Hello Axel . . . consideration of some of the specific requirements of MC vs. MM cartridges was something I spent quite a bit of time with in my own phono stage design, and I very much feel that the usual "change the gain and loading" approach cannot begin to deliver optimum performance from both types. And while my design is specific to lower-output MCs, I designed, measured, and experimented with some MM-specific approaches.
The most important part IMO is the specific noise/impedance relationship between the the input stage itself (NOT just the loading), and the cartridge. MM cartridges have much higher and much more reactive source impedances than MC cartridges, which means they will require an input stage with lower noise current (hence higher noise voltage), than an MC cartridge, which is exactly the opposite. JFETs and vacuum-tubes are the traditional choices here for such an application, though a lightly-biased bipolar can sometimes work well too.
The second input-stage issue for reactive sources like MMs is capacitance, which can be static or dynamic (changing capacitance with signal level), the latter of which will have a big effect on the ultrasonic behavior of the cartridge. Vacuum tubes are pretty good here, as their input capacitance is usually mostly static (like cable capacitance) for the signal levels we're talking about. With JFETs dynamic capacitance is a big issue, and cascoding a JFET input stage is IMO mandatory to keep this under control for an MM cartridge. Bipolars have such a high transconductance that the capacitance is rarely a problem, but they in turn require special attention to input-bias current so a tiny DC current isn't drawn through the cartridge itself.
For the devices themselves, I think it's silly to turn one's nose up at modern monolithic opamps, especially if a JFET input is what's decided upon. In the past 15-20 years, most of the innovations in JFET fabrication have occurred as part of improved monolithic processes, to the point where JFET opamps can be had that beat the noise performance of the very best discrete JFETs available.
The other question is whether or not to have a balanced input from the cartridge, or a simple unbalanced input. This is even a separate question from whether or not a differential-amplifier is used as the input pair, as the latter is frequently used simply for the feedback capabilities from an unbalanced input. For MC cartridges, I think a balanced input is optimum, especially with a transformer input - this is in part because the low source impedance of the cartridge can make the whole system (cartridge, cable, and input stage) very effective at eliminating magnetic hum pickup.
But the balanced input almost always means a noise penalty of 3-6dB when implemented with active electronics (transformers are unsuitable for MM cartridges). And I'm not sure that with the higher/reactive source impedances of MM cartridges, combined with the low-capacitance cable that they prefer . . . really translates into improved real-world noise rejection when used with a balanced input. At the very least, it greatly complicates some of the decisions required to make the input stage tolerant of common-mode noise, and provide proper protection against accidental overload (30VRMS common-mode nose from a bad turntable ground, anyone?)
Anyway, these are just a handful of the overall-picture considerations that can be very different between a phono preamp design that's specific to MM cartridges. |
Hi Mark -- yes we're indeed talking about the same thing, its just that the Miller capacitance is usually much lower. The reason I made the (admittedly rather imprecise) statement about vacuum-tube vs. JFET junction capacitance is from some generalizations about exactly how much this capacitance is actually reflected at the input when these parts are used in a circuit.
That is, 2-3 pF from the interelectrode capacitances of i.e. a 12AX7 are significantly less worrisome than the 30-40 pF or so from something like a 2SK389 set up as a diff-amp. I also think that cascoding JFETs (either with bipolars, or more JFETs) is a simpler, more elegant implementation than the vacuum-tube counterparts, which usually involves bootstrapping and multiple/floating heater supplies.
So I would personally always cascode JFETs in this application, but probably wouldn't with most vacuum-tubes. Just my preferances. |
Hi Axel, Some would not agree with this necessarily, since other then in digital designs, harmonic distortion is never buried in the noise floor completely. Sure it is - just about any datasheet-derived single-NE5534 phono stage will do it, for reasonable signal levels and output current. I obviously don't feel that such circuits are ultimate expression of what's possible in a phono preamp, but the only reason to tolerate measureable harmonic distortion in these circuits is if the designer feels that there are benefits to choosing certain types of parts or topologies - and these choices make it impossible to eliminate distortion. I have no problem with that . . . every designer is free to decide what parameters meet their goals - ultra-low distortion just happens to be one of mine. It seem current understanding that harmonic distortion should rather RISE evenly (even- and odd-order equally) with increasing output, rather than then decreasing with higher output. (Output rise as from cart input rise) I believe that what you're referring to is a specific type fault that many feel can occur as a result of crossover distortion in Class B power amplifiers. Again, a phono preamplifier can and should be completely free from these types of anaomolies. I is VERY difficult if not impossible to prevent some potential differences occurring in components (and amongst each other) during all states of operation. So they best possible directed by use of e.g. star-ground schemes. The point is, there are still caps (to ground) involved and caps have power factors, creating a far less 'clean' signal path then the dedicated (-) in a balanced design. If you re-consider Kirkhoff's laws and look at the signal CURRENT, it always must flow between the power-supply rails, period. The purpose of local bypassing capacitors is twofold - first, to remove the effects of power-supply wiring and traces from the circuit, and second, to prevent different stages' current draw from affecting each other. These functions are necessary in both differential and non-differential circuits, and regardless of whether or not the signal VOLTAGE is defined in relation to ground, at least a portion of the signal CURRENT will always flow through the bypass capacitors . . . and that's the way its supposed to be. It is up to the designer to keep these signal and supply currents separate from each other, regardless of whether or not they flow through a node we call "ground". In fact, it can sometimes be more of a problem in differential circuits, where each side of the circuit has separate bypass capacitors to ground . . . in which the signal current has to flow through a minimum of two capacitors and a ground trace to return to the supply. Its not that I don't like the differential approach, in fact my linestage is designed this way. But I will confess that I'm especially proud of the bypassing scheme - there are separate, unambiguous AC return paths for both differential signal current and signal current that flows to ground (i.e. from an impedance imbalance in the output cable or the amplifier that follows it). The trade-off is most always balanced = more dynamic, and 'cleaner' vs. unbalanced = better harmonic completeness, more natural sounding. I firmly believe that as we improve our art . . . it IS possible to have all of what you describe, without tradeoffs. And not everybody will see it the same way, but hopefully we will all end up with a more fulfilling experience from our recorded music. |
Hi Mark - FWIW, my experimentation with discrete JFET MM stages included both the 2SK389 dual, and a pair of 2SK369 singles. I cascoded them with 2N5089 NPN bipolars, and loaded the collectors with 1K . . . there was also a single 2N5089 used as a simple current source for the tail. Rails were +/- 20V, and I put about 8mA through the tail (4mA through each JFET). VDS across the JFETS was about 5V.
I used this circuit in an unbalanced-input feedback amplifier, in the classic "hybrid-amplifier" (JFET-frontend feeds opamp) topology. Instead of a monolithic, I used a 990 discrete opamp, driving a low-impedance EQ network for the 318uS and 3180uS time-constants, with about 20dB of gain at 1KC. This fed a passive 75uS network and another 990 for 20dB more gain, same as my MC design.
I never really put in the time to fully optimize the circuit, but it was very low-distortion, and CRAZY quiet . . . like greater than -90dBV at the output when looking at a 1K source impedance. I think that the 2SK269s had a little better 1/F noise, and even though the offset on the 2SK389 was much better, it didn't really matter for my application. Of course, YMMV. |
Hi Lewm - I will add my vote as well to the higher-loading approach, I think that 100K with minimal capacitance is much better (with virtually every cartridge that I've tried) than the common 47K/150pF.
But as far as commercially-available phono stages go . . . I don't really see enough of them anymore to have really strong opinions. I do think that in general, getting the subtle details right makes more difference than any generalisations about whether it uses JFETs, tubes, bipolars, opamps, etc. etc. There are certainly an unlimited number of ways to screw things up, or to avoid doing so.
I'm currently using my Beogram 4002/MMC20CL with a little phono preamp that's an opamp design, made on a Vectorboard . . . and the case and power-supply from a Lehmann Black Cube. Its a two-stage topology like my MC phono unit, with active 318/3180 in the first stage, passive 75uS, then more gain in the output stage. Opamps are AD745 (input) followed by AD797 (output), both of them use complementary discrete JFET buffers for high-current output, within each feedback loop. |
I just paid too much for a B&O MMC20CL off eBay. I hope I will be happy. I think that I was looking at that one too . . . glad you got it. If it's in good shape, then you didn't overpay. Stellar cartridge. On the AD797 - this opamp is capable of extremely high performance, but it's definately NOT one to swap into an existing circuit without careful consideration, especially if it originally uses TL072s! The AD797 really shines with low source impedances (not MM cartridges), but on the other hand, it can be limited in its useable output current . . . so it doesn't necessarily do the best job of driving the low-impedance feedback network that suits its low eN characteristics. That's why I used the discrete JFETs as a buffer, and why I used it in the second stage, where it can be driven from the low source impedance of the first. The AD745 is FET-input, so it works great with an MM cartridge . . . but it's not stable at lower gains, so stability and phase margin have to be carefully considered in application. Then there's the fact that the TL072 is a dual, and AD797 and AD745 are singles. For general TL072 replacement, try OP249. |
Hi Lewm . . . so for the subject of balanced inputs for phono preamps. First, its important to understand exactly what type of noise a balanced input is capable of rejecting. The overwhelming majority of cartridges and tonearm/turntable cartridge wiring treat the cartridge winding as a balanced source -- this alone is wholly sufficient to avoid hum pickup from ground-currents flowing between the turntable and preamp, provided no mistakes are made. Suceptibility to RF interference is determined by the design of the input stage itself. That leaves magnetic hum pickup as the main type of noise that we're designing the balanced input to reject.
The effectiveness of the hum-rejection in a balanced-input system is directly related to how closely matched the positive and negative conductors are impedance-matched to ground - that is, if the + and - conductors on the input have different impedances, the magnetic field will cause different amounts of interference in each one, which in turn will manifest itself as signal voltage. This is why (IMO all decent) balanced-signal cables use twisted-pair or star-quad configurations - the tight twisting keeps the impedance very much the same between the conductors, for good hum rejection.
But in a tonearm, the wiring is very rarely twisted-pair . . . they're usually just stuck side-by-side through the tonearm tube. And this isn't necessarily a bad thing, because the side-effect of the twisting is an increase in cable capacitance, which is exactly what we DON'T want for our MM cartridge. And if you want to minimise capacitance in the leads between the tonearm and the preamplifier, then you're also probably looking at simple coaxial cable types instead of shielded twisted-pair. Then there's the matter that having carefully-balanced output impedances is probably only very rarely considered in the design of the phono cartridge . . . because the vast majority of phono preamps over the years have unbalanced inputs.
So in any case when we're designing a phono preamplifier, we CANNOT assume that the impedances coming from the + and - leads of the cartridge will be well-balanced, impedance-wise. So if we want magnetic hum rejection, we need to build a balanced input that is fairly insensitive to these imbalances. And the way to do this is to keep the differential-mode (cartridge loading impedance) as low as possible, and keep the common-mode impedance as high as possible, as it's the ratio of these two that determine the effect of the source imbalance.
So for a traditional, say 5-ohm LOMC cartridge driving a 50-ohm transformer input, this is pretty easy to obtain, because the transformer will have a common-mode impedance in the tens of megohms, say 50 Meg. The ratio between the common-mode and differential-mode input impedances is thus 1,000,000,000 . . . and since the cartridge source impedance is so low, the maximum impedance imbalance will be a fraction of an ohm anyway. So hum rejection can be reasonably effective, regardless of the type of wiring used.
If we're to do this with an all-active input, we would still have the same 50-ohm loading resistor, but to effectively manage i.e. input-bias currents and offset, the input impedance of each side to ground would be probably at the highest maybe 470K, making the common-mode input impedance 235K, and our impedance ratio 4700. Not as good as the transformer, but still worth it.
But when we go to an MM cartridge, the source impedance is usually something like 1.5K and rises with frequency, so the impedance imbalance of the cable is then more likely to be a handful of ohms, and also rising with frequency. Per our earlier discussions, the differential-mode input impedance needs to be about 100K. And assuming JFET inputs, the very highest you can probably get away with for common-mode (without having the offset go through the roof, or cap-coupling) is 2 Meg resistors . . . making the common-mode impedance 1 Meg, and our impedance ratio is at 10.
So the endgame: the higher output impedance of an MM cartridge will make the impedance mismatches in the wiring more apparant, and at the same time makes the necessary design criteria in the phono preamp more suceptable to these imbalances. And the most effective way to reduce the imbalances in the wiring (twisted-pair construction) raises the capacitance, which is exactly what we DON'T want for our MM cartridge.
So then there's implementation - for MM cartridges, we can't use transformers, which leaves us with active realizations, which have a couple of major disadvantages over unbalanced inputs. The first is noise . . . you usually end up with twice as many uncorrelated noise sources, and can only make up for it by the fact that each side will see half of the impedance, giving a minimum 3dB noise penalty. The second is that many input stage designs don't work as well in the presence of significant common-mode voltage (which if we're trying to reject it, means it exists), and with high common-mode impedances, some sort of protection diodes, series resistors, etc. will probably be necessary to keep the input stage from getting fried when a ground wire gets disconnected and suddently there's 30V of common-mode voltage.
While I won't pretend that my conclusions on the matter are definitive, all of the above makes me think that for an MM cartridge, active balanced inputs are unlikely to deliver enough hum rejection to be worth the complications. |
For the reasons cited, it is much more difficult to obtain a true balanced signal from an MM cartridge as compared to an MC one. Feeding a signal with an imbalance of noise on one phase vs the other to the balanced gain stage will result in the amplification of that noise, i.e., it will not be rejected because it is not identically present on both phases. Lewm, this is a fair approximation of most of what I was saying. Maybe add to that . . . if the "balanced" input stage doesn't do a bangup job of eliminating magnetically-induced hum in the real world, then does it really need to be "balanced"? This is of course a question that each circuit designer will have to answer for him/herself. Right now, I'm personally leaning toward "no" . . . but it wouldn't suprise me if I changed my mind in the future. Why is the noise to which you refer not similarly amplified by an SE topology? (It's that "3 db" boost of the noise that I don't quite get.) Noise is noise(?) In a good SE phono stage for MM, is the "ground" isolated from chassis or earth ground? That would seem to be a good idea. Well, in this case "noise is not noise". The noise to be avoided in poor grounding is a result of the amplification of ground currents flowing across ground connections of finitely-low resistance (that is, everything except superconductors), and should be completely eliminated in a good design. This is what I am referring to in response to Axel's comment below. But the 3dB minimum noise increase from an actively-realized "balanced" input stage is a different noise source altogether - here I am referring to the (mainly) thermal noise from the input semiconductors/tubes and their associated passive components. In a differential "balanced" input, there are double the number of devices, each producing their own uncorrelated noise, acting in series. When combined, they will produce double the noise voltage, which is 6dB higher. That can usually be offset by the fact that each side of the differential input stage is now looking at half the source impedance, which can reduce the noise by 3dB . . . so that's where the net 3dB noise increase comes from. This is a big reason why the vast majority of low-noise preamplifiers for ANY kind of low-impedance (hence low-noise-voltage) transducer is usually unbalanced in architecture. But IMO there can be plenty of logic to the idea of swallowing 3dB higher thermal noise (especially if the circuit is still very quiet), for improved hum rejection from a balanced input stage. The inevitable and unavoidable 'ground contamination' influences of capacitors etc. that makes the other argument for differential/balanced vs. unbalanced. Hi again Axel :) - There should be no such "ground comtamination", regardless, in a good circuit design and layout. This is "simply" a matter of the designer carefully analyzing the ground current flow in each part of the circuit, and understanding and considering the subtleties of such things as careful local bypassing, power-supply impedances, and ground-trace routing. But judging by many of the commercial products I see, this seems to be a particular challange, and differential-balanced circuits can sometimes be more forgiving of these sorts of faults. Why? It is that the differential circuit also cancels even-order harmonics in the process of common mode rejection, so you wind up with a bias toward odd-order harmonics, and that is not so 'natural' to our ear. I'm of the opinion that while low-distortion is only one of several required characteristics for a good-sounding circuit . . . I feel that in a high-quality phono preamplifier, ALL harmonic and IM distortion should be completely and totally buried in the noise floor, which in itself should be very low. Yes, low-order and even-order products are less disconcerting to the ear . . . but who wants any of it at all? |
As in other audio areas here there are good, regular and bad designs. Damn straight. Couldn't have said it better myself. |
This seemed to work for LOMC cartridges such as the DV1s and Universe (which only produced a slight hum when the volume was turned up considerably), but fails miserably with the MM which produces an incessant hum at any volume. Halcro, please understand that I mean no disrespect to Bruce Candy or the namesake of your moniker . . . but unless the Continuum tonearm's wiring scheme is unconventional (in a connection sense, not just in a twisted-wiring sense) or defective (i.e some metal tonearm parts accidentally ungrounded) . . . then the problem is your preamplifier, and the manner in which its input stage is designed. There seem to be many designers who look at balanced input stages (both line, phono, and microphone) purely as a pair of opposing voltages, not balanced impedances . . . and its amazing how many otherwise top-notch pieces of professional and consumer audio gear are very intolerant of the slight impedance mismatches that I describe. Another example of this would be the work of Douglas Self . . . a designer who I greatly admire and find much of his work invaluable . . . but simply cannot agree with his approch to designing balanced input stages. But hopefully its just something silly like an a missing/loose ground wire somewhere, and my little diatribe is all for nought. |
I think he was referring to the generally much higher and more reactive impedance of MMs and the fact that these parameters may not be identical for each phase of the output in a SINGLE channel, using a balanced circuit. This creates a noise that cannot be cancelled by the balanced topology and is instead amplified. Channel balance has nothing to do with it. If I am full of baloney, perhaps Kirkus will correct me. Actually Lewm, you got it exactly. Now why do you think, do I have NO hum what so ever with a FULLY opened pre, going balanced into a balanced phono-line-pre? Well I'd say that you happen to have good synergy between the cartridge, tonearm wiring, and preamp input stage, at least in the sense that its noise-rejection happens to be sufficient to completely eliminate interference from the particular amount of mains-frequency magnetic flux to which it is being subjected. Now whether or not the ML engineers anticipated/designed for a similar level of performance in other situations or environments, I have no idea . . . but for yours, they got it right. |
Hi Axel . . . well, the first question is understanding as much as possible all of the different factors that change between your comparisons with/without the SUT. I'm speculating that some of them are: -Cartridge loading is slightly different, i.e. more inductive with the transformer -SUT presents a different source impedance to the phono stage than the cartridge directly -Phono stage loading switches/plugs/resistors/caps are different -Phono stage gain is different, likely affecting noise, bandwidth, and distorion -Of course, the SUT itself has a sonic/performance signature And some of the likely causes of what you observe: - more dynamic depth (better hi/low SPL differentiation) I think this usually corresponds to better headroom, and lower noise floor. The SUT will most likely gives a better En/In match to the cartridge, and better RFI rejection. The phono stage may also have more headroom at the lower gain. - more powerful bass Cartridge loading differences, possibly a little bit of low-frequency 3rd-harmonic distortion from the transformer - more hall/room information, stage depth I associate this with more high-frequency extension, or different high-frequency phase response. Cartridge loading differences, the transformer's sonic signiture, or better phono-stage performance at the lower gain Anyway hope this helps a bit, without re-opening the whole SUT/non-SUT debate. |
1) affordable phono-stage with MM (Reflex?)
2) good stage with MM
3) good stage with SUT and MC (Note: not mentioned headamps i.e. Elevator)
4) top stage with MC
2) and 3) maybe on par, and MC = LO MC > 0.4mV " Hi Axel - I think its really difficult to make these kinds of general heirarchical statements about performance, price and topology . . . for a few reasons: - You can't assume that providing the additional required voltage gain is always going to be the biggest challenge when designing a phono preamp for use with an MC cartridge . . . it may be for some designs, but definately not all. - The usual axiom of "price and performance don't always go together" . . . even though they do frequently, we must always be skeptical here. - I think that the implementation of an external SUT will always be a compromise - mainly because the transformer will have to be designed in a specific way to provide a "standard" step-up ratio that works well (in terms of voltage gain and output impedance) with a "standard" MM phono input. In a phono stage such as my own, the transformer (a Jensen JT-346) has a turns ratio that is much lower than a typical SUT (thus much higher performance) and provides a perfect En/In match to the electronics that follow it, which in turn have had their gain adjusted to suit the paramters of the SUT. That is, the transformer and electronics have been designed specifically to work together, not as a "universal" add-on solution. On the subject of accurate RIAA . . . I was actually inspired by Raul's postings to take this matter very seriously in my own design. This isn't an easy thing . . . there are specific challanges to even measuring it to better than +/- 0.01dB across a 40-dB-ish range of signal level, as most test equipment must either generate or measure it through several output/input measurement ranges. For this, I was lucky to have an APx 525 on loan as an extra test set to verify the results against the AP2700's measurements . . . suffice it to say there needs to be some very careful hand-selection of parts to even begin to come close to this level of accuracy. And there's the question of . . . why should the preamp be so precise, when no cartridge is anywhere close to this? Well, error is error . . . and it's always statistically additive in this case. After all, our CD players and amplifiers should have extremely precise response, even though a loudspeaker can't come anywhere close . . . |
As far as SUTs go, yes they need to be matched, well matched! But if done, they have some 'magic' of their own and can make an MC something more full of 'live' and vibrant. Oh yes, absolutely agreed! Even in a purely technical sense, there are many inherent advantages to transformer-coupling a LOMC cartridge . . . and I obviously chose this approach myself. I simply wanted to give counterpoint for those that feel that SUTs are inherently inaccurate and colored - because I understand how one could come to this conclusion . . . as most implementations on the market IMO don't really realize the full performance possible in the transformer approach. Not necessarily since there are errors to the (-) and (+) side of the curve, yes?
And we are actually talking about observed 'tolerances' just not quite the same as ERROR, in my vocab anyway.
Now, tolerances can be additive or subtractive. From a statistical/tolerance standpoint, yes, this is purely additive error. In electronics design, there is a specific procedure called "monte-carlo" analysis . . . which takes the maximum allowable tolerances of each component and combines them all in the absolutely worst-case situations - and this gives you the performance tolerance of the entire circuit. In the audio chain, to define frequency-response as a deviation from perfection . . . keep in mind that in this definition we may have no idea exactly what the particular error is for each part, only that they are within certain limits. So while we can hope for that perfect synergy where every error just happens to cancel each other out . . . if we are to truly take responsibility for the performance, we have to assume the "perfect storm" where all the errors just happen to add up in the worst way. |
