John Larkin <jjlarkin@highland_snip_technology.com> wrote:> > I've been doing picosecond electronics for decades, and sold a ton of > it... more as time goes on. Sometimes things happen on the first PCB > rev of a difficult design. We fix it. No big deal.I've been doing picosecond stuff too. 16,666,666,666 ps, unfortunately.
Fast buffer idea
Started by ●May 14, 2017
Reply by ●May 16, 20172017-05-16
Reply by ●May 16, 20172017-05-16
On 05/16/2017 09:57 AM, John Larkin wrote:> On Tue, 16 May 2017 03:29:30 -0500, "Tim Williams" > <tiwill@seventransistorlabs.com> wrote: > >> This was fun. Not bad for 3904s. >> https://www.seventransistorlabs.com/Images/High_Speed_Buffer.png >> See how it works with 10GHz transistors. :^) >> >> Tim > > You have almost built an opamp! > >It's a little different than a usual op-amp topology - at low frequencies the LTP and mirror are driving the emitter followers through Q6 and Q7, and high frequencies they're bypassed by C1 and C2 to put current directly into the output transistor bases. I think it's trading off open-loop gain at high frequencies for lower overall phase shift and higher slew rate.
Reply by ●May 17, 20172017-05-17
On Tue, 16 May 2017 18:05:10 -0700 (PDT), pcdhobbs@gmail.com wrote:>>I chide Larkin on his "picosecond electronics". Do you accept his claims? > >>I thought not. > >You really shouldn't let your temper lead you into putting words in people's mouths. > >I have a few of JL's products in my lab. My favourite is the P400 digital delay generator, which I use a fair amount--it's my go-to pulse generator for everything. It's similar to an SRS delay generator I used to have except that its jitter is a good 30 dB less (~5 ps vs ~200 ps). That sort of improvement doesn't happen by accident.When we get around to doing a next gen, I'll send you one. The P400 works fine, but it is sort of ugly. A DDG is a wonderful thing to have around, a real Time Machine. -- John Larkin Highland Technology, Inc lunatic fringe electronics
Reply by ●May 17, 20172017-05-17
On Tue, 16 May 2017 21:24:38 +0200, Gerhard Hoffmann <ghf@hoffmann-hochfrequenz.de> wrote:>Am 16.05.2017 um 20:16 schrieb John Larkin: > >>> There are cases where an extra resistor is actually needed, such as a >>> base parasitic resistor or in the gate of a fast MOSFET. These can remind >>> you to include the resistor in the pcb layout, and help determine the >>> value needed in the circuit. >> >> >> I'm not likely to forget, but Spice is useless in determining the >> value that will kill oscillations. It can establish that some value >> won't slow down normal operation too much. I tend to use 33 ohms. >> >> > >Yes. I second that. In simulation, my 4*IF3601 preamp is definitely >stable, but in real live: > >< >https://www.flickr.com/photos/137684711@N07/34701106245/in/album-72157662535945536/ > > > >For most frequencies outside of the Smith diagram, if you look into >the input. Any L on the input with small enough resistance will >oscillate, including ferrite beads. The usual gate stopper is >not possible for obvious noise reasons. > >:-( Gerhard > >(Sorry for the choice of colours.)That's an amazing part, 0.3 nv/rthz at 5 mA. As you say, you can't put anything in series with the gate without adding Johnson noise. Four of them is (calculates furiously) 0.15 nv/rthz! Are the fets hard in parallel? Maybe they could run separately, and sum downstream. I just designed an amp that has two independent gain paths, resistively summed at the very output. -- John Larkin Highland Technology, Inc lunatic fringe electronics
Reply by ●May 17, 20172017-05-17
On Tue, 16 May 2017 22:47:57 -0400, bitrex <bitrex@de.lete.earthlink.net> wrote:>On 05/16/2017 09:57 AM, John Larkin wrote: >> On Tue, 16 May 2017 03:29:30 -0500, "Tim Williams" >> <tiwill@seventransistorlabs.com> wrote: >> >>> This was fun. Not bad for 3904s. >>> https://www.seventransistorlabs.com/Images/High_Speed_Buffer.png >>> See how it works with 10GHz transistors. :^) >>> >>> Tim >> >> You have almost built an opamp! >> >> > >It's a little different than a usual op-amp topology - at low >frequencies the LTP and mirror are driving the emitter followers through >Q6 and Q7, and high frequencies they're bypassed by C1 and C2 to put >current directly into the output transistor bases. I think it's trading >off open-loop gain at high frequencies for lower overall phase shift and >higher slew rate.No, the polarities are wrong for C1 and C2 to be forward gain paths. -- John Larkin Highland Technology, Inc lunatic fringe electronics
Reply by ●May 17, 20172017-05-17
Am 17.05.2017 um 04:26 schrieb Ralph Barone:> I've been doing picosecond stuff too. 16,666,666,666 ps, unfortunately.Yeah, lots of them!
Reply by ●May 17, 20172017-05-17
"John Larkin" <jjlarkin@highland_snip_technology.com> wrote in message news:932nhcdorp9ld65pv8gllc5ornh95t2f6s@4ax.com...> You don't care much about the even mode in Ethernet, since the signals > are transformer coupled and nicely balanced on both ends. > > So a single LT Spice TLINE or LTLINE would be an OK model for UTP. > LTLINE still wouldn't get the skin loss right.If you want a complete model, just crank the CM impedance up to a reasonable value. 150 ohms is ballpark typical of UTP in raceways, intermingling with other cables. This is why they terminate the cable-facing windings with resistors to ground. (Except it's not really ground, it's ESD-sinking-ground, with a 1nF 1.5kV cap. And, guess what an 8kV contact ESD hit divides down to, when connected to 1nF? Right, around 1.5kV -- the system works, and that's where the isolation spec comes from.) Ethernet transformers are largely rated for 30dB of anything (CMRR, balance, isolation). They're hardly perfect*, so the CMC on the outside definitely helps, even if it isn't terminated into much. *The 30dB specs look more like conforming to the standard, than measured characteristics of anyone's actual product. Like a JEDEC 2N2222 spec, you might get something that skirts the standard, or you might get something a whole lot better... who knows. Tim -- Seven Transistor Labs, LLC Electrical Engineering Consultation and Contract Design Website: http://seventransistorlabs.com
Reply by ●May 17, 20172017-05-17
"John Larkin" <jjlarkin@highlandtechnology.com> wrote in message news:74gnhchkpgp02085hlnrhgce6u0u9av9qc@4ax.com...>>It's a little different than a usual op-amp topology - at low >>frequencies the LTP and mirror are driving the emitter followers through >>Q6 and Q7, and high frequencies they're bypassed by C1 and C2 to put >>current directly into the output transistor bases. I think it's trading >>off open-loop gain at high frequencies for lower overall phase shift and >>higher slew rate. > > No, the polarities are wrong for C1 and C2 to be forward gain paths. >Yeah, it wouldn't work so well without current gain from the mirrors; that's just R+C compensation, which gives about a decade past the dominant-pole compensation that is traditional, which is just dumb, really. It's compensated for unity gain stability, and yes, it is a proper op-amp block, with uncommitted inputs and crappy saturation near the rails (for inputs and outputs). :) If you need more GBW and less feedback, just crank the compensation values down. I wouldn't think it would be very useful into the 100s of MHz, because of real delays in a practical circuit. The emitter degeneration helps, but global feedback can only do so much. Tim -- Seven Transistor Labs, LLC Electrical Engineering Consultation and Contract Design Website: http://seventransistorlabs.com
Reply by ●May 17, 20172017-05-17
Gerhard Hoffmann <ghf@hoffmann-hochfrequenz.de> wrote:> Am 17.05.2017 um 04:26 schrieb Ralph Barone: > >> I've been doing picosecond stuff too. 16,666,666,666 ps, unfortunately. > > > Yeah, lots of them! >It gives me more time to think over what I'm doing :-) I used to joke that to the relaying people anything over 60 Hz was RF, and that to the telecom people, anything under 10.7 MHz was DC.
Reply by ●May 17, 20172017-05-17
On 05/17/2017 10:31 AM, Ralph Barone wrote:> Gerhard Hoffmann <ghf@hoffmann-hochfrequenz.de> wrote: >> Am 17.05.2017 um 04:26 schrieb Ralph Barone: >> >>> I've been doing picosecond stuff too. 16,666,666,666 ps, unfortunately. >> >> >> Yeah, lots of them! >> > > It gives me more time to think over what I'm doing :-) > > I used to joke that to the relaying people anything over 60 Hz was RF, and > that to the telecom people, anything under 10.7 MHz was DC. >To optical antenna people, DC goes up to 1 THz. ;) Cheers Phil Hobbs -- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics 160 North State Road #203 Briarcliff Manor NY 10510 hobbs at electrooptical dot net http://electrooptical.net
