A few years ago, Doug Self or Elektor found out that the NE5532 from Fairchild has less distortion than that of Texas Instruments. Meanwhile, Fairchild has been included to OnSemi. It is the question whether this is still so. Does anyone have the ability to measure a NE5532 from OnSemi with an AP device?
An AP is not necessary if you want to investigate differences, distortion magnification techniques like the ones used by Samuel Groner to produce his data compilation would result in the differences - if substantial - becoming visible through measurements with a relatively pedestrian sound card.
I have an APx-555, and while it has a very low residual, there are still devices out there that have distortion levels equal to or lower than the AP residual. To be able to take the numbers from an analyzer and treat them as reliable without interpretation, the device under test ought to have nonlinearities around 10dB higher than the analyzer residual. So yes, even with a great analyzer, one still needs to use distortion magnification techniques, or some sort of comparison techniques to assess nonlinearities for a great many devices.
Back to the 5532, I'd suggest you try the New Japan Radio versions. In my experience, they make some relatively simple amps like the 5532 very well, and they seem to measure and sound very good compared to many other 'gumball' versions of the same chip. They have basically two amps, a BiFET and a bipolar, and offer a wide range of differently tuned versions of nearly the same amps. And, their process seems to be carefully done, easily as good or better than many versions of the same basic amps.
Back to the 5532, I'd suggest you try the New Japan Radio versions. In my experience, they make some relatively simple amps like the 5532 very well, and they seem to measure and sound very good compared to many other 'gumball' versions of the same chip. They have basically two amps, a BiFET and a bipolar, and offer a wide range of differently tuned versions of nearly the same amps. And, their process seems to be carefully done, easily as good or better than many versions of the same basic amps.
Some of the claimed ultra low distortion opamp datasheets show the distortion magnifying circuit they use to measure that ultra low distortion.
Many are measuring sub 1ppm
Many are measuring sub 1ppm
A residual floor below 1 ppm is a must, measured as the actual harmonic levels by FFT, and not THD+N. My old AP-2322, from the mid 90s, has a residual around -130dBc (around 300 ppb), and the new APx-555 has a residual sometimes as good as -150 to -160dBc, depending on which harmonic, which frequency, and the level in question. That's around 30 ppb. But, a number of amplifiers and passives are still better than that, requiring magnification techniques, or simply accepting that a device is at or below measurement residual.
Same person also wrote: " I am not replacing tantalums. these sound great! "
In the mid eighties one of the Signetics 5534 in my riaa blew and i received a Ti replacement. The bass was so much worse i just couldn't live with it. The rest of the presentation was also different and not in a good way.
Yeah, the link I posted is one POV.
I found another discussion saying the original Signetics (UK foundry) were the best.
Also mention of 1/f noise variation and Roland selecting for this, not sure which mfr, might have been JRC.
So, there are variations of preference, BUT it seems clear that their are indeed sonic variations between mfr/foundry.
Dan.
I found another discussion saying the original Signetics (UK foundry) were the best.
Also mention of 1/f noise variation and Roland selecting for this, not sure which mfr, might have been JRC.
So, there are variations of preference, BUT it seems clear that their are indeed sonic variations between mfr/foundry.
Dan.
This perked my interest so I got off my *** and did an hour of measuring (with a sound card and single jig setup), and indeed the NJR comparatively fares quite well in the measurements that I did with the various models in my opamp tray. Rohm being a dark horse jagging decent ranks including one of the best.
The pool I checked includes various brands and models including TI - although point is that with time and care sufficiently rigorous comparative results can be compiled without expensive hardware.
Yesterday I measured NE5532 (TI) and overlayed this with the plot of NE5532 (Original Signetics). Noise level is about 10dB higher, quite similar to TL072. I really doubt these will meet TIs specification.
Attachments
Last edited:
Maybe, these were utterly cheap. Just bought two samples at the local Conrad store for comparision. They are ISO certified and should be a trusted source.
Voltwide,
do you have an explanation why the black plot is showing a lot more "spikyness" than the green plot?
do you have an explanation why the black plot is showing a lot more "spikyness" than the green plot?
That is a point I wondered at as well.
No, I do not have an explanation for these spikes.
Test Setup is a breadboaded analog front end with two DIL8 sockets, powered by 24 V, symm output - which explains the high signal level. I measured the symm output through a 30dB symm attenuator feeding the symm line input of EMU-Tracker.
Last edited:
just measured the 2 samples bought from Conrad. The FFT-plot is overlayed with same measurements of Signetics, date code 9127.
It is hard to discern both plots, THD measures in both cases 0,00034%.
Conclusion: original TI-chips are not discernable from Signetics originals, and if so - they are fakes.
It is hard to discern both plots, THD measures in both cases 0,00034%.
Conclusion: original TI-chips are not discernable from Signetics originals, and if so - they are fakes.
Attachments
TI spec sheet states a max noise rating only for the 5532A's (low-noise grade) voltage noise. Current noise is given as typical value, and for 5532 (standard grade) both types of noise don't have a max spec. Therefore, I would not be surprised to find a chip within a huge pile of TI 5532's that is an outlier and might result in several dB more total noise in a given circuit than a "typical" unit -- unless the test rig masks/dominates the noise. Further, the 1/f corners might have a lot of tolerance which would make a difference at LF even for parts with the same 10kHz noise rating.
Yes, that is true:
NE5532A: What are the differences between the NE5532 and NE5532A - General Purpose Amplifier & Other Linear Forum - General Purpose Amplifier & Other Linear - TI E2E Community
for the records - the ones I measured were not A-grade.
NE5532A: What are the differences between the NE5532 and NE5532A - General Purpose Amplifier & Other Linear Forum - General Purpose Amplifier & Other Linear - TI E2E Community
for the records - the ones I measured were not A-grade.
Attachments
It is important to use the A-type! Chips without A are rubbish. Do not buy at Conrad! The prices are incredible! The NE5532A is available for 30 cent.
Did you read my posting? Just showed non-A-types bought from Conrad that are NOT rubbish. Prices of Conrad are certainly at the upper limit, but being ISO-certified they are at least a trusted source. Otherwise I prefer Mouser and TME.
Feel free to send me 2 samples of your 30 cent NE5532 and we will see whether they are rubbish or not.
- Status
- Not open for further replies.