A ultra low noise Moving Coil Pre Preamplifier with 60x BF861 and Tentlabs embedded Regulator

wow! have you used a battery for the external 15Vdc? if the current draw is small enough, this may work well. The details of your design are way beyond my ability to comprehend.

I also really like the "input switch " that let's you choose between active, passive and no pre-pre.
 
Thanks again for interesting feedback!

@ mbrennwa and analog_se: I am pretty happy with the GSO caps. I am also using other caps in the whole chain to get a mix. I tried many over the years. Probably still personal, but out of interest (without making a capacitor thread here 😉 ), what are your favorites?

@ tombop56 and mark_t: JFE150 is a nice FET as I can read from the datasheet. Was not available when I started design. bit expensive unfortunately. optimum noise figure is about 2mA, so the same I use with the BF861B. So not sure where why you mention it needs less current? also 0.8mV compared to 1nV / SQRT(Hz). Great datasheet though with great application info. Not sure if the extra clamp diodes are helpful. Sure, a FET to keep on the short list! by the way, the Rs of 20Ohm is also contributing, so I do not think with this FET noise would be concededly lower. Or I miss something. How did you calculate less noise with 10 versus 60 FETs?

@ rif: thanks for feedback and noticing the set up with a very easy way to select the way to connect the MC cartridge to the Phonodude RIAA amplifier. Good way to compare everything also. Yes, I used a 15Volt battery as well. Also, a nice option 🙂

@ jackinnj: thanks for the article. Yes, all cooking with water it seems. SRRP is already very high in total. Empirically measured, I had a total figure of far above -110dB. But you triggered my interest in PSRR again. I have new tools to measure this with the AP, so maybe I am up for a more in-depth session measuring around at the real design (board)

anyone has ideas what else to measure or test?
 
Last edited:
@ TNT : your comment surprises me. Did you really read the article ? There is no real distortion, only Tests at very high signal levels which are way beyond normal operation to show spectrum. So please elaborate where your comment comes from and what level of distortion you were reading when you made your comment?
 
How did you calculate less noise with 10 versus 60 FETs?

By assuming that NXP (Philips) datasheet is correct.

BF861noise.png


Noise density for BF861, at 1 kHz, is shown as 2 nV/rtHz. Paralleling 60 devices would bring that down to 260 pV/rtHz. JFE150 has 0.8 nV/rtHz and 10 devices in parallel would yield 253 pV/rtHz.

I can find only BF862 actual noise measurements and they are better with 1/f corner at 100 Hz and noise density about 1.2 nV/rtHz.
Assuming that BF861 has the same noise as BF862, 23x BF862 would be required to reach the same low noise density as 10x JFE150.
 
😳 So, I took finally a look at the schematic.
Only 10 x BF861 are used for gain. I just assumed that 60 x BF 861 were used in parallel.

Anyway, math still holds.
10x BF861 will provide (1 kHz, 2 mA Idss) 1.6 nV/rtHz noise density.
2x JFE150 with 5 mA Idss would provide 570 pV/rtHz.
 
  • Like
Reactions: decramer

Attachments

  • PSRR Tester.png
    PSRR Tester.png
    12.3 KB · Views: 114
@ Tombo: thanks for sharing your thoughts and the link to the FET tests. VERY interesting I must say. Also, the JFE150 is indeed a nice part (understatement). I will dig into my old AP measurements and see if I can dig up a Noise spectrum graph from the output, so we have real data. The BF861 has a higher 1/f point indeed, certainly compared to a few from the website and JFE150. Of course, the higher frequencies are mostly "annoying" in real life (hiss...). If we would "only" consider higher frequency noise, the two are comparable. It is really the lower frequency noise which is not as good in the BF861. But still very interesting exchange. good stuff!

@Jack:
yes, thanks for sharing. a known to me also. I must admit I like the TentLabs Shunt solution. but it is one way to get to Rome...
 
@ Tombo:
OK, found it... In this graph you see the output noise from the DD PRE PRE at a gain of 20x. So, all possible noise contributors are included here. Please ignore (!) the 50 and 100 Hz Hum components. This was a test with too many cables flying around on the test bench...

What I see and conclude is:

1kHz and above moving from 24 to 20 nV/rtHz noise density. So "input" noise is ~ 1.2 - 1.0 nV/rtHz noise density for this range. This including Rs of 20 Ohm (!) It looks like the datasheet matches (at least) the BF861s I bought (NPX original reel; not Chines fakes like the BF862 from the author of the FET Test Website)

Lower frequency noise average is around let's say 35 nV/rtHz noise density, so only 4,8dB worse. To be honest I am not dissatisfied with this result. I was honestly very worried by the measurements of (other) BF861 you found!

Still, the other FETs mentioned are really good, especially now also BF861 is obsolete. 😉

DD MC PRE PRE - Noise density spectrum OUTPUT - Gain = 20x.png
 
I must admit I like the TentLabs Shunt solution. but it is one way to get to Rome...
Maybe another ‘road to Rome’ will be of interest. Shunt power supply with less output noise than your MC pre has - Salas’ UBiB 1.3 shunt regulator. Though, it is more expensive than TentLabs.

Actual measurement of regulator output rail noise, used in my preamplifier:

UBiB noise.png
 
Last edited:
  • Like
Reactions: schultzsch
@dddac - there is an AP macro in your toolkit "FFT Scaling for Noise" which you should use. Read the company's white paper which illustrates how to use the macro. The macro yields noise in dBV which can be rescaled to V per SQRT Hz

FWIW, I have found that the Analog Analyzer has a lower noise threshold than the Digital Analyzer.

Jung ran his tests on an early AP analyzer without the digital functions we have available to us today. Still, his article is mandatory reading: "Regulators for High Performance Audio" https://refsnregs.waltjung.org/Regs_for_High_Perf_Audio_1.pdf
 
  • Like
Reactions: PinkPink
Yes, FFT for a tone is completely different for the noise floor - noise energy fills the FFT bins like water, the level depends on the width of the bin, whereas a tone sits in one bin, unaffected by its width. This is an invaluable resource for using FFTs to take measurements: https://holometer.fnal.gov/GH_FFT.pdf

If you've ever quoted a peak height on an FFT using a window that's not a flat-topped one, or read the noise floor from the same graph as a peak height, you need to read this article!!
 
@planet IX - Yes, I know those as well. I think mbrennwa did very specifically mean KP but cannot find anything like that. Let's wait for him to answer.

@jackinnj - Thanks for the hint, this would be helpful. As you can see from the titel of the graph, I selected Fs and Length of the FFT such, that it is almost nV / rtHz. I also used the BW/4 function (lower bandwidth of FFT with lower noise floor)

@Mark Tillotson - Same comment as above. I am aware of this - Thanks for the hint anyway, as it is an important topic. The document you shared is very worthwhile! Thanks! I wrote a (les scientific) blog-topic on this subject some time ago as well. Blog Post FFT Noise floor