You need to be careful reading the low end of an FFT plot when presented on a log scale. IME you need a couple of decades or so slack to get a good read at the low end. On a Quantech I showed 2.5nV @ 10Hz, I took a photo of the face plate and posted it here years ago. With the noise unspecified one would expect some occasional lots to require some testing, the yield has been high. The BF862's that I have measured have very little GR noise, while I also posted a picture of genuine 2SK170's that were not. Two or three of Dimitri's examples have dramatic GR noise.
You are correct Mr. Wurcer, I should look more carefully, the ED graph is in dB re: 1nV/rtHz. The BF862 results in ED @10Hz are therefore in line with what you measured (2.5nV/rtHz) but are worse than the datasheet @ 1KHz (1.1nV/rtHz).
After paralleling 8 such devices, you would still get about 2x the noise reported by Mr. Groner @10Hz, in his paper. It could be that Mr. Groner got an exceptionally good BF862 batch. If so, then I would be very careful in claiming record numbers for the noise performance of BF862 jfets running at 1mA.
Just curious, does the jfet G-R noise scale (when paralleling devices) the same way as the Johnson noise, with SQR
?
Last edited:
Horowitz and Hill 3rd show in a table that certain Zetex/Diodes Inc. bipolars have adequately-low rbb' at least. I got some of them but haven't tested them as yet to see if the beta holds up to the specification. I believe Winfield avers that they should, based on the trending for incomplete measurements.
Meanwhile, I have bags of the discontinued Toshiba low rbb' parts, (2SC3329, 2SA1316) and despite the highest beta bracketing the beta is slightly out-of-spec (low). Disappointing, but not an issue for most MC.
There are ways to make the input current of such bipolar stages null, which ought to satisfy the purists. But capacitors sneak in somewhere
That concern about MC cartridges getting a little bias current reminds me of counseling a friend about using a super-triode at low grid-cathode voltage, where the grid current is fairly high. He felt he ought to have a cathode resistor, despite my measurements showing it didn't help much and of course added noise when unbypassed. He said Would you be willing to hook this up to X's 5 thousand dollar cartridge? I said Sure. But I can well imagine that he would say that afterwards the cartridge never quite sounded the same
Do you? Then the comment below is invalid due to the 5534 input current noise 3-4.5pA/rtHz @ 30Hz, rendering about 4nV/rtHz for a modest 1kohm impedance @1KHz MM, therefore doubling the 5534 input voltage noise, to a total of about 6-7nV/rtHz. You would be much better even with a good jfet input opamp.
Last edited:
Just curious, does the jfet G-R noise scale (when paralleling devices) the same way as the Johnson noise, with SQR
Same but different in the sense that it will rss add but unlike the white Johnson noise floor they are not virtually all the same and a really bad one could "poison the pot".
There are ways to make the input current of such bipolar stages null, which ought to satisfy the purists. But capacitors sneak in somewhere
Yeah, not a big deal, but in a discrete implementation would require, I guess, either manual adjustment (or beta matching).
The only problem I could imagine with a good film cap at the input of a MC preamp is it's property to act as a good antenna for all kind of EM perturbations. How much would that matter in audio, I don't know.
Yeah, different specs from different manufacturers. Anyway, a typical MM has 500mH inductance, which is ~3kohm @ 1KHz, and that's on top of the DC resistance. I would still not use a 5534 with such a cartridge, although if your noise performance reference is the needle on the vinyl noise, then you may get along with much worse than a 5534.
P.S. Obviously, "typical" # "specified"
Last edited:
I did not, but based on the data sheet the 2SA872 is absolutely average, many other pnp devices in current production are at par or better.
The BFU760F (fT=45GHz) is indeed low noise at GHz, but usually microwave devices are a disappointment at low frequencies, in particular in audio. Plus that a Vcemax=2V would not make it easy for audio applications.
I have tried it on JPEG, TIFF, PNG, WAV, FLAC and MP3 files. I found that it took rather large blocks of altered data to notice anything. One bit? Can't see it, can't hear it. No way.
Yes I agree---one gets down to MM/MI cartridge thermal pretty quickly with a single modern JFET. Although, as a matter of mostly-marketing principles I would use something other than a 5534, unless the design were severely cost-challenged.
I happened very recently on a preamp article that used a 2SK117 prefacing a 5534 in a 1985 Audio Amateur. There was one glaring typo on the schematic (not uncommon for AA, alas, even then) that led me to suppose the author had erred. But the BOM showed a thousand-times higher value, which validated his stated assertion that the impedance at the JFET drain was high and could be disregarded in terms of loading the subsequent network, which realized the 75us tau. The 1k resistor in the JFET source rather spoiled the voltage noise performance, but it still wasn't that bad (no measurements were provided). I couldn't see anything else that was suspect.
However the author had a generally mediocre-to-poor response from his listeners, both one in that issue and in a later "preamp shootout", and was royally pi$$ed off. I believe it scarred him irrevocably, as while we were still conversing, he spent a good deal of energy trying to persuade me to abandon audio products, at least without applying a large amount of cynicism.
I'm glad I did not. Aside from paying the bills, recent efforts have been very rewarding.
Last edited:
Not to mention the near-impossibility of getting any circuit other that an RF one to be stable with macroscopic supporting components.
But it is intriguing that one can buy discrete Si-Ge parts now. I ordered a few just for the hell of it.
Depends on the bit error. Substitution, not surprising. An insertion/deletion error would jostle things around a bit. Pun intended.
Should be interesting. If you get interesting data/results, I'll buy some also.... Usually, these uwave devices are optimized for noise/performance with 50 ohm I/O circuits.
http://cache.nxp.com/documents/leaf...ORMAT=pdf&WT_ASSET=Documentation&fileExt=.pdf
[but I can only measure to 6GHz]
THx-RNMarsh
Last edited:
- Status
- Not open for further replies.