On Tue, 31 Jan 2017 10:51:11 +0100, Gerhard Hoffmann <ghf@hoffmann-hochfrequenz.de> wrote:>Am 31.01.2017 um 05:08 schrieb John Larkin: >> >> >> I'm doing a fast thing and thought I might go all CMOS, instead of >> traditional expensive ECL. We have the Fairchild NC7SV74 tiny logic >> flipflop in stock, 16 cents each. I thought I'd play with one. >> >> Here are the rise and fall times at the Q output, 3.3 volt supply: >> >> https://dl.dropboxusercontent.com/u/53724080/Parts/Logic/NC7SV74_2.JPG >> >> That fall time is amazing, but the rise time is hard to believe. I >> can't see that I'm doing anything wrong. >> >> The clock rise to Q rise prop delay is 0.8 ns. Its temperature >> coefficient was hard to measure, but it looks like maybe +0.7 ps per >> degree C. Vcc delay coefficient is so close to zero that it doesn't >> matter. >> >> >Looks interesting! > >This is what I got from a pair of SN74LVC1G04DCKR, 50 Ohm source and >load terminated, running on 6V, still legal. I wanted acceptable 3V3 >CMOS levels @ 50R input. >It varies somewhat from batch to batch. > >< >https://www.flickr.com/photos/137684711@N07/32245910100/in/album-72157662535945536/lightbox/ > > and left/right > >Scope rise time is 145 ps.That's pretty clean.> >cheers, GerhardThe Onsemi NL37WZ16US buffer is a nice part. Put all three gates in parallel, run at abs max Vcc, source terminate at maybe 40 ohms, and it's a vicious sub-ns output driver. -- John Larkin Highland Technology, Inc picosecond timing precision measurement jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
insane flipflop measurements
Started by ●January 31, 2017
Reply by ●January 31, 20172017-01-31
Reply by ●January 31, 20172017-01-31
On Tue, 31 Jan 2017 16:22:37 -0000 (UTC), Frank Miles <fpm@u.washington.edu> wrote:>On Mon, 30 Jan 2017 20:08:11 -0800, John Larkin wrote: > >> I'm doing a fast thing and thought I might go all CMOS, instead of >> traditional expensive ECL. We have the Fairchild NC7SV74 tiny logic >> flipflop in stock, 16 cents each. I thought I'd play with one. >> >> Here are the rise and fall times at the Q output, 3.3 volt supply: >> >> https://dl.dropboxusercontent.com/u/53724080/Parts/Logic/NC7SV74_2.JPG >> >> That fall time is amazing, but the rise time is hard to believe. I >> can't see that I'm doing anything wrong. >> >> The clock rise to Q rise prop delay is 0.8 ns. Its temperature >> coefficient was hard to measure, but it looks like maybe +0.7 ps per >> degree C. Vcc delay coefficient is so close to zero that it doesn't >> matter. > >From the sticky-note, the risetime is from the clock; the falltime from >Clear. Does the fall-time change if the device is configured to toggle? >{WAG not what you want in your application}It looks faster on the falling edge if I only use the clock. In toggle mode, rise is maybe 150ish, fall maybe 250. There is some overshoot and ringing, and that fools the scope rise/fall measurements, but it's plenty fast. I'll be using clock and clear to make my pulses. -- John Larkin Highland Technology, Inc picosecond timing precision measurement jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
Reply by ●January 31, 20172017-01-31
Am 31.01.2017 um 23:31 schrieb John Larkin:> The Onsemi NL37WZ16US buffer is a nice part. Put all three gates in > parallel, run at abs max Vcc, source terminate at maybe 40 ohms, and > it's a vicious sub-ns output driver.Mouser says it is obsolete. Maybe reels of 6000 still available. Sigh, Gerhard
Reply by ●January 31, 20172017-01-31
Am 31.01.2017 um 23:57 schrieb Gerhard Hoffmann:> Mouser says it is obsolete. Maybe reels of 6000 still available. > > Sigh, Gerhard >No, the -usg version is still available. They are waiting for the 6K devices.
Reply by ●January 31, 20172017-01-31
On 01/02/17 09:31, John Larkin wrote:> The Onsemi NL37WZ16US buffer is a nice part. Put all three gates in > parallel, run at abs max Vcc, source terminate at maybe 40 ohms, and > it's a vicious sub-ns output driver.What's the best edge-time you've seen from these TinyLogic parts? Got a plot handy? Better than an LVDS driver? I want to build (hobby) a cheap TDR circuit, what's the best driver approach to get into the 100ps vicinity? (similar speed sampler needed, of course). I keep collecting ideas from you and Phil, but haven't built anything yet... I lack test equipment to look at this stuff, so need something that can verify its own behaviour. A little more help would be welcome. Clifford Heath
Reply by ●January 31, 20172017-01-31
"Jon Elson" <jmelson@wustl.edu> wrote in message news:oMWdnUjCoLgObg3FnZ2dnUU7-bnNnZ2d@giganews.com...> Watch out for the shoot-through current on fast CMOS parts. I did a > design > some years ago, mixed-signal stuff, and had insideous crosstalk issues. > After realizing what it might be, I discovered the single-gate and > single-FF > parts had a shoot-through of about 0.5 - 1 A that lasted about 3 ns (might > have been shorter, that was the best I could resolve with the scope and > probes I was using.)Hrm, curious how you measured that? 1A/3ns is at least 1.7V dropped across the package inductance alone. At 3.3 or 5V supply and very nearby bypassing, I can see it being possible, at least. The impedance of such a [presumed] signal is, at most, 10 ohms: considerably less than the Zo of any reasonable structures in the vicinity (bondwires, pins, PCB pads, even shunt resistors if not done carefully), so the stray inductance will be quite dominant. Definitely a challenge to measure accurately. (I wonder if anyone makes a Rogowski coil small enough to slip around an SOIC pin...) Tim -- Seven Transistor Labs, LLC Electrical Engineering Consultation and Contract Design Website: http://seventransistorlabs.com
Reply by ●January 31, 20172017-01-31
On Wed, 1 Feb 2017 00:02:35 +0100, Gerhard Hoffmann <ghf@hoffmann-hochfrequenz.de> wrote:>Am 31.01.2017 um 23:57 schrieb Gerhard Hoffmann: > >> Mouser says it is obsolete. Maybe reels of 6000 still available. >> >> Sigh, Gerhard >> >No, the -usg version is still available. They are waiting for the 6K >devices.Yikes, don't scare me like that! http://www.digikey.com/products/en?keywords=NL37WZ16US There are other, drop-in parts, but the Onsemi version is the best. -- John Larkin Highland Technology, Inc picosecond timing precision measurement jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
Reply by ●January 31, 20172017-01-31
On Wed, 1 Feb 2017 10:31:28 +1100, Clifford Heath <no.spam@please.net> wrote:>On 01/02/17 09:31, John Larkin wrote: >> The Onsemi NL37WZ16US buffer is a nice part. Put all three gates in >> parallel, run at abs max Vcc, source terminate at maybe 40 ohms, and >> it's a vicious sub-ns output driver. > >What's the best edge-time you've seen from these TinyLogic parts? >Got a plot handy? Better than an LVDS driver?This flipflop rise time is the fastest I've seen, 150-200 ps sort of range. LVDS edges are fast, but the voltage swing is small.> >I want to build (hobby) a cheap TDR circuit, what's the best >driver approach to get into the 100ps vicinity? (similar speed >sampler needed, of course).A step-recovery diode is the conventional way. Some are cheap. I think MA44769 is still available, SOT23, under a dollar. Phil and Joerg have made fast pulses with cheap transistors too.> >I keep collecting ideas from you and Phil, but haven't built >anything yet... I lack test equipment to look at this stuff, so >need something that can verify its own behaviour. A little more >help would be welcome.Picosecond electronics is fun. If you really want to get into that, get a sampling scope from ebay. The investment is low hundreds to low thousands of dollars, depending on bandwidth. Old Tek and HP manuals are enormously educational. -- John Larkin Highland Technology, Inc picosecond timing precision measurement jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
Reply by ●January 31, 20172017-01-31
On Tue, 31 Jan 2017 14:56:29 -0800, John Larkin wrote:> On Tue, 31 Jan 2017 16:22:37 -0000 (UTC), Frank Miles > <fpm@u.washington.edu> wrote: > >>On Mon, 30 Jan 2017 20:08:11 -0800, John Larkin wrote: >> >>> I'm doing a fast thing and thought I might go all CMOS, instead of >>> traditional expensive ECL. We have the Fairchild NC7SV74 tiny logic >>> flipflop in stock, 16 cents each. I thought I'd play with one. >>> >>> Here are the rise and fall times at the Q output, 3.3 volt supply: >>> >>> https://dl.dropboxusercontent.com/u/53724080/Parts/Logic/NC7SV74_2.JPG >>> >>> That fall time is amazing, but the rise time is hard to believe. I >>> can't see that I'm doing anything wrong. >>> >>> The clock rise to Q rise prop delay is 0.8 ns. Its temperature >>> coefficient was hard to measure, but it looks like maybe +0.7 ps per >>> degree C. Vcc delay coefficient is so close to zero that it doesn't >>> matter. >> >>From the sticky-note, the risetime is from the clock; the falltime from >>Clear. Does the fall-time change if the device is configured to toggle? >>{WAG not what you want in your application} > > It looks faster on the falling edge if I only use the clock. In toggle > mode, rise is maybe 150ish, fall maybe 250. There is some overshoot > and ringing, and that fools the scope rise/fall measurements, but it's > plenty fast. > > I'll be using clock and clear to make my pulses.Interesting, thanks.
Reply by ●January 31, 20172017-01-31
"John Larkin" <jjlarkinxyxy@highlandtechnology.com> wrote in message news:17929cpj3lgerjo5mmfnjm0dvbn5ee1i9g@4ax.com...> On Wed, 1 Feb 2017 10:31:28 +1100, Clifford Heath <no.spam@please.net> > wrote: > >>On 01/02/17 09:31, John Larkin wrote: >>> The Onsemi NL37WZ16US buffer is a nice part. Put all three gates in >>> parallel, run at abs max Vcc, source terminate at maybe 40 ohms, and >>> it's a vicious sub-ns output driver. >> >>What's the best edge-time you've seen from these TinyLogic parts? >>Got a plot handy? Better than an LVDS driver? > > This flipflop rise time is the fastest I've seen, 150-200 ps sort of > range. LVDS edges are fast, but the voltage swing is small. > >> >>I want to build (hobby) a cheap TDR circuit, what's the best >>driver approach to get into the 100ps vicinity? (similar speed >>sampler needed, of course). > > A step-recovery diode is the conventional way. Some are cheap. I think > MA44769 is still available, SOT23, under a dollar. Phil and Joerg have > made fast pulses with cheap transistors too. > >> >>I keep collecting ideas from you and Phil, but haven't built >>anything yet... I lack test equipment to look at this stuff, so >>need something that can verify its own behaviour. A little more >>help would be welcome. > > Picosecond electronics is fun. If you really want to get into that, > get a sampling scope from ebay. The investment is low hundreds to low > thousands of dollars, depending on bandwidth. > > Old Tek and HP manuals are enormously educational. > > > -- >The Tek 11801/11802 sampling scopes can be had for very low cost. Here is one for $150 plus $111 shipping that is showing errors. The repair is almost certainly a pair of dead memory NVRAM packages and would run about $30 to restore. http://www.ebay.com/itm/H133984-Tektronix-11802-Digital-Sampling-Oscilloscope-/391662629720?hash=item5b30e97358:g:1KkAAOSwEzxYQgcd These are really amazing scopes if you are interested in the ps world. With a SD24 sampler module, you can do TDR in the millimeter range. And two channels for differential measurements.
