Robert Baer wrote:> Bill Sloman wrote: >> On 7/24/2012 6:31 AM, Robert Baer wrote: >>> Bill Sloman wrote: >>>> On 7/23/2012 10:16 AM, Robert Baer wrote: >>>>> Zeners all seem to behave: quiet, no oscillation IF run at or near >>>>> the spec current (usually in the tens of mA). >>>>> But, if one needs to run them at low currents (say 10uA to 500uA) >>>>> then _watch_out_! >>>>> Some brands at certain voltages are extremely noisy and subject to >>>>> oscillation. >>>>> The waveforms seen at low currents (2-50uA) look like randumb >>>>> sawtooth generation, and as the current is increased the amplitude and >>>>> probability increases (do NOT want to say "frequency" as that gives >>>>> wrong impression). >>>>> Then this "converts" to random noise. >>>>> Eventually, there are "bursts" of NO noise. Then at higher current, >>>>> noise bursts decrease in "frequency" and amplitude. >>>>> >>>>> That is what i see in general. >>>>> However, i have found two zeners that, on a curve tracer, one sees >>>>> little if any noise or spikes mentioned. >>>>> NOW for the question: >>>>> How can one test these particular zeners to ensure they do not or >>>>> will not oscillate? >>>>> * >>>>> Yes, there are zeners made for low noise and those have a much lower >>>>> spec current, BUT i cannot afford to have the factory dig the silicon >>>>> mines, etc much less the high $$. >>>> >>>> There was a nice - and tolerably long - thread on the subject "Zener >>>> Diode Oscillation" - here back in 1997. Winfield Hill and the late >>>> Tony Williams posted a lot of interesting measurements and some >>>> interesting references from the semiconductor physics literature. >>>> >>> I remember a while back references to that; sounded worthy of a >>> textbook. >>> Is that thread or (one) book available,and where? >> >> Google groups still has it. Searching on "Zener diode oscillation" in >> google groups advanced search found it immediately. 96 posts, last post >> July 20, 1997. >> >> https://groups.google.com/forum/?fromgroups#!topic/sci.electronics.design/OZtqC_bPpPU/overview >> >> >> > Thanks! > Noticed the (first?) posting by Mike 7/20/97 where he was rather adamant > about "No one can produce a sample of a zener that oscillates." and (to > me) rather flippantly states "The person who first demonstrated the > existence of oscillations in zeners would be up for the Nobel prize." > > I must state "hogwash", in all that one needs for an oscillator is a > device that exhibits negative resistance (note he _does_ mention Dr. Leo > Esaki immediately afterwards). > eXplicitly to the point, buy some Vishay MMBZ5266C-V-GS08 68V SOT-23 > zeners and put them on a curve tracer; very nice negR jump in the > 40-60uA region where the voltage DECREASES as the current INCREASES. > I think the same die is available in the DO-35 package by Vishay if you > need better power dissipation for your oscillator. > > >I have posted a 7CT1N curve tracer screen photo under the heading "ref:Zener noise / oscillations". Mike can send me a cash equivalent of the Nobel Prize any day now; i will be glad to share it with all those here that know better (ie: beans from duck-eggs).
Zener noise / oscillations
Started by ●July 23, 2012
Reply by ●July 26, 20122012-07-26
Reply by ●July 26, 20122012-07-26
"Robert Baer" <robertbaer@localnet.com> wrote in message news:IpqdnflGK8W9KY3NnZ2dnUVZ_uWdnZ2d@posted.localnet...> I need a 2-terminal device so tat i can put hundreds of them in series > for 25KV. > So far, only one person divined that and his comment was trounced, and > not politely.Like I said before (though possibly not in this thread), cascode some HV MOSFETs. Only way you'll get the stability and noise you demand, without stacking a bajillion 431's. Tim -- Deep Friar: a very philosophical monk. Website: http://webpages.charter.net/dawill/tmoranwms
Reply by ●July 26, 20122012-07-26
On a sunny day (Wed, 25 Jul 2012 20:06:05 -0700) it happened Robert Baer <robertbaer@localnet.com> wrote in <IpqdnflGK8W9KY3NnZ2dnUVZ_uWdnZ2d@posted.localnet>:>Jan Panteltje wrote: >> On a sunny day (Tue, 24 Jul 2012 23:39:15 -0700) it happened Robert Baer >> <robertbaer@localnet.com> wrote in >> <WeudnYHn_bcLCZLNnZ2dnUVZ_jOdnZ2d@posted.localnet>: >> >>> Tim Williams wrote: >>>> "Robert Baer"<robertbaer@localnet.com> wrote in message >>>> news:y4CdnbWbSZcVtpPNnZ2dnUVZ_rGdnZ2d@posted.localnet... >>>>> Yes, a TL431 is cheap but the minimal standing current is way above >>>>> what is needed. >>>> >>>> 1mA? TLV431 is 100uA and 1.240V, much improved. >>>> >>>> Tim >>>> >>> Range needed: 10uA-500uA. >> >> It is interesting to note that the PIC 18F14K22 is specified with FVR only running at: >> 1.8V 28 uA >> 3V 65 uA >> 5V 101 uA >> For a reference output of 1.024V >> >> I would think that if you put it into sleep mode, that would make your specs... >> Datasheet page 337 >> It is possible to send the Vref output to a pin in that mode AFAIK. > I need a 2-terminal device so tat i can put hundreds of them in >series for 25KV. > So far, only one person divined that and his comment was trounced, >and not politely.OK, so I did not see you mention 25kV @ say <100uA. That will not be easy with zeners. You also need to mention stabilisation factor in percent for example, for anybody to even think about it. I could easily generate 25kV locally with a transistor circuit and even some feedback turns, without a bleeder on the whole HV. Any decent last century CRT solid state color TV does it that way. Zeners do not make that much sense here to me.
Reply by ●July 26, 20122012-07-26
On a sunny day (Thu, 26 Jul 2012 12:01:13 +1000) it happened "David Eather" <eather@tpg.com.au> wrote in <op.wh06kbmvwei6gd@phenom-pc>:>> The PIC 18F... has an internal programable bandgap reference, >> you can select for 1.024V, and 2 or 3, or 4 x that, >> so up to 4.096 V, and uses very low power. >> It can probably replace the rest of your circuit too ;-) >> >> > >7% accuracy too!Absolute accuracy as to 1.024 perhaps? I have used that reference in at least 3 projects now, also very long term, and also in temp stabilised (in the tritium decay experiment it is constantly compared against an other external reference), data shows it is still within 1 part in 1000 after more than 70 days (sampled each hour), Absolute accuracy is often not that important, as it can be calculated out, simply a #define in the software.
Reply by ●July 26, 20122012-07-26
Ah, so suppose I were to order 10000 uCs from Digikey, preprogrammed. HOw shall I tell them they each need a unique #define? Tim -- Deep Friar: a very philosophical monk. Website: http://webpages.charter.net/dawill/tmoranwms "Jan Panteltje" <pNaonStpealmtje@yahoo.com> wrote in message news:juqkbp$mo2$1@news.albasani.net...> On a sunny day (Thu, 26 Jul 2012 12:01:13 +1000) it happened "David > Eather" > <eather@tpg.com.au> wrote in <op.wh06kbmvwei6gd@phenom-pc>: > >>> The PIC 18F... has an internal programable bandgap reference, >>> you can select for 1.024V, and 2 or 3, or 4 x that, >>> so up to 4.096 V, and uses very low power. >>> It can probably replace the rest of your circuit too ;-) >>> >>> >> >>7% accuracy too! > > Absolute accuracy as to 1.024 perhaps? > I have used that reference in at least 3 projects now, also very long > term, > and also in temp stabilised (in the tritium decay experiment it is > constantly compared > against an other external reference), data shows it is still within 1 part > in 1000 after > more than 70 days (sampled each hour), > Absolute accuracy is often not that important, as it can be calculated > out, simply a #define in the software.
Reply by ●July 26, 20122012-07-26
Tim Williams a �crit :> Ah, so suppose I were to order 10000 uCs from Digikey, preprogrammed. > HOw shall I tell them they each need a unique #define? > > Tim >Nah, have it programmed with a C compiler and small bootstrap that calibrate the #define generate the source, compile the program and then flash it :-) -- Thanks, Fred.
Reply by ●July 26, 20122012-07-26
On a sunny day (Thu, 26 Jul 2012 07:49:00 -0500) it happened "Tim Williams" <tmoranwms@charter.net> wrote in <jurec9$hke$1@dont-email.me>:>Ah, so suppose I were to order 10000 uCs from Digikey, preprogrammed. HOw >shall I tell them they each need a unique #define? > >TimThats is silly, for normal applications the reference is accurate enough, only for super precise things and test equipment where calibration is needed would you use the #define method. And I could even automate that, see how? Same goes for internal oscillator settings, it [some PIC} has a factory adjusted byte that presets it to +-1%, but you can, in your production test, easily include an auto-calibration against some super accurate source you have. If you do not test you products at all yes then you make X % crap anyways. Was not that saying: Those who say it cannot be done should not interrupt person doing it?
Reply by ●July 26, 20122012-07-26
On Wed, 25 Jul 2012 20:50:26 -0700, Robert Baer wrote:> I have posted a 7CT1N curve tracer screen photo under the heading > "ref:Zener noise / oscillations".Which 7000 series mainframe has a *red* phosphor? -- "For a successful technology, reality must take precedence over public relations, for nature cannot be fooled." (Richard Feynman)
Reply by ●July 26, 20122012-07-26
On Thu, 26 Jul 2012 13:02:58 -0700, Fred Abse wrote:> On Wed, 25 Jul 2012 20:50:26 -0700, Robert Baer wrote: > >> I have posted a 7CT1N curve tracer screen photo under the heading >> "ref:Zener noise / oscillations". > > Which 7000 series mainframe has a *red* phosphor?I forgot to ask: What zener? What voltage and current scales? What series resistance? I'd like to try that. -- "For a successful technology, reality must take precedence over public relations, for nature cannot be fooled." (Richard Feynman)
Reply by ●July 26, 20122012-07-26
Ok then. Provide an example wherein a PIC is used, alone, with no external reference, as a voltmeter with guaranteed 0.1% accuracy. The design must be DFM-ready, meaning, it must use a single code version, so that all components can be purchased en masse, assembled by a contract assembler, and shipped direct to customers. No measurement step, no calibration step, no reprogramming allowed (other than what it can do internally, which won't be too useful). If the internal reference has the stated accuracy (~0.1%, i.e., "1.024"), this should be trivial. If not, then don't call it what it clearly is not. This goes double for the manufacturer, to whom less attention should be paid as a result. Note that "1.024" at 7% is 0.952 to 1.096. It's closer to "1.0", which implies 10% error, than it is to "1.02", which implies 1%. Tim -- Deep Friar: a very philosophical monk. Website: http://webpages.charter.net/dawill/tmoranwms "Jan Panteltje" <pNaonStpealmtje@yahoo.com> wrote in message news:juri6h$j3c$1@news.albasani.net...> On a sunny day (Thu, 26 Jul 2012 07:49:00 -0500) it happened "Tim > Williams" > <tmoranwms@charter.net> wrote in <jurec9$hke$1@dont-email.me>: > >>Ah, so suppose I were to order 10000 uCs from Digikey, preprogrammed. HOw >>shall I tell them they each need a unique #define? >> >>Tim > > > Thats is silly, for normal applications the reference is accurate enough, > only for super precise things and test equipment where calibration is > needed > would you use the #define method. > And I could even automate that, see how? > Same goes for internal oscillator settings, it [some PIC} has a factory > adjusted byte > that presets it to +-1%, but you can, in your production test, easily > include an auto-calibration against some super accurate source you have. > If you do not test you products at all yes then you make X % crap anyways. > Was not that saying: > Those who say it cannot be done should not interrupt person doing it? > >
