Hi Brad,
I assume you are talking about the CA3096 as well for noise. I am using the CA3102 (500 MHz) for RF work. They have about 4.6 dB noise figure. Not fantastic, but could be worse too.
Messy and slightly less time consuming.
I hadn't considered the K170. I was using the various Jxxx North American numbers since they are of known lineage around here. I think I used Allied for those. I also have a bunch of J110 J-FETs. The high IDSS makes them naturals for cascode transistors. I have been playing with J201 and J202 J-FETs as well for instrumentation projects. You know what's really nice about sm parts? They all have the same damn pinout! Finally, some common sense. Of course, they did have to make the J-FET and BJT parts different. Why? I'll never know.
-Chris
I assume you are talking about the CA3096 as well for noise. I am using the CA3102 (500 MHz) for RF work. They have about 4.6 dB noise figure. Not fantastic, but could be worse too.
Yes, they were lateral devices unless Intersil changed them (for the better). The CA3096 is one of them. Someone also has NPN and PNP diff pairs - separate parts.
How about a cup of ice water, soldering iron and a "K" thermocouple?
Messy and slightly less time consuming.I hadn't considered the K170. I was using the various Jxxx North American numbers since they are of known lineage around here. I think I used Allied for those. I also have a bunch of J110 J-FETs. The high IDSS makes them naturals for cascode transistors. I have been playing with J201 and J202 J-FETs as well for instrumentation projects. You know what's really nice about sm parts? They all have the same damn pinout! Finally, some common sense. Of course, they did have to make the J-FET and BJT parts different. Why? I'll never know.
Batteries can be noisy as well. A proper power supply can be less noisy than batteries, so they are really on even footing here. Each requires a lot of effort to keep noise down, then shielding against external noise sources. The box within a box plan is your friend here. As for NIST time. OUCH! For DC specs, just use a Fluke enhanced Keysight 3458A. For AC, you're looking at the usual suspects, and I'll add in the current Keysight signal and network analyzers. You will get good numbers and I doubt that NIST would give you anything more except a fresh calibration on the test instrument of your choice. Keithly also makes some very nice (quiet) instrumentation. If you aren't careful, you may find yourself at the bottom of mine shaft with air conditioning to keep your lab environment electrically quiet enough to chase the real capabilities of today's equipment and circuit design. I suspect that you know very well how to create very quiet power sources and linear amplifiers.
-Chris
Last edited:
The SPX431 is through-hole and actually less noisy than your average 431. Might be the same die as the SPX2431, didn't check the datasheets.
I got one of them borrowed by a member here to test it's performance together with an LM317. The attached example shows the Sipex part vs. an Estek el-cheapo. Schematic and more info here: LM317 regulator under test preamp.org
Attachments
Thanks! Yes, oddly the datasheets are not that similar, and in particular there is no noise spec for the SOT-89 and TO-92. Even the highly simplified equivalent circuits are different. A puzzle.
As a reference with little current pulled I thought at least some batteries were pretty quiet. I think Wurcer referenced some "lowest-noise-ever" apparatus at NIST that was needed to see the actual noise.
What is amusing though is that people will imagine that batteries are truly isolated from everything --- but they still have coupling to the environment, so indeed do need carerful shielding.
The JFETs I mentioned, the obsolete 2SK381, were lower current than I had remembered for zero tempco, with two of them in parallel only about 350uA each. But they are quiet. I'd prefer having five mA or more for applications like these moderate-current shunt regs. But enough in parallel will work. I cascoded the two (I had socketing at hand for them and the cascode device) and developed about 15V across a 21.5k R in parallel with 1uF. An optimized-from-a-previous-experiment ballasting resistor worked well for low tempco. The noise across the 21.5k was indistinguishable from the Ap self noise, which was as I hoped. But 21.5k is larger than I would like, particularly considering the noise in the transistor base current, which is always more than shot noise and rises at low frequencies.
In any event, compared to band gaps and zeners, this approach for references is much lower noise. How much lower requires more measurements to say. It's impractical for a lot of things, as we need some decent voltage across for them to work well. But the dual polarity approach of Preamp's and the ~15V magnitudes fits very nicely. The cascode device was a J112.
Hi Brad,
I did some little work on battery powered low noise circuits a long while ago. I was surprised by how noisy some batteries could be. Placing capacitors in parallel with batteries is always a good idea since the internal resistance tends to rise as some batteries discharge.
-Chris
I missed that post. Darn.
I did some little work on battery powered low noise circuits a long while ago. I was surprised by how noisy some batteries could be. Placing capacitors in parallel with batteries is always a good idea since the internal resistance tends to rise as some batteries discharge.
Not bad Brad. That's a pretty decent result.
I'll agree with you on this.
-Chris
That's of course with the 1uF across the 21.5k, hence a NBW of about 11.6 Hz. It could be still lower with a larger cap.
The major noise source is the current noise of the source resistor, about 2.67k, determined by the ~zero tempco conditions. With more FETs and higher currents the noise will fall further.
Yes a film 10uF is what I had in mind in fact. I might use a 'lytic temporarily.
Mouser had stock, although they keep trying to sell me TI --- insisting that it is a drop-in. Certainly not based on noise for the SPX2431 anyway!
A pic I posted on FB for a friend, of another regulator, with the stacked TL431, I referred to as line dancing.
I didn't. I only remembered some old reads here. Its was supposed to be doing much better at 1mA bias when the gain resistors were used but was not changing noise with current tweaking when running at no gain just on its own 2.5V. Since Brad has an AP and knows metrology it will be nice to confirm or not. And useful for his designs. Because DIY preamps & methods were not that controlled back in the day.
Heh!
My personal favorite using BF862: I stacked about ten of them using a tiny bit of adhesive from package to package, then ran a bus wire along the gate leads as both a composite gate connection and a heat conductor (the chip sits on the gate lead), then made similar connections to drains and sources. It worked, and had about the lowest parasitic inductance short of some hybrid using raw chips.
My personal favorite using BF862: I stacked about ten of them using a tiny bit of adhesive from package to package, then ran a bus wire along the gate leads as both a composite gate connection and a heat conductor (the chip sits on the gate lead), then made similar connections to drains and sources. It worked, and had about the lowest parasitic inductance short of some hybrid using raw chips.
Hi Brad,
Hi Salas,
The easy way out is the worst case numbers with the equipment at the same temperature as it was during calibration. You still have to figure out the other error terms from your test setup and other equipment.
I have the Fluke calibration handbook, it has a blue cover. It can get quite involved if you like minutiae and numbers. Something I'm not into.
-Chris
LOL! That's pretty funny.
Hi Salas,
Interesting. That's right around it's minimum recommended current range I think.
The AP would be nice. The Metrology involved is more a painstaking error budget combined with the expected error limits of the equipment at the current temperature and x number of days since the last calibration. If you had enough cal data, you could work with the known trends and arrive at a lower error term. Most of the time, doing all the work to figure out what numbers you can trust from your equipment and test setup simply depresses you!
The easy way out is the worst case numbers with the equipment at the same temperature as it was during calibration. You still have to figure out the other error terms from your test setup and other equipment.I have the Fluke calibration handbook, it has a blue cover. It can get quite involved if you like minutiae and numbers. Something I'm not into.
-Chris
Hi Preamp,
How do you feed all those little mouths? You have a small army lined up in formation - and yes, many antennas!
Hi Brad,
Did you just stick the assembly onto a board using one set of pads? You could use some kynar wire as guys!
Did you use round or flat wire? I'm just curious because you used the gate lead wire as a heat radiator (I think that's what you were getting at).
-Chris
How do you feed all those little mouths? You have a small army lined up in formation - and yes, many antennas!
Hi Brad,
I wouldn't have known that, thanks.
Did you just stick the assembly onto a board using one set of pads? You could use some kynar wire as guys!
Did you use round or flat wire? I'm just curious because you used the gate lead wire as a heat radiator (I think that's what you were getting at).
-Chris
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.