From what I've seen of Benchmark, I'd "guess" (I know, I know) it was performance increase.
Unless they actually advertise that they made this op amp change.
mlloyd1
So you haven't measured it. The 4562 has more current noise, which might be an issue, or not, depending on the circuit details and whether noise or distortion dominate. Apart from this its slightly better than 5532, but more expensive - however its getting cheaper and is only about 3 times the cost of the 5532.
For many DACs the choice of opamp is more convenient if 5V-capable and rail-to-rail, saving the need for extra power supplies. There are some excellent performers out there, not normally on the audio radar (since not +/-18V), I've played with the AD8656 dual which is genuinely rail to rail, can easily drive 16ohm headphones directly (+/-220mA), 2.7nV voltage noise, CMOS input (negligable current noise), typical offset 50µV and 0.0007% SiNaD. Great match for a headphone DAC. One downside is relatively high flicker noise knee though, but it looks promising for future CMOS input opamps.
This inspired me to knock up some Python and plot combined noise plots for various opamps, total theoretical voltage noise due to opamp and a single resistance R (Johnson plus current noise).
Clearly, if 5V devices are ignored, NE5534A still wins from 2k to 40k impedance, basically the range of almost all active filters and general signal processing... The 5532 isn't as quiet but cheaper, dual and no compensation caps needed, still good for most of this range.
This is MM cartridge territory.
Going to lower impedances the LM4562 comes into its own, and for even lower the AD797 (MC cartridge territory, dynamic microphones)
But if you are prepared to limit to 5V operation the AD8656 totally trashes all comers above 1k.
I've randomly assigned JFET and CMOS input current noise density as 10fA, which is probably an over-estimate - its effectively zero at any practical impedance.
Clearly, if 5V devices are ignored, NE5534A still wins from 2k to 40k impedance, basically the range of almost all active filters and general signal processing... The 5532 isn't as quiet but cheaper, dual and no compensation caps needed, still good for most of this range.
This is MM cartridge territory.
Going to lower impedances the LM4562 comes into its own, and for even lower the AD797 (MC cartridge territory, dynamic microphones)
But if you are prepared to limit to 5V operation the AD8656 totally trashes all comers above 1k.
I've randomly assigned JFET and CMOS input current noise density as 10fA, which is probably an over-estimate - its effectively zero at any practical impedance.
Code:
from math import pi, sqrt
import numpy as np
import matplotlib.pyplot as plt
from matplotlib.ticker import FormatStrFormatter
# data for opamps, name, current noise density, voltage noise density, plot colour
opamps = [ ('NE5534A', 0.4e-12, 3.5e-9, 'r'),
('NE5532', 0.7e-12, 5.0e-9, 'g'),
('LM4562', 1.6e-12, 2.7e-9, 'y'),
('TL072', 1e-14, 18e-9, 'b'),
('AD8656 (5V)', 1e-14, 2.7e-9, 'k'),
('AD797', 2e-12, 0.9e-9, 'c'),
('uA741', 8e-12, 80e-9, 'm') ]
# Boltzmann's constant and Johnson noise factor
k = 1.38e-23
T = 300
factor = 4*k*T
# setup plot
fig = plt.figure()
plt.title ("noise")
ax = fig.add_subplot (111)
ax.set_xlabel ('R')
ax.set_ylabel ('nV')
# vector of resistances
Rvec = np.logspace (1.3, 6, 100)
# Johnson noise in nV
Rnoise = 1e9 * np.sqrt (factor * Rvec)
ax.loglog (Rvec, Rnoise, 'k', label='Johnson noise')
for device, CN, VN, colour in opamps:
Vnoise = 1e9 * np.ones ([Rvec.size], float)*VN # voltage noise of opamp in nV
Inoise = 1e9 * CN * Rvec # voltage noise due to current noise into R
ax.loglog (Rvec, Vnoise, colour)
ax.loglog (Rvec, Inoise, colour)
total_noise = np.sqrt (np.square(Rnoise) + np.square(Vnoise) + np.square (Inoise))
# plot total noise with circle
ax.loglog (Rvec, total_noise, colour+'o', label=device)
# Make axes tidy and set limits
ax.grid(which = 'both')
ax.set_ylim (1, 200)
for tick in ax.get_xticklabels(which = 'both'):
tick.set_rotation(90)
ax.xaxis.set_minor_formatter(FormatStrFormatter("%i"))
ax.xaxis.set_major_formatter(FormatStrFormatter("%i"))
ax.legend(loc='upper left', shadow=True, fontsize='x-large')
plt.show()Hi Mark,
Limit this to dual devices. The NE5534A is a single op amp. You can't put it in the same place as a dual. Also, TL072 isn't a particularly great sounding op amp. Looking at your information I can see why someone would try it and leave disappointed. It is a Fet input op amp as well.
Try modeling an OPA2134 (another FET) and the OPA2132 for giggles.
-Chris
Limit this to dual devices. The NE5534A is a single op amp. You can't put it in the same place as a dual. Also, TL072 isn't a particularly great sounding op amp. Looking at your information I can see why someone would try it and leave disappointed. It is a Fet input op amp as well.
Try modeling an OPA2134 (another FET) and the OPA2132 for giggles.
-Chris
Noise levels only matter for low level parts of the signal chain, otherwise all opamps will sound the same (under double-blind conditions). TL072s were/are commonly used in tone controls and EQs to avoid DC offsets causing potentiometer noise - for this application lack of potentiometer noise is what constitutes sounding good!
This isn't modelling, its calculating from the published noise specifications. Device variation is not allowed for, variations between different manufacturers isn't allowed for, flicker noise knee frequency isn't allowed for.
But it shows that choice of opamp for noise performance is completely dependent on the circuit impedances involved. At 10k for instance changing a 5532 for anything but two NE5534A's (on an adapter!) makes no sense. At 600 ohms changing to an AD797 makes sense. Of course if distortion is also important these priorities will change somewhat.
The OP2134's sweet spot seems to be 20k to 100k, probably a good TL072 replacement for those tone controls or active filter use.
This isn't modelling, its calculating from the published noise specifications. Device variation is not allowed for, variations between different manufacturers isn't allowed for, flicker noise knee frequency isn't allowed for.
But it shows that choice of opamp for noise performance is completely dependent on the circuit impedances involved. At 10k for instance changing a 5532 for anything but two NE5534A's (on an adapter!) makes no sense. At 600 ohms changing to an AD797 makes sense. Of course if distortion is also important these priorities will change somewhat.
The OP2134's sweet spot seems to be 20k to 100k, probably a good TL072 replacement for those tone controls or active filter use.
OP37 only stable for gains >= 5, so rarely useful. The figures are good apart from the input bias cancellation, which means the current noise in the datasheet can only be trusted for equal impedance on both inputs, rarely the case (except for a differential amp circuit).
Probably a good choice for balanced microphone preamp, unknown noise performance in follower, filter, and similar circuits at high impedance due to the unknown current noise in asymmetric configurations.
Figure 7 in the datasheet https://www.analog.com/media/en/technical-documentation/data-sheets/OP37.pdf does show curves for both matched and unmatched input resistors from which one can infer the common mode and differential mode current noises if you want. Its the same type of plot as mine so can be compared.
Probably a good choice for balanced microphone preamp, unknown noise performance in follower, filter, and similar circuits at high impedance due to the unknown current noise in asymmetric configurations.
Figure 7 in the datasheet https://www.analog.com/media/en/technical-documentation/data-sheets/OP37.pdf does show curves for both matched and unmatched input resistors from which one can infer the common mode and differential mode current noises if you want. Its the same type of plot as mine so can be compared.
Noise performance in modern high quality op amps is rarely a valid consideration, as they all have below-audible performance. Phono preamp noise performance is invalid due to the fact that phonographs records' signal-to-noise doesn't exceed 70 db. That leaves only microphone, tape head, and guitar preamps as applications where noise performance counts. Microphones and tape heads are generally low impedance sources, which means that bipolar devices will be better; guitar pickups' high source impedance dictates an FET approach. For DIY use (non-SMD, and reasonable $$), that leaves you either an NE5532 or LM4562 for mics and tapes, and an OP2134 for guitars. That's it!
Hi dotneck335,
While you have provided a list of very good op amps, your argument doesn't hold water. While the s/n of the record is lower than most of the electronics out there, it still matters. The only time you hear record noise is between cuts or in a silent passage. At all other times the music masks the noise of the record surface. However, when the tonearm is cued up, or at rest in it's holder you're going to hear the noise of the electronics.
I'm lucky enough to own an extremely high performing phono EQ amp, some okay ones and some that really should never have seen the light of day. Most of the marginal and poor phono preamps are hiding inside receivers and integrated amplifiers. I'm pretty certain that other iffy circuits live inside sound cards and sold separately as "improvements". In short, they are everywhere and we really should be working to create better than okay phono preamps.
One thing you will notice is any excessive hiss or hum added to the music. You can even hear this in some while there is an audio signal. While the op amps you listed are all good choices (I would go with the NE5534A over the NE5532A), circuit design could create a noisy preamplifier. A poor power supply could do the same thing. PCB design can also bring a good electrical design down, without even oscillation as a cause.
-Chris
While you have provided a list of very good op amps, your argument doesn't hold water. While the s/n of the record is lower than most of the electronics out there, it still matters. The only time you hear record noise is between cuts or in a silent passage. At all other times the music masks the noise of the record surface. However, when the tonearm is cued up, or at rest in it's holder you're going to hear the noise of the electronics.
I'm lucky enough to own an extremely high performing phono EQ amp, some okay ones and some that really should never have seen the light of day. Most of the marginal and poor phono preamps are hiding inside receivers and integrated amplifiers. I'm pretty certain that other iffy circuits live inside sound cards and sold separately as "improvements". In short, they are everywhere and we really should be working to create better than okay phono preamps.
One thing you will notice is any excessive hiss or hum added to the music. You can even hear this in some while there is an audio signal. While the op amps you listed are all good choices (I would go with the NE5534A over the NE5532A), circuit design could create a noisy preamplifier. A poor power supply could do the same thing. PCB design can also bring a good electrical design down, without even oscillation as a cause.
-Chris
op amp differences
Hi,
I must admit to being a little baffled by the performance descriptions of some of the more popular op amps. Surely if they 'colour' the sound in any way this means they are not 'hi-fi'? I always thought that hi-fi meant (more or less) faithful reproduction of sound. How do we know which ones are hi-fi?, and does this also mean the rest aren't?
regards john
Hi,
I must admit to being a little baffled by the performance descriptions of some of the more popular op amps. Surely if they 'colour' the sound in any way this means they are not 'hi-fi'? I always thought that hi-fi meant (more or less) faithful reproduction of sound. How do we know which ones are hi-fi?, and does this also mean the rest aren't?
regards john
The New Japan Radio versions of the 5532 are well built and they perform well. I have not measured them recently, but in the past, they had the lowest distortion compared to TI, Philips, Fairchild, Exar, and Rohm - all 5532 are not the same by any means and NJR has their bipolar complementary process figured out.
The NJM5532DD is the prime grade DIP-8 through-hole version, and the NJM5532MD is the prime grade SO-8 SMD version, both available at Mouser right now.
The NJM5532DD is the prime grade DIP-8 through-hole version, and the NJM5532MD is the prime grade SO-8 SMD version, both available at Mouser right now.
Doctors are fond of saying "show me a drug that has no side effects and I'll show you a drug that does nothing".
Take that thinking into audio and it gives you a more realistic mindset.
Yes, it appears that NJR are graphing THD as ~-107db @ 1KHz. I don't see this spec listed on either T.I. or ON data sheets, but I have it listed in my charts as ~-94db. (not sure where I found this!) So it appears to be an improvement. Interesting, though, that it is also just a BIT slower --8 nv/√Hz on the NJR vs. 9 on the TI and ON chips.
I wonder why NJR doesn't provide this info on many of their other ICs.
Interesting that they send you some 2068's in that bunch. They are low noise, but their output drive isn't as good as the 5532. Was this a regular store, or something along the lines of an Ebay "store"?
Not to be nosy, but how many 5532's did you order? It looks like they simply couldn't fill your order. Are you going to keep the JRC2068D or return them?
-Chris
Not to be nosy, but how many 5532's did you order? It looks like they simply couldn't fill your order. Are you going to keep the JRC2068D or return them?
-Chris