Are you saying Duraglit doesn't behave like the 'usual' common base stage as it doesn't have the above Equivalent Input Current Noise?
Richard do you understand the above? Looking at things from the current perspective (Norton) the noise of a resistor is sqrt(4KT/R) and the collector shot noise sqrt(2qIc). But the gain of the collector shot noise to the output is 1/10 while the gain to the input current is 1 (normalized to the circuit gain of course) so rss'ed together you get sqrt((4KT/R) + (2qIc)/100).
In LTSpice you can click on each component in a noise simulation to observe it's contribution. I don't know how to select elements inside the model so you take transistors as a whole (my simulator pushed as deep into the hierarchy as you wanted). I hope we put to bed the fact that LTSpice has erroneous results for this, it does not.
BTW I forgot to mention the point I picked for Ic (12mA) was an optimum for noise alone and yes overkill. I might re-run the analysis for a 3 Ohm cart.
Last edited:
To view things from the practical side, I have made a graph in the image below where Rcart and Vcart as inputs are telling how much noise RTI the MC amp may produce to achieve a 75dB-A S/N after Riaa correction.
I have not included the additional effect of current noise, which will be negligible in most cases.
The most demanding Cart I can find is the AT36E with 3 Ohm/ 0.1mV. This Cart will need an MC amp with 0.2nV/rtHz RTI.
My 38 Ohm / 0.48mV Cart will be happy with 1.35nV/rtHz RTI.
Hans
I have not included the additional effect of current noise, which will be negligible in most cases.
The most demanding Cart I can find is the AT36E with 3 Ohm/ 0.1mV. This Cart will need an MC amp with 0.2nV/rtHz RTI.
My 38 Ohm / 0.48mV Cart will be happy with 1.35nV/rtHz RTI.
Hans
Attachments
There is somewhere an option to store the voltages and currents inside a subcircuit, but I
did not find a way to make use of the stored data.
Maybe one can circumvent that by using a partly idealized transistor and drawing Rbb, Re
etc explicitly, or drawing the subcircuit instead of reading it from an include file.
did not find a way to make use of the stored data.
Maybe one can circumvent that by using a partly idealized transistor and drawing Rbb, Re
etc explicitly, or drawing the subcircuit instead of reading it from an include file.
That would be a work around for most of it, not sure how to drag beta outside. Our simulator was command line driven so simply "display noise <Q1>" printed out the extracted model with the noise contribution of each element.
I'm running the Ortofon 2M Black stylus (Shibata IIRC) on the 2M Red body as you suggested and seems to work well. My son has a stock 2M and I have given absolutely nothing away in sound quality in my jigged cart<>stylus combo. I'm not familiar BTW with the stylus 40 - is that a 3rd party stylus?
I have another candidate for a lowest noise MC amp: Dynavector DV23R. Despite it has a nominal sensitivity of 0.2mV it has an unusual high impedance of 35 Ohms for such an output. Extrapolating your graph up to 35 Ohms will end up below 0.2nV/rtHz RTI to reach the practical 75 dB-A S/N level.
Hans this way of looking at it leaves out too many things like MM carts and their loading/ pre-amp topology.
For one thing the curves are asymptotic on the right to simply the fact that the self noise of the cart is already at -75dB losing the point that any pre-amp at 10dB below that adds less than a dB more. Taking away the message that the pre-amp has to be noiseless in order to not increase the net noise one bit is rather pointless.
In this thread we are discussing ultra low noise MC head amps.
This has nothing to do with MM Carts.
That's why I restricted the graph to MC Carts from 0.1mV to 0.5mV to make it more clear in what situation you really need an ultra low noise Head Amp.
Where is the Cart with -75dB and where is the message that the preamp should be noiseless ?
The lowest noise for a preamp in the graph without Cart is 200pV/rtHz and nowhere noiseless.
Could it be that you did not fully get the meaning of the graph and that it concerns S/N after Riaa and after A-Weighting ?
Or maybe you can give an example illustrating what you mean, because I don't get it.
Hans
That would simply mean that you don’t get 75dB-A S/N even with a 0.2 nV/rtHz MC Head Amp.
But the 75dB-A was just meant as a sensible upper limit.
With the Richard Lee Duraglit Special, producing 0.28nV/rtHz without Cart, you will still get an excellent 72.7dB-A, meaning complete silence with the arm in the air and still quite a bit above the surface noise from the LP.
Hans
Last edited:
Where did you say a minimum pre-amp noise was factored in? It makes sense to plot a line with two asymptotes on the left cart with 0 R and on the right where the cart R equals your reference of choice. For MC only which are flat with frequency you could pick spot noise at 1kHz and it would make no difference.
We are still not synchronised along the same line.
A MC Headamp has a flat noise spectrum, but to listen to LP’s, somewhere in the line a Riaa correction will be applied changing the overall noise spectrum.
But a noise spectrum has to be weighted to correspond to what we actually hear.
In this case I have used an A-weighting.
In the given graph you can enter Rcart on the horizontal axis. Go upwards from there to the curve that represents the output voltage of the Cart.
Go from here horizontal to the Y axis and you will find what equivalent input noise without Cart the MC Headamp should produce to get a S/N of 75dB-A after Riaa and after A-weighting.
This figure differs significantly from the S/N without Riaa and without A-weighting by almost a factor 2.5 !
Hans
The formula is correct, and is exactly SQRT(2*q*Ic) (well, 17.88) I posted above. However it does not apply to the complementary common base circuit. For that, the collector shot noise current for each half of the circuit is divided by a current divider, one side is 1/gm (input impedance of the other half) and the other side is Rc (the cartridge impedance). Since only the current through Rc is input noise current, the total equivalent input current noise is [SQRT(2)/(1+gm*Rc)]*SQRT(2*q*Ic). Since gm=Ic/(kT/q) grows faster than SQRT(Ic), the equivalent input current noise actually decreases with Ic! For low collector currents and/or very low cartridge impedances the equivalent input current noise approaches asymptotically SQRT(2) times your formula, so the two circuit halves equivalent input current simply adds up.
Last edited:
Yes. Mea maxima culpa but do we now have a formal expression for Duraglit Equivalent Input Noise current? assume 'perfect' BJTs, small rbb' bla bla
That would be a major contribution to the fount of human knowledge
___________________
BTW, for common or garden common base, are my Jurassic
18.24 * SQRT(Ic) pA/rt(Hz) Ic in mA numbers correct, Guru Wurcer?
I got this via a roundabout route via 4kTRB applied to Rni & Rnv which is how GG Baxandall preferred to look at noise. I'm much less familiar with 2qIc ... and in fact am having problems figuring out the correct value for 'q' . Don't laugh!
___________________
.. back to our programme
Is your 1/10 factor cos you are running such a high current such that the other transistor is a bigger load on the current noise than the source? Please excuse my newbie questions.Scott Wurcer said:
Thanks for this syn08. May I very humbly ask what is 'q' actually? I know I should consult GG Baxandall & others but I can't do that simply at present. (This is my coffee break from beach bum stuff.) Googling 'q' isn't helpful. Please don't laughsyn08 said:
Thanks for this syn08. May I very humbly ask what is 'q' actually? I know I should consult GG Baxandall & others but I can't do that simply at present. (This is my coffee break from beach bum stuff.) Googling 'q' isn't helpful. Please don't laugh
Ahem, the electron charge, 1.6E-19 coulomb
Yes, see the explanation above, with the current divider in complementary common base.
Last edited: