Reply by Chris Jones April 15, 20142014-04-15
On 15/04/2014 10:09, Clifford Heath wrote:
> On 05/04/14 00:25, Chris Jones wrote: >> On 04/04/2014 19:44, Clifford Heath wrote: >> [snip] >> > <http://cjh.polyplex.org/electronics/fox/Oscillator.brd.png> > > > <http://cjh.polyplex.org/electronics/fox/Oscillator.sch.png> > >> I note that you are taking the output via C9 from the top of the tank, >> which is a very sensitive node. I hope it is going to something with >> really low loss, or that could be a cause for reducing the Q and >> tendency to oscillate. I would suggest taking the output from some >> low-impedance node, such as a tap on the inductor, or the top of R6 or >> something like that. > > I was taking signal off the tank because the signal level is 20dB better > than at the emitter, while the spurious levels are the same.
Yes, fair enough. If you have a load with very high impedance (the gate of a very tiny fet perhaps) then taking the output from the large swing of the tank can give you the best phase noise floor at high offsets, but that load needs to be very high impedance. With a low impedance load it can stop it oscillating.
> > I've switched the design to a tap off the inductor, which is still clean > but much lower impedance (less tank load). I need to re-do the layout > before retesting at a higher frequency, but all the symptoms I am seeing > are explained by excess loading (or various sorts) on the tank. > > I just got bitten yesterday by the very high Cbe of the 2N2222. 20pF > indeed! I'll be more careful using those in future :) > > Thanks everyone for your help.
Reply by Clifford Heath April 14, 20142014-04-14
On 05/04/14 00:25, Chris Jones wrote:
> On 04/04/2014 19:44, Clifford Heath wrote: > [snip] > > <http://cjh.polyplex.org/electronics/fox/Oscillator.brd.png>
> > <http://cjh.polyplex.org/electronics/fox/Oscillator.sch.png>
> I note that you are taking the output via C9 from the top of the tank, > which is a very sensitive node. I hope it is going to something with > really low loss, or that could be a cause for reducing the Q and > tendency to oscillate. I would suggest taking the output from some > low-impedance node, such as a tap on the inductor, or the top of R6 or > something like that.
I was taking signal off the tank because the signal level is 20dB better than at the emitter, while the spurious levels are the same. I've switched the design to a tap off the inductor, which is still clean but much lower impedance (less tank load). I need to re-do the layout before retesting at a higher frequency, but all the symptoms I am seeing are explained by excess loading (or various sorts) on the tank. I just got bitten yesterday by the very high Cbe of the 2N2222. 20pF indeed! I'll be more careful using those in future :) Thanks everyone for your help.
Reply by Clifford Heath April 5, 20142014-04-05
On 05/04/14 11:49, RobertMacy wrote:
> On Fri, 04 Apr 2014 13:58:46 -0700, Clifford Heath <no.spam@please.net> > wrote: >> ...snip... >> Thanks, I will do. >>> ...snip... > I just tested it in a sample circuit, that new model is 1/3 the speed of > the other model! > Envision grounded base with current source(s) to emitter. collector has > 1k load with about 5mA bias current. > Also substrate voltage at -8Vdc. > > Roll off of old model approx 190MHz, new model 64.5MHz! significant > differences there.
Wow, yes, big difference. However the Ft of the CA3046 is meant to be around 300MHz, whereas the LM3046 is min 300, typical 550. I don't really know how much gain your test circuit was asking for, but they do say "DC to 120MHz" on the National data sheet... Maybe I'm asking too much to push to 150MHz, but surely not to 30? That said, I was just playing with parallel R loading the tank, and it doesn't take much to drop the amplitude way down... it seems likely that my original problem was excessive tank loading.
> Plus has a bit more noise, if that's important.
Make it a bit more likely to start perhaps :P -- Clifford Heath
Reply by RobertMacy April 4, 20142014-04-04
On Fri, 04 Apr 2014 14:28:00 -0700, Jim Thompson  
<To-Email-Use-The-Envelope-Icon@on-my-web-site.com> wrote:

> On Fri, 04 Apr 2014 14:03:58 +1100, Clifford Heath > <no.spam@please.net> wrote: > > [snip] > > You apparently have multiple line-wrap issues in your .asc file. > > Try not putting everything in one line ;-) > > ...Jim Thompson
You have no idea how long I waited to be able to finally say this, "I had no problems with it." ;)
Reply by RobertMacy April 4, 20142014-04-04
On Fri, 04 Apr 2014 13:58:46 -0700, Clifford Heath <no.spam@please.net>  
wrote:

> ...snip... > > Thanks, I will do. > >> ...snip...
I just tested it in a sample circuit, that new model is 1/3 the speed of the other model! Envision grounded base with current source(s) to emitter. collector has 1k load with about 5mA bias current. Also substrate voltage at -8Vdc. Roll off of old model approx 190MHz, new model 64.5MHz! significant differences there. Plus has a bit more noise, if that's important.
Reply by Jim Thompson April 4, 20142014-04-04
On Fri, 04 Apr 2014 14:03:58 +1100, Clifford Heath
<no.spam@please.net> wrote:

[snip]

You apparently have multiple line-wrap issues in your .asc file.

Try not putting everything in one line ;-)
		
                                        ...Jim Thompson
-- 
| James E.Thompson                                 |    mens     |
| Analog Innovations                               |     et      |
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    |
| San Tan Valley, AZ 85142   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 Clifford Heath April 4, 20142014-04-04
On 04/04/14 22:49, Chris Jones wrote:
> On 04/04/2014 14:03, Clifford Heath wrote: >> Thanks to some good help here in the past, I have a nice clean >> oscillator (for my new fox transmitters), but I'm having trouble with >> the LTSpice simulation at higher frequencies. I'm assuming that the >> Intersil Spice model for the CA3046 is bad, and I know JT has a better >> one. >> >> Anyhow, I have two questions about the attached LTSpice schematic. >> >> 1) When I set it to 30MHz, LTSpice says the oscillator will start nicely >> (even up to 150MHz in fact). Now I know that the (calculated) Colpitts >> capacitance ratio is bad, as is the base capacitor, but with the >> physical circuit built in a tight layout in SMD and a range of more >> sensible capacitances substituted, I can't get it to oscillate above >> about 12MHz. Why does it not want to run in real life, when Spice says >> it should? Can someone provide me a better CA3046 model please, or say >> what else might be going on? > [snip] > I would suggest estimating the length of some of the more critical > physical wires on the hardware version, and putting appropriate > parasitics back into the simulation model. Particularly in series with > the emitter of transistors, it can affect gain at RF. I would suggest > modelling the emitter pin and leadframe and the wiring as 1nH per > millimetre of length, as a first guess, plus another nanohenry for the > bondwire (unless it is already in your transistor model). Whilst this > probably isn't the cause, it might start to expain something.
The whole oscillator fits in 1cmx2cm - leads are really short. I'd be surprised if parasitics are a problem as low as 30MHz. Fair enough to worry at UHF, but even at 150MHz I don't think my layout is bad. I need find out though! I think I could pretty easily switch to discretes (say MMBT2222A) which have a similar-ish Ft etc, and that could improve the layout a little. Especially by getting pin 8 (collector of the phase-splitter) away from the oscillator device.
> Another thing that may be worth trying, though it isn't usually > necessary until you get to GHz frequencies: The capacitance across > unwanted junctions on chips (perhaps collector to substrate in your > case) can be lossy.
What sort of capacitance would I expect between two SSOP pins? and for a 10mil trace, what capacitance through ordinary 1.6mm FR4? If you model it as a pure capacitor then you can
> simulate better Q or gain or noise figure than reality. Of course if you > add an infinite value resistor in series with the parasitic capacitances > then that also gives zero losses and artificially good performance. > There is some intermediate value of resistance to put in the model in > series with the parasitic capacitance, between zero and infinity Ohms, > that will result in the worst possible losses. There is some other value > of series resistance that is the best model of reality. Often the most > representative value is close to the worst case losses value, except > when you don't want things to oscillate in which case it is very different. > > It is important to put some losses in the model of your inductor also, > as another poster mentioned. Beware of metal very close to the inductor. > I sometimes like to put inductors in screened cans, but all metalwork > should ideally be spaced away from the turns of the inductor by at least > the inductor diameter, to avoid causing more losses.
I planned to box the whole oscillator, but it's hard to develop that way :P.
> Hopefully you have some low-ESR decoupling (e.g. 100n chip ceramic) > capacitor(s) across your supply right near the oscillator. Otherwise the > impedance of the supply wiring might stop things from working, in a way > that would not be simulated.
Yes, the collector of both oscillator and AGC transistor have 0805 caps directly to a ground via. I'm using caps from a cheap Chinese SMD book... I hope they're ok.
> I suggest you get one of those label machines, and affix a label to your > prototype that says "unconditionally stable amplifier". It will surely > oscillate then.
:)
> Are you able to verify the DC collector current and Vce of your > not-oscillating transistor? I would check that the collector current is > close to (but slightly lower than) the value that gives the peak Ft > value.
I don't even know what current that is. Will check the data sheet.
>(I have been warned that above the current that gives peak Ft, > models are often not very good and/or device performance is more > unpredictable, and to therefore stay somewhat below that current if good > Ft is important). Also make sure the Vce is well over a volt,
Ahhh, ok, it's just over a volt - about 1.7v I think. Supply is only 3.2V. and more
> if the transistor has a poorly contacted collector with lots of > distributed resistance (which could cause parts of the device to > saturate before other parts, and wouldn't usually be modelled). > > For oscillators, I like to pick a transistor with an Ft that is about 10 > times higher than the highest desired oscillation frequency. Something > like an old BFY90, in your case, or one of the many surface mount RF > transistors that you can get these days.
Thought I had some BFR93A's here and wanted to splice one in to the existing layout (cut pins and hot-wire the SOT23 on top of the LM3046), but I can't find them. Maybe too fast, but still...
>Actually I prefer FET LC > oscillators but none of this is really relevant to the simulation problem.
Well, I'll experiment with BF862's for VHF work. Not sure how well that will go with a 3.2V supply (Vgs is a bit unpredictable). This existing design is useful even if I can't get it to run at VHF, but I really want to learn what I've done wrong anyhow. -- Clifford Heath
Reply by Clifford Heath April 4, 20142014-04-04
On 05/04/14 02:45, RobertMacy wrote:
> On Fri, 04 Apr 2014 01:28:23 -0700, Clifford Heath <no.spam@please.net> > wrote: > >>> ...snip, way too much .... > > Alex Bordordunov [incredibly prolific producer of effective models and > able electronic designer!] just sent me this model:
Fantastic, I'll try it later today.
> "Hi Robert, > I have found the more neat (exact) model a model transistor CA3046. > .MODEL CA3046 NPN(IS=10f BF=145.76 VAF=100 IKF=46.747m ISE=114.23f > NE=1.483 BR=100.1m VAR=100 IKR=10.01m ISC=10f RC=10 CJE=1.026p > MJE=333.33m CJC=991.79f MJC=333.33m TF=277.09p XTF=309.38 VTF=16.364 > ITF=1.7597 TR=10n CJS=6.3P VJS=0.749 MJS=0.5 VCEO=20 ICRATING=50m MFG=RCA) > To LTspice used the additional parameters to models (CJS, VJS, MJS) > necessary to use the symbol NPN4. > > Alex." > He has over 35MB of models, which he posted to the LTspice group. > I am trying to put the two of you in contact. send me your email > address.
firstname.lastnmae@gmail.com
>Meanwhile, try his model and let us know if it acts much > differently.
Thanks, I will do.
> Don't be offended, ...wrap the lines properly.
Do you mean "let the reader wrap the lines"? Yes, ok, a lot of people read on displays narrower than 80 columns now. Really? I have a 40-year habit of wrapping text, which I'm trying to break... bear with me :) Clifford Heath.
Reply by Clifford Heath April 4, 20142014-04-04
On 05/04/14 00:25, Chris Jones wrote:
> On 04/04/2014 19:44, Clifford Heath wrote: > [snip] >>> any photos? of the circuitry? Don't wish to cast aspersions, but layout >>> above 10 MHz starts to become important. up there a short connection is >>> no longer a connection, >> >> Yes, I'm aware of that. I've routed it single-sided on a copper rear >> plane, using 0805 components throughout. Within that constraint, I'll be >> very surprised if you can find a way to substantially improve it... but >> here are the snapshots of this part of the circuit (from Eagle): >> <http://cjh.polyplex.org/electronics/fox/Oscillator.brd.png> >> <http://cjh.polyplex.org/electronics/fox/Oscillator.sch.png> > [snip] > > Thanks for posting the diagrams. > > Some loss in the inductor would certainly be more realistic. I hope the > inductor is not sitting right on the PCB unless you make a big hole in > the ground plane under the inductor, at least 2x the diameter of the > inductor.
I wasn't aware that much flux will escape the toroid? I'm sure I've seen this done successfully elsewhere. In any case in my prototype the toroid is floating on 2cm leads, so could flop into different places (this is the low-HF version that works fine) and at VHF I have an air coil mounted a few mm above the board. Otherwise it is going to reduce the Q like a shorted turn. As
> long the hole is big enough so that the ground plane is spaced far > enough from the turns, it is no longer a problem and actually helps with > shielding the inductor. > > I note that you are taking the output via C9 from the top of the tank, > which is a very sensitive node. I hope it is going to something with > really low loss, or that could be a cause for reducing the Q and > tendency to oscillate. I would suggest taking the output from some > low-impedance node, such as a tap on the inductor, or the top of R6 or > something like that.
I have a very small coupling cap to a BJT buffer/phase-splitter node with pretty high impedance. The simulator shows the tank tap is cleaner than the signal at the emitter, and I was trying to keep things pure and linear as possible to avoid later filtering (note it's a VCO). I'll keep that in mind at higher frequencies though, perhaps switch to a JFET buffer/splitter.
> Also not related to the present problem, I think you could probably get > rid of C6 and R7 - the DC voltage on each side of C6 is the same, > assuming the varactors stay reverse biased.
Ahh, good point. I cargo-culted that from a circuit where the tank was hot. Thanks for your thoughts. Clifford Heath.
Reply by Jim Thompson April 4, 20142014-04-04
On Sat, 05 Apr 2014 07:37:05 +1100, Clifford Heath
<no.spam@please.net> wrote:

>On 05/04/14 02:38, Jim Thompson wrote: >> [snip] >> Have you figured in the pin-to-pin capacitances of your package? DIL's >> can run as high as 4pF. > >No - but it's an SSOP. Pin capacitances? No idea.
Some of those package are as low as 0.15 to 0.2pF between pins. Add pin-to-pin capacitance to your model as a parameter and play with it to see what matters. ...Jim Thompson -- | James E.Thompson | mens | | Analog Innovations | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | San Tan Valley, AZ 85142 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.