And yet very very few people address this noise source. I know it makes life harder/circuits more complex but it is low hanging fruit.
Thank you very much jackinnj....I'll replace the existing components with what you show. Also thanks for the "headsup" on the Pearl II preamp. If there were a kit available I would make one 🙂
Why do you think a lower "buildout resistor" value is a good thing?
According to the equations for a passive composite RIAA network ... what it does is increase the RIAA cap values - which increases the build cost.
Andy
I was right! The noise calculator was on Bonsai's site . Here's the link to it https://hifisonix.com/so-just-how-quiet-is-your-phono-stage/
Has anyone tried the OPA210/OPA2210?
I just made some simulations in TINA TI with various op-amps and found the OPA210 to give the lowest noise, except for the ideal op-amp used for comparison. I simulated a circuit from Douglas Self (figure 8.41 in "Small Signal Audio Design", 2nd edition, without the subsonic filter). The noise over a 22 kHz BW was 1.5 dB higher with the OPA210 compared to the ideal op-amp. The NE5534 was 4.5 dB higher than the ideal op-amp.
I used the model of 610 ohm + 470 mH for the cartridge.
In the noise calculator from above the OPA210 also shows a good performance. Only the BF862 and LSK389 perform marginally better (around 0.5 dB on the total SNR).
If simplicity is preferred, avoiding the more complex circuits involved with the discrete JFETs, the OPA210/OP2210 looks like a good candidate. The OP209/OP2209 are not far from the performance of the OPA210/OP2210.
I just made some simulations in TINA TI with various op-amps and found the OPA210 to give the lowest noise, except for the ideal op-amp used for comparison. I simulated a circuit from Douglas Self (figure 8.41 in "Small Signal Audio Design", 2nd edition, without the subsonic filter). The noise over a 22 kHz BW was 1.5 dB higher with the OPA210 compared to the ideal op-amp. The NE5534 was 4.5 dB higher than the ideal op-amp.
I used the model of 610 ohm + 470 mH for the cartridge.
In the noise calculator from above the OPA210 also shows a good performance. Only the BF862 and LSK389 perform marginally better (around 0.5 dB on the total SNR).
If simplicity is preferred, avoiding the more complex circuits involved with the discrete JFETs, the OPA210/OP2210 looks like a good candidate. The OP209/OP2209 are not far from the performance of the OPA210/OP2210.
Last edited:
Please tell us also how easy or difficult to get for the part mentioned, some are really difficult.
OPA209 and OPA210 are available in SO8 and SOT23-5 from TI, in SO8 from Mouser.
The duals are perhaps a bigger issue. TI lists the OPA2210 SO8 as "More units week of 27 Feb 2022", so with some luck they should be available soon.
The duals are perhaps a bigger issue. TI lists the OPA2210 SO8 as "More units week of 27 Feb 2022", so with some luck they should be available soon.
With the Riaa circuit diagram as such, little is wrong.
A bit more gain might be welcome by increasing the second amps 3k32 , but that doesn't change the sonic character in any way.
Lowering the Riaa's network resistor values and increasing the cap values, makes hardly any sense, best case only some 0,2dB S/N will be gained.
Using a 1K - 500mH MM Cart, I simulated A-weighted noise for 3 different op-amps, 1) a LM4562 with 2.7nV/rtHz and 1.6pA/rtHz, 2) a 5532 with 5nV/rtHz and 0.7pA/rtHz such as TI specifies and 3) for a super 5532 with 3.5nV/rtHz and 0.4pA/rtHz.
Equivalent A-weighted input noise from 20Hz to 20Khz with these 3 amps was resp. 1.2uV, 793nV and 630nV.
This results resp. in a A-Weighted S/N of 72.4dB, 76.0dB and 78.0dB ref 5mV@1Khz.
Even the 72.4dB S/N is quite a bit above the LP's surface noise and isn't that bad at all.
Obvious is that current noise is an important parameter for MM Carts when looking for a better S/N.
To give a more exact insight for the OP's situation, the Cart in use should be specified, i.e. gain at 5cm/sec@1Khz plus the Cart's R and L.
Hans
P.S. I used 150pF at the input to simulate the cable capacitance.
A bit more gain might be welcome by increasing the second amps 3k32 , but that doesn't change the sonic character in any way.
Lowering the Riaa's network resistor values and increasing the cap values, makes hardly any sense, best case only some 0,2dB S/N will be gained.
Using a 1K - 500mH MM Cart, I simulated A-weighted noise for 3 different op-amps, 1) a LM4562 with 2.7nV/rtHz and 1.6pA/rtHz, 2) a 5532 with 5nV/rtHz and 0.7pA/rtHz such as TI specifies and 3) for a super 5532 with 3.5nV/rtHz and 0.4pA/rtHz.
Equivalent A-weighted input noise from 20Hz to 20Khz with these 3 amps was resp. 1.2uV, 793nV and 630nV.
This results resp. in a A-Weighted S/N of 72.4dB, 76.0dB and 78.0dB ref 5mV@1Khz.
Even the 72.4dB S/N is quite a bit above the LP's surface noise and isn't that bad at all.
Obvious is that current noise is an important parameter for MM Carts when looking for a better S/N.
To give a more exact insight for the OP's situation, the Cart in use should be specified, i.e. gain at 5cm/sec@1Khz plus the Cart's R and L.
Hans
P.S. I used 150pF at the input to simulate the cable capacitance.
That’s because the OP has to replace the LM4562, a dual amp and the better specced NE5534A is a single one.
But it seems that there are other manufacturers promising better NE5532 specs, that’s what I called this a super 5532.
Hans
Actually the noise data in the OPA210/OPA2210 datasheet are inconsistent. Either someone made a big mistake making figure 7-2 of the OPA210/OPA2210 datasheet, or TI tries to pull one of the oldest spec tricks for op-amps with base current compensation.
Manufacturers of op-amps with bipolar input stages with base current compensation often specify the equivalent input noise current with exactly equal impedances driving the positive and negative inputs. No-one ever uses them like that in a real-life application, but it results in nice, low datasheet numbers. That's because the noise of the part of the base current compensation circuit that is common to the positive and negative inputs cancels when the impedances driving the positive and negative inputs are equal.
A typical example is the LT1028: its equivalent input noise current is 1 pA/sqrt(Hz) according to a table in the datasheet, but when you read all the small print and the typical performance graphs, you see it is actually closer to 3.25 pA/sqrt(Hz) under realistic conditions. Hans once measured it and even found 3.4 pA/sqrt(Hz), which is also the value one can calculate by adding up all the major shot noise sources in the schematic diagram at the end of the datasheet.
Now the OPA210/OPA2210. The table in section 6.6 of https://www.ti.com/lit/ds/symlink/opa210.pdf specifies 400 fA/sqrt(Hz) = 0.4 pA/sqrt(Hz) as the equivalent input noise current. That would make it outperform the NE5534A, as the NE5534A also has 0.4 pA/sqrt(Hz) of noise current and a higher noise voltage.
The measuring conditions for the OPA210's 0.4 pA/sqrt(Hz) are not clear. It has an NPN input stage and an input bias current that can be positive and negative, implying that there must be base current compensation.
Section 8.1.1 about noise calculations indicates that that 0.4 pA/sqrt(Hz) applies with and without matched impedances driving the positive and negative inputs. However, Figure 7-2 doesn't. Assuming that the unit for the vertical axis is nV/sqrt(Hz), it indicates a noise level of about 1.6 uV/sqrt(Hz) for a voltage follower driven from a 1 Mohm source resistance. That corresponds to almost 1.6 pA/sqrt(Hz) rather than 0.4 pA/sqrt(Hz)!
If it is really 1.6 pA/sqrt(Hz), then it is not very suitable for moving-magnet amplifiers. Besides, a noise measurement I did a few weeks ago on a voltage reference involving OPA2210's is then all wrong. I'll do some more measurements to find out what is true.
I just did some noise measurements on this circuit
and the results are consistent with 0.4 pA/sqrt(Hz). Apparently figure 7-2 of the datasheet is wrong. Regarding noise, the OPA2210 is a good choice for MM then.
To be precise, I measured the RMS noise over the bandwidth of a Fostex FR2-LE recorder with 96 kHz sample rate, presumably a few Hz to about 44 kHz, with 12.4 kohm or 12.4 kohm//1.13 kohm at the input of the op-amp. There were two 47 ohm resistors and two 47 uF bipolar electrolytic capacitors between the circuit and the microphone input of the Fostex. Correcting for the measured noise of the Fostex itself (including 47 ohm resistors and bipolar capacitors), I found these ratios between the RMS noise levels:
Measured: 0.33048
Calculated, assuming 0.4 pA/sqrt(Hz): 0.32616
Calculated, assuming 1.6 pA/sqrt(Hz): 0.21176
All in all, with very different impedances at the positive and negative inputs, the 0.4 pA/sqrt(Hz) is spot on.
and the results are consistent with 0.4 pA/sqrt(Hz). Apparently figure 7-2 of the datasheet is wrong. Regarding noise, the OPA2210 is a good choice for MM then.
To be precise, I measured the RMS noise over the bandwidth of a Fostex FR2-LE recorder with 96 kHz sample rate, presumably a few Hz to about 44 kHz, with 12.4 kohm or 12.4 kohm//1.13 kohm at the input of the op-amp. There were two 47 ohm resistors and two 47 uF bipolar electrolytic capacitors between the circuit and the microphone input of the Fostex. Correcting for the measured noise of the Fostex itself (including 47 ohm resistors and bipolar capacitors), I found these ratios between the RMS noise levels:
Measured: 0.33048
Calculated, assuming 0.4 pA/sqrt(Hz): 0.32616
Calculated, assuming 1.6 pA/sqrt(Hz): 0.21176
All in all, with very different impedances at the positive and negative inputs, the 0.4 pA/sqrt(Hz) is spot on.
Last edited:
You are right, figure 7-2 does indeed indicate a noise current of around 1.6 pA/sqrt(Hz).
For some reason the OPA1612 seems to have a very similar noise voltage with a 1 Mohm source resistance, but in this case the current noise is specified as 1.7 pA/sqrt(Hz), which fits better with the curve. With the OPA1612 there is also a bias current, which can be either polarity, implying that also in this case base current compensation is used. In this case the bias currents are much higher than with the OPA210 though (+/-60 nA typical versus +/-0.3 nA typical).
For some reason the OPA1612 seems to have a very similar noise voltage with a 1 Mohm source resistance, but in this case the current noise is specified as 1.7 pA/sqrt(Hz), which fits better with the curve. With the OPA1612 there is also a bias current, which can be either polarity, implying that also in this case base current compensation is used. In this case the bias currents are much higher than with the OPA210 though (+/-60 nA typical versus +/-0.3 nA typical).
I just saw your latest post. Great to get a confirmation of the low value and an apparent error in the datasheet. Thanks.
I filled in TI's documentation feedback form and emphasized that this mistake makes their op-amp seem worse than it is. Hopefully they will correct it.
The text along the graph in question mentions “Votlage noise” instead of “Voltage noise” and as you already mentioned it should be in nV/rtHz instead of Vrms.
Added to the misleading graph itself, one might wonder why they are being so inaccurate.
Good that you had a OPA2210 at hand to measure.
This dual amp is a worthy successor of the single NE5534A.
Hans
Marcel,
Not doubting for even a second about the correctness of your measurements, I did the identical exercise but now in LTSpice for the same 44K BW.
Ratio's that I got where of course almost identical with 0.326 and 0.216 for resp. 0.4pA/rtHz and 1.6pA/rtHz, proving that LTSpice is quite reliable with its calculated noise figures.
Hans
Not doubting for even a second about the correctness of your measurements, I did the identical exercise but now in LTSpice for the same 44K BW.
Ratio's that I got where of course almost identical with 0.326 and 0.216 for resp. 0.4pA/rtHz and 1.6pA/rtHz, proving that LTSpice is quite reliable with its calculated noise figures.
Hans