linear ramp

A really good Spice model would help here... or some measurements! Your circuit is essentially jamming current into the emitter, so Vbe is not fixed. With a high-beta transistor (BCX71K maybe) and lots of C-E voltage, it should be pretty good.

I wonder what the effective source impedance would be for a current source like this woth, say, 5 volts Vce and 5 volts across the emitter resistor. Spare time project.

John

hairy,

Thanks, that's interesting.

I see that the curves all cross near 10 mA, but it's hard to tell how much V_F moves with temperature near there. I'd expect some quadratic-looking curve, but who knows, we might get lucky.

Cheers

Phil Hobbs

The emitter resistor provides negative feedback, which stiffens the collector a lot, against both V_BE and Early effect. How much depends on how much feedback you apply, and as John points out, the nonlinear capacitance is much harder to compensate for. Using a really small transistor (e.g. a BFG25AW) helps a lot because the capacitance is very small anyway, but unfortunately there aren't any good PNP candidates that I know of.

Cheers

Phil Hobbs

Looks good, but... why not just a transistor with a resistor to base and add-on Miller capacitor from base to collector? It turns a step into a ramp.

And, there's less power pins to bypass against high frequency currents.

bs

=A0 =A0|__|

=A0--| |----

|--|

=A0 =A0|

///

ost

I don't see how your bootstrap can be improved upon- all the parasitics are effectively filtered with the bootstrapped constant voltage derived from a heavily bypassed precision active zener and the buffer is there too. You're not even close to stressing the AD8014 working at less than 4% its slew rate and 5% of its BW.

OK, ok, I'll reveal a secret. Put a 1K ferrite bead in the collector of a slow, high-beta transistor like a BCX71K. That will decouple the collector capacitance from the ramp cap. Really helps.

John

I need serious linearity and repeatability, and low temperature effects, from this ramp. It's being used to control some downstream timing that has to be accurate to 200 ps or better. So everything has to be very quantitative.

John

So, instead of 'a resistor', use a very quantitative resistor. And instead of 'a capacitor' use a very quantitative capacitor. Was that answer supposed to mean something?

The problem is to get a very quantitative transistor.

John

Sort of like a video peaking inductor?

What do you mean 1k bead?

Did you get my email?

John

Only more so. The more impedance, the better.

Ferrite beads are usually specified by their impedance at 100 MHz. Values range from 10s of ohms to a couple K. They are magical parts, sort of like putting your finger on a circuit.

Yes. You don't owe me anything. But you don't look anything like I expected.

John

Pretty high. I used a circuit like this for an ADC.. I was prepared to provide a parabolic linearization correction, but it wasn't necessary.

Best regards, Spehro Pefhany

IIRC, the voltage reference noise is a few uV RMS 0.1~10Hz .. but it's not supplying much current, and into not much resistance.

Best regards, Spehro Pefhany

=A0|__|

=A0--| |----

--|

=A0|

It's mostly a conductance/resistance effect... as opposed to a capacitance variation (as the length gets longer)?

I've not knowingly bumped into the Early effect.

George H.

t- Hide quoted text -

It's a DC effect--if you look at a transistor datasheet, you'll see a family of curves of I_C vs V_CE for given values of I_B and/or V_BE.

An ideal transistor would have I_C independent of V_CE (i.e. the curves would be flat) but real ones have some tilt--I_C increases when you jack up V_CE. That's due to Early effect.

Cheers

Phil Hobbs

bs

--

=A0 =A0|__|

=A0 =A0--| |----

=A0 |--|

=A0 =A0|

=A0///

ot

most

t- Hide quoted text -

Yeah, Thanks Phil.

George H.

=A0|__|

=A0--| |----

--|

=A0|

Early effect is best visualized on a curve tracer. All the traces intersect at the Early voltage (negative for a NPN). An you correctly capitalized Early.

John Larkin a écrit :

A quantitative transistor is an opamp...

There're some fast VFB opamps too that you can use as integrators. Will get rid of the opamp input capacitance non linearity for the higher linearity.

On a sunny day (Fri, 08 Apr 2011 17:09:02 -0400) it happened Phil Hobbs wrote in :

OK, so we use the circuit upside down with a negative ramp and a NPN :-)

Here is an other ramp generator, but this one is for only 50 Hz,