I'm just sayin' that the best bang for the buck MM phono is with a 5534 Mark. If a builder feels another device is better and costs $30, so be it.
For ultra low noise, I would go with parallel BF862's in a single ended configuration.
Last edited:
What you saying Mark is certainly true when it comes to the English speaking world, but the rest of the planet is loaded with poor diyers ,who would be hard hit with a $30 increase on any parts.
Thanks Jan for the offer. I have in fact, 2 copies of Linear Audio 3 here. One that I bought from you, and one that you sent me, and I just looked up the article. It doesn't look like a 'few pennies' cost to me and more complex that I realized. I just made the example up on the spot, but Groner's approach is more thorough. Bcarso, it does become a PRACTICAL barrier to get significantly better than 5 ohms equivalent noise, but it can be done, so long as you throw away any universality regarding the input loading of the design. Ed Simon achieved this already.
For my next, and perhaps 'final' Vendetta phono stage, I will go for 0.3nV/rt Hz. For most phono needs, it should be 'good enough', and still low enough with sources under 50 ohms to not be terribly bothered by the nonlinear capacitance. Pennies for the input stage? More like $100, doing it my way, but it is true that the 862 approach is a cheaper approach.
When it comes to Groner's article. I would elect to add 9 more feedback resistors from the output and individually bias each 862 with its own 10 ohm resistor. Perhaps that would eliminate the need for the individual inductors at each jfet input. I would hope so.
For my next, and perhaps 'final' Vendetta phono stage, I will go for 0.3nV/rt Hz. For most phono needs, it should be 'good enough', and still low enough with sources under 50 ohms to not be terribly bothered by the nonlinear capacitance. Pennies for the input stage? More like $100, doing it my way, but it is true that the 862 approach is a cheaper approach.
When it comes to Groner's article. I would elect to add 9 more feedback resistors from the output and individually bias each 862 with its own 10 ohm resistor. Perhaps that would eliminate the need for the individual inductors at each jfet input. I would hope so.
The 5534 is a venerable part, and amazing that it has lasted so long. However the current noise is pretty high for cartridges with inductances around 500mH. For example, if the graphs are examined and the current noise density is around 600fA/sq rt Hz, that will produce 18.8nV/sq rt Hz at 10kHz from the L alone. The midband noise voltage density is comparable to the cartridge thermal noise density.I'm just sayin' that the best bang for the buck MM phono is with a 5534 Mark. If a builder feels another device is better and costs $30, so be it.
For ultra low noise, I would go with parallel BF862's in a single ended configuration.
Prefacing with even a single decent JFET adds little to e sub n and virtually eliminates the parallel noise contribution. A synthetic termination to realize the nominal 47k damping resistor gets rid of the majority of that contribution, if done well.
For what it's worth, my recently completed RIAA preamp has a place for an inductor at the input, but hasn't required one. There was some compensation needed in the drain of the upper cascode JFET, but once that was worked out there was no need for the input inductor.
Richard, as others have remarked, I followed the Quasimodo snubber design procedure as described in threads here on diyAudio. The procedure is:
In step #2 I selected a damping factor zeta=1.5, as discussed on p.103 in the magazine, just above Figure 6. Since zeta>1, the second order linear system is slightly over-damped. The remainder of the design procedure is simple plug and chug arithmetic on a calculator with scientific notation.
_
Attachments
You have seen the number of posts on this site where people are asking if $$$$ hand made capacitors are worth the money over mass produced (and consistent) components, convinced that spending huge amounts on passives will somehow tune the performance?
You have seen the number of posts on this site where people are asking if $$$$ hand made capacitors are worth the money over mass produced (and consistent) components, convinced that spending huge amounts on passives will somehow tune the performance?
But of course. And, yes, I do roll my eyes when someone says, "you must _____ as it is clearly better" versus "you know, it's my money and I enjoy ______".
There are some awfully decent passives out there now. Although they are not dirt cheap they are a modest cost adder.
Among active components, if 8-legged beasties with great performance are either too expensive or lacking sufficient drive capability or noise performance, hybrids or even all-discrete circuits are potentially somewhat less expensive. But one has to know what one is doing.
Among active components, if 8-legged beasties with great performance are either too expensive or lacking sufficient drive capability or noise performance, hybrids or even all-discrete circuits are potentially somewhat less expensive. But one has to know what one is doing.
I think Bill's point was more to the "all hype--no material benefit" (if not actual circuit detriment) type parts that fall far within the fashion end of the spectrum, rather than, say, bumping up to use a large-value C0G capacitor in a location where a cheaper part might work only marginally worse.
As to "one must know what one's doing", that's exactly it.
As to "one must know what one's doing", that's exactly it.
I do keep my eye on passive components, particularly resistors and capacitors. I started out about 35 years ago, when I was able to compare polypropylene with solid tantalum caps (my standard at the time). It forced me to do an all new cap multiplier as well, as well as a larger case. Yes, I heard the difference. I like to find out what all my serious competition uses, if I can, because there is often a consensus as to what sounds best with people who listen to differences. Some people even cryo their passive components. I don't but maybe I should.
Ed Simon has done some great measurements on resistors, and over the decades many people listened for differences in resistors and rated them accordingly. The results from listening have paralleled or even matched what Ed has measured. It isn't all over the map, or anything.
Of course there are some exceptions in passive parts, and some people LIKE sound changing components like oil caps, and such. I do not go along with this, and although it can sometimes make pleasant sounding products, it is not accurate, and I go for accuracy first. Objectivity? I will leave that to SY. '-)
Ed Simon has done some great measurements on resistors, and over the decades many people listened for differences in resistors and rated them accordingly. The results from listening have paralleled or even matched what Ed has measured. It isn't all over the map, or anything.
Of course there are some exceptions in passive parts, and some people LIKE sound changing components like oil caps, and such. I do not go along with this, and although it can sometimes make pleasant sounding products, it is not accurate, and I go for accuracy first. Objectivity? I will leave that to SY. '-)
When you dig deeper into a specific application there are many second order issues such as induced gate current noise due to the high capacitances which can not be eliminated by neutralization or bootstrapping. For MM or MI phono stages I don't see anything to gain noise wise over the old trick of a JFET input added to a 5534.
BTW the gate tied to the can on those Interfet parts can be a pain.
And all these at a drain current of ~1mA/device. I've read the LA article, I can somehow buy into the first number (although it appears to significantly depart from the 1/4 power law of noise dependence with the drain current - with any extra correction for large current effects). But the second number (combined with a corner frequency of well under 100Hz) is beyond anything that I have ever read or measured about low frequency noise in jfets. The number is extraordinary even if any G-R noise component would be missing (which is, in a certain amount, an rather unavoidable component of the jfet low frequency noise) and also conflicts with any data I have found about low noise jfets, and in particular the BF862 (latest - Dmitri Danyuk (I think he's also a member here) in ED Measurements Rate SMT Low-Voltage n-JFETs Under Consistent Conditions | Power content from Electronic Design indicating some 4nV/rtHz @ 10Hz; even assuming the G-R noise also scales as 1/SQRT
, which is not obvious to me, this will lead to an about x3 the reported noise performance @10Hz). To add insult to injury, another member here built the preamp on the original PCB and got exactly what I would expect for 8 devices in parallel @1mA: ~0.45-0.5nV/rtHz @ 1KHz and ~1.5-2nV/rtHz @ 10Hz http://www.diyaudio.com/forums/equi...-audio-vol-3-spare-boards-11.html#post4726473
I am also having a very hard time understanding why the measured noise curves are completely missing any harmonic noise artifacts (like mains frequency harmonics, CFLs, CRT line frequency, etc...) To my experience, at these low levels, even if you short/terminate the BNC input of the best piece of signal analyzer (I tried on an 89410A we have in our lab) you will still get at least the 1st and 3rd harmonics of mains frequency clearly displayed on the PSD graph, even if the equipment has an input referred noise of 7-10nV/rtHz. And of course, a sound card is usually much worse.
I have no idea how to explain the reported numbers; I tried to discuss these issues in the thread mentioned above, but Mr. Groner is apparently not willing to discuss such matters. I suggested a possible effect of sorting the devices (which could perhaps select a population closer to an ideal jfet without G-R LF noise, and Johnson noise at the lower limit of the datasheet) but, although the LA article mentions "sorting", the idea of tying the performance to sorted devices was brushed off by Mr. Groner http://www.diyaudio.com/forums/equi...r-audio-vol-3-spare-boards-9.html#post4721683 . A friend of mine, an ex-member here, which I would rather consider much more skilled in noise physics and practice than myself, also investigated and queried about the LA article, with no clear explanations or justifications for the truly amazing reported results.
Last edited:
That should be obvious, you could probably find a MC step-up transformer that does that.
No, for truly extreme low voltage noise you can't beat low noise bipolar devices (unfortunately, now a pretty extinct species). However, the overall noise performance with bipolars strongly depends on the source impedance (because of the rather large noise current ~2*q*Ib). Therefore, jfets are ideal for high impedance MM cartridges, while for low impedance MC cartridges you can get better noise results with bipolars. That's of course if you can ignore the "evil" cartridge DC current, or chose to block DC by using an no less "evil" (for the purists) input capacitor.
Current power devices may have a very low voltage noise, but their limited beta makes the input current noise usually inacceptable high.
- Status
- Not open for further replies.