As Dimitri has recently published again the BF862 is completely free of GR noise and the capacitances are extremely low for its gm. Dimitri's plots also show classic GR spectra on several of the parts.
Hmmm funny. Maybe a comment you made following another post, and I associated the two. I'm pretty sure it wasn't JC. The takeaway was that I thought from the description there was an excess of low-frequency energy, which could conceivably be expected from the far-larger L/R time constant. Well, maybe it is worth my doing a search just to reconcile what I recalled. My apologies for the misattribution.
Hi Chris,
The stylus was in free air for these tests.
You are exactly right about the relative dominance of the cantilever resonance. It is not always the big elephant in the room.
My Shure V15 Type V, which I believe has the Beryllium cantilever, has a reasonably well-damped cantilever resonance up at about 28kHz. On the other hand, many MM cartridges have cantilever resonances in the range of 20kHz, and manufacturers try to work the electrical resonance against the cantilever resonance to achieve a reasonably flat response. This often does not work out well, and is a significant shortcoming of the MM technology. Two wrongs don't make a right, and two resonances do not make a non-resonant response.
I generally believe that, if you can, get rid of the MM electrical resonance and live with the cantilever resonance.
MCs are not troubled by an electrical resonance, but still have the cantilever resonance. Depending on the construction and cantilever material, MC cantilever resonance is not always necessarily higher in frequency than that of a very good MM, like the Shure V15 V. It would be interesting to see a compilation of MC cantilever resonance frequencies by brand and model.
Cheers,
Bob
Mr Cordell,
you're a clearvoyant as well, just as I was about to ask you about the swap from LS844 to LSK489.
(maybe Linear Systems can make it a tradition to bring out a new part a couple of months after you put out a design)
Thank you for the phono article+design, and the application note.
Yes, I can't find fault with the BF862. Long live glorious AM Radio in China! 😀 Provided the drain-gate voltages are kept low, and the stray inductances are minimized so things don't take off at UHF, they seem to be about as good as things are likely to get with everyday silicon. I paralleled ten of them as a stack using copper wire to connect the leads, and as a heatsink, without issues. The only drawback is the need to do some rudimentary matching, and I appreciate the LIS parts for their availability as duals.
The Toshiba 2SK2145 is a decent dual but has roughly the same transconductance to input capacitance ratio of their other old low-noise JFETs --- it's about a half-sized 2SK170. And it has the disadvantage of a hard-wired channel connection. One more lead coming out would have made it a lot more versatile, although there are several configurations possible, assisted by probable dielectric isolation otherwise, or possibly adjacent die selection. Another pleasant surprise is the gate leakage, which for samples I was kindly sent by Wayne and Dimitri was well south of a picoampere.
I would take again the opportunity to bring to your (and other) attention this little critter from Sanyo (actually OnSemi).
It's a 2SK3557 low noise jfet (take a look at Dmitri's measurements, noise wise it's the second best after BF862, has also 35mS transconductance (45mS for the BF862) and the same 10pF Ciss), cascoded on chip with a 2SC4639 50V npn. As the input jfets need almost always to be cascoded, it's hard to beat this combination that costs pennies.
Given the current prices, I wouldn't touch the Linear System jfets with a ten foot pole. Lacking the p channel, other than being available in leaded versions, they have virtually no advantage over other n-ch low noise jfets (BF862, 2SK3557, etc...) while being x10 times more expensive.
Last edited:
I'm interested in the details of what we're calling cantilever resonance, and maybe trying to tease out the relative contributions of stylus mass x vinyl compliance resonance and the cantilever's beam resonances. We lump these together but I tend to believe the British view that the mass x vinyl compliance is dominant.
Tough to tease these apart by test though. The test itself is entangled in the related issue that we've been calling reciprocity here. Thinking out loud, we can either drive the stylus and look at the electrical output (but, drive it with what *believable* source?) or we can (?) drive the electrical output and measure the stylus' motion (maybe sit it on a microphone capsule). The latter fersure doesn't include stylus mass x vinyl compliance resonance.
Maybe we could get Scott Dorsey (at Kludge Audio) to cut us a special test record. What would we want on it?
Much thanks, as always,
Chris
Chris, I think it is more complicated: We also have to put the 'effective contact area' that the stylus actually does, in the equation. It is well known that the SHIBATA stylus has a much better 'effective' frequency response, due to raising the mechanical resonant frequency, all else being equal.
One REAL problem is getting made test records that have an extended response to 100KHz or more. We really need them to really know and understand what the cartridge is really doing, ultrasonically. 20-20K is not enough.
One REAL problem is getting made test records that have an extended response to 100KHz or more. We really need them to really know and understand what the cartridge is really doing, ultrasonically. 20-20K is not enough.
Hi Wally,
Interesting part, but its not a dual. OK for single-ended front ends. I didn't see a spec on input voltage noise.
I do wish the Linear Systems parts were less expensive.
Still cheap compared to the going rate for the obsolete Toshibas 🙂.
I don't think there will ever be a dual p-channel, but p-channels are not needed to build top-notch circuits.
Cheers,
Bob
Hi John,
I agree regarding the unavailability of good test records. Might there still be some around that were used for the JVC discrete 4-channel phono system?
I have found that one of my old Credence Clearwater Revival records had a few ticks and pops that were good for inferring HF response when used with a DSO 🙂.
Cheers,
Bob
Wow, A 417A triode just beats out the LSK170 in noise resistance. Who knew!
http://frank.pocnet.net/sheets/185/5/5842.pdf
See Dmitri's measurements quoted above, you can also easily calculate the equivalent input noise from the noise figure of 1dB (it's easy for jfets).
Dual jfets matching is IMO way overrated. Matching (specified at 10% in the datasheet) is not that good.
If matched jfets are really required, then one can get a much better matching by sorting individual parts. Out of curiosity, I have hand matched jfets to 0.1mA (~1%) Idss and got 20 pairs out of 100 parts of BF862.
For SMD parts, thermal coupling is also not a big issue. Thermal capacity of the SOT23 case is so small that two parts placed on a common PCB copper island are almost ideally tracking temperature, with a rather small time constant. And fwiw, onsemi also sells a dual 2SK3557 (not monolithic though, but I doubt matching is much worse than 10%).
Jfets do not hog current, and a servo is always included to take care of the offset and, lacking p-ch, distortion cancellation is an illusion with duals only, so I can barely see why matched jfets are that critical, anyway. I would think that Linear Systems is targeting the fashion market rather than a real engineering need (hence the price-no-objection tag).
Last edited:
Yes, the contact area helps to determine the effective vinyl compliance (and vice verse!). This is one place where there's no substitute for a test record. If I could get one cut to spec, what would we like on it? What kind of step/ impulse/ tick could be processed best in the face of expected noise levels?
Thanks, as always,
Chris
Last edited:
???
diff pair distortion cancellation works fine with matched square law devices of the same polarity
it is with complementary input that you don't get as big an improvement since N vs P doped Si mobility difference makes N to P ch exact matching impossible
diff pair distortion cancellation works fine with matched square law devices of the same polarity
it is with complementary input that you don't get as big an improvement since N vs P doped Si mobility difference makes N to P ch exact matching impossible
The bugaboo with triodes is the unpredictable 1/f corner. And it's not something that has a well-determined best-biasing strategy either. EDIT: Yes, what SY said.Wow, A 417A triode just beats out the LSK170 in noise resistance. Who knew!
http://frank.pocnet.net/sheets/185/5/5842.pdf
Last edited:
How much time did you spend getting those 20 parts?
I like duals - just my preference.
Not sure what you mean about the "illusion". Can you clarify your assertion?
Cheers,
Bob
- Status
- Not open for further replies.