John Curl's Blowtorch preamplifier part II

Status
Not open for further replies.
Disabled Account
Joined 2012
I have not been following this issue... just noticed it here..... I am not going to attempt using such physically small parts if there are alternatives I can make work as well.

But, I was not surprised to find out that a jFET designed for Rf didnt spec the noise at audio feqs. I was surprised to find that a 16mA Idss deice was being used at 1mA operating current and expecting best performance. I just checked the spec sheet and the Idss range is 10-20mA. So I tested for noise just below Idss and found a sweet spot at 6mA. Going below 5mA, the low freq noise goes up a lot. Not suitable for use at 1mA IMO. But great at 6mA or a little more. The device C is very low which is a good thing for lowest distortion potential.

THx-RNMarsh
 
Last edited:
Member
Joined 2002
Paid Member
Wayne, great job on the Pass phono preamp. You will probably win the award this year. We now have a contest, worthy of the Blowtorch thread. '-)

Thanks John I am pretty happy with it lots of work as you know. I wish those awards came with big checks attached.:D

We displayed across from Constellation at CES a good destination floor for top notch SS with Ayre and Boulder also.
 
www.hifisonix.com
Joined 2003
Paid Member
Another measurement... same 5V/6mA and same p/n:

10nV 2.7 .8 .7 .5

[Approx 30nV at 1mA at 10Hz ]

Fantastic device above 100Hz. Even for this device and others which are worse, the noise rise at 10Hz might best be rolled off/attenuated if going into a sensitive ADC.



THx-RNMarsh

All NXP BF862 JFET's are fab'd at one location in Europe (and always have been). We used to assemble them at multiple sites in China and HK.

Assembly is now consolidated in 1 location in China. If you are getting parts marked HK, they will be older stock. There are lots of these things floating around and NXP still ships 'em in volume. Thats all I am saying.
 
Disabled Account
Joined 2012
Hi Bonsai,

It isnt practical for me to do that number of tests.... just because of the time involved in doing it with the QuanTech. The worst part for me is I also have to solder leads onto each jFET to insert into the test machine.

But I did run the current up and down and voltage up and down.... little noise change with voltage. The current below 5mA causes the noise at 100 and 10Hz in particular to increase significantly, as I said.

I would have to look into the testing differences between others' test results and the QuanTech results to figure out why the difference shown at low freqs. If there are any such test differences.

THx-RNMarsh
 
Last edited:
I have not been following this issue... just noticed it here..... I am not going to attempt using such physically small parts if there are alternatives I can make work as well.

But, I was not surprised to find out that a jFET designed for Rf didnt spec the noise at audio feqs. I was surprised to find that a 16mA Idss deice was being used at 1mA operating current and expecting best performance.

THx-RNMarsh

I have not found this to be true and have made many measurements. Many IC's run FET's at a tiny fraction of Idss. Noise goes as the fourth root of Id so is a VERY weak function of current.
 

Attachments

  • 2sk170.PNG
    2sk170.PNG
    33.6 KB · Views: 331
Disabled Account
Joined 2012
I have not found this to be true and have made many measurements. Many IC's run FET's at a tiny fraction of Idss. Noise goes as the fourth root of Id so is a VERY weak function of current.

In the context I put it.... RF transistors, i have found it true. Transistors optimized for low noise at RF freqs dont do as well as those optimized for LF noise -- at LF. There are some potential audio advantages to RF devices...... low C and low distortion when used in Class A. Some RF devices are intended to be used single-ended class A and are very linear. And, can have low noise when used with low z circuits ( 50 Ohms for example).

With the QuanTech it isnt always a weak function of current.... its easy to vary current with a turn of the knob and watch the noise change. Theory aside, of course.


THx-RNMarsh
 
Last edited:
Theory aside, of course.


THx-RNMarsh

Apparently datasheets too. I can't get excited about 10Hz noise in light of the fact that genuine 2SK170's often have a touch of excess noise themselves. The price differential is approaching more than an order of magnitude and there is no hard yield data yet. The availability issue is only going to get worse. BTW I was told Rode uses RF FET's (J305?) in their ultra low noise microphones. Same issue of some batches must not pass muster.
 
Last edited:
Disabled Account
Joined 2012
I have a bunch of new original 2SK170's and comps... all sorted by Idss. Bought them many, many years ago. Even have some of the original duals. The duals are used (maybe made for) ShibaSoku in thier input for their distortion analyzer. But, now I wont use them, as I dont have a MC phono cart. And dont want the extra bother of cascoding them. Maybe the new Linear Tech jFET for MC. Lower C, too.

J305? Must be really low C for mic use. I dont have... I have J310 though... for some forgotten reason. [btw - Are there any mosfets which are low C and low noise for high Z mic capsule use]?

I have asked Waly for his test conditions... esp what current was used.

Granted, even 20nV isnt a big deal with line levels and low gain. However, running that noise into the front-end of a sensitive, high resolution ADC 'might' cause poor results at low input signal levels. But, just use the BF type at 6mA. Fantastic all the way down to 10Hz.

-Richard
 
Last edited:
Disabled Account
Joined 2012
Device noise update -

Apparently datasheets too.

That part bothers me.....data sheet data....... I noticed that the noise tests are averaged data. So, we dont see the p-p levels that occure randomly. I see them and resembles popcorn noise. If I ignore those peak levels, this device would appear to be about 5nV at 10Hz. The peaks or 'hits' are quit frequent and regular enough to not ignore. And the 'hits' go to above 30nV.... this is with a heavily damped meter. A scope would tell us how high those hits go.

I read that the device is symetrical with respect to D and S. First report was with S-D reversed.... as it didnt matter. But, it does seem to matter. Maybe a regisration shift or similar affect in QC/QA? This time I reversed the leads and get similar numbers at all freq bands but better at down to 1mA.

So, maybe it is OK after-all ! I'll have to double check this.

Other little differences to note--- the temperature was not at a standard test level but rather cool 55-60F.... today slight ly warmer -60-65F.

Also, a very large noise increase occures only at 10Hz band when you blow across the device. So a room with no breeze (A/C etc) has to be used.

Data averaging, use correct device polarity and standard temp/humidy environment with no air movement across the device show some of the differences compared to published noise data.

That really only leaves the random popcorn hits at 10Hz to deal with. That is a mfring issue and should be addressed by mfr but not for an RF transistor.


THx-RNMarsh
 
Last edited:
That part bothers me.....data sheet data....... I noticed that the noise tests are averaged data. So, we dont see the p-p levels that occure randomly. I see them and resembles popcorn noise. If I ignore those peak levels, this device would appear to be about 5nV at 10Hz. The peaks or 'hits' are quit frequent and regular enough to not ignore. And the 'hits' go to above 30nV.... this is with a heavily damped meter. A scope would tell us how high those hits go.

I read that the device is symetrical with respect to D and S. First report was with S-D reversed.... as it didnt matter. But, it does seem to matter. Maybe a regisration shift or similar affect in QC/QA? This time I reversed the leads and get similar numbers at all freq bands but better at down to 1mA.

So, maybe it is OK after-all ! I'll have to double check this.

Other little differences to note--- the temperature was not at a standard test level but rather cool 55-60F.... today slight ly warmer -60-65F.

Also, a very large noise increase occures only at 10Hz band when you blow across the device. So a room with no breeze (A/C etc) has to be used.

Data averaging, use correct device polarity and standard temp/humidy environment with no air movement across the device show some of the differences compared to published noise data.

That really only leaves the random popcorn hits at 10Hz to deal with. That is a mfring issue and should be addressed by mfr but not for an RF transistor.


THx-RNMarsh

Hi Richard
I was going to ask about temperature.

I fiddle with telescopes and CCD cameras, a hobby that lets you feed mosquitoes in the summer and wear snowmobile pants in the winter.

For CCD cameras, the noise the ccd generates is strongly tied to temperature and falls in half every so many degrees C.
As a result the sensitive bits are thermoelectrically cooled, I have two cameras that can regulate at -40 degrees C below ambient.

Now, to keep the ccd from getting covered with frost, it’s behind a window, in an chamber filled with dry inert gas (Helium and Argon thanks to the tig welder).
But, as thermoelectric coolers are common, would they be of use quieting down the noise by lowering the operating temp of the devices here?
Best,
Tom
 
Status
Not open for further replies.