Spell it out for us please. At the 100pV noise level it's easy to mistakenly let the gain-setting components add to the measured noise. You're claiming under 50pV of test-setup generated noise? That requires 0.16-ohm resistors, or more likely none at all, like my HP transistor noise meter, which runs the JFET flat out with a grounded source, and separately measures its gain, to determine the divisor required to obtain the narrowband input noise. Tell us more about your measurement technique.
--
Thanks,
- Win
gm = 123mS is not so high, for a large JFET. Remember, a BJT has 200mS at 5mA. InterFET says the IF3601 has gm = 750mS at Idss (Vgs = 0) = 30mA min. Note, 750/6 = 125mS, and that may be conservative, because gm/Id drops as Id approaches Idss.
--
Thanks,
- Win
Low g_m, ouch! Too bad, that does mean higher e_n.
--
Thanks,
- Win
I've finally got to test some of those huge Interfet jfets for their noise. Impressive, but... they are spec'd at 300pV/rtHz @ 5mA and 100Hz and should get down near 260pV @ 100kHz.
Well for the IF3601 and both side of the IF3602 dual, I get 400pV/rtHz @
5mA, and the noise PSD is almost flat.I have to rise Id to 20mA (x4) to get almost the low Id noise specs (get
290pV/rtHz at 100kHz).Before I get some other parts, I'd like to know if someone managed to obtain the figures they claim.
Both samples have the same date code, so it might be just a bad process issue... or not.
-- Thanks, Fred.
Top post corrected for clarity.
"Joe G (Home)" a écrit dans le message de news:43b55d60$0$18200$ snipped-for-privacy@news.optusnet.com.au...
(not
Of course I can. I duplicate the working conditions of the datasheet (5mA/3V) in a test setup that's designed to add not more than 2% to the measured noise. Not a big deal really. If the final result is out by a 4/3 factor I can tell the part is out by a
4/3 factor, at least. Interfet's test setup surely won't substract noise so... either the parts are off, or the datasheet is off, or their setup doesn't measure the full noise which I translate to a case of wrong datasheet.But we're not there now. I was just asking if someone managed to get the speced figures.
Yes, that's what they wrote and what it seemed at first contact. Then I'm still waiting answers for some questions I asked a while ago. I had not much time to reactivate them, so we'll see... but I'm still waiting.
-- Thanks, Fred.
"Fred Bartoli" schrieb im Newsbeitrag news:43b558ab$0$21719$ snipped-for-privacy@news.free.fr...
noise.
260pVobtain
Hello Fred,
I have looked in a book about JFET voltage noise.
Un^2 = 8/3 * k * T / Gfs
->
Gfs = 8/3 * k * T / Un^2 = 8/3*1.38e-23*300/(0.3e-9)^2 Gfs = 123mS = 123mA/V
I doubt that you have such a high Gfs at Id=5mA, Vds=3V. If you have less Gfs then you never can reach the 0.3nV at Id=5mA. You have to run the JFET then with a higher current. I am curious about your Gfs measurement result.
Best regards, Helmut
Gfs is the forward transconductance.
Yup, that's easy. Grounded source and gate, like this:
| .-. | |470 | | '-' | || 2200u +---||--. | || | .-. === | |1K GND | | '-' 2.2n | | || +------> to low noise preamp AC source >----||-----. | for gain || | | meas. | |-+ ___ | | bias >--|___|----+----->|-+ 100meg | | | | --- | 2.2u --- | | | === === GND GND
Thanks to the large gm, the gain is about 100 and the 1K noise contributes less than 1% for 300pV input noise. This allow enough margin for the 2.2uF esr.
Anyway it seems that Helmut has a point. I first measured noise with a setup more close to the expected design but got too much noise, so I folded back to this more simple setup, where I incorrectly attributed the lower than expected gain to the jfet Ro. I've just reinserted it into the cascoded setup and it measures at 88mS @
5mA and 130mS @ 10mA, closer to the measured results but I still miss 190pV somewhere if I stick to the 300pV given at 100Hz, not speaking of its announced specs at 50 or 100kHz where I get my figures.And this lower gm than expected isn't good news for me (more fets, more parasitics) since I counted on running them at 20mA. Now I have to run them at 20mA just to meet the specs :-( We'll see with Interfet next week... year.
-- Thanks, Fred.
In article , Joe G \\(Home\\) wrote: [...]
I've often found that, with Interfet, merely calling them up to ask a question will get you a full answer. They will also sort to a tighter spec. if you offer money for it.
I haven't used their bigger FETs but the 2N4867s they make seem to be what they say they are so a don't think they are in the habit of squinting while they test.
-- -- kensmith@rahul.net forging knowledge
"Fred Bartoli" schrieb im Newsbeitrag news:43b59f8e$0$21714$ snipped-for-privacy@news.free.fr...
Hello Fred,
Have you done a reference measurement before you inserted the JFET? How much noise have you measured in this pre-measurement? I just want be sure that you haven't overlooked something.
Is your 1k resistor ok regarding extra noise? I have read that thickfilm resistors have more noise than you would expect from the basic noise formula. This extra noise may depend from the voltage across the resistor and the current flowing through it. Who knows more about this?
Best regards, Helmut
190pVthem
Perhaps someone else offered the money first, and now he has what's left :)
-- John Devereux
What about paralleling lots of (and I really mean lots of) BF862?
-- Uwe Bonnes bon@elektron.ikp.physik.tu-darmstadt.de Institut fuer Kernphysik Schlossgartenstrasse 9 64289 Darmstadt --------- Tel. 06151 162516 -------- Fax. 06151 164321 ----------
"Helmut Sennewald" a écrit dans le message de news:dp49qm$4ne$02$ snipped-for-privacy@news.t-online.com...
Of course. When doing such delicate measurements you want to carefully double check all the parameters. I made measurements with differents methods which all agree pretty well if you're carefull enough. I've now spent quite some time chasing all errors and I'm pretty confident in the results... unfortunatly. Anyway I'll probably make another last try next week, because I'm not satisfied with the remaining discrepancy between the predicted noise from gm and the measured one.
Excess noise in resistors comes from resistors fluctuations and has a
1/sqrt(f) dependency. It can only be noticed at low frequency and vanishes under Johnson noise above some (low) frequency. The measurements I have made are at much higher frequencies to keep excess noise well below thermal noise. As a result the noise PSD is fully flat, as expected.Excess noise might be a problem in the final preamplifier design (BW extends down to 0.1Hz) but that's another story.
-- Thanks, Fred.
Yep. Ouch, as you say.
Especially when considering all the incredible sum of constrains the other requirements put on the design and how damaging the parasitics are: measured Cgs=310p, Cds=180p per fet.
-- Thanks, Fred.
may
Well, at 40Eur each (hi Joerg) they're not specially cheap, even more when considering I expected to put there 4 to 6 and the count is now inflating.
I'm even ready to pay more if I can have the right parts. But they do have to exist.
-- Thanks, Fred.
"Fred Bartoli" schrieb im Newsbeitrag news:43b5c21c$0$21690$ snipped-for-privacy@news.free.fr...
de
im
de
methods
quite
gm
as
extends
Hello Fred,
I looked a little bit with Google for "thick film noise smt".
Even for a 1206 resistor they show about 1uV noise for 5V across a 1k-resistor. The question is now about the used noise measurement method.
From this paper one could think it's a measurement at a center frequency of 1kHz with 1 decade bandwidth. I am not sure it's really the same measurement method as used in the datasheet above.
I would be happy if somebody could reveal this standard noise measurement method used for this extra noise of thickfilm SMT resistors. The result is given in uV/V or in dB.
Best regards, Helmut
In the old days - silicon manufacturers used to publish the test setups to acheive the data sheet spec's.
These days there is less information on how the data sheet values are acheived.
Some times the only hint from the data sheet is the "recommend values" (not limits).
If you have some weight..... (meaning.... you are going to purchase significat volumes) you can lean on the distributors of the InterFET or InterFET themselves to get info on how the results are acheived.... No weight (no $$$ ) may mean no support.
You can't say the FET doesn't meet spec's until you repoduce the mfg'ers test setup... and even then they may have something special which helps reduce the noise during testing - that is not documented.
Regards JG
Hmmm the front page of their data books says
InterFET is big enough to competitively supply over 400 standard
JFET types. Small enough to craft JFETs to your exact specifications.
Big enough to supply national and international leading-edge electronic
manufacturers and laboratories. Small enough for you to talk directly
with the people that engineer and manufacture the product.
Every business has a culture, a way of doing business. Ours isÉ
How we serve each customer is more important than how many
customers we serve.
"Fred Bartoli" wrote in message news:43b558ab$0$21719$ snipped-for-privacy@news.free.fr...
In message , Winfield Hill writes
That 30mA is Idss min, with no max given. However, data on the source chip is given on sheet NJ3600L, which shows a possible Idss range of 50mA to 1000mA.
NJ3600L also has graphs of the manufacturing spread of Gfs versus Vgs(off) and Idss, measured at 10V and 20mA. The Gfs figures given there are compatible with those being seen by Fred.B.
That Interfet data sheet is reminiscent of an old Siliconix barn door.
-- Tony Williams
How are you determining your noise density calculation brick-wall bandwidth? You didn't forget the pi/2 = 1.57 term did you (see AoE page 453)? That would lower your measured -> calculated noise from 407nV to 325nV/Hz^1/2.
--
Thanks,
- Win
Not to be difficult, but scope-based FFTs usually fail to yield a properly-calibrated spectral density. Commercial instruments, like HP and SRS machines generally do well. Ability to see fine detail, like CRT refresh interference, is not by itself an indication the root-Herz density value is correct.
--
Thanks,
- Win
"Tony Williams" a écrit dans le message de news: snipped-for-privacy@fromthelaptop.yuk...
More on this. I've made changes on my cascoded setup and now the noise readings agree within 1% on both setups which gives me great confidence that the figures are now accurate.
I've made measurements on 1 IF3601 (simple fet) and both sides of an IF3602, this at 5mA and 10mA drain current.
.--------------.---------------. @Id=5mA | gm(mA/V) | en(pV/rtHz) |
---------+--------------+---------------. IF3601 | 82 | 407 |
---------+--------------+---------------. IF3602/A | 84.7 | 479 |
---------+--------------+---------------. IF3602/B | 85.2 | 478 |
---------+--------------+---------------'
.--------------.---------------. @Id=10mA | gm(mA/V) | en(pV/rtHz) |
---------+--------------+---------------. IF3601 | 124.6 | 338 |
---------+--------------+---------------. IF3602/A | 129 | 427 |
---------+--------------+---------------. IF3602/B | 127.3 | 434 |
---------+--------------+---------------'
If I take the best case (IF3601) then at 10mA theoretical noise should be
298pV. That's still 160pV (1.5R) that slipped somewhere. For the IF3602@10mA we have a 310pV/rtHz excess noise source somewhere. More than the typical spec alone at 5mA!Now at 5mA, the excess sources are respectively 175pV/rtHz & 315pV/rtHz. Almost the same than at 10mA.
As for the datasheet, there are some incoherencies. The 300pV/rtHz @ 5mA should be at least 123mA/V gm. This translates to
250mA/V @ Id=20mA which is never reached on the distribution curves.I guess it's time to see what Interfet have to say on this now.
What do you mean?
-- Thanks, Fred.
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.