove that every normally functioning op-amp pairing must sound different. It proves something narrower:
Successful mechanical and pin compatibility do not guarantee identical electrical operation.
And therefore:
The behaviour of a functioning pairing must be established in that specific circuit rather than inferred from the op-amp’s name alone.
What the failed state does—and does not—prove
I want to be very careful here.
The failed-state example does not prove that small audible differences exist between every two compatible op-amps.
It does not prove that my listening impressions were caused by the mechanism I presently suspect.
It does not invalidate measurements.
What it does establish is that the op-amp and host circuit interact strongly enough for the range of outcomes to include complete silence.
That makes universal claims such as “op-amp changes cannot affect the result” difficult to defend without qualification.
The correct conclusion is not:
“Because total failure is possible, every successful substitution must sound different.”
The correct conclusion is:
“Because circuit-dependent behaviour clearly exists, smaller differences between functioning combinations cannot be dismissed in advance. They must be investigated in the specific circuit.”
That is the distinction I want to explore.
A possible explanation for the great divide
There is an enormous rift in HiFi discussion around op-amps.
One group reports:
My working assumption is that people are reporting their experiences truthfully.
The disagreement may arise because they are testing different op-amp and circuit combinations, inside systems with radically different abilities to preserve and reveal small changes.
A person who hears no difference may be completely correct about their amplifier and system.
A person who hears a substantial difference may also be completely correct about theirs.
The error occurs when either local result is promoted into a universal rule.
“I heard no difference” does not become “there can be no difference.”
Likewise:
“I heard a large improvement” does not become “this op-amp is always a large improvement.”
Both statements exceed the evidence.
A system-resolution hypothesis
My present hypothesis is that op-amp detectability depends upon at least three layers.
1. Circuit compatibility
Can the op-amp operate correctly in the host circuit?
This includes:
2. Circuit utilisation
Does the circuit make meaningful use of the op-amp’s particular strengths or weaknesses?
A high-output-current device may offer no advantage into an easy load.
A FET-input device may provide no useful benefit where source impedance is low.
A faster device may be wasted in one circuit and unstable in another.
A superior specification is only useful if the surrounding circuit allows it to matter.
3. System audibility
If the electrical behaviour changes, does that change survive the rest of the reproduction chain strongly enough to be detected?
This may depend on:
The circuit determines what changes.
The system determines what survives.
The listener and listening method determine what is detected.
Emotional transparency may not equal analytical transparency
For many years, I have treated uncontrollable emotional response—tears, frisson, complete musical “transportation” —as the highest possible standard for a system.
I still consider it extremely important.
But I now suspect a system can be profoundly emotionally moving without being maximally revealing.
A system may possess enough tonal, dynamic, rhythmic, and spatial coherence to communicate music beautifully, while still concealing small differences in:
The D1 and EX-A5 combination sounds wonderful. It is musically satisfying and emotionally convincing. Yet it revealed almost no describable difference between SX52B and NE5532.
My quad-mono ZA3/RX1 system may simply exceed a different threshold of resolution or temporal clarity.
That does not make one system musically valid and the other invalid. It may mean they reveal different classes of change.
Speakers may be a major part of the answer
Speakers are the largest source of variation in most playback systems.
They impose their own:
This suggests two direct tests:
If the same op-amp contrast becomes obvious through one speaker and negligible through another, then op-amp audibility is partly a property of the complete musical translation chain.
Mono testing may be especially useful
The changes I heard with the OPA1622s were not primarily changes in soundstage width or stereo image placement.
They were changes to the sounds themselves:
If the difference remains obvious through a single speaker, then the investigation can focus on timbre, envelopes, decay, and transient structure rather than stereo imaging.
This may also help distinguish actual changes in translated (not reproduced!) sound from changes caused by stereo geometry.
What I intend to test
The amount of work required is rather daunting, but the project now appears worthwhile.
The programme will include:
The stability work matters because a sonic change caused by oscillation or abnormal behaviour cannot automatically be called an improvement.
A scope will therefore be used not to replace listening, but to qualify each combination before listening conclusions are taken seriously.
What would falsify the hypothesis?
For this project to be useful, the hypothesis must be allowed to fail.
Examples which would weaken or falsify parts of it include:
I do not want to prove that op-amps are wonderful.
I want to understand when they matter, when they do not, and why.
The claim I am making
My claim is deliberately limited:
An op-amp is a system component. Replacing it changes the physical system, but the resulting effect may range from undetectable, to measurably different, to audibly different, to improved, degraded, unstable, or completely non-functional. The result belongs to the op-amp–circuit–system interaction, not to the op-amp in isolation.
This framework allows both “nothing happened” and “the change was obvious” reports to be truthful.
It does not grant either report universal authority.
The mistake is not necessarily in the observation. It is in the generalisation.
Measurements remain essential. Controlled listening remains essential. Electrical qualification remains essential. Honest subjective description remains essential.
None of these should be treated as enemies.
I am still confused. But I think I may now have a useful (and painful!) way to become less confused. The work required to test it properly is substantial.
That is also what makes it interesting.
I'd like to hear your thoughts and feedback on my hypothesis and my proposed experiments. My apologies if you are tired of op-amp discussions. I hope my post brings something new to the topic.
In my study system, I replaced the NE5532 op-amps in the XLR and left/right signal-path positions of two Fosi ZA3 amplifiers, run as monoblocks, with OPA1622s. The change was clear, immediate, and easy to describe. It was not merely a vague preference. Individual sounds changed in character: metallic percussion became more convincingly metallic, reverb developed more clearly, transients became more compelling, and bass decays and tails became easier to follow.
The magnitude was comparable to a substantial improvement in speaker orientation. Not a “new amplifier” transformation, but far beyond something I had to strain to detect.
I then performed another experiment in a different system. At my store, I replaced Audience SX52Bs in a Douk D1 with the stock NE5532s. This system is reasonably well optimised, and I know its character very well. It currently drives mass-damped and high-passed JVC EX-A5 wood cone speakers, which I think are remarkable, and a single properly integrated mass-damped passive subwoofer.
The result was almost nothing.
I retained a very slight preference for the SX52Bs, but I could not describe the difference reliably, and I am confident I would not identify them consistently in a blind test.
These two observations led me away from the usual question:
- Do op-amps sound different?
I now think that question is too crude to be useful. The more useful question may be:
- Under what circuit and system conditions do op-amp substitutions produce detectable changes, and under what conditions do those changes disappear?
Op-amps are system components
An op-amp does not operate in isolation. It operates inside a circuit which determines:
- supply voltage
- closed-loop gain
- feedback network
- source impedance
- output load
- capacitive loading
- bias conditions
- decoupling
- stability margin
- physical layout
- the function being performed
After that comes the rest of the playback chain:
- amplifier topology
- gain structure
- loudspeaker crossover
- driver behaviour
- cabinet behaviour
- speaker orientation
- listening distance
- room reflections
- subwoofer integration
- phase and timing
- source quality
- the listener’s familiarity with the system
- Op-amp behaviour is a property of the op-amp–circuit–system interaction, not of the op-amp alone.
This seems obvious once stated, but much discussion treats op-amps as if they possess fixed sonic personalities which should follow them from product to product.
I now doubt that this is a useful model.
The failed-state observation
There is another piece of evidence which I think deserves much more attention.
The Audience SX52B produces no usable audio when installed in the Fosi ZA3 and V3 Stereo. Yet the same SX52B works normally in other products.
The failure is non-destructive. The op-amp is not damaged. The amplifier is not damaged. Reinstalling a compatible op-amp restores normal operation.
That means silence is not merely an absence of evidence. It is an actual operational outcome of the pairing.
The system enters a failed state without permanent damage.
This gives us a range of possible results from an op-amp substitution:
- No detectable effect
- A subtle change
- An obvious change
- A noticeable change with no clear improvement or degradation
- Degraded performance
- Marginal stability, distortion, noise, or oscillation
- Complete non-destructive functional failure
- Damage to the op-amp
- Damage to both the op-amp and host equipment
The same op-amp can function normally in one circuit and fail to pass audio in another.
This does not prove that every normally functioning op-amp pairing must sound different. It proves something narrower:
Successful mechanical and pin compatibility do not guarantee identical electrical operation.
And therefore:
The behaviour of a functioning pairing must be established in that specific circuit rather than inferred from the op-amp’s name alone.
What the failed state does—and does not—prove
I want to be very careful here.
The failed-state example does not prove that small audible differences exist between every two compatible op-amps.
It does not prove that my listening impressions were caused by the mechanism I presently suspect.
It does not invalidate measurements.
What it does establish is that the op-amp and host circuit interact strongly enough for the range of outcomes to include complete silence.
That makes universal claims such as “op-amp changes cannot affect the result” difficult to defend without qualification.
The correct conclusion is not:
“Because total failure is possible, every successful substitution must sound different.”
The correct conclusion is:
“Because circuit-dependent behaviour clearly exists, smaller differences between functioning combinations cannot be dismissed in advance. They must be investigated in the specific circuit.”
That is the distinction I want to explore.
A possible explanation for the great divide
There is an enormous rift in HiFi discussion around op-amps.
One group reports:
- no difference
- no improvement
- no reason to roll
- competent devices behaving transparently
- greater clarity
- altered tone
- more air
- improved transients
- better separation
- changed bass behaviour
- transformed musical engagement
My working assumption is that people are reporting their experiences truthfully.
The disagreement may arise because they are testing different op-amp and circuit combinations, inside systems with radically different abilities to preserve and reveal small changes.
A person who hears no difference may be completely correct about their amplifier and system.
A person who hears a substantial difference may also be completely correct about theirs.
The error occurs when either local result is promoted into a universal rule.
“I heard no difference” does not become “there can be no difference.”
Likewise:
“I heard a large improvement” does not become “this op-amp is always a large improvement.”
Both statements exceed the evidence.
A system-resolution hypothesis
My present hypothesis is that op-amp detectability depends upon at least three layers.
1. Circuit compatibility
Can the op-amp operate correctly in the host circuit?
This includes:
- supply range
- pinout
- single versus dual configuration
- input common-mode range
- output swing
- current demand
- unity-gain or minimum-gain stability
- capacitive-load tolerance
- local decoupling
- physical mounting and added parasitics
2. Circuit utilisation
Does the circuit make meaningful use of the op-amp’s particular strengths or weaknesses?
A high-output-current device may offer no advantage into an easy load.
A FET-input device may provide no useful benefit where source impedance is low.
A faster device may be wasted in one circuit and unstable in another.
A superior specification is only useful if the surrounding circuit allows it to matter.
3. System audibility
If the electrical behaviour changes, does that change survive the rest of the reproduction chain strongly enough to be detected?
This may depend on:
- speaker resolution
- crossover behaviour
- phase coherence
- cabinet movement
- early reflections
- listening distance
- direct-to-reflected ratio
- subwoofer integration
- amplifier topology
- source quality
- level matching
- switching speed
- listener familiarity
- programme material
The circuit determines what changes.
The system determines what survives.
The listener and listening method determine what is detected.
Emotional transparency may not equal analytical transparency
For many years, I have treated uncontrollable emotional response—tears, frisson, complete musical “transportation” —as the highest possible standard for a system.
I still consider it extremely important.
But I now suspect a system can be profoundly emotionally moving without being maximally revealing.
A system may possess enough tonal, dynamic, rhythmic, and spatial coherence to communicate music beautifully, while still concealing small differences in:
- transient shape
- harmonic texture
- decay
- reverberation
- bass envelope
- low-level ambience
- timing and phase behaviour
The D1 and EX-A5 combination sounds wonderful. It is musically satisfying and emotionally convincing. Yet it revealed almost no describable difference between SX52B and NE5532.
My quad-mono ZA3/RX1 system may simply exceed a different threshold of resolution or temporal clarity.
That does not make one system musically valid and the other invalid. It may mean they reveal different classes of change.
Speakers may be a major part of the answer
Speakers are the largest source of variation in most playback systems.
They impose their own:
- frequency response
- directivity
- crossover integration
- phase behaviour
- stored energy
- cabinet radiation
- distortion
- compression
- transient behaviour
- room interaction
This suggests two direct tests:
- Does replacing the speakers on the ZA3 monoblocks with a less explicit design reduce the previously obvious OPA1622 effect?
- Does replacing the EX-A5s on the D1 with the more forward and explicit RX1s make the SX52B/NE5532 difference easier to hear?
If the same op-amp contrast becomes obvious through one speaker and negligible through another, then op-amp audibility is partly a property of the complete musical translation chain.
Mono testing may be especially useful
The changes I heard with the OPA1622s were not primarily changes in soundstage width or stereo image placement.
They were changes to the sounds themselves:
- metallic tone
- transient impact
- reverb development
- decay structure
- bass tails
- apparent texture
If the difference remains obvious through a single speaker, then the investigation can focus on timbre, envelopes, decay, and transient structure rather than stereo imaging.
This may also help distinguish actual changes in translated (not reproduced!) sound from changes caused by stereo geometry.
What I intend to test
The amount of work required is rather daunting, but the project now appears worthwhile.
The programme will include:
- NE5532, OPA1622, OPA1656, SX52B, MUSES02, OPA2134, Sparkos SS3602, various Burson items, and other compatible devices
- multiple amplifier platforms
- multiple op-amp positions and functions
- stereo and mono amplifier configurations
- several loudspeaker types (I have ~50 pairs)
- one-way and multiway speakers
- nearfield and longer-distance listening
- Hot-swapping of identical amplifiers (with different opamps) and also speakers using electrically neutral switching
- careful level matching
- mono and stereo programme material
- familiar reference tracks
- detection separated from preference
- clear reporting of null results
- electrical stability checks
- temperature monitoring
- no-input and square-wave scope testing
- documentation of mounting method, including socket extenders or adapters
| Amplifier | Op-amp | Position | Mounting | Stable | Audible result |
| ZA3 | NE5532 | XLR/L-R | Direct | Yes | Baseline |
| ZA3 | OPA1622 | XLR/L-R | Adapter | To verify | Large subjective change |
| ZA3 | SX52B | Various | Direct | Non-functional | No usable output |
| D1 | NE5532 | Main | Direct | To verify | Near-identical to SX52B |
| D1 | SX52B | Main | Direct | To verify | Slight preference only |
The stability work matters because a sonic change caused by oscillation or abnormal behaviour cannot automatically be called an improvement.
A scope will therefore be used not to replace listening, but to qualify each combination before listening conclusions are taken seriously.
What would falsify the hypothesis?
For this project to be useful, the hypothesis must be allowed to fail.
Examples which would weaken or falsify parts of it include:
- the original OPA1622 result cannot be repeated
- the effect disappears under instant and level-matched comparison
- changing the speakers produces no change in detectability
- stereo versus mono makes no meaningful difference
- the supposedly superior combination is found to oscillate
- the audible change follows level mismatch rather than op-amp selection
- all stable combinations become indistinguishable under controlled conditions
- system optimisation fails to predict detectability
I do not want to prove that op-amps are wonderful.
I want to understand when they matter, when they do not, and why.
The claim I am making
My claim is deliberately limited:
An op-amp is a system component. Replacing it changes the physical system, but the resulting effect may range from undetectable, to measurably different, to audibly different, to improved, degraded, unstable, or completely non-functional. The result belongs to the op-amp–circuit–system interaction, not to the op-amp in isolation.
This framework allows both “nothing happened” and “the change was obvious” reports to be truthful.
It does not grant either report universal authority.
The mistake is not necessarily in the observation. It is in the generalisation.
Measurements remain essential. Controlled listening remains essential. Electrical qualification remains essential. Honest subjective description remains essential.
None of these should be treated as enemies.
I am still confused. But I think I may now have a useful (and painful!) way to become less confused. The work required to test it properly is substantial.
That is also what makes it interesting.
I'd like to hear your thoughts and feedback on my hypothesis and my proposed experiments. My apologies if you are tired of op-amp discussions. I hope my post brings something new to the topic.
Last edited:
