Why don't you take a modern mixer which will have zillions of 5532/4s and replace them all with your 'better choices'.
I am willing to bet loadsa $$$ that your THD/noise bla bla will become a LOT worse ... IF the thing stays stable.
If someone who knows what he is doing spends a HUGE amount of work & $$$ on your Frankenstein mixer, you might get back the performance of the cheapo 5532/4s version but probably no better.
How do I know this? Kingston, in GroupDIY did just this and demonstrated how OPA rolling comes a VERY POOR second to proper earthing and decoupling.
And guess what? Practically ALL your 'better choices' are MUCH fussier than cheapo 5532/4s so hardly ever give better performance except when only 1 OPA is used.
If a modern mixer is expected to output 18 to 20 bits, it's hard to imagine getting the job done with 5534s and such. Even properly dithered operations in 24 bit or 32 or whatever can keep to a noise floor standard that's really hard to get with all the otherwise required analog stages. For a simpler four/ 8/ 16 channel mixer, clean, no special effects in-line, a classic analog mixer works great, and would be appropriate for most DIYers. Something like a small Mackie or Yamaha stage mixer is very perfected, but anything bigger (Modern) is a real challenge for DIY.
Much thanks, as always,
Chris
Much thanks, as always,
Chris
You are of course right Marcel. Even this Jurassic analogue dodo, whose heyday was broadcast desks in the early 80s, has contributed to the march of evil digits. Simple Arbitrary IIRs
Just the cost of precision pots and capacitors make it worthwhile to do as much as much processing digitally as possible today as computing power is MUCH cheaper ... let alone the simple act of mixing a zillion channels
thisusername, I apologise for my rant about replacing 553x. Today, there are many devices which are better for some purpose than 553x.
I've mentioned NJM2068 for MM preamps. I like MAX410 for when Vcc is limited eg LNprimer There's a number of low current OPAs that now make it sensible to incorporate inside a P48V microphone eg Charis
I am not averse to using a not-553x ... when its appropriate
But as a general purpose OPA .... Self looks at various uber OPAs but NONE of them is a replacement for 553x. https://www.diyaudio.com/community/...st-possible-thd-n-really-the-best-way.367692/
In fact only OPA2134, which he damns with faint praise, might give better performance if swapped for 5532 cos it is the least fussy of decoupling and it has excellent RFI immunity. All the other $$$ candidates, LM4562, OPA 627 bla bla have 'features' which usually give worse performance as drop in replacements for 5532.
Last edited:
I've quantified the equivalent Rnv for a 556 mH 1342Ω dummy cart in both the RAW (flat) and RIAA-EQ'd domains by measurement and found the "near-typical" MM cart to have an Rnv of about 30 KΩ (20 kHz BW) in the flat domain and somewhere in the 8-10KΩ range when RIAA EQ is applied. https://proaudiodesignforum.com/forum/php/viewtopic.php?t=1354
The FFTs text field state 534 mH but I realized later I was using the right channel which measures about 556 mH.
Here are some of the earlier measurements for an op amp-based MM instrumentation amp input. This may be what Richard is referring to. My transistor comparisons are for MC preamps and linked-to elsewhere. These are the MM op amp comparisons.
Noise Test Using The NJM2114, 2068, 5532, 33078 and LME49720: https://proaudiodesignforum.com/forum/php/viewtopic.php?p=8665&#p8665
Stanton 681 mounted on the tonearm comparing the noise performance of the OPA1612 to the NJM2068DD and OPA2134: https://proaudiodesignforum.com/forum/php/viewtopic.php?p=9140#p9140
I didn't expect the bias-current-compensated amps to do well but I wanted to see just how bad they were. One Ib-compensated that did do well was the OP270.
I have some 70s-era gold lead 2N440X and 2N5088 I ought to measure.
Has anyone tried the ZTX1053A at lower collector currents? I'm having great results with the ZTX851 in an MC design and have looked at the ZTX1053A and intend to use it in a mic preamp. Never tried it at low Ic though but it has high gain and a lower estimated rbb than a 2N4403. The ZTX1053A might be a good candidate for an MM design.
BTW I don' think the 5532 is going EOL though the sources for it in PDIP are diminishing. TI is the last 5534 producer and it is EOL. The OPA2134 is also EOL.
Last edited:
Richard Visee also found that the equivalent series resistance of his moving-magnet cartridge increases substantially over the audio band.
Nice to see noise density plots with real cartridges! I don't quite follow the first post you linked to, though. If I understand it correctly, you try to find a resistor (a real physical resistor that generates thermal noise) that gives the same integrated output noise as you get with a cartridge.
Doesn't the result depend on the voltage and current noise of the phono amplifier? If you had a hypothetical amplifier with only voltage noise, the effect of the cartridge impedance would be that it adds its own thermal noise voltage. The value of the resistor would then be a (weighted) average of the equivalent series resistance (real part of the impedance) of the cartridge.
If you had an amplifier with lots of current noise, it would be mostly the magnitude rather than the real part of the impedance that matters, so you would end up with a larger resistor.
Yes and yes.
I was looking for a Rnv I could use for test with that particular topology and IC.
The equivalent Rnv was significantly higher than the 2K-ish value JRC uses to grade the NJM2068 DD rank which is why I looked at it. 2K seemed way too low and it was.
I did consider the results instructive however and typical of what we'd see with bipolar input op amps we'd typically use.
Your mileage may vary.
I'm confident FET input op amps are going to show a lower equivalent Rnv but I haven't had a chance to make comparisons since I was running it on a production board and didn't want to roll op amps.
Last edited:
I'm a bit late to the party, but I've recently been measuring an RIAA preamp spectrum with a dummy cartridge consisting of 5 100mH inductors in series, making approx 500mH (discounting mutual inductances between them). Series resistance is close to 820 ohms.
To get clean(ish) spectra I had to go far away from mains electricity and test on battery power, the inductors are unshielded and very sensitive to EMI...
All plots are for 5mV rms 1kHz input, preamp gain about 38.5dB, classic single opamp non-inv RIAA circuit.
But here are the 3 plots for NE5532 (~5nV/√Hz, 0.5pA/√Hz):
for OPA1612 (~1nV/√Hz, ~1.7pA/√Hz):
and for OPA1652 (~5nV/√Hz, 0pA/√Hz):
Clearly the OPA1612 fares very poorly despite its very low voltage noise (its by far the quietest with shorted input). The current noise of the 47k load resistor actually masks the difference between the NE5532 and OPA1652 (which has negligible current noise). Short of electronically cooling the 47k load, there is little point moving to JFET from NE5532, and thus even less point moving from NE5534A to JFET for lower noise (although EMI susceptability would be a valid reason).
In defence of the OPA1612 is actually wins below about 40Hz...
If I had a version of that circuit with an NE5534A I'd have tested it, but its a very compact circuit using dual opamp on the inputs. There is a rumble filter that causes the roll off below 20Hz.
I realize I'd not actually seen graphs of the effect of current noise in an MM phono preamp before, it makes the theory spring to life to see the 10kHz hump!
For comparison finally a graph with just the an 820R resistor without inductance for NE5532:
Here the RIAA response shows up nicely in the noise floor without the inductance to complicate things.
To get clean(ish) spectra I had to go far away from mains electricity and test on battery power, the inductors are unshielded and very sensitive to EMI...
All plots are for 5mV rms 1kHz input, preamp gain about 38.5dB, classic single opamp non-inv RIAA circuit.
But here are the 3 plots for NE5532 (~5nV/√Hz, 0.5pA/√Hz):
for OPA1612 (~1nV/√Hz, ~1.7pA/√Hz):
and for OPA1652 (~5nV/√Hz, 0pA/√Hz):
Clearly the OPA1612 fares very poorly despite its very low voltage noise (its by far the quietest with shorted input). The current noise of the 47k load resistor actually masks the difference between the NE5532 and OPA1652 (which has negligible current noise). Short of electronically cooling the 47k load, there is little point moving to JFET from NE5532, and thus even less point moving from NE5534A to JFET for lower noise (although EMI susceptability would be a valid reason).
In defence of the OPA1612 is actually wins below about 40Hz...
If I had a version of that circuit with an NE5534A I'd have tested it, but its a very compact circuit using dual opamp on the inputs. There is a rumble filter that causes the roll off below 20Hz.
I realize I'd not actually seen graphs of the effect of current noise in an MM phono preamp before, it makes the theory spring to life to see the 10kHz hump!
For comparison finally a graph with just the an 820R resistor without inductance for NE5532:
Here the RIAA response shows up nicely in the noise floor without the inductance to complicate things.
Last edited:
There is something odd about the NE5532A equivalent input noise current specification.
As the NE5532A has a differential pair input stage without base current compensation, it seems reasonable to assume that its noise current mostly comes from base current shot noise and base current 1/f noise. The white noise level (shot noise) should then actually be 0.2532 pA/√Hz, as this is the value corresponding to its 200 nA bias (base) current. That would mean that the specified noise current at 1 kHz is actually largely 1/f, and the noise current at 10 kHz is far lower.
As the NE5532A has a differential pair input stage without base current compensation, it seems reasonable to assume that its noise current mostly comes from base current shot noise and base current 1/f noise. The white noise level (shot noise) should then actually be 0.2532 pA/√Hz, as this is the value corresponding to its 200 nA bias (base) current. That would mean that the specified noise current at 1 kHz is actually largely 1/f, and the noise current at 10 kHz is far lower.
Back in the early 2000’s, the then Phillips was a big supplier of 5532 and 5534. The dirty secret was that for both devices back then, they were selling in volume to distribution for under 10 $ cents and terrible margins. The guy that ran that business killed both parts off and the company stepped out of the business. Another little factoid is that Philips was for a while the biggest analog IC company in the world driven by consumer and TV - they supplied all the TV chip sets for the Japanese and European brands.
I suspect part of the reason TI have stepped out is the NE553x devices aren’t making good margins and the process is legacy which is a PITA for the fab guys.
I suspect part of the reason TI have stepped out is the NE553x devices aren’t making good margins and the process is legacy which is a PITA for the fab guys.
Rumors of the NE5532P and NE5532AP's demise are premature as both are still available in PDIP from TI and Nisshinbo. Nisshinbo also make on a modern process the NJM5532C in SMT. At Mouser, Nisshinbo are charging a premium for the 5532 compared to TI.
Finding a low-noise bipolar input dual without bias current compensation is indeed a challenge these days with only the 5532 and NJM2068 coming to mind.
Mark's measurement of the OPA1612 is very similar to mine. The LME49720 equally sucks in an MM preamp even when you find a quiet one without burst noise or EMI.
I think it’s the process mix at a manufacturer that really dictates what they will hang onto and what they will let go, unless they can outsource it, but that comes with its own can of worms and costs that require payback in 2 yrs or better- hard to do on parts making low margins. Semi companies also build die stocks ahead of end of life. At Philips, the BF862 ran for about 3 or 4 yrs from die stock because the process they were made on was shut down.
They are also available from many sources referenced on LCSC (not much from TI...)
I do not have the test equipments able to measure the performance of those eastern chips, but as a DIYer I think we should consider those suppliers as very interesting sources for 'old' chips in DIY friendly packages... : HGSemi make TDA2050, LM1875, LM3815, MAX7219, uPC1237, NE5532 & 34, etc..
Maybe someone here could measure those 5532 and 5534 chips and avoid never ending chat regarding eastern chip quality (...I mean in Audio use cases)
I do not have the test equipments able to measure the performance of those eastern chips, but as a DIYer I think we should consider those suppliers as very interesting sources for 'old' chips in DIY friendly packages... : HGSemi make TDA2050, LM1875, LM3815, MAX7219, uPC1237, NE5532 & 34, etc..
Maybe someone here could measure those 5532 and 5534 chips and avoid never ending chat regarding eastern chip quality (...I mean in Audio use cases)
I used the OPA1612 in an MC front-end amp, and it was pretty ok - not too dissimilar to an AD797 (0.9nV/rt/Hz IIRC) - nowhere near a discrete input amp which is 4x better on spot noise. I guess the OPA1612 bias compensation is not a problem with very low Rsource transducers.
I've gone over to using the OPA1641/2 for general purpose line level applications - the LM4562 and its siblings have indeed gotten a bad rap for burst noise, although I never had a problem using them as low noise high PSRR regulators - but the output is followed with a large LP filter (47-220 Ohms with 1000uF). The PSRR is superb all the way out to 100 kHz.
I've gone over to using the OPA1641/2 for general purpose line level applications - the LM4562 and its siblings have indeed gotten a bad rap for burst noise, although I never had a problem using them as low noise high PSRR regulators - but the output is followed with a large LP filter (47-220 Ohms with 1000uF). The PSRR is superb all the way out to 100 kHz.
And that's why BJTs are often chosen as input stage devices for MC amps. You can get the rbb down below 3 ohms with suitable device selection and parallelism, so the enoise becomes tiny. At the same time the BJT's inoise is pretty much immaterial, as this quote points out.