fast ramp follies

OK, I need to charge a capacitor with a stable constant current. The desired slope is about a volt per nanosecond.

So I did this:

formatting link

Things like this tend to oscillate, so I used a fairly slow, high-beta transistor, BCX71K. The ferrite in the collector is supposed to isolate the ramp cap from the transistor capacitance and make the ramp linear.

Well, the ramp looked not much better than an R-C curve, and the BCX71 oscillates at 80 MHz. Increasing R1 from 50 to 150 ohms kills the oscillation and makes the curvature worse.

So we went to the opposite extreme, a BFT92, a 5 GHz PNP. Typ beta is

50! The ramp is now visually linear, and the oscillation frequency went up some. Tried a ferrite instead of R1, and it *really* oscillates. A 100 ohm base resistor seems to work.
--
John Larkin                  Highland Technology Inc
www.highlandtechnology.com   jlarkin at highlandtechnology dot com   
 Click to see the full signature
Reply to
John Larkin
Loading thread data ...

I thought you said that "resetting a filter" worked?

You'd be better off with a simple R-C (or R-L-C), and do the compensation with the following comparator reference. ...Jim Thompson

--
| James E.Thompson, CTO                            |    mens     |
| Analog Innovations, Inc.                         |     et      |
 Click to see the full signature
Reply to
Jim Thompson

I'd be more inclined to cascode the current source output rather than add an inductor.

Reply to
miso

Would that help? The beta error would increase, and cascodes can oscillate too. There's hardly any difference.

--
John Larkin                  Highland Technology Inc
www.highlandtechnology.com   jlarkin at highlandtechnology dot com   
 Click to see the full signature
Reply to
John Larkin

Cascodes are the classic way to isolate circuits, which in this case is basically improving the reverse parameters. [Think of how they kill the MIller effect in input states.] The would also increase the output impedance of the circuit, which would make the ramp straighter.

Since the circuit uses feedback, you don't need the precision (i.e. predictability) of a bipolar. Maybe a P-fet can be used, then there is no beta issue.

Reply to
miso

Yup. Beads aren't that nice for this sort of job. If you make the BFT92 into a series-shunt pair, i.e. put a resistor in series with its collector and wrap a super-hot NPN around it, 0 | *-----------* BFT92 | | |< | 0---------| | |\ | | | | |/ *--BEAD---| BFP640 | |> R | R | R | | | *-----------* | 0

you get higher collector impedance since the BFT92 runs at lower and constant I_C, and the BFP640's VAF is effectively infinite. Also, since the BFP640 is so fast, you can push the unity gain cross of the feedback pair out far enough for a 10- or 30-ohm bead to do a good job of stabilizing it, without causing your ramp to roll over.

Since your barefoot BFT92 is working, there's no reason to do this on this particular board, of course, but it's a cute trick. (I've been investigating all sorts of these kinds of local feedback things recently.)

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
 Click to see the full signature
Reply to
Phil Hobbs

We found that something between 22R and 33R right up against the base was what it took to stop a BFT92 from oscillating.

Is L1 - I presume FB means "ferrite bead" - actually doing anything useful? All it seems to be doing is eating up some of the collector- emitter bias on Q1, increasing the collector-base and collector- emitter capacitances, which isn't in any way a useful contribution.

Putting it in series with the emitter of Q1 might make more sense (above the feedback connection to the op amp rather than after it).

If you wanted to go nuts, you could replace Q1 with a complementary Darlington, with a BFR92 delivering 9mA of your 10mA current - you'd need to put at least 1mA through the BFT93 to keep it quick, and the BFR92 needs a 22R base-stopper, just like the BFT93 - but it would reduce the base current error by a factor of ten, it it worked.

In real life they tend to oscillate, but the parts are cheap so it's easy enough to find out.

-- Bill Sloman, Nijmegen

Reply to
Bill Sloman

a
p
1

Or, damp L1 with a parallel resistor.

Right now it looks an awful lot like the classic UHF tuner oscillator:

Vcc --- | L1 | +------. | | |/ | Vbias-+---| Q1 --- Cfb | |>. --- --- | | --- +------+ | .-. | =3D=3D=3D | | --- C1 | | --- '-' | | =3D=3D=3D =3D=3D=3D

This is a common-base amplifier stage with feedback (parasitics are enough) and tuned loads. Damping L1 kills the gain.

-- Cheers, James Arthur

Reply to
dagmargoodboat

e

eta

amp

X71

is

n

Hmmm. L1 + series resistor might work too. That's a little dicier, but still lets L1 act as a super-fast CCS.

-- Cheers, James Arthur

Reply to
dagmargoodboat

You haven't specified exactly how linear it _has_ to be, nor how accurate. What kind of reset restrictions are there (e.g. recovery). And we're guessing that you don't have any higher voltage supplies. All these transistor topologies are fun, but unless we know what's needed...

Reply to
Frank Miles

A couple per cent linearity would be OK. I'd like a percent or so over temperature too, but the beta-50 RF transistor hurts there. There's no temperature sensor on this board, otherwise I could do some compensation in software, crude shot at beta compensation. I really need some day to find a fast, low-capacitance PNP with more beta. Mosfets tend to have a lot of capacitance, and they don't make p-channel PHEMTS.

What kind of reset restrictions are there (e.g. recovery).

Well, the fet shorts out the cap, so it discharges pretty fast. That's not a current source issue.

The box runs off 5 volts from USB, which can actually be 4.5 for a rotten laptop and a cheap cable. I made a little charge pump from an HC14 and a dual diode, to make +8 or so. That seems OK.

--
John Larkin                  Highland Technology Inc
www.highlandtechnology.com   jlarkin at highlandtechnology dot com   
 Click to see the full signature
Reply to
John Larkin

I wonder if you could (ab)use one side of a LVDS driver as a CCS?

-Lasse

Reply to
langwadt

What's missing that is clearly obvious to the most casual observer?

Larkin will now invect (my latest invented word :-)

Folks in the Phoenix area are invited to come by (*), and I'll show how easy it is to do. But I'll never post it here. I want everyone to observe how profoundly ignorant Larkin is ;-)

(*) Not between 12 and 2 today... big pow-wow with the patent attorneys. ...Jim Thompson

--
| James E.Thompson, CTO                            |    mens     |
| Analog Innovations, Inc.                         |     et      |
 Click to see the full signature
Reply to
Jim Thompson

Maybe it's just timing.

Reset always on. Start current source. delay release reset.....

RL

Reply to
legg

I think some more voltage would help - looking at the Ccb vs. voltage for PNPs looks pretty bad. With a node capacitance of 10pF, 2% linearity doesn't give you much leeway there. How far does this sweep have to go? A bootstrap "current source" may be better in this instance. Alternatively, could your topology allow inverting the ramp, so you could use NPNs? Yes, you'd have to move your switch but that may be easier than a positive current source. {I asked about the switch due to ramp- start issues, not ramp-end.}

Reply to
Frank Miles

?

RL

Reply to
legg

There are junction capacitances (aka varicaps) everywhere in a circuit like this. One prays that some of them will sort of cancel.

How far does this sweep have to go?

It's driving some comparators to make delays. The ECL comparators have a max common-mode voltage spec of +3, and they seem to be serious about that. When USB drops to 4.5, it's even less. So, say we want a max of +2.5 volts.

Probably. A follower opamp, a level shifting bandgap, and a resistor back down.

Something like this:

formatting link

The switch is a PHEMT, fast and low capacitance, but definitely n-channel.

--
John Larkin         Highland Technology, Inc

jlarkin at highlandtechnology dot com
 Click to see the full signature
Reply to
John Larkin

And leave the mecca of enlightenment: the SF Bay Area! No thanks.

Reply to
miso

ta

mp

71
s

on

o
a

=A0All

..

r
d
-

This ckt from Mr. Bloggs has pretty high cuteness:

Vcc | .----+----. | | | [120] |LM385adj | .-----. | | + |FB | | |------+ | - | | '-----' | | |< +-------| 3906 | |\ | | [15K] | | V | =3D=3D=3D

That gets you an accurate CCS that's slow, all it needs is speed.

-- Cheers, James Arthur

Reply to
dagmargoodboat

That is elegant. A couple more resistors would isolate the fast stuff and maybe keep the PNP from oscillating.

and maybe a cap across the LM385?

The LM385-adj data sheet has a similar circuit, with a Darlington.

--
John Larkin         Highland Technology, Inc

jlarkin at highlandtechnology dot com
 Click to see the full signature
Reply to
John Larkin

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.