John Curl's Blowtorch preamplifier part II

Status
Not open for further replies.
Thank you for looking Steve,

No problem, Anatech. I didn't even have to move, just glance up from where I am sitting. I have almost full sets of the technical books from Toshiba, Motorola, Texas Instrument and several other companies. It is not easy for me to read a book anymore, so I guess I should consider letting them go. I keep putting off selling off my equipment and supplies. I guess I don't want to admit it"s over for me.
 
Administrator
Joined 2004
Paid Member
Hi Steve,
You make me a jealous man! A great collection there.

I should remind you that it's not over until you can't think. This is not your problem and you'll need those references until that time comes. I'm thinking that's a long time off.

So, in no way is "it" over for you sir!

My collection has suffered three complete set backs now. Meaning, I have rebuilt my "library" three times from close to scratch. The things I really can not replace are the eight filing cabinets full of service manuals I collected over a 16 year period. That really hurts. So does the loss of many vacuum tube related material.

So hang on there Steve. Your mind is alive and you think clearly. Keep your printed material. I have to say I'm still a bit envious of your material though! ;)

Best always Steve, Chris
 
Administrator
Joined 2004
Paid Member
Hi Salas,
I'm sorry, I forgot to answer you.

Adding a series resistance is perfectly reasonable - it works. It will also increase the separation between the two channels. Don't forget to stagger the capacitor values to maintain a low AC impedance across the band (up to 1 MHz at least). You just have to adjust the voltage before the resistor drop to end up where you want to be.

The question of how much hiss you'll end up with is more complicated, and probably something that others with more experience than I have with very low level signals. Once you have eliminated the supply noise, you still have to contend with air currents and thermal noise. When you are down that far in signal, even air moving over the transistor legs will generate noise. Your solution will therefore be possibly more mechanical than it is electronic. Same goes for damping mechanical vibrations out.

Bet you didn't want to know about the rest of that stuff. :spin:

-Chris
 
Why are we even considering Vgs = 0? I thought that was not recommended to build a decent ccs.

What happens if Vgs is not zero? In particular, if we use a resistor value between gate and source such that Id is about 1/5 Idss.

That would essentially significantly increase the noise. The Johnson noise of that resistor should be added to the JFET noise.
 
Administrator
Joined 2004
Paid Member
Hi Salas,
Everything in design is a compromise of some sort.

When is the power supply clean enough? When do I have enough shielding, damping, hair gel? These are all things that only you can decide. In fact, you may not get any more improvements once you pass some point. You have to recognize that point. In fact, this gives you valuable information for the next time. Of course, how much money you can spend for materials is another factor, but not as great as someone going into production commercially.

Why not prototype and experiment? Do your final build once you're happy with things. The stuff I design and build normally works first time. It's the improvements that is never ending. :eek:

-Chris
 
Here are the results for the Norton reference. All measurements are at 24V from a HP6624 low noise power supply. For each device, I measured the Idss and chosed a resistor R so that the output voltage is around 19V. Noise was measured with an HP89410A VSA, in 10Hz-1KHz, 1601 points. The first picture in this message is the analyzer noise floor (input shorted). It can be eventually extracted from the measured noise data (I have all the traces) but this won't dramatically impact the conclusions. No extra filtering (electrolytic or other) was employed in these measurements.

Here's the devices I had at hand (datasheet specs and my measurements). Measurement results are also attached, each picture is identified by the device name.

2N5458
------
Datasheet:
Idss=6mA
Vt=2V
Vn@100Hz=2nV/rtHz

Measured:
Idss=7mA
R=2.7k
Vo=18.9V
Noise=224.8nV/rtHz

PN4392
------
Datasheet:
Idss=25...75mA
Vt=2...5V
Vn@100Hz=Unspecified

Measured:
Idss=47mA
R=390ohm
Vo=18.33V
Noise=109.4nV/rtHz


2SK246Y
-------
Datasheet:
Idss=1.2...3mA
Vt=0.7...6V
Vn@100Hz=Unspecified

Measured:
Idss=1.3mA
R=15k
Vo=19.5V
Noise=124.7nV/rtHz


2SK170BL
--------
Datasheet:
Idss=6...12mA
Vt=0.2...1.5V
Vn@100Hz=1nV/rtHz

Measured:
Idss=9.7mA
R=2k
Vo=19.4V
Noise=302.4nV/rtHz


As you see, my previous comment regarding the JFET Gm and the output noise is totally confirmed. Gues who's the worse contender? The lowest noise device, having an average transconductance, the 2SK170BL. Why? because it has by far the lowest pinch voltage!

And the best contender? The monster Gm device, with unspecified noise, but high pinch voltage, the PN4392 from Fairchild. This exercise is a good example on how intuition can trick us...

One to another, my results show that it is possible to get around 100nV/rtHz with a Norton reference. We'll see in the next days how's with LM329 and LEDs.

P.S. I see the file names are not displayed. The pictures are in the same order as the devices above.
 

Attachments

  • 89410.jpg
    89410.jpg
    21.5 KB · Views: 383
  • 2N5458.jpg
    2N5458.jpg
    21 KB · Views: 384
  • PN4392.jpg
    PN4392.jpg
    21.3 KB · Views: 385
  • 2SK246Y.jpg
    2SK246Y.jpg
    20.4 KB · Views: 372
  • 2SK170.jpg
    2SK170.jpg
    20.9 KB · Views: 374
Last edited:
diyAudio Chief Moderator
Joined 2002
Paid Member
Why? because it has by far the lowest pinch voltage!

And the best contender? The monster Gm device, with unspecified noise, but high pinch voltage, the PN4392 from Fairchild.

One to another, my results show that it is possible to get around 100nV/rtHz with a Norton reference. We'll see in the next days how's with LM329 and LEDs.

Yes lets see. I have some 5459s and yummy 1.6V leds in 3 colors:). I hope I can spare some 2SKs if it works good with the 5459s. Not when the Jfet is residing under just Vbe margin in some power supplies unfortunately, where I need the low Vt.
 

iko

Ex-Moderator
Joined 2008
That would essentially significantly increase the noise. The Johnson noise of that resistor should be added to the JFET noise.

OK, I'm not disagreeing with you, but I'm trying to make sense of this. The same formulas don't apply, I think.

The output conductance when a resistor Rs is used between G and S is given by
go = gos^2 / (1 + Rs (gos + gfs))

Where for the j201 gos = 1uS and gfs = 0.5mS. Let's say Rs = 2k, then
go = (0.000001)^2/(1+2000*(0.000001+0.0005))
or
go = 0.00000000000049950049 S

Would the self noise amplify differently?
 
And here's a quick pass on a randomly selected LM329. Anode current is 1mA, reference voltage was measured as Vref=7.1V You would of course note the missing 60Hz peak, a good sign that the circuit has a low impedance. Also no filtering (electrolytic or otherwise) was used.

It happens that the noise is, at 108nV/rtHz, as good (or bad) as the best JFET Norton reference...

To follow in the next days, time permitting...
 

Attachments

  • LM329.jpg
    LM329.jpg
    22.7 KB · Views: 302
OK, I'm not disagreeing with you, but I'm trying to make sense of this. The same formulas don't apply, I think.

The output conductance when a resistor Rs is used between G and S is given by
go = gos^2 / (1 + Rs (gos + gfs))

Where for the j201 gos = 1uS and gfs = 0.5mS. Let's say Rs = 2k, then
go = (0.000001)^2/(1+2000*(0.000001+0.0005))
or
go = 0.00000000000049950049 S

Would the self noise amplify differently?

On top of my head, I don't know, you can pick up, do yourself the math and make sense. I chosed to do the measurements at Vgs=0 because that's how JC decided to implement the Norton reference in the power supply buffers in his Vendetta Research (and Blowtorch) RIAA preamps. Also, the ~19V output voltage is not a coincidence.
 
Last edited:
diyAudio Chief Moderator
Joined 2002
Paid Member
And here's a quick pass on a randomly selected LM329. Anode current is 1mA, reference voltage was measured as Vref=7.1V You would of course note the missing 60Hz peak, a good sign that the circuit has a low impedance. Also no filtering (electrolytic or otherwise) was used.

It happens that the noise is, at 108nV/rtHz, as good (or bad) as the best JFET Norton reference...

To follow in the next days, time permitting...

When a leds fast pass?. Low impedance we like.:) We want leds, we want leds, we want leds. :D
 
When a leds fast pass?. Low impedance we like.:) We want leds, we want leds, we want leds. :D

Ok, family is going to kill me anyway:

Red LED, manufacturer is unknown (got them from a Chinatown store) @5mA. Voltage across the LED is 1.86V No filtering (electrolytic or other) and the LED is fed via a 2.7kOhm resistor. I'll do a JFET current source next time, but I'm not exepecting to be anytime better.

About 50nV/rtHz and now I really have to substract the HP89410A noise. SQRT(50^2-35^2)=35.7nV/rtHz You need 4pcs to get to 7V, this will make it to a total of about 70nV/rtHz So far, the best of all, but I wouldn't hold my breath. As I've already said, LEDs have huge variations from manufacturer to manufacturer and from batch to batch. Also, the output impedance is not that low...

Gotta go now...
 

Attachments

  • RedLED.jpg
    RedLED.jpg
    20.7 KB · Views: 298
Last edited:

iko

Ex-Moderator
Joined 2008
syn08, it is very nice of you to take time to run these experiments. I hereby volunteer a case of your favorite beer (if you drink beer), which is too little a reward, you deserve more, but hey, that's my offer :)

I think this would be very valuable for a lot of people. Would you agree to test j201 (low Idss, pinch-off and, go) and 2n3819 (high Idss, pinch-off, and go). I can drop a few of them off anywhere downtown TO around lunch time.
 

iko

Ex-Moderator
Joined 2008
On top of my head, I don't know, you can pick up, do yourself the math and make sense. I chosed to do the measurements at Vgs=0 because that's how JC decided to implement the Norton reference in the power supply buffers in his Vendetta Research (and Blowtorch) RIAA preamps. Also, the ~19V output voltage is not a coincidence.

Fair enough. Unless I made a mistake, my calculation shows output noise of about 456nV/sqrtHz for the j201.
 
Status
Not open for further replies.