John Larkin wrote:> I'll be timing edges to picoseconds so I don't want unpredictable > offsets.Given that a picosecond is just enough to travel 0.3mm at light speed in vacuum, maybe 0.2mm on FR4, how are you going to check if all the timings and delays are correct? Best regards, Piotr
really fast buffers
Started by ●October 3, 2018
Reply by ●October 8, 20182018-10-08
Reply by ●October 8, 20182018-10-08
On Sunday, 7 October 2018 16:24:45 UTC+1, John Larkin wrote:> On Sat, 06 Oct 2018 21:02:09 GMT, Steve Wilson <no@spam.com> wrote:> >Oscillators work on positive feedback. If the loop gain is positive and in > >phase, oscillations ensue. > > Not always. Most positive feedback loops just latch in one state.And at the other end of the scale some can produce a gain of far over 1,000 and be somewhat stable. That isn't the once round loop gain of course, it's V_out over V_in. NT
Reply by ●October 8, 20182018-10-08
On Mon, 8 Oct 2018 13:19:01 +0200, Piotr Wyderski <peter.pan@neverland.mil> wrote:>John Larkin wrote: > >> I'll be timing edges to picoseconds so I don't want unpredictable >> offsets. > >Given that a picosecond is just enough to travel 0.3mm at light speed >in vacuum, maybe 0.2mm on FR4, how are you going to check if all the >timings and delays are correct? > > Best regards, PiotrThere might be some mild hand-waving involved. The customer's machine has, in some boxes, microsecond delays that drift around, and maybe a nanosecond or two of of jitter, and all sorts electrical and optical propagation delays, and they want 1 ps jitter and 100 ps accuracy. That's meaningless, but I need to keep them happy. I can use an 11801 sampling probe and poke around, and make a list of time fudge factors, to make up for cable and pcb and part delays. That can go into a cal table for the test set. Once that's done, I need stability. CMOS gates typically drift positive a couple ps per degree C. That BUF602 will help at least to get the fudge factors measured before everything drifts too far. Buffering and fanning out the fast signals allows better PCB geometry. -- John Larkin Highland Technology, Inc lunatic fringe electronics
Reply by ●October 9, 20182018-10-09
On Sat, 06 Oct 2018 20:11:58 GMT, Steve Wilson <no@spam.com> wrote:>George Herold <gherold@teachspin.com> wrote: > >> My first (or second) cap multiplier oscillated and I added a F-bead and >> 100 ohm resistor on the base lead, through hole hair-pin. >> No problems since then. > >Beads are bad. They can crack or break. The 100 ohm resistor is all that is >needed. > >> George H. >We use lots of surface-mount beads, no problems. A bead can kill RF oscillation but not contribute low-frequency noise like a resistor would. -- John Larkin Highland Technology, Inc lunatic fringe electronics
Reply by ●October 9, 20182018-10-09
On Sat, 6 Oct 2018 14:45:19 -0400, Phil Hobbs <pcdhSpamMeSenseless@electrooptical.net> wrote:>On 10/5/18 9:36 PM, George Herold wrote: >> On Friday, October 5, 2018 at 5:26:37 PM UTC-4, John Larkin wrote: >>> On Fri, 5 Oct 2018 10:40:12 -0400, Phil Hobbs >>> <pcdhSpamMeSenseless@electrooptical.net> wrote: >>> >>>> On 10/4/18 11:09 PM, Steve Wilson wrote: >>>>> John Larkin <jjlarkin@highland_snip_technology.com> wrote: >>>>> >>>>>> On Thu, 4 Oct 2018 08:39:11 -0700 (PDT), Steve Wilson >>>>>> <9fe142ac@gmail.com> wrote: >>>>>>> I hope that anyone playing with 42GHz GBW would automatically add a base >>>>>>> resistor to kill parasitic oscillations. Even a lowly 2N3904 can go into >>>>>>> parasitics with a long base lead. >>>>> >>>>>> Or as an emitter follower with the base well bypassed to ground. The >>>>>> wire bonds inside are nice high-Q inductors. >>>>> >>>>> How do you make an emitter follower when the base is bypassed to ground? >>>>> Where do you put the signal? >>>>> >>>> >>>> The classical example is the one-transistor !RESET generator, which is a >>>> slow RC with an emitter follower. >>> >> It oscillates because of a slow edge and no hysteresis?>If you put a resistor in series with the emitter, the reflected >impedance at the base is a negative resistance.I don't understand that. At low frequencies, as the base voltage goes up, the base current goes up. The base looks ohmic.> If that becomes larger >than the extrinsic base resistance, the thing becomes unstable. The >parasitic Ls and Cs help set the oscillation frequency, but even without >them, the transistor becomes a relaxation oscillator.Can you Spice such a relaxation oscillator? At high frequencies, the base can look like a negative resistance; some oscillators work that way. -- John Larkin Highland Technology, Inc lunatic fringe electronics
Reply by ●October 9, 20182018-10-09
Am 09.10.2018 um 06:00 schrieb John Larkin:> On Sat, 6 Oct 2018 14:45:19 -0400, Phil Hobbs > <pcdhSpamMeSenseless@electrooptical.net> wrote: > >> If you put a resistor in series with the emitter, the reflected >> impedance at the base is a negative resistance. > > I don't understand that. At low frequencies, as the base voltage goes > up, the base current goes up. The base looks ohmic.That does not work at low frequencies. It takes takes transistor- internal parasitic capacitance to have an effect. The effect gets more pronounced for high emitter resistance and high emitter-GND capacitance. In the pages from Rhea that I have referenced, there is a transformation that converts such a loaded follower to the usual feedback loop by re-assigning GND to the emitter itself and redrawing the circuit. If you measure into base-gnd at RF, you see some pF in series with, say -50 Ohms. That is used in virtually all VHF/UHF VCOs from Z-Comm and others. I have once tried to pull down a Z-Comm VCXO by adding more varicaps since I had a few of them and needed lower frequencies. I didn't get very far because of the few pF that are in series to the base at RF. Disappointing, but finally understood. In the case of my baseband amplifier with the 2 IF3602 pairs, there are nF capacitances, and the amplifier goes negative impedance below 100 KHz. Ferrite chokes have quite a good Q at these frequencies, and one can tune the oscillation frequency with them :-( At higher frequencies where they work like resistors (as intended) they also generate voltage noise. You can't have your Kate, and Edith, too. regards, Gerhard
Reply by ●October 9, 20182018-10-09
On 09/10/2018 16:11, Gerhard Hoffmann wrote:> Am 09.10.2018 um 06:00 schrieb John Larkin: >> On Sat, 6 Oct 2018 14:45:19 -0400, Phil Hobbs >> <pcdhSpamMeSenseless@electrooptical.net> wrote: >> >>> If you put a resistor in series with the emitter, the reflected >>> impedance at the base is a negative resistance. >> >> I don't understand that. At low frequencies, as the base voltage goes >> up, the base current goes up. The base looks ohmic. > > That does not work at low frequencies. It takes takes transistor- > internal parasitic capacitance to have an effect. The effect gets > more pronounced for high emitter resistance and high emitter-GND > capacitance. In the pages from Rhea that I have referenced, there is > a transformation that converts such a loaded follower to the usual > feedback loop by re-assigning GND to the emitter itself and redrawing > the circuit. > > If you measure into base-gnd at RF, you see some pF in series > with, say -50 Ohms. That is used in virtually all VHF/UHF > VCOs from Z-Comm and others. I have once tried to pull down > a Z-Comm VCXO by adding more varicaps since I had a few of them > and needed lower frequencies. I didn't get very far because of > the few pF that are in series to the base at RF. Disappointing, > but finally understood. > > In the case of my baseband amplifier with the 2 IF3602 pairs, > there are nF capacitances, and the amplifier goes negative > impedance below 100 KHz. Ferrite chokes have quite a good Q > at these frequencies, and one can tune the oscillation frequency > with them :-( > > At higher frequencies where they work like resistors (as intended) > they also generate voltage noise. You can't have your Kate, and > Edith, too. > > regards, > Gerhard > >I found that a unity gain buffer built from an AD8066 opamp also seems to have a negative resistance at its input. If you decide that you want to measure e.g. the DC offset, and short the input to ground with a 40cm jumper lead, it will oscillate at about 40MHz, and with shorter wires it will oscillate up to about 116MHz. Putting a 560R resistor in series with the input connector on the board, and a 3.3pF capacitor from the non-inverting input pin of the opamp to ground fixed it, but made it otherwise worse. I recall that there is nothing much in the datasheet to warn about this sort of problem.
Reply by ●October 9, 20182018-10-09
Chris Jones wrote:> On 09/10/2018 16:11, Gerhard Hoffmann wrote: >> Am 09.10.2018 um 06:00 schrieb John Larkin: >>> On Sat, 6 Oct 2018 14:45:19 -0400, Phil Hobbs >>> <pcdhSpamMeSenseless@electrooptical.net> wrote: >>> >>>> If you put a resistor in series with the emitter, the reflected >>>> impedance at the base is a negative resistance. >>> >>> I don't understand that. At low frequencies, as the base voltage goes >>> up, the base current goes up. The base looks ohmic. >> >> That does not work at low frequencies. It takes takes transistor- >> internal parasitic capacitance to have an effect. The effect gets >> more pronounced for high emitter resistance and high emitter-GND >> capacitance. In the pages from Rhea that I have referenced, there is >> a transformation that converts such a loaded follower to the usual >> feedback loop by re-assigning GND to the emitter itself and redrawing >> the circuit. >> >> If you measure into base-gnd at RF, you see some pF in series >> with, say -50 Ohms. That is used in virtually all VHF/UHF >> VCOs from Z-Comm and others. I have once tried to pull down >> a Z-Comm VCXO by adding more varicaps since I had a few of them >> and needed lower frequencies. I didn't get very far because of >> the few pF that are in series to the base at RF. Disappointing, >> but finally understood. >> >> In the case of my baseband amplifier with the 2 IF3602 pairs, >> there are nF capacitances, and the amplifier goes negative >> impedance below 100 KHz. Ferrite chokes have quite a good Q >> at these frequencies, and one can tune the oscillation frequency >> with them :-( >> >> At higher frequencies where they work like resistors (as intended) >> they also generate voltage noise. You can't have your Kate, and >> Edith, too. >> >> regards, >> Gerhard >> >> > > I found that a unity gain buffer built from an AD8066 opamp also seems > to have a negative resistance at its input. If you decide that you want > to measure e.g. the DC offset, and short the input to ground with a 40cm > jumper lead, it will oscillate at about 40MHz, and with shorter wires it > will oscillate up to about 116MHz. Putting a 560R resistor in series > with the input connector on the board, and a 3.3pF capacitor from the > non-inverting input pin of the opamp to ground fixed it, but made it > otherwise worse. I recall that there is nothing much in the datasheet to > warn about this sort of problem. >I use discrete amplifiers with feedback to set the input impedance to 50 Ohms, because that's quieter than using an actual terminating resistor. These also have a negative input impedance in a certain frequency range, which causes instability with reactive sources. Fortunately --well no, by design-- this happens outside of my signal bandwidth, so I can muffle that with some filtering at the input. Of course, like Gerhard says, the muffler produces thermal noise, but since that's outside my signal bandwidth, it doesn't matter. Jeroen Belleman
Reply by ●October 10, 20182018-10-10
Winfield Hill wrote:> Gerhard Hoffmann wrote... >> >> BTW, I have built a variation of Winfield's 70 pV/rtHz amplifier >> from AoE3. The non-differential version, since my source is >> never centered around 0 Vdc. Works as promised. :-) >> >> Yes, it could use even more 'lytics, but that was a test balloon for >> a chopper amplifier. At 500 kHz the caps are no more that critical. >> >> https://www.flickr.com/photos/137684711@N07/45141749941/in/album-72157662535945536/ > > Nice! Please post your complete schematic. We only show a > common-emitter amplifying stage, with millions of paralleled > transistors, and our measurements, pages 506-509. > >In certain cases, one can achieve what LOOKS like negative time delay. Input pulse has a (relatively) slow rise time, and feeds a fast op-amp set to sense low threshold of the rise...output pops up beyond logic threshold BEFORE input.
Reply by ●October 10, 20182018-10-10
On Wednesday, 10 October 2018 08:38:16 UTC+1, Robert Baer wrote:> In certain cases, one can achieve what LOOKS like negative time delay. > Input pulse has a (relatively) slow rise time, and feeds a fast > op-amp set to sense low threshold of the rise...output pops up beyond > logic threshold BEFORE input.What would happen if a computer were built of negative time delay gates? The output would appear before the input occurred. Since the input would sometimes be undecided when the output occurred, the output would determine the input. So take a standard computer, swap labels on inputs & outputs, and there you have it, a faster than light future predicting computer. In fact the slower it is, the faster it is :) NT
