OP Amps - do they sound different?

Ok lets add a proviso - assuming the circuit allows them to operate cleanly without any sign of instability.

Do you believe that different chips have their own 'sound'?

I often read that they do but what is the experience of users?

I have some interest in this as my mega Quad 34 upgrade has hit a hitch - long story buy basically I'm going over to DIL sockets for all op amps plus the CMOS switches (Martin I've sourced some goodies).

The science indicates that if THD is low, there is no instability and no clipping then results shouldn't really differ between these things.

Over to you.

 

They certainly can, but it's all about suitability and application.

The Behringer mods I do is a good example. The op-amps they use in the analog circuits are stable, but they just happen to have rather high IMD. Replace with a simple NE5532 and the sound is audible less hasshy and hard because IMD drops massively. However I think this is an unusual case. The majority of audio equipment uses just a handful of op-amps found to work very well for audio applications. Swapping one known good performer for another isn't likely to yield audible changes IMO.

Having said that I use the TL074 in the active filter of my Advantage bass speaker. It's one of the most common J-FET op-amp used in audio becase it is a reliable good performer. None the less, I'm going to try the OPA4134 op-amp. At the moment if you put your ear right next to the woofers you can hear a small hum and hiss. I'll be interested to see if the OPA has less audible noise.

 

Substrates sound different, everyone knows that!

 

I'm with Andrew here, different op-amps 'can' sound different even when implemented to specification. There could be many reasons for this. Implementing them 'beyond' spec ie with even better supplies, decoupling etc, might make them sound even better than expected.

As Andrew mentions different substrates could sound different. I certainly have experience of different silicon performing differently even though it was the same design just extracted from different wafers or made at different plants.

it could just be that the layour of the op-amp itself is enough to make them sound different, to give them a signature. Different switches have different effects, EMI/RF etc, just look at a wifi chip, that is just made of gates switching in sync.

I think most differences people hear are more than likely down to plain old poor implementation in certain circuits. An LM4562 requires very different decoupling to a 5532 for example, and has quite different out of band performance.

 

There's a lot on AOS about op-amp rolling in the Caiman. Stan Beresford seems to believe there is something in it - or does he only believe his customers so believe?

 

Totally agree with that.

Paraphrasing stuff I've spouted before, if you swap opamps - even those that are notionally compatible - you are not just changng one specification, but a whole bunch at the same time, for example:

-Slew rate
-PSRR, of both rails
-where the internal integrator is referenced (can fundamentally change how attached circuit works, and optimal decoupling routes)
- Decoupling requirements - both the above affect layout and values required.
-Input-referenced noise level due to bias currents (may or may not matter)
- Phase margin (stability) owing to GBWproduct limitation, requiring adjusted external compensation to match performance of the opamp you are replacing (level teh playing field)
- ditto owing to loading effects both in and out-of-band
-Tolerance of loading applied (depends on several things!)

... and that's just the obvious ones for a start. No wonder different opamps 'sound' different; most DIYers don't bother to try to experiment let alone optimise any of the above each time between swaps. Was is being played with isn't the difference between opamps, but the difference between opamp susceptibilities in the target circuit as a whole. Opamp swaps a level playing field? No, very far from it, even if the %thd numbers are 'beyond reproach' as most appear these days (hint: these things are always measured into resistive loads)

For example - an opamp that has high bandwidth or very high open-loop output impedance or massive HF feedback or restricted output drive current could be poor choices in a filter circuit, or wherever the load is reactive (even a few pF). A circuit with a noisy raw supply could easily sound better with a 'worse' opamp with great PSRR than a better opamp with poor PSRR.. and so on and so on. Every pin on an opamp is effectively an input at some point in its operation.

This is why I don't recommend generic swaps anymore - it's a total lottery without rather more effort than 'unplug-plug'. Get the circuit working as well as you can via attention to PSU and layout and maybe parts/value optimisation, and then - only if you must - try rolling opamps.; IMO there is little there is to be gained when other more obvious problems are fixed.



[ - and still I cannot get LM4562s sound worth having, nor the single LME4xxxx versions ]

 

Doug self makes an interesting point about the good old 5532/4 and how that gets a regular slagging because people just fling them into any application without real understanding of the circuit conditions. There are often do decoupling caps on them for starters.

My position is that any sound difference is solely due to effects such as Martin mentions, in other words get the operating conditions right for a pair of fairly modern op amps and you'll get them sounding the same. Reading something like DIY Audio is frightening with people just throwing in one after another.

Going back to my Quad project, that hit a nasty hitch that could so easily have been avoided. One of the reasons that OPA134s were selected was that they are close enough in specification to stand-in for TL071s without circuit changes.
They aren't super fast, have similar JFET input conditions, aren't critical of PSU decoupling and don't draw too much current. The primary differences are lower noise and distortion under just about any circuit condition compared with the veteran chips.

However, cleanliness is next to godliness as they say. When stripping the Quad board I quickly unsoldered the old chips as usual and popped in the new ones. I didn't take too much care over cleaning the board and it would seem that the flux residue around the op amps had markedly worsened the HF crosstalk. There was now significant bleed at HF. Channel balance on two of the inputs was also now slightly out - only a dB and that's nothing to some well known amps I've seen (no names) but it was perfect pre modification. These things matter less on low impedance circuits but were quite audible here. So today the thing is stripped of active stages and CMOS switches - thoroughly de-soldered using the correct tools and the board cleaned with resin remover. DIL sockets are in place this time for both the op amps and perhaps just as importantly, the CMOS ICs. I've ordered some much higher quality versions of the 4066 which can run on the Quad voltages without PSU changes.

Anyway, back to the point - this leaves open the option to use different op amps if necessary. Little point straying from 134s in the main unit IMO but we could certainly try something else in the RIAA section, suitably stabilised. Not just swaps for the sake of it but parts likely to work well in the circuit. Anyone interested can pull-up the schematic. I'll experiment myself over the next few weeks but if anyone thinks something might work especially well we can try it - needledrops always an option.

 

See this
http://www.audiokarma.org/forums/showthread.php?t=334562&highlight=Catman

maybe Felix could help with some needledrops if asked

 

Familiarity and perhaps over-active imagination.