linear ramp

Nope. With no DC bias they look more like an inductor with a parallel resistor and cap. Add a series resistor to get a bit closer. Here's such a simple model for a 1K @ 100MHz part:

___ +--------UUU----+ | 1.65uH | | ___ | ___ o---+-------|___|---+--|___|----o | 1.35K | | || | 0.6 ohm +------ -||-----+ ||

(from Spice model below.. this works out to |Z| = 969 ohms at 100Mhz)

Data sheet here, see curves:

More complex models abound.

• SPICE Netlist Generated by TDK Corporation
• Copyright(C) 2011 TDK Corporation.
• All Rights Reserved.

• TDK P/N: MMZ1005Y102C (Chip Bead)
• Property: |Z|=1000ohm at 100MHz
• Size(LxWxT): 1x0.5x0.5mm, 0.039x0.02x0.02inches
• Model Type: Simple Model
• Model Generated on Feb. 23, 2011

• External Node Assignments:

• 1 ---@@@--- 2

• Applicable Conditions:
• Temperature = 25 degC
• DC Bias Current = 0 A
• Small Signal Operation

Best regards, Spehro Pefhany

On a sunny day (Sat, 09 Apr 2011 06:28:13 -0400) it happened Spehro Pefhany wrote in :

Very nice, thank you. I just learned something, this can be very useful.

Neat circuit, but...

Why does it need the op amp, an expensive solution ?. What's wrong with a single pnp transistor emitter follower, with collector to -5v and emitter to +15 via resistor ?. Also, if the refresh rate is only

Regards,

Chris

There's always room for a bit of politics, but seems to have been a lot of that recently, including brainless slanging matches, which is the point at which I start to lose interest.

Tech stuff and discussion is what this group is so good at :-)...

Regards,

Chris

It's that old hang-dog look, I guess.

John

It would have the 2.5 mv/K Vbe drift, and I need the ramp to be precise. And base current would affect things, too. In this application, performance is way more important than parts cost.

Also, if the refresh rate is only

An open collector would have a capacitance that varies with voltage, which would curl my ramp. The schottky diode has nonlinear C, but not much of it... maybe a tenth of a pF over the active range of my ramp.

There really aren't any fast open-collector comparators anyhow. LVDS receivers are fast and cheap, and make excellent r-r comparators.

John

Excellent post, Spehro. Thanks for that.

John

Most of the surface-mount ones we use seem to look like a few uH at low frequencies, not particularly correlated to the "ohms at 100 MHz" value.

They are great for EMI. They kill Q beautifully.

ftp://jjlarkin.lmi.net/Ferrite.JPG

John

I didn't know John had a rusty barbeque. Mine is a commercial grade stainless steel DCS :-) ...Jim Thompson

LVDS isn't RFI-prone... it's differential.

There are better ways, though I hesitate to post something Larkin would steal.

Anyone needing such a ramp contact me privately. ...Jim Thompson

A 100-ohm bead in the base usually prevents transistors from oscillating, and does it without adding noise at low frequency.

Nice trick for isolating the nonlinear capacitance.

Cheers

Phil Hobbs

The single point of intersection is closer to the truth when you're using old, slow, high voltage transistors. I suspect that this is because they act much more like a 1-D model with a small perturbation. Small, fast, lower voltage transistors have a lot more curvature to their I_C vs V_CE plots.

Cheers

Phil Hobbs

If you ever wanted to go crazy you could feed the schottky with a step

• cheesy ramp. That eliminates the capacitance and leakage modulations.

-- Cheers, James Arthur

y,

e.g., like this:

.-------< from ramp | V1 .-. -+- | | R1 |\ | | | | \ ------ '-' D1

--| \ | | | schottky | >--' '----o-----|

-o| / Q1 | | / | |/ .-. | | R2 | | (optional) '-' | =3D=3D=3D

-- Cheers, James Arthur

Thompson was born an asshole, and has spent the rest of his life building on that.

A gutless asshole, as his posts demonstrate.

John

Based on some admittedly delicate measurements that we made on this diode, the delta-C of the schottky is probably around 20 fF in the active part of the ramp. I'd be more concerned with the nonlinear input capacitance of the opamp... if it weren't for the fact that I can get away with 1% linearity in this application.

One way to get a more linear ramp is to keep all the junctions back-biased as much as possible (to linearize the c-v curve as much as possible) and then apply a square-law or exponential correction (pretty much the same thing for small numbers) by ramping the current source.

If this is being used in a delay vernier or in a time stamper (using DACS or ADCs respectively) and you can afford to buy more bits, or miss some codes, then digital correction is the way to go.

You could play with this for a long time.

John

airy,

with

R2 allows for a little of that--it'll bend the ramp's tail down a little.

Use a PIC! (or a 555)

Cheers, James

=A0 =A0...Jim Thompson

A gutless asshole does seem to be a self-contradicting notion - the asshole is one end of the gut, after all.

There's no doubt that Thompson is an unsavoury character, and his habit of kill-filling people who post stuff with which he fails to agree is the action of someone who retreats from conflict, but John Larkin should have been able to come up with a more felicitous label.

-- Bill Sloman, Nijmegen

He's responded to my posts recently, so he lied about kill-filing me.

He insulted my wife, which makes him an asshole.

He's gutless, because he keeps referring to wonderful circuits he won't post.

Let's be charitable and assign some part of his idioticy to senility.

John