RF Transistor Noise Source
in [Design]
|
Prev: shoppingreps
Next: Power factor correction
From: Jim Thompson on 29 Dec 2009 11:10 On Mon, 28 Dec 2009 18:27:50 -0800, John Larkin <jjlarkin(a)highNOTlandTHIStechnologyPART.com> wrote: >On Mon, 28 Dec 2009 18:17:39 -0800 (PST), "miso(a)sushi.com" ><miso(a)sushi.com> wrote: > >>On Dec 28, 1:49�pm, RST Engineering <jwei...(a)gmail.com> wrote: >>> . >>> . >>> There was a general discussion in this NG a couple of weeks ago about >>> using a lightly-biased zener as a noise source. �There was no clear >>> definition about how flat or to what frequency the noise was useful. >>> >>> It got me to thinking and I'll do the experiment as soon as I can >>> clean off my bench, but what do you think I'm going to see for >>> reasonable noise bandwidth if I use a small signal (like a 2N5770 or >>> 918) and use the emitter-base junction as the zener. �Most of them >>> zener somewhere around 5 volts and that should be reasonable. >>> >>> Most of the comments regarding bandwidth using a "regular" zener >>> centered around the rather large junction area necessary to carry some >>> decent current; the junction of an RF transistor ought to be at least >>> an order of magnitude (several??) smaller than that. >>> >>> Thoughts? >>> >>> Jim >> >>One thing to keep in mind is the transistor is easily damaged when >>zenered. You should do this with a current limited supply, say 100ua. > >It ruins their beta, but if the transistor will never be used as >anything but a zener, who cares? I suppose the issue is whether the >*zener* properties will change over time. I'm guessing that a >transistor, especially an RF transistor, will have a much higher >current density than a part designed to be a zener. > >This suggests some interesting experiments. > >John Indeed. Within a transistor the current density is much greater, and the zener voltage degrades... accelerated by high temperature, such as under the hood of a car... noted while developing integrated alternator regulators in the mid '60's. Also accelerated by mounting in cheap phenolic packaging. ...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 | Help save the environment! Please dispose of socialism responsibly!
From: John Larkin on 29 Dec 2009 12:40 On Tue, 29 Dec 2009 02:59:51 -0500, Phil Hobbs <pcdhSpamMeSenseless(a)electrooptical.net> wrote: >On 12/29/2009 12:07 AM, Tim Wescott wrote: >> On Mon, 28 Dec 2009 19:08:29 -0500, Phil Hobbs wrote: >> >>> On 12/28/2009 6:59 PM, Tim Wescott wrote: >>>> On Mon, 28 Dec 2009 14:49:06 -0800, RST Engineering wrote: >>>> >>>>> On Mon, 28 Dec 2009 16:07:24 -0600, Tim Wescott<tim(a)seemywebsite.com> >>>>> wrote: >>>>> >>>>>> On Mon, 28 Dec 2009 13:49:05 -0800, RST Engineering wrote: >>>>> >>>>> >>>>>> It'll be good to know what your results are. >>>>>> >>>>>> Twenty years ago you could buy a noise diode from MA-COM (IIRC; it >>>>>> may have been some other company), home-brew your own circuit to hold >>>>>> it, then send it back to MA-COM for calibration. I don't know if you >>>>>> still can. >>>>>> >>>>>> A noise diode was, of course, 'just a zener', optimized for use at >>>>>> microwave frequencies. >>>>> >>>>> Noisecom and Micronetics are the only two I know of. Noisecom used to >>>>> sell "factory seconds" to hams for pennies on the dollar but that >>>>> practice seems to have gone by the wayside. They, as you noted, would >>>>> also do a calibration of your design for a few bucks. Gone also. >>>>> >>>>> It will be fun to get back to experimenting with something where I >>>>> don't have a real good idea what the answer is going to be. >>>>> >>>>> Jim >>>> >>>> I think it's Noisecom that I was thinking of. Dang; I should have >>>> taken advantage while I could. >>>> >>>> I have thought that if you were building something low-noise enough you >>>> could measure the noise figure with a pair of transmission lines >>>> terminated in resistors: drop one into ice water (or dry-ice/acetone, >>>> or LN2), and heat the other one up (boiling water, or a >>>> not-quite-melted- solder heat furnace). Then switch between them. >>>> With no current flowing through the resistors, you'd certainly know >>>> their noise temperatures! >>>> >>>> >>> A common approach in physics labs is to terminate the input with a 300 >>> kelvin resistor, measure the noise, dunk the resistor in liquid >>> nitrogen, and measure it again. Works great. >>> >> Goodness you keep the heat turned up -- or is that in the summer? >> >> It's about 293K in here now, because I can get comfort cheaper with a >> sweater than by turning up the heat. >> > >Nope, we reduce waste by keeping the house about 59 F in the winter (55 >at night). So since it's 3 AM here, I'll see your 293 and raise you -8 >kelvins. ;) That sort of frugality would get me divorced. You've met Mo and probably noticed her unfavorable mass/surface area ratio. Car seat warmers are another great marriage-saving invention. All of which explains this week's automation project. John
From: Jim Thompson on 29 Dec 2009 12:54 On Tue, 29 Dec 2009 09:40:44 -0800, John Larkin <jjlarkin(a)highNOTlandTHIStechnologyPART.com> wrote: >On Tue, 29 Dec 2009 02:59:51 -0500, Phil Hobbs ><pcdhSpamMeSenseless(a)electrooptical.net> wrote: > >>On 12/29/2009 12:07 AM, Tim Wescott wrote: >>> On Mon, 28 Dec 2009 19:08:29 -0500, Phil Hobbs wrote: >>> >>>> On 12/28/2009 6:59 PM, Tim Wescott wrote: >>>>> On Mon, 28 Dec 2009 14:49:06 -0800, RST Engineering wrote: >>>>> >>>>>> On Mon, 28 Dec 2009 16:07:24 -0600, Tim Wescott<tim(a)seemywebsite.com> >>>>>> wrote: >>>>>> >>>>>>> On Mon, 28 Dec 2009 13:49:05 -0800, RST Engineering wrote: >>>>>> >>>>>> >>>>>>> It'll be good to know what your results are. >>>>>>> >>>>>>> Twenty years ago you could buy a noise diode from MA-COM (IIRC; it >>>>>>> may have been some other company), home-brew your own circuit to hold >>>>>>> it, then send it back to MA-COM for calibration. I don't know if you >>>>>>> still can. >>>>>>> >>>>>>> A noise diode was, of course, 'just a zener', optimized for use at >>>>>>> microwave frequencies. >>>>>> >>>>>> Noisecom and Micronetics are the only two I know of. Noisecom used to >>>>>> sell "factory seconds" to hams for pennies on the dollar but that >>>>>> practice seems to have gone by the wayside. They, as you noted, would >>>>>> also do a calibration of your design for a few bucks. Gone also. >>>>>> >>>>>> It will be fun to get back to experimenting with something where I >>>>>> don't have a real good idea what the answer is going to be. >>>>>> >>>>>> Jim >>>>> >>>>> I think it's Noisecom that I was thinking of. Dang; I should have >>>>> taken advantage while I could. >>>>> >>>>> I have thought that if you were building something low-noise enough you >>>>> could measure the noise figure with a pair of transmission lines >>>>> terminated in resistors: drop one into ice water (or dry-ice/acetone, >>>>> or LN2), and heat the other one up (boiling water, or a >>>>> not-quite-melted- solder heat furnace). Then switch between them. >>>>> With no current flowing through the resistors, you'd certainly know >>>>> their noise temperatures! >>>>> >>>>> >>>> A common approach in physics labs is to terminate the input with a 300 >>>> kelvin resistor, measure the noise, dunk the resistor in liquid >>>> nitrogen, and measure it again. Works great. >>>> >>> Goodness you keep the heat turned up -- or is that in the summer? >>> >>> It's about 293K in here now, because I can get comfort cheaper with a >>> sweater than by turning up the heat. >>> >> >>Nope, we reduce waste by keeping the house about 59 F in the winter (55 >>at night). So since it's 3 AM here, I'll see your 293 and raise you -8 >>kelvins. ;) > >That sort of frugality would get me divorced. You've met Mo and >probably noticed her unfavorable mass/surface area ratio. > >Car seat warmers are another great marriage-saving invention. > >All of which explains this week's automation project. > >John And Joerg was coming unglued because I said we have seat warmers/coolers in our car ;-) ...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 | Help save the environment! Please dispose of socialism responsibly!
From: John Larkin on 29 Dec 2009 13:02 On Tue, 29 Dec 2009 02:27:37 -0800 (PST), Wimpie <wimabctel(a)tetech.nl> wrote: >On 29 dic, 09:56, Robert Baer <robertb...(a)localnet.com> wrote: >> m...(a)sushi.com wrote: >> > On Dec 28, 1:49 pm, RST Engineering <jwei...(a)gmail.com> wrote: >> >> . >> >> . >> >> There was a general discussion in this NG a couple of weeks ago about >> >> using a lightly-biased zener as a noise source. �There was no clear >> >> definition about how flat or to what frequency the noise was useful. >> >> >> It got me to thinking and I'll do the experiment as soon as I can >> >> clean off my bench, but what do you think I'm going to see for >> >> reasonable noise bandwidth if I use a small signal (like a 2N5770 or >> >> 918) and use the emitter-base junction as the zener. �Most of them >> >> zener somewhere around 5 volts and that should be reasonable. >> >> >> Most of the comments regarding bandwidth using a "regular" zener >> >> centered around the rather large junction area necessary to carry some >> >> decent current; the junction of an RF transistor ought to be at least >> >> an order of magnitude (several??) smaller than that. >> >> >> Thoughts? >> >> >> Jim >> >> > One thing to keep in mind is the transistor is easily damaged when >> > zenered. You should do this with a current limited supply, say 100ua. >> >> > If you've ever done ESD testing or fuse testing, invariably the >> > reverse biased diode is the thing that is easily fried. When popping >> > metal fuses, you need to insure that the inductive kick of the zapper >> > is such that after popping the fuse, the diode junction gets forward >> > biased. This does dump current into the device, but that can be >> > controlled by the size of the capacitor used in zapping. For ESD >> > structures where there is no diode to the positive rail, the snap back >> > of the "off" nfet saves the parasitic diode junction. >> >> � �Be advised that the damage to a zenered transistor E-B junction is a >> time * current or dosage product, exactly as if it got radiation damage. >> � �Total dosage: a little over a long time = = a lot over a short time. >> � �Fairchild uA709s used in the Apollo got "nailed" by that. >> � �Turns out the company hired to test and burn them in did not know >> what an op amp was or how to test them or even burn them in despite a >> burn-in circuit in the data sheet! >> � �Their circuit zenered the inputs and that caused a failure mode >> during a mission. >> � �Fairchild engineers had to teach some basic electronics, and then >> advance to op amps and test methods as part of proof the 709s were not >> initially bad. >> � �The other part was a setup burning in NIB same date lot code parts in >> 2 batches: one using the nasty circuit and the other using the datasheet >> circuit. >> � �Oh yes; the cure is to anneal out the damage in an oven. > >Hello Robert, > >I did some experiments with low current reversed bias to the BE >junction of BC847. Even at low reverse current (10uA for several >hours), the HFE at low collector current (<10uA) drops significantly >after applying the reverse current. > >It looks like adding a resistor parallel to the BE junction as the HFE >at high current did not drop significantly. > >Does such dosage degradation to junctions also occur when reverse >biasing microwave schottky mixer diodes, or PN junction diodes? > >Best regards, > >Wim >PA3DJS >www.tetech.nl >please remove abc in case of PM. "Reference zeners" like the 1N935 sort of guys, are remarkably stable over time. They are usually a stack of a zener in series with one or two forward diodes, giving a TC that is zero at some current around 7.5 mA, usually. You can tune the current to hit zero TC. http://www.microsemi.com/datasheets/SA6-7.PDF The best way to get a really good current source is to use the reference zener to make its own current source, using a simple bootstrap circuit. Just make sure it starts up! I once bought a 6.2 volt zener from Motorola for $35, which was a lot of money then. It came in a presentation-quality tube with a 1000-hour graph of stability, signed by all sorts of important people. Some good stuff here: http://focus.ti.com/lit/an/slyt183/slyt183.pdf John
From: John Larkin on 29 Dec 2009 13:05
On Tue, 29 Dec 2009 00:43:19 -0800, Robert Baer <robertbaer(a)localnet.com> wrote: >RST Engineering wrote: >> . >> . >> There was a general discussion in this NG a couple of weeks ago about >> using a lightly-biased zener as a noise source. There was no clear >> definition about how flat or to what frequency the noise was useful. >> >> It got me to thinking and I'll do the experiment as soon as I can >> clean off my bench, but what do you think I'm going to see for >> reasonable noise bandwidth if I use a small signal (like a 2N5770 or >> 918) and use the emitter-base junction as the zener. Most of them >> zener somewhere around 5 volts and that should be reasonable. >> >> Most of the comments regarding bandwidth using a "regular" zener >> centered around the rather large junction area necessary to carry some >> decent current; the junction of an RF transistor ought to be at least >> an order of magnitude (several??) smaller than that. >> >> Thoughts? >> >> Jim > Well, all of the bipolar transistors seem to have the_specification_ >of a max reverse VBE of 5 volts, but in fact they all zener in the 8 >volt region. Lots of NPNs zener around 5 volts. If you use the collector and emitter, you get a "reference zener", a zener in series with a forware-biased diode, around 6.2 volts with a very low TC. For some reason, PNP transistors sometines have higher zener voltages, 10-12 volts maybe. John |