Ring of two current source

Er... if dV/dT = 0.001 V/K, then how does dT < 0.01 K help you?

Indeed, highly insensitive circuits need extremely _large_ inputs to register useful outputs. More along the lines of a cardboard box with a resistor and fan inside. Now *that's* engineering.

Tim

Using diode bridges to mimic variable resistance has indeed been done before. In the 1960s some audio compressors and tape recorder automatic level controls used them and searching online should bring up circuits. The term "variolosser" was also used.

Some designs bypassed the need for opposing variable control current sources by using single ended current drive and floating the signal path using transformer in and out. Cannot say anything about temperature stability though.

piglet

Avoid the cardboard box with a heat lamp inside. *That's* experience. ;-)

The measuring instruments were going in the thermostat box, not the DUT (device under test).

I was thinking of a FET phase splitter, with equal resistances in series with the source and drain, and positive and negative FET current sources with a chopamp (chopper stabilized op amp) each, forcing the drop across their source resistors to equal that across the phase splitter resistors. All four resistors should be from the same array.

Since it sounds like you care more about balance than about the absolute current accuracy, the array doesn't have to be anything too gold-plated.

Cheers

Phil Hobbs

I just got it and skimmed through it.

It's certainly grades below AoE. I'm glad I only paid 30 bucks for it

Cheers

Klaus

I've sketched out this scheme, it benefits from an instrument amp with good CMRR to regulate the phase splitter. But, of course, there's another way: just make a battery-powered current source, that floats. The current source pin is the current source, and the Vee (negative battery terminal) is the sink.

Those two currents, balance as well as the best insulation you can arrange for the circuit board and battery box. Kirchoff insists on this!

That's inside out from the case I was considering, which is like an OTA. Yo ur method can't slew an output without being unbalanced, e.g. by putting a resistor in one side.

Cheers

Phil Hobbs

John, can you tell us more about how you came up with the tempco data? E.g., 84 ppm/deg-C, etc. It's the resulting current-output tempco, right?

Yes, that's the collector current. I just measured the TC; cardboard box, heat gun, thermocouple. Not extremely accurate measurement, but good enough for my application. With a better emitter resistor, one might optimize the LED current for min TC.

The emitter resistor was a cheap axial 1% metal film part, so that may dominate the TC that I measured.

I think a resistor from emitter to ground might null out most of the power supply sensitivity. I didn't try that.

I could have done it better, sorry, but I was in sort of a hurry and it was good enough. That current charges a capacitor in a fast ramp+comparator adjustable pulse width

I don't know if the base resistor is necessary, so I left it in on the actual PCB.

Can you tell us what transistor type you used?

NO, no, that's encrypted!

How am I going to make a linear picosecond ramp if I don't know what kind of transistor to use?

Spoiler alert! MMBHT81 == MPSH81

piglet

Close guess, but wrong!

Den onsdag den 13. maj 2015 kl. 00.51.50 UTC+2 skrev John Larkin:

sot23 vs. to92

-Lasse

MMBTH81 ^^

Nice little part, 20 volt PNP with Ft around 1 GHz. Kinda like a BFS17, fast but not insanely fast.

I'd use a higher beta part in the current source when speed didn't matter, BCX71 or some such.

I just bought a whole bunch of PN5179s, which are the same sort of part except in TO-92 so I can prototype with them more easily.

Cheers

Phil Hobbs

For sot-23 npn, how about BFT25, 0.3pF 5GHz but beta = 80.

But I'm still curious about your pnp choice. It wasn't H81 was it? There's gotta be a better choice than that.

BFQ149, BFT92, BFT93 and the W versions ... poor beta. Maybe the NE97833, NE97733-A, etc. Nah... OK, how about Intersil's awesome HFA3135, 7GHz, beta = 57.