How big a deal is third order intercept point going to be?

[Xnke]

Working on an updated mixer/filter board for my SI5351 synth based transceiver.

Currently, I have an MC1496 gilbert cell mixer for the 1st converter down to 9Mhz IF, that gets switched to modulate audio up to 9mhz on transmit. Switching is done by small DIP style relays.

From there, a 9mhz diplexer terminates the mixer, and another relay switches the 8-pole crystal filter. An MC1350P IF amp is used, and drives another MC1496 gilbert cell mixer for transmit output and product detection.

There is no LNA in front of the 1st converter, just the bandpass filter.

I have options-all of which have been shown to work well for decades, just as the gilbert cell mixers have.

Option 1: Leave this alone. It works. But the mixer's third order intercept point is about +15dB with 8mA through the gilbert cell. It's much lower at the datasheet's 1mA, and I can get it up to +20dB if I bust the specs at 13mA.

Option 2: Switch to the minicircuits RMS-2 diode ring mixer. IIP3 is typically +17dB according to the datasheet.

Option 3: Switch to the H-mode mixer with FSA3157 SPDT silicon switches. With careful tweaks,  +35dB IIP3 or more can be achieved.

Now, it seems that I'd need to upgrade my IF amp to something better if I want to really push this IIP3 thing. The MC1350P is pretty good but I have no idea what it's IIP3 performance looks like, or how much it really matters.

How much does this really matter? I know it's directly governing the dynamic range of the receiver, and that's always a positive thing to have, but in reality, how much is that going to matter? Should I damn the torpedos, and build for +30dB IIP3 through all stages of the transceiver? Or is +15, +18 good enough for the girls I date?

[vk4ffab]

I am from the AVE school of thought. GOOD ENOUGH FOR THE KINDS OF GIRLS I GO OUT WITH. In other words, if it works and does what its meant to, do not fix it until its broken. Don't over think things.

Now the fan boys will tell you that you are doing it all wrong and you need this and that, when in fact what you have is perfectly adequate for your situation. If you have no strong strong signals near by that are causing issues then dont bother with changing anything. Also, by changing to a ring mixer you will than also need to add 15x gain back into the chain that the Gilbert Cell is providing, you will probably need to amplify and buffer the LO signal because the SI5351 is marginal for 7dmn mixer as well as port termination matching on the mixer. In other words, you will be changing a lot of stuff for what might be marginal improvement. Sure you can wank on that your IP3 is better, but is it really fixing anything in your current working receive?

Rob. 

[Xnke]

Considering that I'm 490 meters from the next ham, I essentially can't operate when he is. He's also a former CB'er and subscribes to the power at all costs, damn the regulations mentality, so my equipment needs to handle his 4kW PEP blasts of "CQ CONTEST" for hours at a time on HF.

The current thing works, but receive is just OK. Let's say I'm on 14mhz, and he's decided he's gonna operate on 14.2 today, well I might as well turn it off. I can still hear some strong signals, but you can literally hear him fire up just by watching the S-meter-you'll hear it hit the pin! 

That's the only reason for pondering dynamic range-although I don't really thing it'll help much either. If I'm operating on a lower frequency than he is, I'm *usually* ok, but not always.

[vk4ffab]

I would start by beafing up the IF filter, you really should not be hearing him accross the whole 20m band or within say 50khz of his tx frequency. Steeper skirts and deeper stopband will help out a lot. I have 4 guys closer to me than that and i dont hear them unless im trying to sit 2khz away.

[PA0PBZ]

I would start by beafing up the IF filter, you really should not be hearing him accross the whole 20m band or within say 50khz of his tx frequency. Steeper skirts and deeper stopband will help out a lot. I have 4 guys closer to me than that and i dont hear them unless im trying to sit 2khz away.

It depends on what signal he (the neighbor) is producing, if he is over-driving his PA and producing a 100KHz wide signal there's nothing you can do on the receiver side.

[vk4ffab]

I would start by beafing up the IF filter, you really should not be hearing him accross the whole 20m band or within say 50khz of his tx frequency. Steeper skirts and deeper stopband will help out a lot. I have 4 guys closer to me than that and i dont hear them unless im trying to sit 2khz away.

It depends on what signal he (the neighbor) is producing, if he is over-driving his PA and producing a 100KHz wide signal there's nothing you can do on the receiver side.

This is true, but in that situation you can get the authorities involved to solve that part of the problem. But even so, 1/2 a Km is some distance, so you should be able to operate in the CW portion when he is operating SSB assuming the filter situation is up to the task. If the OP is using the typical slightly wide 4 pole crystal filter most of us use in our homebrew receivers, well, its not going to be up to the task, he will need something more like 8 poles or even 10 and to be quite narrow 2.2K for SSB to get the rejection he needs for that strong adjacent signal. After that i would worry about intermod in the first mixer.

[Xnke]

8 pole crystal filters. KVG X-9 S-44, S-43, and S42 (The telrad filters)

When someone is pumping 4kw of overdriven CB linear amp, he's whirlwide, ya know. And no, there are no authorities going to come to his house, we've tried for years. Every year, two or three times, someone local goes over and pins his coax when he pisses someone off enough. That helps for a few weeks, till his payday comes and he buys replacements.

The current filter arrangement is selection of a 7.2khz AM filter, or a 3.6khz filter for USB or LSB. LO is 9.0Mhz, so the IF falls centered on the AM filter, or at the edge of either sideband filter.

[vk4ffab]

8 pole crystal filters. KVG X-9 S-44, S-43, and S42 (The telrad filters)

When someone is pumping 4kw of overdriven CB linear amp, he's whirlwide, ya know. And no, there are no authorities going to come to his house, we've tried for years. Every year, two or three times, someone local goes over and pins his coax when he pisses someone off enough. That helps for a few weeks, till his payday comes and he buys replacements.

The current filter arrangement is selection of a 7.2khz AM filter, or a 3.6khz filter for USB or LSB. LO is 9.0Mhz, so the IF falls centered on the AM filter, or at the edge of either sideband filter.

How hard would it be to reconfigure your filter stackup? 3.6k is wide for SSB, I do not use anything wider than 2.2K. Anyway, if possible can you move the AM filter and put it ahead of the SSB/CW filters and use it as a roofing filter. So your stackup would look something like this:

BPF->RF Preamp/Attenuator->MIXER->Roofing Filter->IF AMP->SSB/CW Filter->IF AMP->Mixer->AF Stages

If we assume that these filters give 80db (I cant find a spec sheet) of attenuation im not sure of their shape factor so will ignore it, you now have 160db of attenuation 7khz away from your operating frequency. If you can find a spec sheet and shoot for 200db of filtering in your IF, it wont matter what he does, you will sit 10khz away from him on any band and not hear a peep from him. Well, assuming the first mixer is up to the task, but then, you will know the mixer is the issue and not the filters.

I am not sure the telrad filters are all that good with respect to shape and stop band, I do recall reading about someone cutting them open to remove the 9mhz xtals from them to make filters with a better shape factor and deeper stopband. I would have to google and find the source again, so dont quote this as being true. If you have a VNA I would not mind seeing a plot of the filters to compare against my homebrew ones.

[Georg - PY5ZSE]

IMHO what you experience with your neighbour's strong signal is not intermodulation, but blocking / reciprocal mixing.

To generate IM, 2 signals are needed who's fundamental or harmonics combine to a mixing product that falls into your reception bandwidth, i.e.

n*f₁ +- m*f₂ = f_r,  with n, m = 1, 2, ....

For third order IM products, usually n=2, m=1 and sign = -. The typical IM situation is 40m at night, where the strong broadcast signals directly above the ham band result in a multitude of intermodulation products that make the band feel "full".

You have only one strong signal, but a local oscillator with a lousy phase noise performance. What happens, is that your mixer "reverses" it's function, the strong signal works as LO and transfers the part of your LO's noise sidebands that is f_r - f_i away from your LO's carrier into your reception channel. What you observe, is not an identifiable signal, but just that the noise level of your RX goes significantly up when the "nice guy" transmits.

The SI5351 is a clock generator. It never has been conceived as a local oscillator in high performance receivers, und thus it's phase noise is not specified. The only value concerning short term stability that appears in the data sheet is "jitter < 70 ps"

While it is rather straightforward (albeit tedious) to calculate jitter from a given phase noise plot, the inverse is not possible, or only under a lot of assumptions (corner frequencies of the phase noise plot's segments). One on line calculator https://www.changpuak.ch/electronics/phase_noise_jitter_conversion_reverse.php I found converts the 70 ps into -107 dBc at 100 kHz. Considering what I just mentioned, I read this as "somewhere between -120 and -95 dBc.

An AD9951 DDS chip, run from a clean clock source, should at 23MHz have a phase noise of around -150 dBc, -145 at least.

To find out, if what I briefly described is the source of your problems, disconnect your SI5351 oscillator from the mixer's LO input, and replace it with the output of a known low phase noise signal generator. An old HP8640 will perfectly do. If you are then able to receive weak signals some 200 kHz beside your neigbour, you should seriously rethink using the SI5351.

"normal" short wave receivers have Mixer IIP'3's somewhere between 10 (rather bad) and 30 dBm ( rather good). So, when you get 20 dBm out of your MC1496, you are somewhere in the middle, and that's not too bad. Actually, I am surprized this chip is so good, I still have a couple of them laying around...

With H-mode Mixers IIP3's of over 40dBm are possible. To achieve this in a receiver, not only the mixer itself on the test bench, EVERYTHING must be carefully selected. Besides the mixer, these are: The input band selection filters, the roofing filter, the diplexer terminating the mixer, up to the IF strip to avoid the in-channel IM behaviour being much worse than the out-channel values. You probably know martein, PA3AKE's work on this topic, if not, here is the reference:https://martein.home.xs4all.nl/pa3ake/hmode/

[Xnke]

Here's a plot of the USB filter from the Telrad set:

http://f5mww.blogspot.com/2017/08/testing-kvg-filter-with-vnwa.html

https://4.bp.blogspot.com/-UaauMYnDpwM/WYcq0qeSu7I/AAAAAAAAGeE/d_N-yrRw7vkSHtWr9LSKn3mPu30lqrA_gCLcBGAs/s640/60dB.PNG

I have been considering putting the AM filter in as a roofer-but I was concerned about total IF gain needed. I guess more AGC-controlled stages would mean wider usable dynamic range, but also a higher noise figure. Trade-offs all around.

As to the phase noise-no. The SI5351 isn't bad at all, compared to some other options. Both the SI5351 and the SI570 are PLL clock sources, not DDS, so any phase noise issue is mostly determined by the division chain needed to create the output. The SI570 only allows integer division, while the SI5351 allows integer and 20-bit numerator/denominator fractional division. The 70ps jitter is measured at the worst case condition, using a 25Mhz external non-TXCO clock, followed by a fractional only, multiple-step divisor chain. By using the integer-only divisor chain methods, the jitter is very, very much reduced. The phase noise resulting isn't as good as a crystal, but it's not much different from any of the Analog Devices parts when properly programmed.

Now, I am aware that the SI570 is actually spec'ed for phase jitter at 0.3ps. But, when asking Silabs about it, I was sent the following plot below:



In the plot, the SI5351A, properly programmed, shows a jitter of 2.7ps, vs 0.3 ps of the SI570. Yeah, the SI570 is better. But 2.7ps isn't NEARLY the 70ps shown as the worst possible condition in the datasheet. And at the 9Mhz IF frequency, using integer only division from the 27Mhz TXCO used in my radio, I should see phase noise in the region of -134 dbc/hz. Also note the spikes in the phase noise plot above-those are frequencies where integer-only division isn't possible, and fractional divisors are used. Big difference!

Also, the Elecraft KX2 uses it, so it can't be *that* terrible. The KX2 is a decent radio, IMO.

One other thing that is often neglected with regard to phase noise and the SI5351, is that there is some trickery involved in getting multiple outputs on the same chip at the same frequency to maintain an accurate, constant, phase offset (like if you needed two outputs at 9Mhz in quadrature, for example!) you have to make sure you do a PLL reset:

1) Every time you change the *MultiSynth* divider (whether integer or fractional). Not the PLL feedback divider, which is opposite to the documentation...

2) Every time you switch the outputs on/off using the Clock Control registers e.g. Register 16. If you are about phase relationships, INCLUDING if you have simply set a 180-degree phase relationship to another clock output using the clock invert bit in the clock control register, even then, you have to do a PLL Reset.

If you don't, then the phase offset will be random-even though the frequency is accurate.

which brings up another point-I have considered using the SI570 as the "tuning" LO, for use with the first mixer on receive and the final transmit mixer, and using a 5351A as a quadrature generator for I/Q modulation and demodulation schemes. Haven't gotten that far along yet-there is much to improve on my more traditional radio first!