"bitrex" wrote in message news:EEiSA.74451$bl5.68044@fx25.iad... On 05/15/2017 10:01 AM, John Larkin wrote:> BW is 230 MHz as a unity-gain buffer. Each half of the dual will > source and sink 88 mA. And there are lots of faster THS-series parts. > >> >> But indeed you can do that with a few NPNs of the old days, >> > > Post it.>With discretes you can have:>1) wide bandwidth>2) low parts count>3) low quiescent current drawActually, you missed accuracy. The key technical limit is the triple of speed, accuracy, and power. You can't chose them independently. Typical example, to get lower distortion in a feedback system, gain might be increased. This means you have to slow it down with compensation to make it stable. You can increase he BW back up by increasing the current. Increasing current reduces collector output resistance by Va/Ic, but the gain stays the same because gm has gone up. The lower ro.Cl increases the BW of the stage. -- Kevin Aylward http://www.anasoft.co.uk - SuperSpice http://www.kevinaylward.co.uk/ee/index.html
Fast buffer idea
Started by ●May 14, 2017
Reply by ●May 15, 20172017-05-15
Reply by ●May 15, 20172017-05-15
>"bitrex" wrote in message news:xW2SA.99882$jo6.643@fx33.iad...it...> >> Is your object to drive a length of coax? If it's terminated, you can >> just ignore the cable entirely. Or poke in a Spice transmission line >> to explore mismatch consequences. >>Ya, exploring the "mismatch" consequences I'd like to do. How do I do that? >I haven't used Spice transmission line models before...Well... in SS, you just place a transmission line symbol from the lib, and double-click on it to pop up its setup box. :-) -- Kevin Aylward http://www.anasoft.co.uk - SuperSpice http://www.kevinaylward.co.uk/ee/index.html
Reply by ●May 15, 20172017-05-15
On a sunny day (Mon, 15 May 2017 13:15:52 -0400) it happened bitrex <bitrex@de.lete.earthlink.net> wrote in <cnlSA.15485$Qg.3753@fx14.iad>:>> >> Actually when I did read that OP, I scribbled this on my 'desk': >> http://panteltje.com/pub/maybe_this_amp_IMG_6183.JPG >> that is a dual complementary emitter follower. >> >> The reason I did not jump into the discussion with that, is that _in this form_ >> I am not sure about zero temp drift, it may need some resistors. >> But it should have zero output offset versus input, low output impedance, >> high input impedance, temp tracking, and I do not have spice running >> to give an idea, so not a tested thing. >> So if it melts you mega $ trannies so be it. >> > >It's a regular "diamond buffer" topology. Without emitter degeneration >resistors in the output emitters it's asking for thermal runaway, and >the output waveform won't be particularly linear. > >Try adding 10 ohm degeneration resistors, and check the quiescent >current on that bad boy with 10/-10 rails. Yikes! > >Taking feedback from the tops of the degeneration resistors to the input >collectors makes it more linear by reducing Early effect variations. >It's an even better idea with a MOSFET input stage, as ideally if the >Vds is near constant you wont have the gm compromised due to >channel-length modulation.OK, challenge? I have the transistors: http://panteltje.com/pub/have_those_transistors_in_quantity_IMG_6184.JPG So build it, used a 10K pot (single 10 V supply) to put the input at half way (5 V) 47k in series, poly coupling cap, 1000Hz audio from eeePC: http://panteltje.com/pub/build_it_and_testing_at_10V_IMG_6185.JPG NO resistors here without flash so you can read the mA meter: http://panteltje.com/pub/buid_it_and_testing_no_flash_IMG_6186.JPG Sorry for the crowded place, lots of things multitasking here. I did not load it, but it sure wanted to beep on my FM radio as I did not decouple the supply lines... Old Trio Kenwood scope does not go that high. Never underestimate El Pante Current does not increase from 5 to 10 V, did not try higher,, The audio generator is Linux sgen. Now for the spice by those who have it!
Reply by ●May 15, 20172017-05-15
On 05/15/2017 02:29 PM, John Larkin wrote:> On Mon, 15 May 2017 13:38:11 -0400, bitrex > <bitrex@de.lete.earthlink.net> wrote: > >> On 05/15/2017 01:28 PM, John Larkin wrote: >> >>>>> Snoop the input current of that buffer. At 20 MHz, the overall current >>>>> gain is less than 1. At 100 MHz, the current gain is about 0.1. You'd >>>>> be better off with a piece of wire. >>>> >>>> I'm not sure what "input current" you're looking at. I'm seeing an >>>> average current of around 600uA into the MOSFET gates at 100MHz. >>> >>> I put a 1 mohm resistor in series with the signal generator and probed >>> the current, with a sine wave signal input. Try it and see what you >>> get. >> >> How are the gate capacitances of the input MOSFETs supposed to be >> charged up and down with that fuckin' thing sitting in the way?! >> >> Fuck the circuit to hell and then complain it's bad?! Shit! :O > > mohm means milliohm, in SI units and in Spice.Ok, I see what you mean; forgive my, uh, outburst. And there is a bug with sine wave drive that isn't apparent with a square drive. Problem seems to be that with these slow transistors, I optimized R9 and R10 to make a nice square wave. But those resistors aren't optimal for a sine-wave of say 1V P2P; the boost transistors never turn on hard enough, and if the output transistors can't get the current they need they're going to try to pull it from somewhere, namely from the MOSFETs. Or since Vds is fixed it can probably try to pull it right from the input source. That's probably fixable.
Reply by ●May 15, 20172017-05-15
On a sunny day (Mon, 15 May 2017 17:18:17 GMT) it happened Steve Wilson <no@spam.com> wrote in <XnsA776875EB8A7idtokenpost@69.16.179.23>:>Jan Panteltje <pNa0nStpealmtje@yahoo.com> wrote: > >> On a sunny day (Mon, 15 May 2017 16:24:56 GMT) it happened Steve >> Wilson <no@spam.com> wrote in >> <XnsA7767E5052407idtokenpost@69.16.179.22>: >> >>>bitrex <bitrex@de.lete.earthlink.net> wrote: >>> >>>> On 05/15/2017 10:09 AM, bitrex wrote: >>>>> On 05/15/2017 10:01 AM, John Larkin wrote: >>>>> >>>>>> BW is 230 MHz as a unity-gain buffer. Each half of the dual will >>>>>> source and sink 88 mA. And there are lots of faster THS-series >>>>>> parts. >>>>>> >>>>>>> >>>>>>> But indeed you can do that with a few NPNs of the old days, >>>>>>> >>>>>> >>>>>> Post it. >>>>> >>>>> With discretes you can have: >>>>> >>>>> 1) wide bandwidth >>>>> >>>>> 2) low parts count >>>>> >>>>> 3) low quiescent current draw >>>>> >>>>> Pick two. >>>> >>>> The two transistor circuit Jan posted does have a pretty wide >>>> bandwidth and a low parts count. It has a pretty nice small signal >>>> output impedance, but its large-signal output impedance is bad. It >>>> also draws a relatively large amount of quiescent current. >>> >>>I somehow missed the two transistor circuit that Jan posted. Can you >>>please repost the link? >>> >>>Thanks >> >> From: >> http://panteltje.com/panteltje/quadcopter/hud.html >> the circuit diagram: >> http://panteltje.com/panteltje/quadcopter/hud-0.1_circuit_diagram_IMG >> _5984.JPG bottom left: >> that has a gain of 2, so can drive a 75 Ohm cable with reasonable Zi. >> The DC path requires an other 75 Ohm at the end of the cable. > >OK, thanks. > >That is a Sziklai Pair with a voltage divider in the feedback path: > > https://en.wikipedia.org/wiki/Sziklai_pair > >It would normally have the same thermal drift as a simple base-emitter >junction, but the divide by two may increase it. > >The output impedance may be increased due to the low loop gain. The >bandwidth is determined by the transistors you use. > >But it is difficult to tell what the transistors are due to the poor >clarity of the handwritten numbers. I can't read them even when I expand >the drawing. > >I really wish you could convert some of your critical circuits to LTspice >so we can see the operation a bit better. That would have a far greater >impact for the value of your work.Well OK, it is BC547-B / BC557-B (B is the beta range) Sometimes I wonder what sort of monitor or viewer people use... It may be that in some cases LTspice can not do my circuits :-)
Reply by ●May 15, 20172017-05-15
On a sunny day (Mon, 15 May 2017 13:30:20 -0400) it happened bitrex <bitrex@de.lete.earthlink.net> wrote in <MAlSA.118066$ef.9536@fx34.iad>:>On 05/15/2017 01:18 PM, Steve Wilson wrote: >> Jan Panteltje <pNa0nStpealmtje@yahoo.com> wrote: >> >>> On a sunny day (Mon, 15 May 2017 16:24:56 GMT) it happened Steve >>> Wilson <no@spam.com> wrote in >>> <XnsA7767E5052407idtokenpost@69.16.179.22>: >>> >>>> bitrex <bitrex@de.lete.earthlink.net> wrote: >>>> >>>>> On 05/15/2017 10:09 AM, bitrex wrote: >>>>>> On 05/15/2017 10:01 AM, John Larkin wrote: >>>>>> >>>>>>> BW is 230 MHz as a unity-gain buffer. Each half of the dual will >>>>>>> source and sink 88 mA. And there are lots of faster THS-series >>>>>>> parts. >>>>>>> >>>>>>>> >>>>>>>> But indeed you can do that with a few NPNs of the old days, >>>>>>>> >>>>>>> >>>>>>> Post it. >>>>>> >>>>>> With discretes you can have: >>>>>> >>>>>> 1) wide bandwidth >>>>>> >>>>>> 2) low parts count >>>>>> >>>>>> 3) low quiescent current draw >>>>>> >>>>>> Pick two. >>>>> >>>>> The two transistor circuit Jan posted does have a pretty wide >>>>> bandwidth and a low parts count. It has a pretty nice small signal >>>>> output impedance, but its large-signal output impedance is bad. It >>>>> also draws a relatively large amount of quiescent current. >>>> >>>> I somehow missed the two transistor circuit that Jan posted. Can you >>>> please repost the link? >>>> >>>> Thanks >>> >>> From: >>> http://panteltje.com/panteltje/quadcopter/hud.html >>> the circuit diagram: >>> http://panteltje.com/panteltje/quadcopter/hud-0.1_circuit_diagram_IMG >>> _5984.JPG bottom left: >>> that has a gain of 2, so can drive a 75 Ohm cable with reasonable Zi. >>> The DC path requires an other 75 Ohm at the end of the cable. >> >> OK, thanks. >> >> That is a Sziklai Pair with a voltage divider in the feedback path: >> >> https://en.wikipedia.org/wiki/Sziklai_pair >> >> It would normally have the same thermal drift as a simple base-emitter >> junction, but the divide by two may increase it. >> >> The output impedance may be increased due to the low loop gain. The >> bandwidth is determined by the transistors you use. > >Its output impedance for large negative-going signals is _dogshit_, >regardless of what the small-signal analysis feedback-loop-gain etc. >calculations indicate. It can source current okay. It definitely can't >sink it. > >It doesn't really matter for a composite video signal, though, because >it's not really that wide a bandwidth signal. The only potentially >difficult fast full-level transition I think of would be if you have to >go from the "back porch" immediately to "full luminance."Well that is same as from black to white, works fine tough. that is just the PNP conducting pulling the output. This circuit has been used for video by me many many times. Never ever failed.
Reply by ●May 15, 20172017-05-15
On 05/15/2017 03:05 PM, bitrex wrote:> On 05/15/2017 02:29 PM, John Larkin wrote: >> On Mon, 15 May 2017 13:38:11 -0400, bitrex >> <bitrex@de.lete.earthlink.net> wrote: >> >>> On 05/15/2017 01:28 PM, John Larkin wrote: >>> >>>>>> Snoop the input current of that buffer. At 20 MHz, the overall >>>>>> current >>>>>> gain is less than 1. At 100 MHz, the current gain is about 0.1. You'd >>>>>> be better off with a piece of wire. >>>>> >>>>> I'm not sure what "input current" you're looking at. I'm seeing an >>>>> average current of around 600uA into the MOSFET gates at 100MHz. >>>> >>>> I put a 1 mohm resistor in series with the signal generator and probed >>>> the current, with a sine wave signal input. Try it and see what you >>>> get. >>> >>> How are the gate capacitances of the input MOSFETs supposed to be >>> charged up and down with that fuckin' thing sitting in the way?! >>> >>> Fuck the circuit to hell and then complain it's bad?! Shit! :O >> >> mohm means milliohm, in SI units and in Spice. > > Ok, I see what you mean; forgive my, uh, outburst. And there is a bug > with sine wave drive that isn't apparent with a square drive. > > Problem seems to be that with these slow transistors, I optimized R9 and > R10 to make a nice square wave. But those resistors aren't optimal for a > sine-wave of say 1V P2P; the boost transistors never turn on hard > enough, and if the output transistors can't get the current they need > they're going to try to pull it from somewhere, namely from the MOSFETs. > Or since Vds is fixed it can probably try to pull it right from the > input source. > > That's probably fixable.But it doesn't necessarily make sense to view a square wave as a Fourier series and judge the performance of a square-wave amplifier based on how well it amplifies a sine wave at the fundamental frequency if the system's parameters are not essentially LTI. If the bias of the amplifier is changing an enormous amount in response to certain signal amplitudes then it isn't really LTI. If one were clever or use even more parts I think it would be possible to slide the bias in a more uniform fashion to eliminate the non-linearity, but as it stands now the shift looks pretty bang-bang.
Reply by ●May 15, 20172017-05-15
On a sunny day (Mon, 15 May 2017 18:07:57 GMT) it happened Steve Wilson <no@spam.com> wrote in <XnsA7768FC3A2F00idtokenpost@69.16.179.23>:>bitrex <bitrex@de.lete.earthlink.net> wrote: > >> On 05/15/2017 01:18 PM, Steve Wilson wrote: >>> That is a Sziklai Pair with a voltage divider in the feedback path: > >>> https://en.wikipedia.org/wiki/Sziklai_pair > >>> The output impedance may be increased due to the low loop gain. The >>> bandwidth is determined by the transistors you use. > >> Its output impedance for large negative-going signals is _dogshit_, >> regardless of what the small-signal analysis feedback-loop-gain etc. >> calculations indicate. It can source current okay. It definitely can't >> sink it. > >The negative swing depends on the pulldown resistor, same as any single- >ended emitter follower. That should be obvious. > >These issues would be immediately apparent in a proper LTspice analysis. > >Unfortunately, with some users still devoted to ASCII-art or hand-drawn >schematics, it is difficult or impossible to do a reasonable preliminary >analysis.Man you are way way out in the boonies, I can tell from just a few strokes if it wrks or not, Van Gogh could too. [My]Tronix is art, you are no art knower, just a wall painter using spice as paintbrush. ;-) THIS is why US can no longer send humming beans to space, You live in a virtual world. Its all over. The Matrix has you, no way back. ....
Reply by ●May 15, 20172017-05-15
On a sunny day (Mon, 15 May 2017 19:50:40 +0100) it happened "Kevin Aylward" <kevinRemovAT@kevinaylward.co.uk> wrote in <Ca6dnRH1w-4cZYTEnZ2dnUU7-dfNnZ2d@giganews.com>:>"bitrex" wrote in message news:EEiSA.74451$bl5.68044@fx25.iad... > >On 05/15/2017 10:01 AM, John Larkin wrote: > >> BW is 230 MHz as a unity-gain buffer. Each half of the dual will >> source and sink 88 mA. And there are lots of faster THS-series parts. >> >>> >>> But indeed you can do that with a few NPNs of the old days, >>> >> >> Post it. > >>With discretes you can have: > >>1) wide bandwidth > >>2) low parts count > >>3) low quiescent current draw > >Actually, you missed accuracy. > >The key technical limit is the triple of speed, accuracy, and power. You >can't chose them independently. > >Typical example, to get lower distortion in a feedback system, gain might be >increased. This means you have to slow it down with compensation to make it >stable. You can increase he BW back up by increasing the current. Increasing >current reduces collector output resistance by Va/Ic, but the gain stays the >same because gm has gone up. The lower ro.Cl increases the BW of the stage.Hi Kevin, just wanted to ask, I was following some public lectures by several professors last week, and the last one was about the Higgs boson. In that space is not empty and filled with the Higgs field. Today I have been studying that stuff.. and found some interesting mass related problems. Did you have thoughts about that? Ever met the man?
Reply by ●May 15, 20172017-05-15
On 05/15/2017 03:13 PM, Jan Panteltje wrote:> Well that is same as from black to white, works fine tough. > that is just the PNP conducting pulling the output. > This circuit has been used for video by me many many times. > Never ever failed.Ya I'm sure it works fine for composite video. If the risetime off the porches is a little sluggish I don't think it matters much; there isn't anything timing-critical to that transition as its just setting up the luminance "data." Might make the edges a little fuzzy? Like "vintage warmth"
