Reply to post #119:
Those typical characteristics make no sense at all. For one thing, the 1/f-noise corner frequency is normally higher for the noise current than for the noise voltage of a bipolar transistor.
Judging by the input bias current and the fact that there is no base current compensation, I would expect the input noise current density of the NE5532 to drop to √(2q • 200 nA) ~= 0.2532 pA/√Hz at high frequencies. I never measured if it really does, though.
Those typical characteristics make no sense at all. For one thing, the 1/f-noise corner frequency is normally higher for the noise current than for the noise voltage of a bipolar transistor.
Judging by the input bias current and the fact that there is no base current compensation, I would expect the input noise current density of the NE5532 to drop to √(2q • 200 nA) ~= 0.2532 pA/√Hz at high frequencies. I never measured if it really does, though.
I would be more worried about your voltage divider to bias the input at 7.5Volt.
As you can see, PSRR is quite miserable probably the reason for your unexpected noise spectrum.
You better use a high grade voltage regulator like the LT1763 or one from the LT3045 serie.
Hans
As you can see, PSRR is quite miserable probably the reason for your unexpected noise spectrum.
You better use a high grade voltage regulator like the LT1763 or one from the LT3045 serie.
Hans
Attachments
I agree, IMHO better use not unipolar but bipolar power supply, as in https://www.diyaudio.com/community/threads/my-opa1656-riaa-preamp-one-more.425815/post-8000058 This eliminates the need for power supply filtering - the operational amplifier's 120...140 dB PSRR is much better than any integrated regulator in the case of a single-polar supply. In addition, there is no need for a bunch of separating electrolytes (with too long charging time at power on state).
Last edited:
Are you comparing cartridge versus reverse RIAA circuit or cartridge versus open input?
The point is that a preamplifier has both equivalent input noise voltage and equivalent input noise current. The effect of the current increases with source impedance. When you measure without the cartridge inductance, the source impedance is too low and you underestimate the effect of the equivalent input noise current.
In a typical moving-magnet phono preamplifier with a valve or a FET as input device, it's the thermal noise of the 47 kohm input parallel resistor that dominates the equivalent input noise current. The resistor produces about 0.5869 pA/√Hz of thermal noise current at 20 degrees Celsius. Hence, when you want to compare phono preamplifiers in this category and use a too low source impedance, the one with the lowest noise voltage will perform best, and as they all have the same noise current, that is also the best with a real cartridge. That is, the measurements are inaccurate, but the ranking is still correct.
It gets worse when some of the preamplifiers have a bipolar input stage. The shot noise of the base current then adds to the equivalent input noise current. With a bipolar op-amp with base current compensation, the base current compensation also adds to the equivalent input noise current, sometimes quite dramatically. When you then measure with a too low source impedance, it could very well be that the amplifier that performs worst with the cartridge as source gives the best result in the measurement.
(Since 1939, techniques have been known to reduce the equivalent input noise current below 0.5869 pA/√Hz while still having a 47 kohm input resistance and without needing to cool the input resistor with liquid nitrogen or helium. The inventor called it an electrically "cold" resistance, but there are other terms in use for the exact same thing. It is not often used for phono preamplifiers, but it is used all over the place in all sorts of radio receivers.)
That's nice to read!
Last edited:
I know that, but my turntable has an unipolar supply only, and I wanted to see how far I could get by utilizing that and not going the route of converting unipolar supply to bipolar supply, for instance with a charge pump. I ran the pre-amp externally with a bipolar supply as well (after making the neccesairy changes to the pre-amp) and the difference was not such that I felt I needed to use a bipolar supply. I must admit at the time of that test the pre-amp wasn't shielded, so I might have been listening to hum pick-up rather than supply induced hum. I need to say though that neither noise nor hum is audible at normal listening distances in my current unipolar supply, shielded, pre-amp set-up.
I tried comparing noise with the cartridge inserted versus open input (with the cartridge removed from the tone arm)
Thanks for the explanation. I have also read the post on audiosciencereview linked in a previous post.
That same article tested the NE5534 in an MM set-up, and it faired quite well. I am using the NE5532, which I think is slightly worse, but the NE5534 came in a close second after a JFET input Opamp and was a better performer than the OPA134. I also read that D. Self wrote that the NE5532/34 is a sensible choice for an MM pre-amp.
I might experiment further with the power supply. The two options I see are bipolar regulated (via a charge pump) or perhaps I will go the battery route. I have measured the current consumption at 7 mA idle. Allowing 2-3 mA for the drive signal, and I can expect 10 mA current draw, which I think is fine for a battery supply.
The amplifier is dead quiet according to Arjen, so I wouldn't worry about it at all. As you can see in your own plot or by inspecting the circuit, power supply rejection gets better with frequency, so power supply noise can at most affect the noise at very low frequencies somewhat.
What you simulated is the transfer from the supply to the output (while what is usually called the PSRR is referred to the input). At audio frequencies up to 50 Hz, the gain from input to output is roughly 300 times, so the 5 nV/√Hz voltage noise of the NE5532 results in 1.5 μV/√Hz at the output. The gain you simulated from the supply to the output drops from -7 dB to -24 dB from 20 Hz to 50 Hz.
The supply is supposed to be regulated. Assuming a really poor regulator with 1 μV/√Hz of output noise, the noise contribution of the supply at the output is more than 10 dB below the effect of the voltage noise of the NE5532 at 20 Hz, and more than 20 dB at 50 Hz.
At all frequencies between 150 Hz and 20 kHz, the simulated gain from supply to output is below -50 dB and the gain from input to output is above 10 dB. That means that 1 μV/√Hz at the supply corresponds to less than 1 nV/√Hz equivalent at the input.
Regarding hum, suppose the unregulated supply has 1 V RMS of ripple at either 50 Hz or 100 Hz and the poor regulator has 60 dB of ripple rejection. That boils down to 1 mV RMS at the regulated supply. With the simulated gain of -24 dB at 50 Hz, -40 dB at 100 Hz from supply to output, the resulting ripple at the output would be either -84 dBV at 50 Hz or -100 dBV at 100 Hz. With a nominal level of about 160 mV (30 dB above 5 mV), so -16 dBV, the signal-to-hum ratio would be 68 dB (50 Hz) or 84 dB (100 Hz). The latter case seems more likely, as single-phase rectification is not often used. If needed, the suppression could always be improved by using two RC sections instead of the 4.7 kΩ - 47 μF that is now used. No need to use expensive and almost impossible to hand solder regulators, as far as I can see.