You are right, I intended this more in general 🙂 and not specifically for this device.
When looking at opamp noise, you really have to consider the source impedance - it makes a significant impact in the overall noise peformance.
I took Steve Hagman's very good 'Noise Visualizer' spread sheet from his EDN article and modded it so that it could accept 6 op amp noise inputs - I've attached the plot which covers LME49990, LM4562, NE5534, AD844, OPA2134and AD797.
Below 400 Ohms, the AD797, LME49990 and the LT1115(*) are the clear winners. Great for mic amps, buffers etc.
In the 400-2k Ohm range, its still the AD797, LME49990 and the LT1115. Good for buffering volume control pots of up to 10k Ohms (at mid resistance position, the source impedance seen by the opamp is 1/4 of the element resistance assuming you are using a low driving source). For a 10k pot, this is 2.5k.
2000 Ohms to 10k, surprisingly it's the NE5534 - the very low input noise current of 0.4pA/rtHz (is that perhaps a misprint in the TI data sheet?) looks like a winner. Doug Self will be sitting in his his chair right now, nodding his head saying 'I told you so' 😉. Its a great pity that the IP bias currents on the 5534 are so high - 200nA typical and as high as 1uA. Given its age, it is a remarkable opamp though.
Above about 10k, the OPA2134, due to its very low noise current, looks best. It does not perform well compared to the others with low source resistances, because its input noise voltage is high at 8nV/rtHz. Its noise current is 3fA - 3 orders of magnitide lower than the bipolar opamps.
From the above, you can see why for volume control pots of 10k and below, the AD797, LME49990 and the LT1115 are a good choice, while for a 50k volume pot, an OP2134 might make a better choice. The NE5534 still makes a fine RIAA equalizer amp, because the low input noise current contributes less noise betwen 10 kHz and 20 kHz than competing designs (the cartidge inductance causes noise to increase). Add to this the RIAA equalization in an all active design, and you have an optimum solution wrt noise - another reason DS likes them. But, you will need to couple the cartidge to the opamp through a cap because of the high bias current.
(* I am not sure if the LT1115 uses the bias current cancellation that DS mentions in his book - same used on the LT1028 - that actually causes problems so the noise figure numbers published have some caveats attached to them)
There are some nice AD7xx and AD8xx opamps that I have not shown - some more reseach is probably called for to get an even broader picture.
😎
PS - Anyone know how to get the damn x-axis tick labels to the bottom of the graph?
I took Steve Hagman's very good 'Noise Visualizer' spread sheet from his EDN article and modded it so that it could accept 6 op amp noise inputs - I've attached the plot which covers LME49990, LM4562, NE5534, AD844, OPA2134and AD797.
Below 400 Ohms, the AD797, LME49990 and the LT1115(*) are the clear winners. Great for mic amps, buffers etc.
In the 400-2k Ohm range, its still the AD797, LME49990 and the LT1115. Good for buffering volume control pots of up to 10k Ohms (at mid resistance position, the source impedance seen by the opamp is 1/4 of the element resistance assuming you are using a low driving source). For a 10k pot, this is 2.5k.
2000 Ohms to 10k, surprisingly it's the NE5534 - the very low input noise current of 0.4pA/rtHz (is that perhaps a misprint in the TI data sheet?) looks like a winner. Doug Self will be sitting in his his chair right now, nodding his head saying 'I told you so' 😉. Its a great pity that the IP bias currents on the 5534 are so high - 200nA typical and as high as 1uA. Given its age, it is a remarkable opamp though.
Above about 10k, the OPA2134, due to its very low noise current, looks best. It does not perform well compared to the others with low source resistances, because its input noise voltage is high at 8nV/rtHz. Its noise current is 3fA - 3 orders of magnitide lower than the bipolar opamps.
From the above, you can see why for volume control pots of 10k and below, the AD797, LME49990 and the LT1115 are a good choice, while for a 50k volume pot, an OP2134 might make a better choice. The NE5534 still makes a fine RIAA equalizer amp, because the low input noise current contributes less noise betwen 10 kHz and 20 kHz than competing designs (the cartidge inductance causes noise to increase). Add to this the RIAA equalization in an all active design, and you have an optimum solution wrt noise - another reason DS likes them. But, you will need to couple the cartidge to the opamp through a cap because of the high bias current.
(* I am not sure if the LT1115 uses the bias current cancellation that DS mentions in his book - same used on the LT1028 - that actually causes problems so the noise figure numbers published have some caveats attached to them)
There are some nice AD7xx and AD8xx opamps that I have not shown - some more reseach is probably called for to get an even broader picture.
😎
PS - Anyone know how to get the damn x-axis tick labels to the bottom of the graph?
Last edited:
Kingston had an excellent thread on what is REALLY important for noise & THD;
opamps and local decoupling of rails, some questions
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.
____________
Only a literal handful of people on this forum can extract the maximum performance from 5532. Da unwashed masses are unlikely to benefit from true 5532 let alone da new uber OPAs. Even Self's new favourite LM4562/LME family has gotcha's he's unaware of .. especially when doing LF filter circuits and with single rails.
However, OPAx134 is a good replacement for TL07x. It is the least fussy of the uber OPAs and hence usually has better 'real life' performance than its more highly spec'd but also highly strung competitors.
It also has very high immunity to RFI which is why it appears in loadsa real professional stuff (as opposed to toys or Golden Pinnae audiofool sh*t).
____________
Certain applications need better than 5532. eg amps for Lo source Zs benefit from AD797. I like the MAX410 family (even though it has evil input bias cancellation) rather than LM4562 etc family for these impedances. Not as LN as AD797 (though practical differences are small) but better than LM(E)xxxxx and without the faults.
Note the LN OPAs for various Zs that Bonsai mentioned. Horses for courses.
____________
Andrew, LT1115 does indeed have evil input bias cancellation. Some of the earlier LN OPAs like OP27 & LT1028 are sorta 5532/4 with input bias cancellation. In most cases, their 'real life' noise performance is worse than 5532/4 though they have slightly better spec.
opamps and local decoupling of rails, some questions
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.
____________
Only a literal handful of people on this forum can extract the maximum performance from 5532. Da unwashed masses are unlikely to benefit from true 5532 let alone da new uber OPAs. Even Self's new favourite LM4562/LME family has gotcha's he's unaware of .. especially when doing LF filter circuits and with single rails.
However, OPAx134 is a good replacement for TL07x. It is the least fussy of the uber OPAs and hence usually has better 'real life' performance than its more highly spec'd but also highly strung competitors.
It also has very high immunity to RFI which is why it appears in loadsa real professional stuff (as opposed to toys or Golden Pinnae audiofool sh*t).
____________
Certain applications need better than 5532. eg amps for Lo source Zs benefit from AD797. I like the MAX410 family (even though it has evil input bias cancellation) rather than LM4562 etc family for these impedances. Not as LN as AD797 (though practical differences are small) but better than LM(E)xxxxx and without the faults.
Note the LN OPAs for various Zs that Bonsai mentioned. Horses for courses.
____________
Andrew, LT1115 does indeed have evil input bias cancellation. Some of the earlier LN OPAs like OP27 & LT1028 are sorta 5532/4 with input bias cancellation. In most cases, their 'real life' noise performance is worse than 5532/4 though they have slightly better spec.
Last edited:
Richard, you raise some good points.
There have been a few cases on this forum where opamp rolling has led to less than expected results - Mooly showed one example with oscillation . . .
LM4562 (LME49710, LME49990 also) - handle with care. Very high open loop (and hence loop) gains mean you have to be extraordinarily careful about layout - and especially stray capacitance to the inverting input. They dont like being overdriven (PMA did some tests . . . it damaged the device) and the low input stage current does mean its more susceptable to RFI - whereas the 5534 with its high input stage current is less so. But, the 4562 has a magic sound if you take care of RFI and layout - and its no slouch in the noise department.
The OPAxx34's are good where you have a high source impedance - Ropt is about 270k ! or where you want to avoid coupling caps on the input - but I see no point if you use decent film devices.
JFET input opamps cannot compete for the most part with bipolars below about 2000 ohms, but between 2k and 10k the shift is towards JFET, and beyond 10k, JFETs are very clear winners in the noise department.
LT1115 - I suspected it . . . pity
There have been a few cases on this forum where opamp rolling has led to less than expected results - Mooly showed one example with oscillation . . .
LM4562 (LME49710, LME49990 also) - handle with care. Very high open loop (and hence loop) gains mean you have to be extraordinarily careful about layout - and especially stray capacitance to the inverting input. They dont like being overdriven (PMA did some tests . . . it damaged the device) and the low input stage current does mean its more susceptable to RFI - whereas the 5534 with its high input stage current is less so. But, the 4562 has a magic sound if you take care of RFI and layout - and its no slouch in the noise department.
The OPAxx34's are good where you have a high source impedance - Ropt is about 270k ! or where you want to avoid coupling caps on the input - but I see no point if you use decent film devices.
JFET input opamps cannot compete for the most part with bipolars below about 2000 ohms, but between 2k and 10k the shift is towards JFET, and beyond 10k, JFETs are very clear winners in the noise department.
LT1115 - I suspected it . . . pity
Last edited:
" . . . No more, not ever again will I make this mistake. I can also lay to rest this blind swapping madness and concentrate on design and data."
LOL
Quote from the link you posted.
LOL
Quote from the link you posted.
Yes, much more expensive that 25$, but perhaps in any moment you can think in use discrete opamps:
Discrete Op Amps | Sparkos Labs.comSparkos Labs.com
http://sparkoslabs.com/wp-content/uploads/2013/12/SS3601_SS3602.pdf
Discrete Op Amps | Sparkos Labs.comSparkos Labs.com
http://sparkoslabs.com/wp-content/uploads/2013/12/SS3601_SS3602.pdf
Yeah, gotta take account of the source impedance ! That's where some people go wrong, just looking at the noise specs.
Interestingly the 5534 is quoted as having between only 300k - 500k input resistance. This could make a difference to the overall circuit input resistance/impedance, if loaded with a too high source !
I've used OP27/37's many times over the years, & always had good results.
Interestingly the 5534 is quoted as having between only 300k - 500k input resistance. This could make a difference to the overall circuit input resistance/impedance, if loaded with a too high source !
I've used OP27/37's many times over the years, & always had good results.
From the data sheet it would be typ. 1.2pA/rt-Hz (4.5uA Ib). JC's favorite kind of op-amp huge Aol and two pole roll-off.
Last edited:
AFAIK the LT1115 is an underspeced LT1028, the AD797 also uses Ib comp we both had to to make parts for the rest of the world.
I found the claims that there is noise cancellation via corellation in a split collector lateral pnp to be only partially true. The corellation coefficient is pretty low in reality.
😕
Duu.uh! Scott, the few times I've used AD797 in the last Millenium, it exhibited the +ve input bias current shown in the Datasheet and typical of a npn input at high current.
My experience was a 1980's production commercial RIAA stage, I think OP27 . We had loadsa variation in input noise. Some makers were better than others but swapping to a Mullard/Philips 5534 cured the production variations and was better than the best OP27 samples.
RIAA is of course the worst case for showing up evil Ib cancelling as the impedances to +ve & -ve input are large and about as different as possible.
Another useful dodge was direct connection of the 5534 to the MM cartridge. 😱
You can work out the changes in mechanical & electrical characteristics of the cartridge due to this evil trick and they are negligible. We never managed to measure ANY deterioration in cartridge performance from direct coupling .. but saw improvements in 1/f noise.
You will have to fill me in on the perceived evils of Ib comp. the OP27 also uses it. In many cases it is fully bootstrapped and sees no supply or input voltage modulation. It increases the Ib noise by 3dB in general and not much else.
been a while - but I plotted a few fet and cmos op amp noise curves http://www.diyaudio.com/forums/solid-state/68793-new-interesting-cmos-op-amps-2.html#post843381
which brings up another dimension of noise - it is frequency dependent, 1/f noise corner can be a limiting audio performance spec, particularly for RIAA
other recent smt only parts are TI's OPA164x fet input op amps - you could parallel the quad for lower noise in less space than a dip 5534
which brings up another dimension of noise - it is frequency dependent, 1/f noise corner can be a limiting audio performance spec, particularly for RIAA
other recent smt only parts are TI's OPA164x fet input op amps - you could parallel the quad for lower noise in less space than a dip 5534
Last edited:
That's what i did when i designed the RIAA stage of my PreAmp with OP27's many moons ago. Quiet as a mouse, & doesn't squeak either 🙂
Scott,
Re the input bias current, I only raised if because DS mentions in his book. I've not used (ever) the OP27, 1028 or the LT1115 - maybe it's a non issue. I'll have to dig the specifics out tonight.
jcx,
I agree 1/f is an issue in RIAA. The plots I posted up are the en and ien worst case figures at 1 kHz.
Re the input bias current, I only raised if because DS mentions in his book. I've not used (ever) the OP27, 1028 or the LT1115 - maybe it's a non issue. I'll have to dig the specifics out tonight.
jcx,
I agree 1/f is an issue in RIAA. The plots I posted up are the en and ien worst case figures at 1 kHz.
Last edited:
No, not needed due to the low IB 😉 Never had any issues with it 🙂
http://www.analog.com/static/imported-files/data_sheets/OP27.pdf I didn't use the RIAA circuit in Fig 41, but designed my own.
Lets get this straight Mr. Wurcer.
Are you claiming your AD797 uses input bias current cancellation? The AD Datasheet implies not and IIRC so did my Jurassic tests.
3dB is what I'd expect and what we saw on our Jurassic RIAA MM preamp with good OP27 wrt to 5534 using CCIR/ARM and also true CCIR 468 measurements. Some OP27 makes were worse. The inductance of the typical MM cartridge makes In at circa 4kHz the major audible noise factor.
Bottom line is that for 2 OPAs with the same noise spec., one with and one without Ib cancelling, on many circuits, you can expect the one without Ib cancelling to have better 'real life' noise.
Of course there are applications where the difference is small (equal Z on both inputs or LoZ) 🙂
Andrew, we use to design & sell cartridges, turntables & arms so had all the toys to look at vinyl playback performance. I can quote da theory but I know of NO tests which show worse MM performance due to evil 5532 Ib.
In fact, even in MCs, (which in theory should be more affected cos higher efficiency) the effects are still well below that of other evils.
- Status
- Not open for further replies.