Explanations not needed, I've done all this before. Or, not sure how familiar you are with my posting history and topics of interest, but, suffice it to say I know how to design a network?
Some concerns:
It is too low for the 3.5MHz band (OP is experimenting at) because when creating a transformer (or a choke) you would need a lot of turns, that increases the parasitic capacitancies (creating issues like lower Q, self resonance somewhere within you freqs of interest etc).
Capacitance is irrelevant
at a given center frequency Fc until such point as the parallel-resonant (impedance peak) frequency drops below Fc. The impedance just keeps going up and up until then.
Whether one is interested in harmonics, is another matter, but generally in radio, it's desired to filter those out, for which the capacitance would even seem desirable.
(My experience is primarily SMPS and wideband, and SMPS is generally quite wide. Last year I characterized a PFC choke up into the 200MHz range; modern SJ MOSFETs happily excite ringing this far up, and following such ringing through the circuit is critical for EMC emissions. Don't let the "10s of kHz" fool you
)
I would say wideband network design is far more comprehensive than narrowband, but NB design can still be done as a primary path; just be sure to check for spurs once in a while.
As I wrote above the rule of thumb in this business is to have the inductive reactance of primary/secondary windings (the source or the load see) at least 5x higher than the impedance/reactance of the source/load at the frequency of interest (or at the lowest freq of interest when we talking the broadband transformers)..
You're defining a Q less than 0.2 already; I don't see what the point is. That is, for XL > 5 RL, the system-embedded inductor Q is less than 1/5
by definition. That's pretty damn low to start with. The Q merely reduces to 0 if the core is 100% lossy, in which case efficiency drops by merely a few percentage points / insertion loss goes up a dB or so.
Unless you're doing a high-power transmitter on battery power and have extremely high pressure to optimize battery life... it seems a difference unlikely to matter? Or compact parts where power density is high, so that temperature rise would be an issue.
Indeed the rule is most often applied to pulse and isolation transformers, baluns, etc., where the
signal loading due to core loss should be kept acceptably low, i.e. insertion loss of a dB or so. Reactance isn't important, just that the transformer can be treated as a more-or-less ideal transformer over the frequency range of interest. It's a rule of expediency -- or perhaps laziness, but one must not be completely lazy and forget why such a rule exists, under what assumptions it was created, and should be applied.
Ferrites are often predominantly resistive in the VHF range, so the raw signal loss is important. Inductance is irrelevant or inapplicable in that case. Note that larger and higher-mu ferrites can be resistive even down into the MF range, i.e. 1MHz or below.
Only the low-frequency cutoff case is given by XL and RL, and whatever dB attenuation/reflection is acceptable there.
Powdered iron cores are typically inductive up to some MHz, except for the worst (#52 and #26) types of course. They would seem a suitable basis for an RFC here, and Micrometals' publications concur.
You wouldn't normally choose a powdered iron core for a transformer
per se, because of the low inductivity, but you would when the DC bias is considerable, and making it a transformer isn't abjectly worse than doing a parafeed connection with a ferrite transformer. Or putting it another way: a single-ended tube amplifier output transformer, is still a transformer even if its effective core permeability is ballpark 60 or so. It just means you don't get the same LF cutoff you would for the same number of turns on an ungapped push-pull core. But that push-pull core saturates at some 10s mA if you try using it single-ended, it ain't gonna work, you have to gap it.
OP writes he is using material 67 which is better for higher bands like 14-50MHz, afaik.
For the lower frequencies like the 3.5MHz the material 43 (mu'=850 in his toroid size) is better from my experience.
Assuming of course, it doesn't saturate; typical toroids used in radio saturate in the couple to low tens of At, fine for signal currents, but again, the transistor in question can draw several amperes with very little base bias, so it's worth asking the OP what their bias actually is.
PPS: except the mu' (function of frequency) there is the mu'' which represents the losses (function of frequency) - that is a parameter amateur radio operators/enthusiast are interested in too..
Strange you aren't aware of the applicability of complex permeability to SMPS inductor design. CMCs too!
In fact, I've used an alternative expression of it, the generalized Steinmetz core loss formula (which normally fits ~okay~ (+/-20%?) to common power ferrites), to estimate (via a single closed expression, amazingly enough) the number of turns and air gap required for maximum Q on a given core, for given value and ratings. The aforementioned PFC choke was calculated to have a Q around 350, and was measured a bit over 300 at 100kHz; but still got too hot for comfort and a larger design had to be chosen! Alas!
Tim