The '553x chips derive from an obscure short-production strain-gauge amplifier, 1101?. This differed from other old designs by having quite large device areas; as opposed to old-think which sought to minimize Silicon.
And of course the TL07x was a cheapener BiFET, LM356??
But between pure discrete and hybrid or monolithic was PCB then Burr-Brown module packaging of audio-intended amplifiers. Flickinger made thousands of PCBs. MCI(?) and Jensen made good audio amps on the analog-computer module best known from Burr-Brown. The Jensen 990's descendants are well-used to this day.
DIY JE-990 Discrete Operational Amplifier
Attachments
Last edited:
Yes. The oldest sheets I find for 1034 cite "professional audio". Ben Duncan's words may be illuminating:
https://www.eevblog.com/forum/beginners/classic-chips/?action=dlattach;attach=1128514
https://www.eevblog.com/forum/beginners/classic-chips/?action=dlattach;attach=1128514
Maybe because its discrete build and runs in class A while the 5532 has a PP output stage that runs in class B?
<The signal path consists of two consecutive differential stages, a single-ended common emitter voltage amplification stage, and a class B push-pull output follower with a current-sensing overload protection.>
Measurements are not the whole story of audio sound IMHO.
Last edited:
I worked in the Research labs at EMI where the Mixers for Abbey Road were designed and built, it was all done in house. That was in the early 70s 1971-1973 and all the consoles had discrete bipolar transistors in all areas. There were NO valves/tubes by then. It is a long time ago but I am sure the majority of transistors were BC109s.
I can't say how many N|E/SE553x devices were in later consoles but as was mentioned above, each channel strip contributes to the overal noise noise level. In its time the NE5532 and NE5534 were game changers and in well designed circuits still stand uo well today. Of course there are now opamps that have superior specs in some areas but to paraphrase on old saying You never got fired for using a 5532. I suspect that few of us could tell the difference between a 5532/2132/4562 et al. We all like to think we have golden ears but in truth, we don't.
Why not design devices that can use old and new op amps? That's what I did.
I included a spot for a 22 or 100 pF capacitor right next to the output and - terminals of the op amp in case the LM4562 or some other fast op amp needed compensation, but you don't need to populate it if you use a NE5532 or a slower op-amp.
Also, of course the output should be decoupled using a resistor, something like 100 to 500 ohms, if you are driving a capacitive load like an audio cable. This is generally true of any op-amp anyways.
The LM4562 has 2.7 nV/Sqrt[Hz] vs NE5532 with 5 nV/Sqrt[Hz]. But to take advantage of this, you need to ensure that the Johnson noise of the resistor network you are feeding the input from is small, which for such a low input noise voltage means you are in the low kOhms. And all of the filters and such must be scaled so that their corner frequencies are the same, so the capacitances must increase as the resistances decrease. This is assuming that other sources of noise have been thoroughly quashed, ground loops, RFI and such, and the audio sources themselves have the low SNR to take advantage of this path. Using relays to switch out unneeded audio sources that can contribute noise also can help.
The open-loop bandwidth on LM4562 is 140 dB, vs. the NE5532 which is 100 dB, so it can be expected that the distortion is lower on the LM4562. Unfortunately, the LM4562 is not tested in Samuel Groner's study:
https://www.nanovolt.ch/resources/ic_opamps/pdf/opamp_distortion.pdf
However, the harmonic distortions don't really show up much until the signals being amplified are nearly at the limits of the output and/or the op-amp is very loaded, with for example the NE5532 or LME49860 for example, though the LME49680 is definitely superior it can be made to distort as well near the supply rails. For most line-level signals this is probably not an issue. Slew limiting at audio frequencies, especially for relatively low amplitude line-level signals, is probably not going to be an issue either. If one places the op-amp in the feedback loop of a power amplifier, the voltage swing and sourced output current are much greater but then probably the output stage of the amplifier is going to be far more limiting of performance than the op-amp itself.
There's probably going to be no measurable, much less discernible, difference between NE5532 and LM4562 unless the entire system is thoroughly engineered to eke that little extra noise margin out of the audio path. I don't think just replacing one part with another is going to make a big difference because the op-amp is just not necessarily the limiting factor to the signal fidelity. LM4562 is not very expensive anymore and generally there is not much harm in using it as long as a few precautions are taken so it does not oscillate but it is not necessarily going to be a massive improvement over the NE5532.
One thing I have really appreciated listening to records, especially classical recordings from the 60s and 70s, for example Van Cliburn or Artur Rubinstein and some of those RCA recordings, is how much attention the sound engineers must have paid to getting things right for the recordings to sound that good. Using the better equipment we have today, the music is alive and shines through and really makes the old music sound even better. It really makes you appreciate how critical sound engineering is to making the music. And this was done using tube or discrete transistor equipment.
I included a spot for a 22 or 100 pF capacitor right next to the output and - terminals of the op amp in case the LM4562 or some other fast op amp needed compensation, but you don't need to populate it if you use a NE5532 or a slower op-amp.
Also, of course the output should be decoupled using a resistor, something like 100 to 500 ohms, if you are driving a capacitive load like an audio cable. This is generally true of any op-amp anyways.
The LM4562 has 2.7 nV/Sqrt[Hz] vs NE5532 with 5 nV/Sqrt[Hz]. But to take advantage of this, you need to ensure that the Johnson noise of the resistor network you are feeding the input from is small, which for such a low input noise voltage means you are in the low kOhms. And all of the filters and such must be scaled so that their corner frequencies are the same, so the capacitances must increase as the resistances decrease. This is assuming that other sources of noise have been thoroughly quashed, ground loops, RFI and such, and the audio sources themselves have the low SNR to take advantage of this path. Using relays to switch out unneeded audio sources that can contribute noise also can help.
The open-loop bandwidth on LM4562 is 140 dB, vs. the NE5532 which is 100 dB, so it can be expected that the distortion is lower on the LM4562. Unfortunately, the LM4562 is not tested in Samuel Groner's study:
https://www.nanovolt.ch/resources/ic_opamps/pdf/opamp_distortion.pdf
However, the harmonic distortions don't really show up much until the signals being amplified are nearly at the limits of the output and/or the op-amp is very loaded, with for example the NE5532 or LME49860 for example, though the LME49680 is definitely superior it can be made to distort as well near the supply rails. For most line-level signals this is probably not an issue. Slew limiting at audio frequencies, especially for relatively low amplitude line-level signals, is probably not going to be an issue either. If one places the op-amp in the feedback loop of a power amplifier, the voltage swing and sourced output current are much greater but then probably the output stage of the amplifier is going to be far more limiting of performance than the op-amp itself.
There's probably going to be no measurable, much less discernible, difference between NE5532 and LM4562 unless the entire system is thoroughly engineered to eke that little extra noise margin out of the audio path. I don't think just replacing one part with another is going to make a big difference because the op-amp is just not necessarily the limiting factor to the signal fidelity. LM4562 is not very expensive anymore and generally there is not much harm in using it as long as a few precautions are taken so it does not oscillate but it is not necessarily going to be a massive improvement over the NE5532.
One thing I have really appreciated listening to records, especially classical recordings from the 60s and 70s, for example Van Cliburn or Artur Rubinstein and some of those RCA recordings, is how much attention the sound engineers must have paid to getting things right for the recordings to sound that good. Using the better equipment we have today, the music is alive and shines through and really makes the old music sound even better. It really makes you appreciate how critical sound engineering is to making the music. And this was done using tube or discrete transistor equipment.
My Studer A810 runs with a dozens of 5532, the Revox B77 was build using discrete transistors. For my ears, the Studer sounds slightly different. Maybe this is because the circuits are discrete vs. integrated or other differences. Studer has a preferable sound to me, but its build with much more effort for perfection, too. I wouldn't say that both have complete different sonic characters, but a slight tendency to typical OP amp sound could be spotted with the Studer. Not bad in my ears, but noticeable.
Btw, Revox A77 and B77 had different sound characters, too and both use the same circuits, just different transistors.
I would rate the pre- transistor era at Abbey Road much higher in terms of sonic excellence. Hearing the famous Beatles album with the same name, it has a transparency, drive, energy and airyness the later 1970's productions were not able to catch up with.
This technology change from tube to transistor is clearly audible in the audio productions, and not for the favor of the transistor.
The 5532 could have been a game changer for the integrated circuits audio qualities, but a very good discrete amp circuit still would outperform it. I think its a very good IC to sound that close to a good discrete circuit as it does.
Btw, Revox A77 and B77 had different sound characters, too and both use the same circuits, just different transistors.
I would rate the pre- transistor era at Abbey Road much higher in terms of sonic excellence. Hearing the famous Beatles album with the same name, it has a transparency, drive, energy and airyness the later 1970's productions were not able to catch up with.
This technology change from tube to transistor is clearly audible in the audio productions, and not for the favor of the transistor.
The 5532 could have been a game changer for the integrated circuits audio qualities, but a very good discrete amp circuit still would outperform it. I think its a very good IC to sound that close to a good discrete circuit as it does.
Last edited:
You DO realize that the 990 was designed by the brilliant Deane Jensen FORTY years ago, don't you?
I think the first tube to transistor transition was much smoother due to germanium transistors being the first transistors used with their inherent lower crossover distortions while input , output transformers were still the norm for impedance convertion with lower noise than the best semiconductor available even today...and input -output transformers also made possible for all NPN transistor designs to be used all throughout allowing for even lower crossover distortions. lower H2 and variable damping and high order harmonic reduction simillar to Dolby principle of sliding band noise reduction due to higher output impedance and transformer saturation curve which made transformers to be very efficient noise reducers.
Single bipolar transistor input circuits are less prone to hard clipping than differential input though...
I only tried lm4562 in I/V converters where the current was big enough to not cause significant current noise and in line out unity gain circuits where it clearly sounds better than ne5532.After reading Salas opinions on lm833 a few years ago i was interested in why Philips and Japanese manufactureres preffered it and UPC4570 for the I/V stage after the DAC over ne5532.Denon used UPC4570 in I/V and ne5532 for line out unity gain in their best cd players as they used low current output Burr Brown DAC's.Salas measured lm833 and observed it has less common mode distortions than ne5532 too.
LM833 made me think a lot with no clear answer on why in the same identical I/V circuit it has very noticeable higher gain than all the other op-amps i tried.
Never tried lm4562 in high gain circuits cause my preffered PHONO preamps need lower current noise , nor did i try ne5534 in phono just because i now preffer opa2228 over anything with a minimum gain of 5x.At least for my ears there's nothing better than opa2228 for higher gains and ne4562 for unity gain and i rarely get out of this combo recipee which i find optimum...lm6172 is another beast i optimised for audio use cause as stock is not able to reproduce the right level under 50Hz,while opa2132 and njm2043 my main workhorses for mosly everything when not too peaky, but since i discovered Technics Class AA i don't really care about op-amps specs too much...Technics class AA made my njm4556 headphones amps sound like TPA6120 on 250 ohms headphones ...LM833 is clearly and audibly better than ne5532 in I/V circuits, but now we have opa1612 and its footstep followers and new dac's that don't even need low current noise inputs and much of the old needs are gone...Me being mostly a phono and tape user i simply ignore all the new op-amps discoveries...
I think of NEVE rebranding NE5532 for high end use in full balanced mode before TPA6120 in its RNHP....
Other than that i can't make any silicon circuit sound like tubes and that might be because i never made a transformer coupled transistor circuit but won't know for sure until i try that too...
Last edited:
