> I am interested in just how you are going to make a really quiet current source with the many ma needed for this circuit.
Many BF862's in parallel, I would.
Perhaps with 10R or so source resistor for each JFET ....
Patrick
Many BF862's in parallel, I would.
Perhaps with 10R or so source resistor for each JFET ....
Patrick
Well the key is that it doesn't have to be that quiet. If they were two independent sources/sinks simply substituted for the existing pullup/pulldown resistors, one would need a good deal more voltage so the active devices would have enough ballasting for low noise.
But to understand the function of the floating source, consider starting with two independent sources. Each has so much noise, and the noise of each is adding at the output node as root-sum-of-squares, just like the resistor noises do.
Now wave a magic wand and make the noise in each current source 100% correlated. When the upper source increases in magnitude a bit, the lower one does as well. But assuming that the folded cascode parts are matched and passing source currents without loss to the drains (and the output R), the cancellation of the fluctuation is perfect. The fluctuation effects are equal and opposite.
Of course the P part has more gate-source C than the N part, so the balance at higher frequencies will deteriorate as some of the fluctuation will be lost to Cgs current. But this is a fairly small effect. There will also be the difference in drain output conductance and capacitance between the two complementary input parts, so the sharing of the fluctuations will be a little different due to that as well.
So, in the absence of magic correlation wands, we get the same cancellation by using a single current source with enough compliance voltage---with the inconvenience that it must float, either with batteries or a well-isolated supply of some other sort. For the 16mA I needed for this experiment I get adequately high Z from a quad of JFETs with about 110 ohms in each source. The spec on the 2SK381 is 10uS of drain conductance at Idss, i.e. a slope resistance of about 100k without external source resistance, so it's already a fairly long-channel device, and the ballasting R's raise the slope resistance some more. 2SK246s would probably work well (except they are obsolete now too).
With a ~22V supply there is 7V left for the I source to operate while the whole thing spans the ~15V between the cascode parts' sources, at +7.5V and -7.5V.
I would stress that the anticipated noise improvement is small. The last sims I did indicated perhaps a 6% reduction, which isn't that much. The other benefits include the slightly higher gain and the reduced sensitivity to noise at the cascode device gates, but whether any of this is worth the trouble is debatable.
I bought a bunch of NiMH AA cells when I was playing with some of EUVL's JFET current conveyors ("I-V converters" when terminated with a resistor).
Li-ion rechargables are a bit harder to get as naked cells, owing to their propensity for starting fires. Since I have proud flesh from a couple of disasters involving other components, I tend to give Li a miss.
The batteries are helpful for a floating source, and they usually are low noise as well. But even batteries are not truly floating as they have capacitance to the outside world, so there is a tradeoff between operating time motivating larger cells, and better isolation helped by smaller ones. I believe Patrick mentioned using groups of AAA cells with welded leads for his circuits.
I'm not building a product here btw. This was just a response to the question posed a while back of whether there might be any performance improvements to be had, starting from the simplified Vendetta-style circuit. I had been working on an earlier version of a phono pre some time ago which had some places where the floating source appeared to be something that would help, but I've since moved on to what I think will be quite adequate that does not require the float.
Bcarso, I hope it works for you. I worry about getting any sort of noise cancellation, and worry also about noise gain of the current source.
It gets converted rather efficiently to triglycerides, alas. So more like a carb. There is some countervailing effect that has been discussed that makes those triglycerides not as deleterious as others, but this was probably wishful thinking among well-informed lushes.
There's another way to enforce correlation between otherwise-independent current generators, using a unity-turns-ratio transformer. But it takes such large magnetization inductances to work down to low frequencies that it's probably worse than the floating supply approach.
Of course absolutely agree. I was referring mainly to the input selection
I have used the PGA2310 and can attest to its superb performance. I have a write- up on my website on a preamp I did using them.
For some reason I got the invite to bid on the Fairchild auction of their 6" wafer line in West Jordan, Utah. Maybe I can bring back the original 741's?
Actually I thought they started on the 1" wafers. But I will defer to you on this issue. (703s 1"?)
They were 2" by 1969 and 741 is circa 1970, can't find any references to 1" used for production of anything but exotic stuff still readily available BTW.
Last edited:
741 very late 60's. I was using 709s in high school. Finished in 71. Had 703's in junior high, not really useful. But I assumed you were right about 2". Back then 741s were state of the art. Getting that compensation capacitor into the chip really was a leap.
1968 is circa 1970, these days proving how old you are is losing its appeal most of my friends that remember 1" wafers are dead or disabled.
- Status
- Not open for further replies.