The inductance of MM/MI and a few high-output MC cartridges is the dominant impedance at higher frequencies. This is what makes current noise, or to avoid conflation with excess noise due to current flow in resistors, parallel noise, significant. It is not usually dominant, but nice to make as small as possible if we can do it fairly easily.
So for testing and characterization of phono preamps we really want to see a spectrum of voltage noise (input shorted) and current noise (input open). If we know that correlation is small (as it usually is at audio frequencies for favored amplifying devices) we can determine what the total noise spectrum will be with a given cartridge, given a good model of the cartridge impedance with frequency. A pretty good model is a resistance in series with an inductance. We can't get any quieter than the thermal noise at the operating temperature of that resistance, appropriately transformed by the entire network when the cartridge is connected to the given preamp.
Short of this amount of detail, we can use some standard series-R-L load and back out the thermal noise of the R. Or, for those wanting something even easier to measure and to avoid the terrors of computation, we can use a synthetic R with much less than 300K thermal noise. If I were making test equipment for audio and wanted to cater to the recent revival of vinyl, I'd either make such a test load available, or have software for doing the computations available, or both.
And yes, we need good test records and good signal averaging for determining frequency and transient response to optimize loading. Part of NRC's work in the old days was to get a better test record, and it was still a royal pain to get anything repeatable. I think this accounts for a near-hostility of some of my friends to vinyl --- that as well as its limitations for greater-than-two channels.
Again someone will remind us that surface noise and other imperfections of the medium will be larger than any but the poorest preamps. But this sort of argument can be extended to abolish all progress on anything in a reproduction chain that will contribute less than a given "weakest link". Now in some areas there is silliness here, and just because we can doesn't mean we should. But when it is relatively easy to do, shouldn't we do it?
So for testing and characterization of phono preamps we really want to see a spectrum of voltage noise (input shorted) and current noise (input open). If we know that correlation is small (as it usually is at audio frequencies for favored amplifying devices) we can determine what the total noise spectrum will be with a given cartridge, given a good model of the cartridge impedance with frequency. A pretty good model is a resistance in series with an inductance. We can't get any quieter than the thermal noise at the operating temperature of that resistance, appropriately transformed by the entire network when the cartridge is connected to the given preamp.
Short of this amount of detail, we can use some standard series-R-L load and back out the thermal noise of the R. Or, for those wanting something even easier to measure and to avoid the terrors of computation, we can use a synthetic R with much less than 300K thermal noise. If I were making test equipment for audio and wanted to cater to the recent revival of vinyl, I'd either make such a test load available, or have software for doing the computations available, or both.
And yes, we need good test records and good signal averaging for determining frequency and transient response to optimize loading. Part of NRC's work in the old days was to get a better test record, and it was still a royal pain to get anything repeatable. I think this accounts for a near-hostility of some of my friends to vinyl --- that as well as its limitations for greater-than-two channels.
Again someone will remind us that surface noise and other imperfections of the medium will be larger than any but the poorest preamps. But this sort of argument can be extended to abolish all progress on anything in a reproduction chain that will contribute less than a given "weakest link". Now in some areas there is silliness here, and just because we can doesn't mean we should. But when it is relatively easy to do, shouldn't we do it?
The 1976 (iirc) and 1980 editions of the National Semiconductor Handbooks have different and equally good models for phono input noise figure calc's. The 1980 edition divides things up into octaves, and adds weighting. SY has a calculator built into his web page, but it requires a MS Office thingy to work (or else I'm just too slow - either is equally likely).
Bcarso's goal of a noise figure is where things really need to be going. RIAA weighted and with-and-without A weighting.
Thanks,
Chris
Bcarso's goal of a noise figure is where things really need to be going. RIAA weighted and with-and-without A weighting.
Thanks,
Chris
It's fairly standard in some books and articles, but for some reason has not become universal. Of course the complementary term for voltage noise is series noise.
It's called "Excel."
There's a version in Open Office for the MS-o-phobes which should work just fine.The funny thing about the old NS Audio Handbook I may have mentioned, is flat-out conflict between the initial section about phono preamp noise and the last article with the octave band calculations. And even after accounting for the 47k load resistor's thermal noise, the last section only mentions that, with amplifier current (i.e., parallel) noise, things will just get worse! My suspicion is that Editor Dennis Bohn insisted on the addition at the end, but didn't want to further delay publication by making the initial article a lot longer. But this is just conjecture, based on knowing how clever Bohn is. The Handbook dates from his pre-Rane days I believe.
I like the last section because it shows how much harder things were in the days before the ubiquity of simulators and spreadsheets. Note the author's comments on how quickly simple networks produce hairy-looking equations.
It was SY who (recently) pointed me to the 1980 version, and I remember having the same (pre-Excel thinking) reaction. Old dog, I guess.
Thanks,
Chris
Excellent post Brad !
May I add something on the cartridge part of it.
I had done impedance measurements on MM and MC by feeding them with a frequency sweeping signal (amplitude matching cartridge nominal output) . Impedance plots were quite different when tip was on the air, resting on a groove of a stationary record and running in an unmodulated groove of a rotating record. (The files have to be in one of my broken laptops
)On the subject of RIAA pre input capacitance documentation that SY mentioned, I think that the optimum would be a complex impedance/freq plot published by the designer.
Last, something that is turning in my mind lately.
A simple circuit is composed of cartridge impedance/interconnect wiring/pre input impedance. Which of the various impedance typical combinations (MM, MI, MC/Common base, common emitter) is intrinsically more vulnerable to RFI interference, especially from wi-fi transmitters that have a good chance to be around?
(assume same minimum loop area. Wi-fi carrier approx. freq: 2.4GHz, 3.6GHz, 5GHz and soon 16GHz. Intercannel freq:2.4MHz)
George
Cartridge loading and noise
I might point out that I touch on many of these noise and cartridge loading issues in my VinylTrak preamp article in Linear Audio volume 4, September 2012. You may find it an interesting read, even if you do not agree with some of the design philosophy.
The VinylTrak uses entirely separate preamps for MM and MC, both with discrete JFET differential input stages. The first stage of these preamps operates without negative feedback. The MC uses 4 paralleled LSK389 cascoded dual JFETs. The MM uses the new LSK489 from Linear Systems. Both are DC coupled to the cartridge and use DC servos.
The MM preamp can operate in the conventional cartridge loading mode or in the "over-damped" mode where the cartridge is deliberately loaded with an unusually low resistance to convert the usual second-order resonant pole-pair to essentially a single-pole roll-off around 8kHz.
Cheers,
Bob
I might point out that I touch on many of these noise and cartridge loading issues in my VinylTrak preamp article in Linear Audio volume 4, September 2012. You may find it an interesting read, even if you do not agree with some of the design philosophy.
The VinylTrak uses entirely separate preamps for MM and MC, both with discrete JFET differential input stages. The first stage of these preamps operates without negative feedback. The MC uses 4 paralleled LSK389 cascoded dual JFETs. The MM uses the new LSK489 from Linear Systems. Both are DC coupled to the cartridge and use DC servos.
The MM preamp can operate in the conventional cartridge loading mode or in the "over-damped" mode where the cartridge is deliberately loaded with an unusually low resistance to convert the usual second-order resonant pole-pair to essentially a single-pole roll-off around 8kHz.
Cheers,
Bob
Excellent post Brad !
May I add something on the cartridge part of it.
I had done impedance measurements on MM and MC by feeding them with a frequency sweeping signal (amplitude matching cartridge nominal output) . Impedance plots were quite different when tip was on the air, resting on a groove of a stationary record and running in an unmodulated groove of a rotating record. (The files have to be in one of my broken laptops
I wish you could find these, we never finished the discussion on if electromecanical reciprocity has any play in this. I find the case of MC hard to believe myself, and the moving/non-moving makes no intuitive sense at all (certainly not at an easily measureable level).
What I really would like to see, is an article on measuring what happens on the digital lines (I2S, SPDIF) using down to earth instruments.
I am working on an almost all digital system now. I even have a spare miniDSP for RIAA implementation (I’ll wait Scott).
OK. I build it, I experiment with the x-over settings, I listen to it.
What I (will) miss is the usual torturing. Measuring, understanding and improving.
George
Excellent point about the cartridge being in the groove for the measurements. After all if there were no mechanical coupling effects we could hardly expect much signal either!May I add something on the cartridge part of it.
I had done impedance measurements on MM and MC by feeding them with a frequency sweeping signal (amplitude matching cartridge nominal output) . Impedance plots were quite different when tip was on the air, resting on a groove of a stationary record and running in an unmodulated groove of a rotating record. (The files have to be in one of my broken laptops
On the subject of RIAA pre input capacitance documentation that SY mentioned, I think that the optimum would be a complex impedance/freq plot published by the designer.
George
I also did an experiment with an old and battered M91E to see how much the inductance changed with tip displacement. It was appreciable if not enormous.
One of the things that strikes me when reading about various preamps is how so many claim they have discovered a better way than dreamt of before for handling the cartridge signal and the damping. It is as if the cartridge designer couldn't have optimized things, despite earnestly wishing to do so. See for example the June Stereophile, where we are again told that somehow treating the output of an MC cartridge as a current rather than a voltage is prima facie advantageous (Analog Corner, A Current Event, pp. 29-35).
Since you can't have currents without voltage and vice versa (superconductors and perfect insulators the exceptions) the distinction is unsupportable. But again we agree to be a little loose about terminology, and when an amplifier's input impedance is small compared to the source impedance, call it a "current" input.
The other side of the coin is when DAC outputs are described as "current outputs", which would imply a very high impedance. But once again nomenclature is deceptive, as terminating such outputs in anything that develops a significant voltage may well invalidate the specs, because the impedance is rarely that high (the nominal ones seen range from 1k to a good deal lower, although there are exceptions) and there will likely be code-dependent modulation of the output resistance.
!!! If the change was enough to measure it would likely be enough to cause appreciable FM/PM. Yet another argument for Bob Cordell's damped approach for MM/MI sources.
Thanks,
Chris
- Status
- Not open for further replies.