opamp rolling the correct way?

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

buschfsu

Well-known member
Joined
Dec 31, 2004
Messages
760
Location
jacksonville FL
guys,

i've doen a bit of subjective opamp rolling and I'm sure i can talk myself into liking this or that.  i'd like to know what measurements make sense to take when evaluating an opamp in a circuit that wasn't designed for. i can measure dc on the output rail but not sure about others.  i have multimeter and oscilloscope.  btw i know that measurements don't make necessarily a great sounding amp in circuit but id like to be able to see what/where oscillation would occur and have also heard that extra components may be needed (like ps decouple) but how would i tell if this is necessary??

thanks
 
You don't specify what you are swapping OPAs on but ..

Kingston has an excellent thread on what is REALLY important for noise & THD; opamps and local decoupling of rails, some questions

http://www.groupdiy.com/index.php?topic=37307.80

Many true gurus chime in.  It proves how OPA rolling takes a VERY poor second place to correct earthing, layout & decoupling.  It’s a long thread but read the whole thing from #41 to find pearls of wisdom.

Be VERY suspicious if a swap results in increased clarity & definition.  Your new uber OPA is probably oscillating.
 
buschfsu said:
i've doen a bit of subjective opamp rolling and I'm sure i can talk myself into liking this or that.  i'd like to know what measurements make sense to take when evaluating an opamp in a circuit that wasn't designed for. i can measure dc on the output rail but not sure about others.  i have multimeter and oscilloscope.  btw i know that measurements don't make necessarily a great sounding amp in circuit but id like to be able to see what/where oscillation would occur and have also heard that extra components may be needed (like ps decouple) but how would i tell if this is necessary??

Ideally would want to measure the distortion, which isn't really possible with those instruments, since you're talking about minute (and possibly fictional  ;) ) differences. However, recovery from overload is one of the main differences between opamps that may be audible, so you could at least look at what happens around clipping, even better if you have a gated oscillator so you can look at what happens when moving between clipping to no clipping.  You could also look for ringing on a square wave, particularly on those devices that have to drive difficult (e.g. capacitive) loads.
 
thanks thats helpful.  ringing on a square wave is distortion of sorts correct?  i also assume oscillation can be seen on my scope provided there is enough bandwidth.  and will oscillation happen as soon as it powers up or is this some sore of event catalyzed by signal being present? 
 
buschfsu said:
thanks thats helpful.  ringing on a square wave is distortion of sorts correct?  i also assume oscillation can be seen on my scope provided there is enough bandwidth.  and will oscillation happen as soon as it powers up or is this some sore of event catalyzed by signal being present?

I am not a big fan of design by ear, but you could possibly isolate subtle differences between otherwise similar circuits, by performing a null test. When subtracting one path output from the other, the null product is the difference between the two. However the simple null test does not tell you which one is causing the difference.

Sounds like a good way to waste a lot of time.

JR
 
yeah i guess im trying to understand what technical considerations should be addressed when looking at replacing opamps other than just pinout.  a null test would tell me that they are or are not affecting audio in the same way but not about the stability of the candidate opamp.  looking to put togeher something like...

When considering replacing a cheap NJM opamp with some 3$ burr brown you should...

1. make sure the pinout is the same
2. make sure the psu voltage is under the max for the new chip
3. make sure the current draw difference won't swamp your psu
4. make sure there are coupling caps on each side of the psu to ground very close to the opamp
5. measure a square wave for increased ringing.
6. measure oscialltion (not sure under what conditions)
7. compare THD specs
8.anything else
9.that ive left out

NOW you can AB the two chips knowing that you at least have a viable candidate and let your ears decided

I've heard some folks warn that oscillation can cause a chip to appear different/better but when stacked like in a mixer it would be terrible.

thanks for the input.  ill go read the thread above.

 
buschfsu said:
4. make sure there are coupling caps on each side of the psu to ground very close to the opamp
Make that

4 electrolytic decoupling caps on each side of the psu to ground very close to the opamp.

Under 8 & 9 add

check the polarity of input/output/feedback electrolytics are correct for the input & output offsets of your new OPA.

pnp/npn/fet input OPAs all have different input & output offsets depending on the EXACT circuit.

Note (as in Kingston's thread) that the uber OPAs with 1ppzillion THD very rarely have the best THD in anything but the simplest 'real life' applications.
 
buschfsu said:
offsets? do you mean dc decoupling caps on the input and output pins?
Yes.

If you replace NE5532 (npn i/ps) with NJM2068 (pnp), the DC voltages at the +ve & -ve input pins will probably reverse.  Output DC may change too depending on the EXACT circuit.

OPA2134 (fet i/p) and uber OPAs with evil input bias cancelling (eg OP270, LT1028, MAX412) are trickier as their input DC voltages are indeterminate.  :eek:

Some uber OPAs like LM4562 also latch up so are inappropriate for HP filter circuits.  Really nasty sounding when it happens.  This is something Self doesn't pick up.
 
Designing electronic circuitry is not like cooking chili, you don't just add or remove new ingredients to see what would happen.

What do you want to improve?

Are you asking us to tell you what to measure that you will hear????

Did i mention I don't care for this game. In some ways it's simple, in other ways it's complicated, but people are quick to draw simple conclusions often incorrectly from slender data.

JR
 
ok.  i know that may companies make tons of money doing just that and i guess it could be snake oil but there are a lot of positive reviews.

i'm trying to improve the overall sound quality.  focus the bass clean up the mid and open the high end (very scientific i know)

no, i want to know what to measure that will indicate that the chip is technically working well in the circuit, then i can use my ears to determine if the working amp actually sounds better.

some if this is a game in that i have mediocre gear and a couple dozen high quality chips so why not, its a hobby.  no one is gonna use this gear for any critical applications i have other stuff for that. also trying to learn about what makes a good robust design.

i know you are in a different place with knowledge and experience and this could be a fools errand but i may learn something along the way.

thanks for the posts

 
buschfsu said:
i want to know what to measure that will indicate that the chip is technically working well in the circuit, then i can use my ears to determine if the working amp actually sounds better.
Note that when Kingston had done ALL the work necessary to make sure his OPAs were "technically working well in the circuit", the audible differences went away.

The work isn't trivial.
 
Op amp rolling - what great fun !

Checking the DC offset is a good place to start to determine if the op amp is ok and working correctly.  A good rule of thumb is that the higher bandwidth that an opamp has, the more fussy it will be.  I wouldn't worry much about adding extra bypass capacitors and stability issues unless you have an op amp with over 10MHz unity gain bandwidth. 

You would need an oscilloscope (with adequate bandwidth!) to really know if your device is oscillating.  If you dont have one, you can check if the device is warm.  Sometimes, (but not always) oscillating devices run hotter than they normally would.  Sometimes they oscillate and remain cool, but if its excessively hot, you likely have a problem. 

Here is a white paper you may be interested in that talks about op amp selection, stability, etc. 

http://sparkoslabs.com/audio-op-amps-gain-seek-bandwidth/
 
buschfsu said:
hmm ok  I'm discouraged.
You shouldn't be.

By proper decoupling & grounding, you are not making the uber OPAs sound worse.  You are making your cheapo NJMs sound as good or better than stuff with big $$$ uber OPAs but badly decoupled.  Kingston has exactly this experience in his thread.  As he shows, even some of the big names aren't aware of this.

I wouldn't worry much about adding extra bypass capacitors and stability issues unless you have an op amp with over 10MHz unity gain bandwidth.
No.  You get audible improvements (certainly measured ones) even with TL07x & 553* circuits by adding 10u electrolytics from rails to an appropriate earth AT the OPAs.  This isn't pontificating.  Its based on 'real life' experience which Kingston only confirms.

But proper decoupling & earthing isn't as simple as swapping OPAs  ;D

You still haven't told us what you are swapping on.
 
Oh, for sure.  Im not suggesting additional bypass capacitors wont have a positive affect, Im just saying they wont necessarily be required for stability unless you have a wide bandwidth device. 
 
buschfsu said:
hmm ok  I'm discouraged.

You should be. If you are not willing to go through the massive ordeal of stabilising the entire psu rail network including rethinking what ground means in these contexts, you are only dealing with a bunch of opamps behaving badly. Note i even ended up replacing the whole psu to succeed, and that thing was nasa grade.  This was after weeks of rewiring power distribution quite literally making an entire new layer to the pcb.

Even if you manage to create such a great equaliser (of opamp audio performance) you will be in for a bit of a disappointment. They pretty much all sound the same and it's likely you will no longer be able to detect differences with even decent measurement rigs.
 

Latest posts

Back
Top