Re: 50 ohm termination

On 2025-03-29 21:59, Gerhard Hoffmann wrote:

Am 30.03.25 um 00:53 schrieb Phil Hobbs:

On 2025-03-19 10:41, john larkin wrote:> On Wed, 19 Mar 2025 06:23:25 -0400, Toaster <toaster@dne3.net> wrote:
 >
 >> On Tue, 18 Mar 2025 19:01:51 -0700
 >> john larkin <jlArbor.com> wrote:
 >>
 >>> On Tue, 18 Mar 2025 18:59:44 -0400, Toaster <toaster@dne3.net> wrote:
 >>>
 >>>> On Tue, 18 Mar 2025 16:02:45 -0700
 >>>> john larkin <jl@glen--canyon.com> wrote:
 >>>>
 >>>>> On Tue, 18 Mar 2025 18:29:42 -0400, Toaster <toaster@dne3.net>
 >>>>> wrote:
 >>>>>
 >>>>>> On Mon, 17 Mar 2025 19:17:13 -0700
 >>>>>> john larkin <jlArbor.com> wrote:
 >>>>>>
 >>>>>>> On Mon, 17 Mar 2025 19:50:17 -0400, Toaster <toaster@dne3.net>
 >>>>>>> wrote:
 >>>>>>>
 >>>>>>>> Thank you for the advice. In my case I have a 10Mhz signal with
 >>>>>>>> very sharp transitions (500ps, 5V) and wanted to make sure I
 >>>>>>>> did things properly.
 >>>>>>>
 >>>>>>> Interesting. What's generating the 5v signal? Lots of AC and
 >>>>>>> Tiny Logic chips are that fast, but might strain to drive 50
 >>>>>>> ohms. We use several tiny triple buffers in parallel sometimes.
 >>>>>>>
 >>>>>>> Regular thick-film surface-mount resistors are fine as
 >>>>>>> terminators at 500 ps.
 >>>>>>>
 >>>>>>> LVDS line receivers are great at the receive end.
 >>>>>>>
 >>>>>>>
 >>>>>>
 >>>>>> I used a THS3111CD. Split up my project into a timing and driver
 >>>>>> board, so i have some 50 ohm BNC cables between and wanted to be
 >>>>>> extra safe about reflections at these higher frequencies.
 >>>>>
 >>>>> Is the signal some analog thing, or a 10 MHz clock? The THS is an
 >>>>> opamp, but they can make good cable drivers too, even for clocks.
 >>>>>
 >>>>> Lately I'm enamored of BUF602, a unity-gain 1 GHz beast.
 >>>>>
 >>>>
 >>>> I had a really hard time finding a good line driver. I might look
 >>>> into this chip.
 >>>>
 >>>> Thank you!
 >>>
 >>> Is your signal analog or digital?
 >>>
 >>
 >> digital
 >
 > I use Tiny Logic triple buffers as line drivers, with all three
 > sections in parallel, and then sometimes two or three chips.
 >
 > NL37WZ16US costs 10 cents.
 >
 > https://www.dropbox.com/scl/fi/gw7wetgtovqc04as2gxol/NL37WZ16_5V_Pulse.JPG?rlkey=2eqbyhds8l1myrzfjsrqwn5b3&raw=1 >
 >
 > That US8 package is nasty to solder or probe.
 >
Just rereading this.  John,  the prop delay spreads in the datasheet are all over the place--almost a factor of 2 from typical to max over temperature.
>
I'd expect the three sections to match OK, but paralleling packages seems quite a lot sportier.  How well does that work in production?
>
Cheers
>
Phil Hobbs

 This here are 2 single TI 74LVC inverters, 100 Ohms on each output,
RG174 / RG188-like Coax into the 50 Ohm of an Agilent 2.4GHz scope.
 < https://www.flickr.com/photos/137684711@N07/32245910240/in/album-72157662535945536/       >    + 2 pics to the right.
 Sorry for the blue on black trace; it is default for input 3 that
happened to be unused. Looks better if downloaded.
Having a GND/VCC pair for each output is friendlier to GND bounce
and temp rise is also smaller in comparison to multichannel gates.
 Gerhard
 

Thanks.  We do similar things, but only with single packages--we often drive fast (500 ps to 5 ns) LEDs with 74AHC04s or 74LVC04s in QFN. (QFNs are dramatically faster for this than TSSOPs or SOICs, on account of the much lower package inductance in the power and ground.)
For that job, the trick is to parallel four sections with one resistor at the output, and drive them with the other two sections in cascade, to get a bunch of voltage gain.  That sharpens up slow input edges very nicely.  (In bad cases, one might want a 1G14 Schmitt in there as well.)
Cheers
Phil Hobbs
--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510
http://electrooptical.net
http://hobbs-eo.com