an mm/mc phono stage

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Member
Joined 2011
Paid Member
Am I correct, that the positive shunt regulator's output voltage is (VBE_Q7 * R20 / (0.5*R22)) = +14 volts?

And if Q7's VBE varies by -2.2mV per degree C, then the regulated output voltage varies by -44mV per degree C. Is that correct? So a temperature change of 10 degrees C causes a supply voltage change of 0.44 volts.

I would imagine it's a good idea to thermally couple Q7 and Q8 so their junction temperatures track each other. This would help ensure that the regulated V+ and regulated V- tracked each other.
 
Administrator
Joined 2004
Paid Member
Scott's comment about surface noise is the key, if you are below it no further practical improvement is necessary. (And still I try for that black background)

You can make this argument for all sorts of ills including hum/buzz below the medium's noise floor, but then you have to deal with audible noises when not playing records which always leaves you with the impression that things are broken.

How many of you think about the equivalent noise resistance of the cartridges you are using?

The 5534 is probably still a good device to consider where voltage noise matters most.

Bill has a great sense of humor, and a turn of phrase that is always entertaining. And his comment is oh, so true.
 
I have often wondered whether the NE5534A was designed with moving magnet cartridges in mind; with 3.5 nV/sqrt(Hz) and 0.4 pA/sqrt(Hz) it seems to be optimized for source impedances around 3.5 nV/sqrt(Hz)/0.4 pA/sqrt(Hz) = 8.75 kohm, which is of the same order of magnitude as my 12 kohm impedance estimate.

Regarding datasheet values versus real-life values for current noise: the problem is typically the base current compensation circuit that some bipolar op-amps (but not the NE5534) have. It usually contains a scaled-down version of one of the input transistors biased at a scaled-down collector bias current. Its base current is then amplified by a scaled PNP current mirror and injected into the positive and negative inputs.

This results in two correlated noise currents being injected into the + and - inputs of the op-amp. When the scaling factor / PNP current mirror gain is large, these noise currents are much greater than the normal base shot noise of the input stage.

As the currents come from the same source and have a large correlation, they drop out of the equation when the + and - inputs are driven from exactly equal impedances - which never happens in any real-life application, but which can easily be done in a test set-up to make datasheet values look better. A typical example is the LT1028: 1 pA/sqrt(Hz) with equal impedances and 3.25 pA/sqrt(Hz) with unequal impedances.
 
Scott's comment about surface noise is the key, if you are below it no further practical improvement is necessary. (And still I try for that black background)

I like the noise of my amplifier to be negligible even when there is no record playing, but maybe that is a bit over the top (though not as far over the top as building an amplifier with a distortion in the ppm range to drive a loudspeaker with a distortion of the order of several percents).
 
A typical example is the LT1028: 1 pA/sqrt(Hz) with equal impedances and 3.25 pA/sqrt(Hz) with unequal impedances.

I would like to see an application that really benefits from this. I'm sure you can find one looking hard enough. Besides a MM cart is a complex impedance not a resistance and a balancing act would be a real kludge.

BTW I was unable to verify the claimed degree of cancellation when I tried years ago, and I have never seen an argument in print that supports the necessary correlation coefficients (assuming the circuit presented in the JSSC).
 

rjm

Member
Joined 2004
Paid Member
Am I correct, that the positive shunt regulator's output voltage is (VBE_Q7 * R20 / (0.5*R22)) = +14 volts?

And if Q7's VBE varies by -2.2mV per degree C, then the regulated output voltage varies by -44mV per degree C. Is that correct? So a temperature change of 10 degrees C causes a supply voltage change of 0.44 volts.

I would imagine it's a good idea to thermally couple Q7 and Q8 so their junction temperatures track each other. This would help ensure that the regulated V+ and regulated V- tracked each other.

Yes, the regulator output is referenced to the emitter diode drop of Q7,8 multiplied by the ratio of R20 or R21 over the relevant section of the trimmer R22. The output tends to be about 12-13 V.

It would be a bad design choice if a) the currents were variable and/or sufficient to raise the temperature above ambient or b) we were concerned about the exact voltage seen by the op amps.

In practice the output voltage may vary a few hundred millivolts depending on ambient room temperature, but this does not influence the performance of the circuit. Positive negative tracking is largely obtained already since the current draw is the same, and the ambient temperature also.

(Or, short answer: there is no advantage to put Q7,8 in direct thermal contact since neither transistor gets warm during operation.)
 
Last edited:

rjm

Member
Joined 2004
Paid Member
The current and voltage noise balance of the NE5534 is pretty much ideal for audio. Not MM specifically, though it is okay for that, but backing any kind of filter or potentiometer you find yourself typically dealing without impedances above 1k but less than 10k. Even for the typical op amp gain stage configuration, impedances seen by the inputs are so often in that range.

No surprise to learn the designers knew what they were doing but ... yeah, the designers knew what they were doing.
 

rjm

Member
Joined 2004
Paid Member
OT

rjm, Last time I was in Kyoto the traditional arts center was on its last week. Has anything replaced it?

If you mean the one that used to be on Shijo street in Gion down from Yasaka Jinga, no, that's gone. The handicraft center on Marutamachi street is still around though, there is this fancy place in Karasuma I haven't been to yet

TASK
 
Last edited:

rjm

Member
Joined 2004
Paid Member
For reference, here are photos of the finished boards.

I have a few spares. pm if interested.
 

Attachments

  • DSC_4812s.jpg
    DSC_4812s.jpg
    272.6 KB · Views: 431
  • DSC_4813s.jpg
    DSC_4813s.jpg
    283.2 KB · Views: 425
Member
Joined 2017
Paid Member
residual noise

This is the reference for those(including me) who are interested in vinyl. The attached is the measurement of MC cart without playing(amplifier noise only) and with playing(surface noise included).

This is simple MC headamp and equalizer. MC headamp is ADA4004(1.8nV/sqrtHz,3.5pa/sqrtHz) with 20dB. Equalizer is also ADA4004 with 34dB. MC cart is MC-L1000(0.3mV). Vinyl is a good condition like new 180gram. The maximum amplitude depends on the recording time of vinyl. 60 minute one has almost half than 30 minutes one. The maximum amplitude of 30 minutes is almost 0dBFS in this measurement.

The White waveform is residual noise,-75.6dB. The yellow is residual and surface noise,-72.5dB. Actual SNR of this circumstance can be 72.5dB with A-filter. But surface noise itself is probably 72.5+3=75.5dB because -72.5 includes both residual amplifier noise and surface noise.

Acoustically speaking, 72.5 is considerably good. You can hardly listen to the surface noise. If you use "deconstruct" process to digitally reduce noise level from 2kHz to 8kHz by 6dB, you can forget the existence of surface noise. 72.5dB is such level.

The blue waveform is the real waveform of MC cart. You can see click pulse.
 

Attachments

  • ADC1.jpg
    ADC1.jpg
    250.2 KB · Views: 404
This is the reference for those(including me) who are interested in vinyl. The attached is the measurement of MC cart without playing(amplifier noise only) and with playing(surface noise included).

This is simple MC headamp and equalizer. MC headamp is ADA4004(1.8nV/sqrtHz,3.5pa/sqrtHz) with 20dB. Equalizer is also ADA4004 with 34dB. MC cart is MC-L1000(0.3mV). Vinyl is a good condition like new 180gram. The maximum amplitude depends on the recording time of vinyl. 60 minute one has almost half than 30 minutes one. The maximum amplitude of 30 minutes is almost 0dBFS in this measurement.

The White waveform is residual noise,-75.6dB. The yellow is residual and surface noise,-72.5dB. Actual SNR of this circumstance can be 72.5dB with A-filter. But surface noise itself is probably 72.5+3=75.5dB because -72.5 includes both residual amplifier noise and surface noise.

Acoustically speaking, 72.5 is considerably good. You can hardly listen to the surface noise. If you use "deconstruct" process to digitally reduce noise level from 2kHz to 8kHz by 6dB, you can forget the existence of surface noise. 72.5dB is such level.

The blue waveform is the real waveform of MC cart. You can see click pulse.
As it seems, you are displaying A-weighted spectra, so far so good, the shape of the curves are as to be expected.
However the fact that both curves are having almost the same level at 1 Khz, is pointing in the direction of a scaling error, since surface noise from LP's can never be the 75dBA that you mention.

To start with, the -75,6 dBA residual noise, most likely relative to 0.5mV @ 1Khz, is too high to be possible.
The 1.8nV/sqrtHz from an ADA400, with the input shortened and without any contribution from any feedback resistors etc., cannot physically exceed -73,8 dBA.
A very good ADA4004 design will have a SNR of at least a few dB less, let's say 71dBA with the input shortened.

In case of having the cart connected, you will lose another few dB's caused by the impedance of Cart and wiring.
But since your element has only 0.3mV @ 1Khz, you will loose another 4.4 dB in SNR (=20log(0.3mV/0.5mV)), thereby still completely neglecting noise pick-up by the cabling !
In effect, without noise pick-up, the resulting SNR will then be like 64dBA ref 0.3mV@1Khz@5cm/sec.

Surface noise in your example is almost the same, which resulted in raising the combined noise level by 3dB.
So this brings the surface noise to a more realistic level of -64dBA, which is still a bit on the optimistic side.

To conclude, try to calibrate your scaling factor.


Hans
 
Member
Joined 2017
Paid Member
SNR was not proper expression in this case because the signal level was not identified. The full scale of many ADCs is 10Vpp(3.5Vrms). The residual noise of -75dBFS means almost 0.62mVrms. This is the correct value. But the signal level usually doesn't mean 0dBFS for vinyl recording by ADC.

Fig 1 is 1kHz of 4.7cm/sec from test record. It shows almost -15dB. If you regard this as the maximum signal, SNR is -15+75=60. But the same test record has larger signal than this like Fig2. The right channel(-4.4dB) is a little bit larger than the left(-6.1dB). Another person who measured DL-103 says MC cart could be larger than 5cm/sec by 10dB. MC-L1000 is same as DL-103.

Digital recording has no margin above 0dBFS. It means you need to have enough headroom below 0dBFS. That's why my 0dBFS is relatively high. I don't want to discuss the definition of SNR. If the traditional definition is based on 5cm/sec, there is no problem to subtract 15dB from the noise level.

What I want to emphasize is that the surface noise of MC-L1000 is almost same as the residual noise of ADA4004. I measured total gain, headamp, equalizer and SE to DIF converter. It's 67dB which means input Referred Noise is 0.62/2200=280nVrms. Additional 2dB is necessary for A-filter compensation. I guess it results in about 330nVrms. This is a little bit good value from 1.8nV/sqrtHz of ADA4004 under 20kHz bandwidth because Johnson noise must be added to the value. I think this comes from equalizer amplifier which suppresses high frequency.
 

Attachments

  • eq1.jpg
    eq1.jpg
    264.8 KB · Views: 350
  • eq2.jpg
    eq2.jpg
    262.4 KB · Views: 342
Fig 1 is 1kHz of 4.7cm/sec from test record. It shows almost -15dB. If you regard this as the maximum signal, SNR is -15+75=60.

Digital recording has no margin above 0dBFS. It means you need to have enough headroom below 0dBFS. That's why my 0dBFS is relatively high. I don't want to discuss the definition of SNR. If the traditional definition is based on 5cm/sec, there is no problem to subtract 15dB from the noise level.



Thanks for your kind explanation.
Starting from different angles we come to the same conclusion, being that based on 1KHz@5cm/sec the surface noise from your LP is ca. 60dBA, a figure well in line with different measurements.
I wish you lots of listening pleasure with your gear.


Hans
 
Member
Joined 2014
Paid Member
I think we are in need of a new thread to discuss the correct way of measuring this stuff and reminders that vinyl 0dB and digital 0dBFS are up to 20dB apart as a useful aide memoir and to stop polluting RJMs thread which now has a signal to noise of just a few dB
 
I think we are in need of a new thread to discuss the correct way of measuring this stuff and reminders that vinyl 0dB and digital 0dBFS are up to 20dB apart as a useful aide memoir and to stop polluting RJMs thread which now has a signal to noise of just a few dB

Hi Bill,

I do not think that we have to define new standards in a new thread, they are already there 😁
Most of the postings in this thread are about: how much noise and SNR ?
Mega high SNR phono stages are making no sense playing LP’s having a relatively moderate SNR in the order of some 60 dBA.
This figure should be kept and used as an important reference.
Like Scott said: “LP surface noise swamps many evils”

Hans
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.