Stereo phono input testing - low freq noise

Since we are measuring phono stages here, here’s a couple of measurements for comparison. RH RIAA 3.0 is my own design that works pretty well but you’ll see that my man-cave is troubled with some 50 Hz noise that is caused by the heat pump outdoor unit just behind the wall.

You can measure MC stages pretty easily but not without adding attenuators in the cables. In the MC mode I use a semi-differential 66 dB attenuating cable that has resistor divider built into the RCA connector that attaches to the MC input.With MM even straight connection gives OK results but attenuator is not a bad idea there either.

Here the attenuator cable is not taken into account in the “gain” reading and that is why it is showing -5.85 dB while the real gain is ~60 dB.

Here’s the RIAA response of the same unit (in MM mode):

The low frequency end has a couple of funny jumps partially caused by the QA403 software or some non-optimal setting caused by me. The latest software releases and the hi-res RIAA weighting files have fixed that.

1 Like

Very cool. I was deciding myself on attenuators for testing MC front ends. Care to share your network & values? 500uV from 1V (66dB)- did you find that was the sweetspot for values/performance? What is the parallel impedance for the stage ~100R and how hard does it drive the QA’s output?

I would be thrilled to have such a response with your little bit of heat pump noise. I rarely measure MC stages with what I do, but wonder how your 66dB attenuating setup works as @restorer-john requested… A friend brought by his misbehaving vintage mcIntosh C22 preamp today for a problem (removed all the tubes and they tested good and it worked fine after that), so I decided to test the phono stage for this thread. This preamp appears to be original except for the tubes and was made from '65-72, so >50yrs old:

1 Like

The values are not critical but I chose the middle resistor that is across the RCA plug terminals to be 10 ohms since many MC cartridges have roughly the same coil resistance. The other two resistors that are connected from the BNC cable center to both sides of the 10 ohm resistor are 20 k. So I’m using both of the QA403 outputs even though the normal single ended MC input does not see the negative side signal. The benefit of this is that it partially cuts the GND loop and gives at least some common mode attenuation.

An added benefit is that the output signal level is easy to remember - just use 0 dBV and you have 500 uV at the MC input. :slight_smile:

The same principle could be applied to a MM measurement cable. I would just replace the middle resistor with 100 ohms to get 46 dB attenuation. Admittedly 100 ohms does not represent a typical cartridge as well as in the MC case though.

I think it is worth mentioning that before this differential -66 dB cable I used a single ended -40 dB attenuator cable that had the divider resistors built in the BNC end. This worked but not nearly as well as the new one since it did not give any common mode attenuation, was generally more sensitive to external disturbance and it caused the QA403 output to distort slightly. I think I used about 900 ohm / 10 ohm divider (I need to check the exact values!) there which gives about -40 dB when the output resistance of QA403 is taken into account.

Clearly 1 k loading at -26 dBV (to achieve 500 uV input signal) is not optimal for the QA403 since I saw roughly one decade higher distortion from the MC RIAA compared to the measurement shown above. I actually thought at first that I was overloading the MC stage output due to rather low feedback chain impedance but it turned out to be the QA403 after all.

1 Like

Thanks for the info. I think I have been dreaming about phono stage measurements.I will look into the MM 100ohm and two 20k ohm method. I am not interested so much in MC as the majority of gear that I look at is vintage with “just” MM phono inputs.

1 Like

I finally got to measuring my tonearm preamp away from mains hum (in the car, using laptop and the fact the prototype is battery-powered currently): Nice clean spectrum allows THD+N to shine:

At 5mV level, no special input network (I am ordering some inductors for trying measurement with more accurate source impedance modelling).

Plotting against level gives an unremarkable plot:

The preamp maxes at -28dBV input or so, which is only 40mV, but that would improve if run from +/-15V rather than the current +/-9V battery supply.


The THD plot looks much better when you go to you car to measure your preamp. Not practical for trying to measure a receiver, though it may be worth a shot to hook up a receiver to a UPS for a few minutes just for grins- possibly to much hash on the lines from the UPS’s inverter, and just measure with a laptop and the QA402… at some point, anyway.

Indeed, you’d need a quiet bench and quality cables to measure the amp’s intrinsic noise level in a building via the THD+N statistic. An inverter produces its own noise and mains too, so I don’t think that’s the way for completely clean spectra, better shielding and separation help, but in a building its hard to avoid.

BTW I didn’t mention it but the preamp of mine uses OPA1612 and NE5532’s

I am not thinking that the UPS is going to make things work better, but I like to experiment here and there. But not enough to design and build my own phono preamp :slight_smile:

So I measured another unit of my tonearm preamp, which has NE5532 input stage, with direct input from the QA403, with 820R of series resistance, and with 500mH of series inductance (which has ~ 820 ohms of resistance). Thus I compare straight dumb testing of the preamp with a model of the cartridge resistance, and with a more realistic model of the cartridge inductance and resistance.

I put the spectra together as an animated GIF to make the differences stand out more - its quite a nice technique that maybe could be an output option for the QA40x software at some point?


The noticable rise in noise floor with the inductance should be the interaction of the input device current noise with that inductive impedance. Cable and stray capacitance puts a ceiling on the frequencies involved, giving the little peak in response…

Even parked well away from any electricity supply the inductors pick up a lot of noise, presumably from the QA403 supplies and laptop, possibly from some of the vehicle circuits (central locking?). Or maybe that grass at the high end is just non-linearity in the inductor?

I hope to compare with the OPA1612 version at some point, which should have considerably more current noise (although without inductance is significantly quieter due to low voltage noise, witnessed by the THD+N figures, here -74.4dB at best, compared to -82.9dB


Hi @MarkT, I think you are right on the inductive impedance. A TI engineer told me about the OPA210, which is very similar to OPA1612, but with super-beta inputs. The high-beta means greatly reduced bias currents and thus greatly reduced current noise in spite of being BJT. The OPA210 current noise is 400fa/rthz, while the OPA1612 is 1.7pA/rthz (both at 1 kHz).

In any case, as folks are developing on spice and/or real boards and wondering just what the current noise contribution might be at higher impedances, it might be useful to drop in a OPA210 to see what changes. Same packages too for single and dual!

The OPA210 clearly has bias current cancellation circuitry so the current noise density figure has to be taken with a big pinch of salt - realistic preamps don’t have equal source impedance on both opamp inputs, which is what the datasheet value assumes, being the condition where the cancellation common-mode noise cancels.

OPA1656 is a FET opamp with 2.9nV/√Hz and genuinely has no issue with current noise (6fA bias)

The OPA1656 if for sure a very good opamp but it is not always the best solution. At frequencies below 1kHz the input voltage noise is significantly rising - this the case with most FET input opamps. For a phono input stage this is something to evaluate carefully because the RIAA curve has higher gain at low frequencies.
Based on some Spice simulation, I found a hybrid amplifier with JFE2140 input stage followed by a OPA1611 provides the best compromise for amplifying signals from medium and higher impedance sources like MM cartridges.

No such thing as a free-lunch for phono preamps really - NE5534A still holds its own after decades for the simple good performance approach, pretty much seems to have been optimized for the task with the then-available technology.


And I’ve tested the OPA1612 version and made a 0R / 820R / 500mH animation as above for it, showing the large increase in noise the OPA1612 creates with 500mH inductive source:


Its a whole 30dB noisier at 10kHz with the 500mH using the OPA1612 (1.7pA/√Hz is translated to 53nV/√Hz…

And direct comparison of NE5532 with OPA1612 (an NE5534A would be somewhat better than the 5532):

[ edited for correct bipolar opamp type ]

1 Like

And I’ve replaced the OPA1612 with a JFET OPA1652 to get this result:


The current noise from the 47k input load resistor is now the only current noise source of note, which is about 0.6pA/√Hz - in fact this opamp performs about the same as the NE5532 as the voltage noise is 4.5nV/√Hz and the 47k resistor dominates the current noise for that too.

1 Like

From a noise perspective, I don’t see a significant difference but the THD is definitely better with OPA16xx.
I assume the 47k resistor is the input impedance of the preamp. This input resistor is in parallel to the generator and therefore not so relevant for the noise. The cartridge network is more relevant.
By the way, there is an interesting section in the NSC Audio Handbook from 1977 on S/N of phono cartridges and a revised and improved
version from TI is available as AN-104.

You can think of the 47k resistor generating current noise of √(4kT/R), which is converted to voltage noise by the cartridge network impedance, as well as the current noise from the opamp input. This only happens effectively at high frequencies where the cartridge is high in impedance and not strongly shorting out the 47k resistor noise.

500mH is 31.4k at 10kHz, the winding resistance in my setup is about 820R, so the 47k resistor dominates at higher frequencies, even if partially shunted.

This is why its worth considering lowering the noise of the 47k load by clever tricks (sometimes called electronic cooling, using an anti-bootstrapped load resistor of high value).

You should rather measure with no signal, then your RMS reading would give the total noise directly.