EEVblog Electronics Community Forum
Electronics => Beginners => Topic started by: rteodor on August 20, 2022, 08:36:43 pm
-
I've had this naive attempt to design and build an FM radio in 2010. Now I want to learn what I can from its failure.
The LO PLL built with LM7001 works in 80-120MHz range (varicap 3-20V) but the Colpitts output is more or less messy.
[attach=1]
First of all I could not find any information if Colpitts oscillators should put out so much harmonics. Does this spectrum for 100MHz look normal ?
[attachimg=2]
Second thing: the phase noise looks terrible (not to mention that frequency is a little bit bigger than what is programmed in LM7001).
[attachimg=3]
Overall along the 80-100MHz range the output is
[attachimg=4]
But at some frequencies weird things happen:
[attachimg=5]
I am especially interested for an answer to #1 (harmonics). The construction is unshielded SMD and not to pleasant to look at.
-
Mosfets with voltage divider biasing as you have are not a great solution. The gate threshold and turn on voltage vary wildly and can't be biased like bipolars in general. You need to adjust bias on these devices (with pots or selected resistors) or better yet with stabilized op-amp or DC feedback type biasing. I would start by looking at the DC levels and bias currents, they're probably near railed and creating a bunch of clipping and distortion. A much better approach would be to use bipolar Q's in a similar discrete cascode type configuration. Something like an MPSH10 has a lot of bandwidth at VHF and is very low cost and available. The biasing for this stuff as is jacked I'm affraid.
-
I don't like the biasing of the BF961 either. The DC operating may or (more likely) may not be correct, typically only one gate would have a bias voltage applied in this kind of mixer circuit. A bigger sin is that there is no decoupling in the bias dividers so any noise on the supply rail is being injected straight into the gates.
-
Also the switcher (the varicaps bias voltage) needs better filtering, imho, I doubt you can filter its noise out without a CLC or something like that..
PS: the simple Colpitts could have that spectra.. Its output is never a nice sine, afaik..
PPS: you may try to feed the mixer from the hot side of the R5..
-
There was an extremely interesting thread on the Peltz oscillator some months back. I suggest you read it, the thread is here:
https://www.eevblog.com/forum/projects/simple-sinusoidal-oscillators/ (https://www.eevblog.com/forum/projects/simple-sinusoidal-oscillators/)
Cheers.
-
Mosfets with voltage divider biasing as you have are not a great solution.
I don't like the biasing of the BF961 either.
The MOSFET amplifier and mixer were next in line after the VCO, but now that you noticed it...
Neither ever worked and one thing I kept in mind about them is that they might not work with such a low supply voltage.
They were biased in DC (Q1: Vds = 3.49V Id = 12.2mA; Q2: Vds = 4.12V Id = 6mA) but they only worked as attenuators :)
Today I put a higher supply of 15V, changed R27 to 1K2 and supplied Q1 gates from battery variable power supplies (did not have any potentiometer). Set input signal to 100MHz / -30dBm then I made a sweep of gate voltages.
About 400 documented measurements later the best I could get was ~2dB amplification out of Q1.
There is an error in the schematic that the gates were inverted so the signal was injected in reality from G1. But I do not think that would make a significant difference.
So yeah, it would be much easier with bipolars. But a little itch still remains: how could I made it work with MOSFET's ?
the simple Colpitts could have that spectra.. Its output is never a nice sine, afaik..
This was along the answer I was seeking: some real life spectrum analyzer image of a proper Colpitts VCO output.
In the meantime I found something in this video https://odysee.com/@AllElectronics:d/microwave-1.7ghz-vco-oscillator:a (https://odysee.com/@AllElectronics:d/microwave-1.7ghz-vco-oscillator:a)
It settles for me the question about harmonics but not for spur and phase noise.
-
I didn't notice that the first cascode was backwards- of course. I would expect a voltage gain of about 1.5 from that first stage, 100 ohm load and 15 mS transconductance. Kind of defeats the purpose of a cascode- there really is no miller multiplication. The power gain could be large though, input impedance is very high. Calculate source impedance. This is value of FET's in front ends. They also have low noise
As far as how to bias with FET's in general. You setup a loop where the drain current rises to satisfy a gate voltage- sometimes called "self biasing". Choose a drain current less than Idss The BF961 is a depletion mode device, it conducts with no gate drive. It has a zero gate drain current of about 8 mA. You DC ground the gate (a large resistor will do since gate current is zero), you put an R in the source lead to give you something reasonable- say for 1 volt, you would use 125 ohms. The BF961 comes in three grades of successively tighter Idss, the cheap ones don't give you much guarantee of your min Idss hence gm. You take the cascode gate to 4v and bypass. You usually also bypass the R in the source. You then get a drain load that give your final voltage gain- gmR. Look at Art of Electronics- they are big fans of JFET's but depletion mode MOSFET's are biased the same and operate similarly.
For enhancement mode devices, you do similar things often using a bipolar Vbe as a voltage standard to set the drain current with a DC loop. You'll see circuits with a bipolar in the source lead circuit with an R dropping a sample of Id across their Vbe, the collector is then used to pull on a bias network to keep the drain current at some DC value.
You can also use current sources to force a drain current and let the gate voltage fall where it may.
Hams and other hobbyists lament the fact that dual gate MOSFET's are uncommon today. They really aren't great parts and there are usually better (lower noise, higher efficiency, cheaper) ways to do things today. There are some clever things they can do like make mixers and VHF gain stages but the performance of these dual gate circuits isn't that stellar compared to other alternatives. Their noise at low frequencies isn't great either though is decent at VHF. JFET's still have some untility- again see A of E.
I know this is experimentation and you've got a pile of these things, I understand. I have a bunch of old LM10 op-amps that I got for cents on the dollar. They're not really stellar but I force them to work in hobby circuits. One amazing trait they have is that they'll operate from a 1.1v supply. Widlar had app notes where he ran them floating on the Vbe drop of a darlington to make a HV supply. Pretty esoteric.
Have Fun.