kb8tq@n1k.org said:

Is that required for real circuits or just for the simulations?

--
These are my opinions.  I hate spam.

There are a plethora of ways to build up a current source. The nice thing
about spice is you can start with a generalized model to see which way you
need to go. For a bipolar current source (sampling current at the emitter)
you are going to achieve a maximum output resistance of beta*ro with an
active servo element as you have included, or with heavy emitter
degeneration. For a basic discrete bjt like a 2n3904, a you will also
include 3-4 pF of output capacitance. For a 20 mA bipolar current source
you are looking an output resistance of 500 kOhm. You can try an ideal 20
mA current source shunted with a 500 kOhm resistor and 4 pF capacitor.

If you need higher output resistance you will have to move to a FET based
approach. If you need less than 1 pF of output capacitance you will need a
better transistor and care in how you physically construct your circuit.

On Fri, Jul 1, 2016 at 2:54 PM, Hal Murray hmurray@megapathdsl.net wrote:

Hi

Works for both.

Bob

On Fri, 1 Jul 2016 15:45:36 -0400
Scott Stobbe scott.j.stobbe@gmail.com wrote:

I do not think that the output resistance is really the limiting effect
here, but rather the speed at which the transistor reacts, ie its f_t.
But I have to admit that I have not fully understood the circuits
behaviour yet. I still lack good understanding of how a transistor
behaves under high frequency conditions :-/

		Attila Kinali

--
Malek's Law:
Any simple idea will be worded in the most complicated way.

The transition frequency of a transistor is more of a ballpark figure of
merit than a device specific constant, as it is also a function of how the
device is biased. Since most amplifiers and oscillators need power gain to
operate, current gain is usually needed (though you could use voltage gain
at the expensive of bandwidth) making ft a useful figure. The transition
frequency is also a good proxy for the effective junction capacitances and
forward transit time of the device.

For a current source, the desired outcome is for the circuit to output one
specific current and no other. So, having a large forward ac-current
bandwidth is not of great importance, as you don't want it to react at all,
you want it to keep sinking/sourcing one current value.

The issue you are having with your circuit is in biasing and loop
compensation. The standard spice model for a transistor does not include
lead inductance (most manufactures will provide a subcircuit model for
their device which includes package parasitics). Referencing the schematic
you previously attached, once your servo amplifier runs out of steam, if Q7
were a 2N3904 looking into the emitter of the device would look like a
10-15 Ohm resistor (1/gm + Rs/beta) in series with a 200 - 250 nH inductor
(tf Rs). When you changed to a BFU520 you lowered the impedance seen at the
base of Q16.

As a test for yourself, If you drive the base of Q16 with an appropriately
chosen voltage source (roughly 3.2 to 3.3 V) you will see a 2n3904 perform
dramatically better.

On Sat, Jul 2, 2016 at 1:04 PM, Attila Kinali attila@kinali.ch wrote: