Richard Lee's Ultra low Noise MC Head Amp

Member
Joined 2014
Paid Member
QUAD used the zetex transistors in their model 34 and 44 preamplifier MC amps, so I would assume they had good noise performance.


Now going on Richard's recollection they did use them, but they had to be selected for 1/f noise back then. However the Quad 44 service manual shows BC440 and BC461. I'm trying to find a 34 manual to compare as the only high res photo I can find shows the metal can BC models again.


Edit: found the manual and the 34 uses ZTX650 and 750
 
Last edited:
If the transistors are well matched the base current of one transistor should flow into the base of the other transistor and the current thru the cartridge should be miniscule. The QUAD 44 service manual I have has an addendum with the revised MC amplifier that uses the zetex transistors. The first model 44s didn't use the zetex transistors. They later revised the MC amplifier.
 
If the transistors are well matched the base current of one transistor should flow into the base of the other transistor and the current thru the cartridge should be miniscule.

Simply relying on base current matching would IMO be a bad design decision; figure out the cartridge current at power on (rail never come up synchronously) or in case of a power supply failure. If the single ended cartridge DC current would matter, then no sane manufacturer would accept the liability. Quad never occured to me as a risk taking company, in particular for the sake of a stinkin' capacitor.

I think it is fair to accept that the scaremongering around cartridge DC current is simply yet another audio myth. Myself, I have never seen any data about, say, a 10uA DC current on the cartridge integrity or performance. The AC current in a typical MC cartridge is about 5uA (0.5mV/100ohm) @1KHz (50uA @ 20Hz) and that's before considering the music high crest factor. I simply cannot imagine any destructive mechanism at these current levels. As of the sound quality, it was long discussed and proved that cartridges are not reciprocal motors.
 
Last edited:
The old 2N4250 PNP low noise transistor, widely used in MM preamps, had a beta of about 400. For low noise the input transistor would be run at about 100 uA. The base current would then be 100/400 uA i.e. one quarter of a microampere. I don't think that is going to damage the cartridge. Professor Leach in his preamp design used direct coupling and I have two of his preamps and never had any problems.
http://leachlegacy.ece.gatech.edu/papers/wbpreamp/feb77article.pdf
 
For low noise the input transistor would be run at about 100 uA. The base current would then be 100/400 uA i.e. one quarter of a microampere.

The subject here is mostly MC cartridges; running a MM cartridge with a bipolar stage is in general a bad idea, since the input current noise of bipolars cannot, in general, be ignored. Depending on a particular MM cartridge impedance, and the available bipolar transistors, it is possible to reach a trade between various conflicting noise, gain, etc... parameters, but the results will change (and in general not for the good) with a different cartridge. Too many things to keep under control, while a single JFET stage (or an JFET input 8 leg monster) would deliver much better performance, anyway.

Because you mentioned it, Ic=100uA is not really for low noise. The collector shot noise equivalent resistance is 1/2gm which makes it inverse proportional with the collector current. Running "low noise" at 100uA makes sense only if Rbb is so large that any collector current shot noise is masked by the Rbb contribution. Which (Rbb) may ruin the noise performance, to start with.

I have no idea how "widely used" is the 2N4250 in MM stages, but if so, it is because there was no better option at the time.
 
If you look at the noise contour plots for the 2N4250 transistor, with a 1K source resistance at 10 Hz the noise figure is at a minimum at about 100 uA collector current. The 2N4250 has an Rbb of about 150 ohms and that was why it was popular as the input transistor in preamps.
The cartridge impedance increase with frequency so current noise increases, but the RIAA de-emphasis mitigates that. If you have looked at a SPICE noise analysis of a simulated MM phono cartridge, 1000 Ohms in series with 0.5 Henry, You will find that half of the cartridge noise is in the 20-200 Hz bandwidth because the RIAA de-emphasis attenuates the high frequency noise. Same thing happens with the current noise.
 
If you look at the noise contour plots for the 2N4250 transistor, with a 1K source resistance at 10 Hz the noise figure is at a minimum at about 100 uA collector current. The 2N4250 has an Rbb of about 150 ohms and that was why it was popular as the input transistor in preamps.
The cartridge impedance increase with frequency so current noise increases, but the RIAA de-emphasis mitigates that. If you have looked at a SPICE noise analysis of a simulated MM phono cartridge, 1000 Ohms in series with 0.5 Henry, You will find that half of the cartridge noise is in the 20-200 Hz bandwidth because the RIAA de-emphasis attenuates the high frequency noise. Same thing happens with the current noise.

At 100uA the collector current shot noise equivalent resistance is 2/gm=500ohm which divided by SQRT(beta) is 50ohm, much lower than Rbb=150ohm (practically negligible from a noise perspective, since noise goes with the square of resistance). Exactly what I said, for such a large Rbb, the collector shot noise contribution can be safely ignored, that's why you can get away with Ic=100uA. However, Rbb=150ohm is rather large for a MC (1.6nV/rtHz equivalent), which again, is largely the topic in this thread. We were talking about Rbb in the 3-5ohm range.

You are of course free to choose any solution that fits your taste, and to claim it's the best since sliced bread. Others may agree (or not).
 
in posting #1576, Richard Lee's Ultra low Noise MC Head Amp, I calculated the equivalent input resistance to be 3.5R based on a measured noise of -191,2dBV.

The calculation itself was correct, but the measurement can't be correct, because this is beyond what's physically possible.
When measuring a sine wave of known amplitude, the FFT is exact, but for noise, for whatever reason, the outcome is too optimistic.
Could this be the averaging proces or the used window, I have no idea.

But careful calculating gave -189.4dBV as the noise level for the ZTX devices with an Rb of resp 1.2R & 1.5R, or 1.8dB above my measurement.
That would result in a noise for the Bare Amp module of 312pV/rtHz.
The equivalent input resistance in that case would be 5.8R instead of 3.5R.

On the other hand, the more important figure is input noise after Riaa and after A-Weighting.
So without taking into account the noise contribution of the equipment behind the Head Amp, this 312pV/rtHz becomes 126pV/rtHz after Riaa and A-weighting, see images below.

This 126pV/rtHz equivalent input noise, is an equivalent input resistor of 0.95R when using the Amp for LP playback!!

Hans
 

Attachments

  • Riaa Noise.jpg
    Riaa Noise.jpg
    362.5 KB · Views: 473
This time the Head Amp as being part of the whole audio chain, I was curious what difference it made to the noise spectrum when powering the Head Amp from either a Power Bank or from a USB mains adapter.
The two images below show the results of the measurements, with absolutely no significant difference.
Inter connections of equipment shown in the third image.


Hans
.
 

Attachments

  • Nokia Chain.jpg
    Nokia Chain.jpg
    257.9 KB · Views: 401
  • Pwr Bank Chain.jpg
    Pwr Bank Chain.jpg
    243.2 KB · Views: 382
  • CONNECTION.jpg
    CONNECTION.jpg
    171.5 KB · Views: 378
I was so euphoric yesterday by the nice results of the noise spectra with the cart connected, that I forgot to check all connections.
This morning I realised that I had forgotten to connect one (isolated) cable shield going to mains gnd.
After having corrected this, I did the spectra again, and this time even the 50Hz pollution is no longer there.


Hans
.
 

Attachments

  • Power Bank Shield.jpg
    Power Bank Shield.jpg
    218.5 KB · Views: 338
  • Power Nokia Shield.jpg
    Power Nokia Shield.jpg
    223.4 KB · Views: 96
in posting #1576, Richard Lee's Ultra low Noise MC Head Amp, I calculated the equivalent input resistance to be 3.5R based on a measured noise of -191,2dBV.

The calculation itself was correct, but the measurement can't be correct, because this is beyond what's physically possible.
When measuring a sine wave of known amplitude, the FFT is exact, but for noise, for whatever reason, the outcome is too optimistic.
Could this be the averaging proces or the used window, I have no idea.
The only thing that could cause such an issue would be an effective FFT length that is higher than displayed, maybe by a factor of 2. Sure enough, they've got 2097148 samples but 1048576 bins, so #bins = N/2.

You get the voltage noise density like this:
e_n = 1 V * 10^(dBV_displayed/20) / √(bin_width)
For -191.2 dBV in 0.9313 Hz, this gives me 0.285 nV/sqrt(Hz).
Assuming, of course, the machine isn't so smartelligent that it already performs this calculation - playing with FFT length while observing noise floor should clear that up.

Anyway, this voltage noise density gives me pretty much exactly 5 ohms @ 295 K, so it looks like your calculation was off after all.
 
The only thing that could cause such an issue would be an effective FFT length that is higher than displayed, maybe by a factor of 2. Sure enough, they've got 2097148 samples but 1048576 bins, so #bins = N/2.
You get the voltage noise density like this:
e_n = 1 V * 10^(dBV_displayed/20) / √(bin_width)
For -191.2 dBV in 0.9313 Hz, this gives me 0.285 nV/sqrt(Hz).
Thank you for your respons and I fully agree.
I got 4.5R instead of 4.9R because I didn't care to correct for the slightly smaller bin width.
Anyway, this voltage noise density gives me pretty much exactly 5 ohms @ 295 K, so it looks like your calculation was off after all.
The -191.2 nV/rtHz was measured including a 1R resistor simulating Rcart, so this value had still to be subtracted from my calculated 4.5R to get the noise for the Bare Amp giving the 3.5R.

One other thing that was also taken into account: without a bandwidth limiting filter, noise may fold back when Fs is too low, resulting in an incorrect higher noise reading.
The 60dB flat amplifier that was used between Head Amp and Scope, has a -3dB bandwidth of 1Mhz. That was exactly the reason to use a 2Mhz sampling rate to prevent any possible folding back.
But as said, when offering a pure Sine wave, reading on my scope is correct, but for some reason noise is too low even when correcting the 0.3dB for the 0.93Hz bin width.

To check my results, I simulated the circuit with 1) generic noise free transistors, 2) with ZTX transistors having an Rb of only resp. 0.12R & 0.15R and 3) with the 1.2R & 1.5R that H&H has measured.
Noise for the Bare Amp in these 3 cases was resp. calculated as:
1) 268nV/rtHz
2) 277nV/rtHz
3) 312nV/rtHz

Using the above correctly calculated (4.9-1)=3.9R would have been 255nV/rtHz, below the 268nV/rtHz of noise-free NPN and NPN transistors, which is of course impossible.
That's why I corrected to 312nV/rtHz or 5.8R for the Bare Amp.

Hans

P.s. with noise-free npn & pnp transistors, I mean no noise other then from 1/2gm.
 
Last edited:
What LNA is this? With 1GHz GBW, it looks like a very impressive implementation. My best 60dB +/-0.2dB LNA with 0.35nV/rtHz never went over 200MHz GBW (-3dB @ 200KHz)
This one has 1nV/rtHz, a -3dB BW from 20Hz to 1Mhz and 1Ghz GBW.
In #1549 I have shown the measured spectrum up to 1Khz, in this case divided by 34dB to show how much its noise contribution was below the 34dB Head Amp under development.
Also shown in this link is the transfer curve as measured with my VNA.
Richard Lee's Ultra low Noise MC Head Amp

Schematic diagram and picture below.
Allard is my second forename, that's where Allard Audio comes from.


Hans
.
 

Attachments

  • Kast.jpg
    Kast.jpg
    101.3 KB · Views: 131
  • 60dB_Flat.jpg
    60dB_Flat.jpg
    190.6 KB · Views: 211
This one has 1nV/rtHz, a -3dB BW from 20Hz to 1Mhz and 1Ghz GBW.
In #1549 I have shown the measured spectrum up to 1Khz, in this case divided by 34dB to show how much its noise contribution was below the 34dB Head Amp under development.
Also shown in this link is the transfer curve as measured with my VNA.
Richard Lee's Ultra low Noise MC Head Amp

Schematic diagram and picture below.
Allard is my second forename, that's where Allard Audio comes from.


Hans
.

Ah, three gain stages, 60dB gain is distributed, so it is not really a 1GHz GBW since halving the gain doesn’t double the BW (and unity gain bandwidth is far from 1GHz, more like 16MHz the OPA627 GBW).
 
Last edited:
Member
Joined 2014
Paid Member
I was so euphoric yesterday by the nice results of the noise spectra with the cart connected, that I forgot to check all connections.
This morning I realised that I had forgotten to connect one (isolated) cable shield going to mains gnd.
After having corrected this, I did the spectra again, and this time even the 50Hz pollution is no longer there.

Hans
.


I've hesitated asking this question as I know I'm missing something obvious I used to know and will feel even more of an idiot than usual when I am told. Your interconnection scheme is basically as advocated by Whitlock, but he uses an RF coupling capacitor between shield far end and chassis. And I understand this for the case of mains powered equipment with PE connection. But for when the headamp is either battery or DI wall wart connected is this actually a benefit over just grounding at the far end?
 
I've hesitated asking this question as I know I'm missing something obvious I used to know and will feel even more of an idiot than usual when I am told. Your interconnection scheme is basically as advocated by Whitlock, but he uses an RF coupling capacitor between shield far end and chassis. And I understand this for the case of mains powered equipment with PE connection. But for when the headamp is either battery or DI wall wart connected is this actually a benefit over just grounding at the far end?

Let me give it a try.
When both enclosures are connected to PE, a coupling capacitor at the far end may be used to suppress RF because the shield could act as an antenna. Although quite possible, I have never experienced any problem without this cap with a GSM or radio transmitter near the cable, so I never apply this somewhat complex cap.

Now for equipment fed from Battery or a wall wart, the interconnect shielding could very well be connected at both ends to the the enclosures, but in case of the very sensible Head Amp, I have chosen not to do, but to give it its own direct PE connection because the wall wart is producing quite some pollution that I didn’t want it to be injected in the next stage.

Without this PE connection to the Head Amp, the wall wart causes an enormous amount of very strong mains harmonics in the spectrum, so these guys are producing quite some dirt. But as shown in the above spectra, with the PE connected to the Head Amp’s analog gnd, in this very case the noise spectrum is as clean as with a battery (Power Bank)

Hans
 
Last edited: