John Larkin <jjlarkin@highnotlandthistechnologypart.com> wrote:
: But if I wanted a negative current source, I'd just cascode into a
: PHEMT. I need positive current, and the people who make gaasfets
: stubbornly refuse to make p-channel ones.
The same with SiGe PNP's - I would have wanted to make a push-pull
for LHe but there are no discretes available. An NPN-NPN push-pull
is a possibility, but awkward. Maybe the discretes will become available
soon, at least there are complementary IC processes producing such
devices as the OPA835/6 (which actually work in LHe but only marginally).
Regards,
Mikko
picosecond test points
Started by ●February 11, 2012
Reply by ●February 13, 20122012-02-13
Reply by ●February 13, 20122012-02-13
On Feb 12, 9:07=A0pm, John Larkin <jjlar...@highNOTlandTHIStechnologyPART.com> wrote:> On Sun, 12 Feb 2012 16:07:51 -0800 (PST), dagmargoodb...@yahoo.com > wrote: > > > > > > > > > > >On Feb 12, 2:15=A0pm, John Larkin > ><jjlar...@highNOTlandTHIStechnologyPART.com> wrote: > >> On Sun, 12 Feb 2012 13:39:37 -0500, Phil Hobbs > > >> <pcdhSpamMeSensel...@electrooptical.net> wrote: > >> >Okkim Atnarivik wrote: > > >> >> Phil Hobbs <pcdhSpamMeSensel...@electrooptical.net> wrote: > >> >> : Looks good. =A0Another approach would be something like a BFP650 =follower,> > >> >> =A0 Phil, I was about to suggest that if you've ended up using the > >> >> BFP650 by following my example, you might be better off with the BF=P640.> >> >> Or more modern ones with a higher hFE (I'm fond of the NESG3031 and > >> >> NESG4030). > > >> >> =A0 Namely, my choice of BFP650 (and not the then more widely avail=able> >> >> BFP640) was driven by the search of the lowest possible R_BB, becau=se> >> >> I was deemed to encounter lower source resistances than the 50ohms =for> >> >> which SiGe devices are generally optimized. I reasoned that the bas=e> >> >> geometry of the BFP650, as a medium power device, is more likely yi=eld> >> >> a low R_BB. In your applications BFP640 may work better with the "m=ore> >> >> ordinary" impedance levels and ambient temperatures. > > >> >> =A0 The "I was about.." part means that I'm somewhat puzzled now th=at I> >> >> looked up the recent Infineon data sheets. The BFP650 is now listed > >> >> as a SiGe:C part (the datasheet 2010-10-22) and its Gummel-Poon > >> >> model reads RB=3D6.376 . The preliminary datasheet I have, dated > >> >> Aug-16-2004, lists it as an ordinary SiGe part with RB=3D1.036 . So=, I'm> >> >> wondering how much the device has changed from the versions I've be=en> >> >> using. I seem to recall that also in some other more recent data sh=eet> >> >> versions it was listed as a SiGe, not SiGe:C . > > >> >> =A0 Given the fact that the most recent BFP640 datasheet (dated 200=7-05-29)> >> >> reads RB=3D3.129, the BFP650 does not seem to have any noise advant=age> >> >> any more, either. > > >> >> =A0 In fact, there was a recent paper in the RSI where > >> >> the Jena group claim to have achieved 15 pV/rtHz with BFP640 in LHe=,> >> >> as compared with my 75 pV/rtHz with BFP650. This is a puzzling resu=lt,> >> >> actually, because it is better than one would expect in the picture=where> >> >> u_N originates from the collector shot noise acting on the r_E - pr=ovided> >> >> that the thermal voltage saturates at ~5..7 meV even when the ambie=nt> >> >> temperature keeps going down. According to my data it does saturate=in> >> >> BFP640 just like it does in all other SiGe's I've tried, when measu=red by> >> >> the transconductance. I'm wondering whether my transistors have alw=ays> >> >> been oscillating so high that I cannot see it (ie. >27 GHz) and the > >> >> V_T saturation is an artefact due to it... > > >> >> =A0 Regards, > >> >> =A0 =A0 =A0 =A0 =A0 =A0Mikko > > >> >I started by taking your advice, which was pretty valuable--thanks > >> >again. =A0I noticed the change in specs as well, when I ordered anoth=er> >> >batch a couple of months ago. =A0They changed the type number to BFP6=50H> >> >instead of BFP650E when they changed the process. =A0(The dogs.) > > >> >I'm using it as a cascode stage for a SKY65050 pHEMT, and it works > >> >great. =A0One of the best things about it is that it has effectively > >> >infinite Early voltage, so you can get a lot of voltage gain out of a > >> >single stage. =A0Even for situations where you don't care so much abo=ut> >> >ultralow noise, the combination of a gigantic f_T with a very high V_=A> >> >is unique in my collection. =A0It does want to oscillate at 14 GHz if=you> >> >look at it crosswise, but a nice 5-ohm bead in series with the base > >> >cleans that right up. > > >> >I'm going to be using it in a new front end design, as a > >> >bootstrapped-bootstrap wrapped round a couple of parallelled BF862s. ==A0DC> >> >coupling the bootstrap and using a current source load gets rid of th=e> >> >thermal tails on the BF862s by keeping their dissipation constant. > >> >(Running them just slightly above I_DSS gives them a zero tempco anyw=ay,> >> >but this one needs really good dc stability.) > > >> Funny you should say "cascode." I was just about to post the > >> following: > > >> If I want a fast, low-capacitance, accurate, positive current source, > >> I could build a fairly slow active current source with a voltage > >> reference, a resistor, an opamp, and a p-fet. But it would have a lot > >> of capacitance and would be fairly slow. So I could cascode that with > >> a microwave-type PNP transistor, with a bit of base resistance or a > >> ferrite bead to keep it stable. But then I'd have the base current > >> error, and the good fast PNPs, what few there are, have mediocre > >> betas. > > >> A PNP darlington is too slow. > > >> So, how to correct for the base current error? > > >> Two ideas so far: make a Darlington, but add a lowpass filter from the > >> higher-current transistor base, into the emitter of the second > >> transistor. The main transistor needed a base resistor anyhow. > > >> Or sense the base current of the PNP, with opamps and such, and > >> increase the current of the precision/slow source to make up for it. > > >It's sure a lot easier if you can flip the circuit into a sink. =A0Then > >you've got more choices than just a BFT92. > > > Sadly, I need a positive current source. My customer wants really > precise laser bias current, which I frankly think is unjustified,If you sense and compensate base current perfectly, the current source will be perfect. Maybe a flying cap sampling the base resistor would do better than Fred's and the EDN ckts? For speed beyond the BFT92, easiest are the hybrids you've already done--outboard additions (L and R) that improve GHz response, the bjt handles mid-range, and the op-amp handles d.c. accuracy. Cascoding two BFT92s helps, eliminating Cfb feedback of the exposed device. To do better than that, bootstrap the exposed BFT92. I can think of a few ways--pretty hairy.> but > unfortunately he's the one with the money.No, it's actually good he's the one with the money. If you had it, you wouldn't need him! -- Cheers, James Arthur
Reply by ●February 13, 20122012-02-13
On Mon, 13 Feb 2012 10:30:41 +0200 (EET), Okkim Atnarivik <Okkim.Atnarivik@twentyfour.fi.invalid> wrote:>John Larkin <jjlarkin@highnotlandthistechnologypart.com> wrote: >: But if I wanted a negative current source, I'd just cascode into a >: PHEMT. I need positive current, and the people who make gaasfets >: stubbornly refuse to make p-channel ones. > > The same with SiGe PNP's - I would have wanted to make a push-pull >for LHe but there are no discretes available. An NPN-NPN push-pull >is a possibility, but awkward. Maybe the discretes will become available >soon, at least there are complementary IC processes producing such >devices as the OPA835/6 (which actually work in LHe but only marginally). > > Regards, > MikkoUnfortunately, most of the good fast parts are designed for RF use, which means we only get n-types, and very badly characterized ones at that. S-params and load pulls don't tell you much about DC or switching behavior. You're lucky if they tell you Idss, very lucky if they give you DC curves, and heaven is getting a Spice model. I asked Mini-Circuits if one of their MMICs inverts the signal or not. They didn't know. Grrrrr. -- John Larkin, President Highland Technology Inc www.highlandtechnology.com jlarkin at highlandtechnology dot com Precision electronic instrumentation Picosecond-resolution Digital Delay and Pulse generators Custom timing and laser controllers Photonics and fiberoptic TTL data links VME analog, thermocouple, LVDT, synchro, tachometer Multichannel arbitrary waveform generators
Reply by ●February 13, 20122012-02-13
On Feb 13, 9:28=A0am, dagmargoodb...@yahoo.com wrote:> On Feb 12, 9:07=A0pm, John Larkin wrote: > > On Sun, 12 Feb 2012 16:07:51 -0800 (PST), dagmargoodb...@yahoo.com > > wrote: > > > >On Feb 12, 2:15=A0pm, John Larkin wrote:> > >> A PNP darlington is too slow. > > > >> So, how to correct for the base current error? > > > >> Two ideas so far: make a Darlington, but add a lowpass filter from t=he> > >> higher-current transistor base, into the emitter of the second > > >> transistor. The main transistor needed a base resistor anyhow. > > > >> Or sense the base current of the PNP, with opamps and such, and > > >> increase the current of the precision/slow source to make up for it. > > > >It's sure a lot easier if you can flip the circuit into a sink. =A0The=n> > >you've got more choices than just a BFT92. > > > Sadly, I need a positive current source. My customer wants really > > precise laser bias current, which I frankly think is unjustified, > > If you sense and compensate base current perfectly, the current source > will be perfect. =A0Maybe a flying cap sampling the base resistor would > do better than Fred's and the EDN ckts? > > For speed beyond the BFT92, easiest are the hybrids you've already > done--outboard additions (L and R) that improve GHz response, the bjt > handles mid-range, and the op-amp handles d.c. accuracy. > > Cascoding two BFT92s helps, eliminating Cfb feedback of the exposed > device. > > To do better than that, bootstrap the exposed BFT92. =A0I can think of a > few ways--pretty hairy. > > > but > > unfortunately he's the one with the money. > > No, it's actually good he's the one with the money. =A0If you had it, > you wouldn't need him!What if you separate the a.c. and d.c. paths at the pass element too? (Maybe this is what you meant by "lowpass filter at the base"?) | |<' low z a.c. >----+---| BFT92 | |\ |-' | d.c. bias >-<-| | |---Z1--+ | .-. | | f.b. | | '-' | Z1 isolates FET capacitance @ rf. James
Reply by ●February 13, 20122012-02-13
On Mon, 13 Feb 2012 08:41:34 -0800 (PST), dagmargoodboat@yahoo.com wrote:>On Feb 13, 9:28�am, dagmargoodb...@yahoo.com wrote: >> On Feb 12, 9:07�pm, John Larkin wrote: >> > On Sun, 12 Feb 2012 16:07:51 -0800 (PST), dagmargoodb...@yahoo.com >> > wrote: >> >> > >On Feb 12, 2:15�pm, John Larkin wrote: > >> > >> A PNP darlington is too slow. >> >> > >> So, how to correct for the base current error? >> >> > >> Two ideas so far: make a Darlington, but add a lowpass filter from the >> > >> higher-current transistor base, into the emitter of the second >> > >> transistor. The main transistor needed a base resistor anyhow. >> >> > >> Or sense the base current of the PNP, with opamps and such, and >> > >> increase the current of the precision/slow source to make up for it. >> >> > >It's sure a lot easier if you can flip the circuit into a sink. �Then >> > >you've got more choices than just a BFT92. >> >> > Sadly, I need a positive current source. My customer wants really >> > precise laser bias current, which I frankly think is unjustified, >> >> If you sense and compensate base current perfectly, the current source >> will be perfect. �Maybe a flying cap sampling the base resistor would >> do better than Fred's and the EDN ckts? >> >> For speed beyond the BFT92, easiest are the hybrids you've already >> done--outboard additions (L and R) that improve GHz response, the bjt >> handles mid-range, and the op-amp handles d.c. accuracy. >> >> Cascoding two BFT92s helps, eliminating Cfb feedback of the exposed >> device. >> >> To do better than that, bootstrap the exposed BFT92. �I can think of a >> few ways--pretty hairy. >> >> > but >> > unfortunately he's the one with the money. >> >> No, it's actually good he's the one with the money. �If you had it, >> you wouldn't need him! > >What if you separate the a.c. and d.c. paths at the pass element too? >(Maybe this is what you meant by "lowpass filter at the base"?) > > | > |<' >low z a.c. >----+---| BFT92 > | |\ > |-' | >d.c. bias >-<-| | > |---Z1--+ > | > .-. > | | f.b. > | | > '-' > | > >Z1 isolates FET capacitance @ rf. > >JamesOne could just use a p-channel mosfet as the cascode. That could get the source capacitance down to 20 pF, maybe better if one found the right fet (I'm talking 50-100 mA maybe.) Then put a good wideband inductor in its drain. "Good wideband inductor" is non-trivial. The Piconics conical things are good, but expensive and hard to handle. Making your own takes a bunch of parts... several different FBs and Ls in series, with damping resistors. -- John Larkin, President Highland Technology Inc www.highlandtechnology.com jlarkin at highlandtechnology dot com Precision electronic instrumentation Picosecond-resolution Digital Delay and Pulse generators Custom timing and laser controllers Photonics and fiberoptic TTL data links VME analog, thermocouple, LVDT, synchro, tachometer Multichannel arbitrary waveform generators
Reply by ●February 13, 20122012-02-13
On 02/13/2012 03:12 AM, Okkim Atnarivik wrote:> Phil Hobbs<pcdhSpamMeSenseless@electrooptical.net> wrote: > : I started by taking your advice, which was pretty valuable--thanks > > Don't mention it. I have already gained more from your, Larkin's, > Win Hill's and many other's posts here in SED. > > : again. I noticed the change in specs as well, when I ordered another > : batch a couple of months ago. They changed the type number to BFP650H > : instead of BFP650E when they changed the process. (The dogs.) > > They did? The datasheet still list the part as BFP650 without suffixes > even though it is the SiGe:C version.The distributors' stock of the E version all ran out a few months ago, and all they have now is the H version. I haven't managed to get the straight story as to just what they did.> > : I'm using it as a cascode stage for a SKY65050 pHEMT, and it works > > Interesting hint, I should give it a try. > > : great. One of the best things about it is that it has effectively > : infinite Early voltage, so you can get a lot of voltage gain out of a > : single stage. Even for situations where you don't care so much about > : ultralow noise, the combination of a gigantic f_T with a very high V_A > : is unique in my collection. It does want to oscillate at 14 GHz if you > : look at it crosswise, but a nice 5-ohm bead in series with the base > : cleans that right up. > > The high Early seems to be characteristic to the SiGe's, it puzzled me > already in the thread http://tinyurl.com/8a44pzn . If you check the > page 4 of http://tinyurl.com/6qmnd4u you see it there too (and additionally > you get the large hFE = 400 ).I've ordered some of those too, thanks. It's a bit worrying that there are seem to be only about 100 of them in the world, according to FindChips. Cheers Phil> > Regards, > Mikko-- 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 845-480-2058 hobbs at electrooptical dot net http://electrooptical.net
Reply by ●February 13, 20122012-02-13
On Feb 12, 4:49=A0pm, Fred Bartoli <" "> wrote:> John Larkin a =E9crit : > > > > > > > On Sun, 12 Feb 2012 13:39:37 -0500, Phil Hobbs > > <pcdhSpamMeSensel...@electrooptical.net> wrote: > > >> Okkim Atnarivik wrote: > >>> Phil Hobbs <pcdhSpamMeSensel...@electrooptical.net> wrote: > >>> : Looks good. =A0Another approach would be something like a BFP650 fo=llower,> > >>> =A0 Phil, I was about to suggest that if you've ended up using the > >>> BFP650 by following my example, you might be better off with the BFP6=40.> >>> Or more modern ones with a higher hFE (I'm fond of the NESG3031 and > >>> NESG4030). > > >>> =A0 Namely, my choice of BFP650 (and not the then more widely availab=le> >>> BFP640) was driven by the search of the lowest possible R_BB, because > >>> I was deemed to encounter lower source resistances than the 50ohms fo=r> >>> which SiGe devices are generally optimized. I reasoned that the base > >>> geometry of the BFP650, as a medium power device, is more likely yiel=d> >>> a low R_BB. In your applications BFP640 may work better with the "mor=e> >>> ordinary" impedance levels and ambient temperatures. > > >>> =A0 The "I was about.." part means that I'm somewhat puzzled now that=I> >>> looked up the recent Infineon data sheets. The BFP650 is now listed > >>> as a SiGe:C part (the datasheet 2010-10-22) and its Gummel-Poon > >>> model reads RB=3D6.376 . The preliminary datasheet I have, dated > >>> Aug-16-2004, lists it as an ordinary SiGe part with RB=3D1.036 . So, =I'm> >>> wondering how much the device has changed from the versions I've been > >>> using. I seem to recall that also in some other more recent data shee=t> >>> versions it was listed as a SiGe, not SiGe:C . > > >>> =A0 Given the fact that the most recent BFP640 datasheet (dated 2007-=05-29)> >>> reads RB=3D3.129, the BFP650 does not seem to have any noise advantag=e> >>> any more, either. > > >>> =A0 In fact, there was a recent paper in the RSI where > >>> the Jena group claim to have achieved 15 pV/rtHz with BFP640 in LHe, > >>> as compared with my 75 pV/rtHz with BFP650. This is a puzzling result=,> >>> actually, because it is better than one would expect in the picture w=here> >>> u_N originates from the collector shot noise acting on the r_E - prov=ided> >>> that the thermal voltage saturates at ~5..7 meV even when the ambient > >>> temperature keeps going down. According to my data it does saturate i=n> >>> BFP640 just like it does in all other SiGe's I've tried, when measure=d by> >>> the transconductance. I'm wondering whether my transistors have alway=s> >>> been oscillating so high that I cannot see it (ie. >27 GHz) and the > >>> V_T saturation is an artefact due to it... > > >>> =A0 Regards, > >>> =A0 =A0 =A0 =A0 =A0 =A0Mikko > >> I started by taking your advice, which was pretty valuable--thanks > >> again. =A0I noticed the change in specs as well, when I ordered anothe=r> >> batch a couple of months ago. =A0They changed the type number to BFP65=0H> >> instead of BFP650E when they changed the process. =A0(The dogs.) > > >> I'm using it as a cascode stage for a SKY65050 pHEMT, and it works > >> great. =A0One of the best things about it is that it has effectively > >> infinite Early voltage, so you can get a lot of voltage gain out of a > >> single stage. =A0Even for situations where you don't care so much abou=t> >> ultralow noise, the combination of a gigantic f_T with a very high V_A > >> is unique in my collection. =A0It does want to oscillate at 14 GHz if =you> >> look at it crosswise, but a nice 5-ohm bead in series with the base > >> cleans that right up. > > >> I'm going to be using it in a new front end design, as a > >> bootstrapped-bootstrap wrapped round a couple of parallelled BF862s. ==A0DC> >> coupling the bootstrap and using a current source load gets rid of the > >> thermal tails on the BF862s by keeping their dissipation constant. > >> (Running them just slightly above I_DSS gives them a zero tempco anywa=y,> >> but this one needs really good dc stability.) > > >> Cheers > > >> Phil Hobbs > > > Funny you should say "cascode." I was just about to post the > > following: > > > If I want a fast, low-capacitance, accurate, positive current source, > > I could build a fairly slow active current source with a voltage > > reference, a resistor, an opamp, and a p-fet. But it would have a lot > > of capacitance and would be fairly slow. So I could cascode that with > > a microwave-type PNP transistor, with a bit of base resistance or a > > ferrite bead to keep it stable. But then I'd have the base current > > error, and the good fast PNPs, what few there are, have mediocre > > betas. > > > A PNP darlington is too slow. > > > So, how to correct for the base current error? > > Try this : > > =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 ==A0 =A0 =A0 =A0 =A0 =A0 =A0Ic =3D 3mA> > =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A01K =A0 =A0==A0 =A0 =A0 =A0 =A0 |> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0___ =A0 ==A0 =A0 =A0 =A0 =A0|/> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 5V>-----|___|----+------| > =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 ==A0 =A0| =A0 =A0 =A0|>> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 ==A0 .-. =A0 =A0 =A0 |> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 ==A0 | | =A0 =A0 =A0 |> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A01K=| | =A0 =A0 =A0 |> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 ==A0 '-' =A0 =A0 =A0 |> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 ==A0 =A0| =A0 =A0 =A0 =A0|> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 |\ =A0 =A0 =A0 =A0 ==A0 | =A0 =A0 ||-+> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A03V>--|+\ =A0 =A0 =A0 =A0 =A0| ==A0 =A0 ||<-> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 | =A0>---------|-----=||-+> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 .-|-/ =A0 =A0 =A0 =A0 =A0=| =A0 =A0 =A0 =A0|> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 | |/ =A0 =A0 =A0 =A0 =A0 =| =A0 ___ =A0|> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 '--------------+--|___|-+ > =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 ==A0 =A0 =A0 =A0 =A0 =A0 |> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 ==A0 =A0 =A0 =A01K =A0.-.> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 ==A0 =A0 =A0 =A0 =A0 =A0| |> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 ==A0 =A0 =A0 =A0 =A0 =A0| |1K> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 ==A0 =A0 =A0 =A0 =A0 =A0'-'> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 ==A0 =A0 =A0 =A0 =A0 =A0 |> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 ==A0 =A0 =A0 =A0 =A0 =A0=3D=3D=3D> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 ==A0 =A0 =A0 =A0 =A0 =A0GND> > Delta Ic < 3nA for 45<Beta<65 for ex. > > -- > Thanks, > Fred.- Hide quoted text - > > - Show quoted text -Would someone be kind enough to help me see how this compensates for the base current? I've done the Vref-opamp-fet-resistor current source. Color me slightly confused. (I'd be happy to pay the, one beer licensing fee,... to understand it.) George H.
Reply by ●February 13, 20122012-02-13
On 02/13/2012 12:10 PM, George Herold wrote:> On Feb 12, 4:49 pm, Fred Bartoli<""> wrote: >> John Larkin a �crit : >> >> >> >> >> >>> On Sun, 12 Feb 2012 13:39:37 -0500, Phil Hobbs >>> <pcdhSpamMeSensel...@electrooptical.net> wrote: >> >>>> Okkim Atnarivik wrote: >>>>> Phil Hobbs<pcdhSpamMeSensel...@electrooptical.net> wrote: >>>>> : Looks good. Another approach would be something like a BFP650 follower, >> >>>>> Phil, I was about to suggest that if you've ended up using the >>>>> BFP650 by following my example, you might be better off with the BFP640. >>>>> Or more modern ones with a higher hFE (I'm fond of the NESG3031 and >>>>> NESG4030). >> >>>>> Namely, my choice of BFP650 (and not the then more widely available >>>>> BFP640) was driven by the search of the lowest possible R_BB, because >>>>> I was deemed to encounter lower source resistances than the 50ohms for >>>>> which SiGe devices are generally optimized. I reasoned that the base >>>>> geometry of the BFP650, as a medium power device, is more likely yield >>>>> a low R_BB. In your applications BFP640 may work better with the "more >>>>> ordinary" impedance levels and ambient temperatures. >> >>>>> The "I was about.." part means that I'm somewhat puzzled now that I >>>>> looked up the recent Infineon data sheets. The BFP650 is now listed >>>>> as a SiGe:C part (the datasheet 2010-10-22) and its Gummel-Poon >>>>> model reads RB=6.376 . The preliminary datasheet I have, dated >>>>> Aug-16-2004, lists it as an ordinary SiGe part with RB=1.036 . So, I'm >>>>> wondering how much the device has changed from the versions I've been >>>>> using. I seem to recall that also in some other more recent data sheet >>>>> versions it was listed as a SiGe, not SiGe:C . >> >>>>> Given the fact that the most recent BFP640 datasheet (dated 2007-05-29) >>>>> reads RB=3.129, the BFP650 does not seem to have any noise advantage >>>>> any more, either. >> >>>>> In fact, there was a recent paper in the RSI where >>>>> the Jena group claim to have achieved 15 pV/rtHz with BFP640 in LHe, >>>>> as compared with my 75 pV/rtHz with BFP650. This is a puzzling result, >>>>> actually, because it is better than one would expect in the picture where >>>>> u_N originates from the collector shot noise acting on the r_E - provided >>>>> that the thermal voltage saturates at ~5..7 meV even when the ambient >>>>> temperature keeps going down. According to my data it does saturate in >>>>> BFP640 just like it does in all other SiGe's I've tried, when measured by >>>>> the transconductance. I'm wondering whether my transistors have always >>>>> been oscillating so high that I cannot see it (ie.>27 GHz) and the >>>>> V_T saturation is an artefact due to it... >> >>>>> Regards, >>>>> Mikko >>>> I started by taking your advice, which was pretty valuable--thanks >>>> again. I noticed the change in specs as well, when I ordered another >>>> batch a couple of months ago. They changed the type number to BFP650H >>>> instead of BFP650E when they changed the process. (The dogs.) >> >>>> I'm using it as a cascode stage for a SKY65050 pHEMT, and it works >>>> great. One of the best things about it is that it has effectively >>>> infinite Early voltage, so you can get a lot of voltage gain out of a >>>> single stage. Even for situations where you don't care so much about >>>> ultralow noise, the combination of a gigantic f_T with a very high V_A >>>> is unique in my collection. It does want to oscillate at 14 GHz if you >>>> look at it crosswise, but a nice 5-ohm bead in series with the base >>>> cleans that right up. >> >>>> I'm going to be using it in a new front end design, as a >>>> bootstrapped-bootstrap wrapped round a couple of parallelled BF862s. DC >>>> coupling the bootstrap and using a current source load gets rid of the >>>> thermal tails on the BF862s by keeping their dissipation constant. >>>> (Running them just slightly above I_DSS gives them a zero tempco anyway, >>>> but this one needs really good dc stability.) >> >>>> Cheers >> >>>> Phil Hobbs >> >>> Funny you should say "cascode." I was just about to post the >>> following: >> >>> If I want a fast, low-capacitance, accurate, positive current source, >>> I could build a fairly slow active current source with a voltage >>> reference, a resistor, an opamp, and a p-fet. But it would have a lot >>> of capacitance and would be fairly slow. So I could cascode that with >>> a microwave-type PNP transistor, with a bit of base resistance or a >>> ferrite bead to keep it stable. But then I'd have the base current >>> error, and the good fast PNPs, what few there are, have mediocre >>> betas. >> >>> A PNP darlington is too slow. >> >>> So, how to correct for the base current error? >> >> Try this : >> >> Ic = 3mA >> >> 1K | >> ___ |/ >> 5V>-----|___|----+------| >> | |> >> .-. | >> | | | >> 1K | | | >> '-' | >> | | >> |\ | ||-+ >> 3V>--|+\ | ||<- >> |>---------|-----||-+ >> .-|-/ | | >> | |/ | ___ | >> '--------------+--|___|-+ >> | >> 1K .-. >> | | >> | |1K >> '-' >> | >> === >> GND >> >> Delta Ic< 3nA for 45<Beta<65 for ex. >> >> -- >> Thanks, >> Fred.- Hide quoted text - >> >> - Show quoted text - > > Would someone be kind enough to help me see how this compensates for > the base current? I've done the Vref-opamp-fet-resistor current > source. > > Color me slightly confused. > (I'd be happy to pay the, one beer licensing fee,... to understand > it.) > > George H.The summing junction is held still by feedback, and we assume the 5V reference is also stable. Any base current thus divides 50:50 between the two 1k resistors. Thus I_B /2 goes to the SJ, pulling it down by (2k)(I_B/2) = 1000 ohms * I_B. To restore balance, the source of the FET has to go up by the same amount, which requires an increase in drain current of 1000*I_B/1000, i.e. exactly I_B. That'll work great at DC, but any fast voltage swing at the collector will make a mess, due to the slow loop dynamics and high impedance at the base. The simple way to fix it is to use a transistor with infinite Early voltage and bypass the daylights out of the base, which is how the BFP650 came up. 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 845-480-2058 hobbs at electrooptical dot net http://electrooptical.net
Reply by ●February 13, 20122012-02-13
On Mon, 13 Feb 2012 09:10:28 -0800 (PST), George Herold <gherold@teachspin.com> wrote:>On Feb 12, 4:49�pm, Fred Bartoli <" "> wrote: >> John Larkin a �crit : >> >> >> >> >> >> > On Sun, 12 Feb 2012 13:39:37 -0500, Phil Hobbs >> > <pcdhSpamMeSensel...@electrooptical.net> wrote: >> >> >> Okkim Atnarivik wrote: >> >>> Phil Hobbs <pcdhSpamMeSensel...@electrooptical.net> wrote: >> >>> : Looks good. �Another approach would be something like a BFP650 follower, >> >> >>> � Phil, I was about to suggest that if you've ended up using the >> >>> BFP650 by following my example, you might be better off with the BFP640. >> >>> Or more modern ones with a higher hFE (I'm fond of the NESG3031 and >> >>> NESG4030). >> >> >>> � Namely, my choice of BFP650 (and not the then more widely available >> >>> BFP640) was driven by the search of the lowest possible R_BB, because >> >>> I was deemed to encounter lower source resistances than the 50ohms for >> >>> which SiGe devices are generally optimized. I reasoned that the base >> >>> geometry of the BFP650, as a medium power device, is more likely yield >> >>> a low R_BB. In your applications BFP640 may work better with the "more >> >>> ordinary" impedance levels and ambient temperatures. >> >> >>> � The "I was about.." part means that I'm somewhat puzzled now that I >> >>> looked up the recent Infineon data sheets. The BFP650 is now listed >> >>> as a SiGe:C part (the datasheet 2010-10-22) and its Gummel-Poon >> >>> model reads RB=6.376 . The preliminary datasheet I have, dated >> >>> Aug-16-2004, lists it as an ordinary SiGe part with RB=1.036 . So, I'm >> >>> wondering how much the device has changed from the versions I've been >> >>> using. I seem to recall that also in some other more recent data sheet >> >>> versions it was listed as a SiGe, not SiGe:C . >> >> >>> � Given the fact that the most recent BFP640 datasheet (dated 2007-05-29) >> >>> reads RB=3.129, the BFP650 does not seem to have any noise advantage >> >>> any more, either. >> >> >>> � In fact, there was a recent paper in the RSI where >> >>> the Jena group claim to have achieved 15 pV/rtHz with BFP640 in LHe, >> >>> as compared with my 75 pV/rtHz with BFP650. This is a puzzling result, >> >>> actually, because it is better than one would expect in the picture where >> >>> u_N originates from the collector shot noise acting on the r_E - provided >> >>> that the thermal voltage saturates at ~5..7 meV even when the ambient >> >>> temperature keeps going down. According to my data it does saturate in >> >>> BFP640 just like it does in all other SiGe's I've tried, when measured by >> >>> the transconductance. I'm wondering whether my transistors have always >> >>> been oscillating so high that I cannot see it (ie. >27 GHz) and the >> >>> V_T saturation is an artefact due to it... >> >> >>> � Regards, >> >>> � � � � � �Mikko >> >> I started by taking your advice, which was pretty valuable--thanks >> >> again. �I noticed the change in specs as well, when I ordered another >> >> batch a couple of months ago. �They changed the type number to BFP650H >> >> instead of BFP650E when they changed the process. �(The dogs.) >> >> >> I'm using it as a cascode stage for a SKY65050 pHEMT, and it works >> >> great. �One of the best things about it is that it has effectively >> >> infinite Early voltage, so you can get a lot of voltage gain out of a >> >> single stage. �Even for situations where you don't care so much about >> >> ultralow noise, the combination of a gigantic f_T with a very high V_A >> >> is unique in my collection. �It does want to oscillate at 14 GHz if you >> >> look at it crosswise, but a nice 5-ohm bead in series with the base >> >> cleans that right up. >> >> >> I'm going to be using it in a new front end design, as a >> >> bootstrapped-bootstrap wrapped round a couple of parallelled BF862s. �DC >> >> coupling the bootstrap and using a current source load gets rid of the >> >> thermal tails on the BF862s by keeping their dissipation constant. >> >> (Running them just slightly above I_DSS gives them a zero tempco anyway, >> >> but this one needs really good dc stability.) >> >> >> Cheers >> >> >> Phil Hobbs >> >> > Funny you should say "cascode." I was just about to post the >> > following: >> >> > If I want a fast, low-capacitance, accurate, positive current source, >> > I could build a fairly slow active current source with a voltage >> > reference, a resistor, an opamp, and a p-fet. But it would have a lot >> > of capacitance and would be fairly slow. So I could cascode that with >> > a microwave-type PNP transistor, with a bit of base resistance or a >> > ferrite bead to keep it stable. But then I'd have the base current >> > error, and the good fast PNPs, what few there are, have mediocre >> > betas. >> >> > A PNP darlington is too slow. >> >> > So, how to correct for the base current error? >> >> Try this : >> >> � � � � � � � � � � � � � � � � � � � � � � � � �Ic = 3mA >> >> � � � � � � � � � � � � � � � �1K � � � � � � � | >> � � � � � � � � � � � � � � � �___ � � � � � �|/ >> � � � � � � � � � � � 5V>-----|___|----+------| >> � � � � � � � � � � � � � � � � � � � �| � � �|> >> � � � � � � � � � � � � � � � � � � � .-. � � � | >> � � � � � � � � � � � � � � � � � � � | | � � � | >> � � � � � � � � � � � � � � � � � �1K | | � � � | >> � � � � � � � � � � � � � � � � � � � '-' � � � | >> � � � � � � � � � � � � � � � � � � � �| � � � �| >> � � � � � � � � � � � � � |\ � � � � � | � � ||-+ >> � � � � � � � � � � �3V>--|+\ � � � � �| � � ||<- >> � � � � � � � � � � � � � | �>---------|-----||-+ >> � � � � � � � � � � � � .-|-/ � � � � �| � � � �| >> � � � � � � � � � � � � | |/ � � � � � | � ___ �| >> � � � � � � � � � � � � '--------------+--|___|-+ >> � � � � � � � � � � � � � � � � � � � � � � � � | >> � � � � � � � � � � � � � � � � � � � � � �1K �.-. >> � � � � � � � � � � � � � � � � � � � � � � � �| | >> � � � � � � � � � � � � � � � � � � � � � � � �| |1K >> � � � � � � � � � � � � � � � � � � � � � � � �'-' >> � � � � � � � � � � � � � � � � � � � � � � � � | >> � � � � � � � � � � � � � � � � � � � � � � � �=== >> � � � � � � � � � � � � � � � � � � � � � � � �GND >> >> Delta Ic < 3nA for 45<Beta<65 for ex. >> >> -- >> Thanks, >> Fred.- Hide quoted text - >> >> - Show quoted text - > >Would someone be kind enough to help me see how this compensates for >the base current? I've done the Vref-opamp-fet-resistor current >source. > >Color me slightly confused. >(I'd be happy to pay the, one beer licensing fee,... to understand >it.) > >George H.Just run the math. Assume Ic is output, base current is Ic/beta, emitter current is Ic*(beta+1)/beta You'll find a need for 2*VR-VH > 0 (VR=3, VH=5 in this example) And Ic can't be 3mA as shown in the example ;-) ...Jim Thompson -- | James E.Thompson, CTO | mens | | Analog Innovations, Inc. | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | Phoenix, Arizona 85048 Skype: Contacts Only | | | Voice:(480)460-2350 Fax: Available upon request | Brass Rat | | E-mail Icon at http://www.analog-innovations.com | 1962 | I love to cook with wine. Sometimes I even put it in the food.
Reply by ●February 13, 20122012-02-13
On Mon, 13 Feb 2012 12:20:39 -0500, Phil Hobbs <pcdhSpamMeSenseless@electrooptical.net> wrote:>On 02/13/2012 12:10 PM, George Herold wrote: >> On Feb 12, 4:49 pm, Fred Bartoli<""> wrote: >>> John Larkin a �crit : >>> >>> >>> >>> >>> >>>> On Sun, 12 Feb 2012 13:39:37 -0500, Phil Hobbs >>>> <pcdhSpamMeSensel...@electrooptical.net> wrote: >>> >>>>> Okkim Atnarivik wrote: >>>>>> Phil Hobbs<pcdhSpamMeSensel...@electrooptical.net> wrote: >>>>>> : Looks good. Another approach would be something like a BFP650 follower, >>> >>>>>> Phil, I was about to suggest that if you've ended up using the >>>>>> BFP650 by following my example, you might be better off with the BFP640. >>>>>> Or more modern ones with a higher hFE (I'm fond of the NESG3031 and >>>>>> NESG4030). >>> >>>>>> Namely, my choice of BFP650 (and not the then more widely available >>>>>> BFP640) was driven by the search of the lowest possible R_BB, because >>>>>> I was deemed to encounter lower source resistances than the 50ohms for >>>>>> which SiGe devices are generally optimized. I reasoned that the base >>>>>> geometry of the BFP650, as a medium power device, is more likely yield >>>>>> a low R_BB. In your applications BFP640 may work better with the "more >>>>>> ordinary" impedance levels and ambient temperatures. >>> >>>>>> The "I was about.." part means that I'm somewhat puzzled now that I >>>>>> looked up the recent Infineon data sheets. The BFP650 is now listed >>>>>> as a SiGe:C part (the datasheet 2010-10-22) and its Gummel-Poon >>>>>> model reads RB=6.376 . The preliminary datasheet I have, dated >>>>>> Aug-16-2004, lists it as an ordinary SiGe part with RB=1.036 . So, I'm >>>>>> wondering how much the device has changed from the versions I've been >>>>>> using. I seem to recall that also in some other more recent data sheet >>>>>> versions it was listed as a SiGe, not SiGe:C . >>> >>>>>> Given the fact that the most recent BFP640 datasheet (dated 2007-05-29) >>>>>> reads RB=3.129, the BFP650 does not seem to have any noise advantage >>>>>> any more, either. >>> >>>>>> In fact, there was a recent paper in the RSI where >>>>>> the Jena group claim to have achieved 15 pV/rtHz with BFP640 in LHe, >>>>>> as compared with my 75 pV/rtHz with BFP650. This is a puzzling result, >>>>>> actually, because it is better than one would expect in the picture where >>>>>> u_N originates from the collector shot noise acting on the r_E - provided >>>>>> that the thermal voltage saturates at ~5..7 meV even when the ambient >>>>>> temperature keeps going down. According to my data it does saturate in >>>>>> BFP640 just like it does in all other SiGe's I've tried, when measured by >>>>>> the transconductance. I'm wondering whether my transistors have always >>>>>> been oscillating so high that I cannot see it (ie.>27 GHz) and the >>>>>> V_T saturation is an artefact due to it... >>> >>>>>> Regards, >>>>>> Mikko >>>>> I started by taking your advice, which was pretty valuable--thanks >>>>> again. I noticed the change in specs as well, when I ordered another >>>>> batch a couple of months ago. They changed the type number to BFP650H >>>>> instead of BFP650E when they changed the process. (The dogs.) >>> >>>>> I'm using it as a cascode stage for a SKY65050 pHEMT, and it works >>>>> great. One of the best things about it is that it has effectively >>>>> infinite Early voltage, so you can get a lot of voltage gain out of a >>>>> single stage. Even for situations where you don't care so much about >>>>> ultralow noise, the combination of a gigantic f_T with a very high V_A >>>>> is unique in my collection. It does want to oscillate at 14 GHz if you >>>>> look at it crosswise, but a nice 5-ohm bead in series with the base >>>>> cleans that right up. >>> >>>>> I'm going to be using it in a new front end design, as a >>>>> bootstrapped-bootstrap wrapped round a couple of parallelled BF862s. DC >>>>> coupling the bootstrap and using a current source load gets rid of the >>>>> thermal tails on the BF862s by keeping their dissipation constant. >>>>> (Running them just slightly above I_DSS gives them a zero tempco anyway, >>>>> but this one needs really good dc stability.) >>> >>>>> Cheers >>> >>>>> Phil Hobbs >>> >>>> Funny you should say "cascode." I was just about to post the >>>> following: >>> >>>> If I want a fast, low-capacitance, accurate, positive current source, >>>> I could build a fairly slow active current source with a voltage >>>> reference, a resistor, an opamp, and a p-fet. But it would have a lot >>>> of capacitance and would be fairly slow. So I could cascode that with >>>> a microwave-type PNP transistor, with a bit of base resistance or a >>>> ferrite bead to keep it stable. But then I'd have the base current >>>> error, and the good fast PNPs, what few there are, have mediocre >>>> betas. >>> >>>> A PNP darlington is too slow. >>> >>>> So, how to correct for the base current error? >>> >>> Try this : >>> >>> Ic = 3mA >>> >>> 1K | >>> ___ |/ >>> 5V>-----|___|----+------| >>> | |> >>> .-. | >>> | | | >>> 1K | | | >>> '-' | >>> | | >>> |\ | ||-+ >>> 3V>--|+\ | ||<- >>> |>---------|-----||-+ >>> .-|-/ | | >>> | |/ | ___ | >>> '--------------+--|___|-+ >>> | >>> 1K .-. >>> | | >>> | |1K >>> '-' >>> | >>> === >>> GND >>> >>> Delta Ic< 3nA for 45<Beta<65 for ex. >>> >>> -- >>> Thanks, >>> Fred.- Hide quoted text - >>> >>> - Show quoted text - >> >> Would someone be kind enough to help me see how this compensates for >> the base current? I've done the Vref-opamp-fet-resistor current >> source. >> >> Color me slightly confused. >> (I'd be happy to pay the, one beer licensing fee,... to understand >> it.) >> >> George H. > >The summing junction is held still by feedback, and we assume the 5V >reference is also stable. Any base current thus divides 50:50 between >the two 1k resistors. Thus I_B /2 goes to the SJ, pulling it down by >(2k)(I_B/2) = 1000 ohms * I_B. To restore balance, the source of the >FET has to go up by the same amount, which requires an increase in drain >current of 1000*I_B/1000, i.e. exactly I_B. > >That'll work great at DC, but any fast voltage swing at the collector >will make a mess, due to the slow loop dynamics and high impedance at >the base. The simple way to fix it is to use a transistor with infinite >Early voltage and bypass the daylights out of the base, which is how the >BFP650 came up. > >Cheers > >Phil HobbsBut that part has negative (positive?) Early voltage, namely negative output impedance. I wonder if that's thermal or wideband. So, it would oscillate with an LC in its collector and no base drive! -- John Larkin, President Highland Technology Inc www.highlandtechnology.com jlarkin at highlandtechnology dot com Precision electronic instrumentation Picosecond-resolution Digital Delay and Pulse generators Custom timing and laser controllers Photonics and fiberoptic TTL data links VME analog, thermocouple, LVDT, synchro, tachometer Multichannel arbitrary waveform generators
