Choosing a mixer for HF transceiver

Hello,

I think nice design, well described. Did you check your frequency plan? I am not sure, but if I understand the solution correctly, 1st IF (8MHz) is in the input RF band. Or is there a gap in the RF received band? If no, you should revise the frequency plan and use e.g. 1st IF above the RF band. You have to decide about the noise figure. Usually the noise figure is requested ~3dB (or less) for the VHF, UHF bands but I don't have experience with the HF band, I assume the noise figure can be higher because of higher noise background. To reach better NF, LNA shall be placed in front of mixer of course at the expense of linearity. Define what you want to reach (NF, IIP3) and calculated level diagram gives you answers whether use or not to use gain blocks somewhere in the RF chain. 

I read SA612 datasheet and noticed that it has just 5dB noise figure which is smaller than all ADI mixers. Now I am very confused.

Ok, understand, so 1st IF is positioned in the unused RF band-gap. 

zoonman said:

I think NF is more important than IIP3.

I can not agree with you. Both parameters are important and specially nowadays the IIP3 is probably more important than NF. Everywhere are different sources of interferers which increase the noise background. I want to say if you design RX with extremely low NF, the receiver will receive and amplify useless noise from different SMPS, industrial instruments, cabling, which generate EMI spectrum mainly in the HF band. As you know, the NF is above all defined by RX front-end and IIP3 by 2nd IF. 

zoonman said:

4 layers board but it is an overkill for ham radio

From my experience, the price difference is not significant between 2L and 4L boards. Especially if you don't have special requirement and you use any pool-service.

I was looking at LNAs and I cannot find anything in a reasonable price range. Probably something like BGA2866 will suit my needs.

Regarding 4 layers PCB here are the quotes:

Almost $60.

$27.

I think 2x price difference is huge. Even $27 for a hobby project is a bit expensive.

I will give a shot for LT5526 as a mixer. Also I recently learned about critical path so if my tracks are short enough impedance mismatch will be negligible.

Also I thought about RX/TX switch and I think HMC544A should get the job done. 

Yes, I was wrong with the PCB price assumption. Usually I use local PCB manufacturer and the prices are (after currency conversion)...2L:39$, 4L: 76$. So you are right, sorry.

LNA seems to a good choice, your design is really well price optimized. On my previous project I used gain blocks from OnSemi, these are also very cheap. Supply goes thru coil (e.g. for SMA3103), it can be benefit for the LNA linearity. But better way is to calculate level diagram first and search the LNA not just for price but also with known gain, NF and IIP3.

I don't recommend you to use switch HMC544A. It is a GaAs switch, so you can expect strong degradation in linearity below ~10MHz. For the RX it is not an issue, but in the TX mode the harmonics distortion will be high. 

OK, lets do some math regarding overloading LNA.
Here, in Boston we have 25kW AM broadcast station working at 1.33 MHz, I think it is the most powerful one here. This is 74dBm. Lets assume that we have perfect antennas on both ends and no losses in transmission lines. The station located approximately 5 km from me which means -48dB free space path signal loss. I used some online calculator to get the estimates.
Normally, Butterworth 3rd order BPF will have loss around 6dB in band pass region and around 40 dB in rejection region at 1.3 MHz.
So, what we have: 74 - 48 - 40 = -14 dBm.
BGA2866 has -18 dBm input power at 1dB gain compression. Given real antennas, losses in cables, connectors, switches we easily loose these 4 dB and fit into desired LNA's Pi(1dB).
Worst case, we can bypass this LNA and +10dBm for LT5526 is totally fine. That translates into about 100W for ham radio transceiver inside pass band with bypassed LNA located 5 km from receiver. Again, no losses.
If we want to have higher dynamics we have to pick LTC5562 which has +15dBm Pi. That's teasing me now :-) Maybe I should use it.
The only caveat - I am unsure how to properly design injection of audio signal for transmission path and acquire audio from second mixer.

What would you recommend instead of HMC544? Relay? My plan was to get rid of relays whenever possible and use latching ones for LPF in transmitter. I know, I still have to switch antennas using relay.

Thank you for your help, Victor! This is incredibly valuable for me.

OK, lets do some math regarding overloading LNA.
Here, in Boston we have 25kW AM broadcast station working at 1.33 MHz, I think it is the most powerful one here. This is 74dBm. Lets assume that we have perfect antennas on both ends and no losses in transmission lines. The station located approximately 5 km from me which means -48dB free space path signal loss. I used some online calculator to get the estimates.
Normally, Butterworth 3rd order BPF will have loss around 6dB in band pass region and around 40 dB in rejection region at 1.3 MHz.
So, what we have: 74 - 48 - 40 = -14 dBm.
BGA2866 has -18 dBm input power at 1dB gain compression. Given real antennas, losses in cables, connectors, switches we easily loose these 4 dB and fit into desired LNA's Pi(1dB).
Worst case, we can bypass this LNA and +10dBm for LT5526 is totally fine. That translates into about 100W for ham radio transceiver inside pass band with bypassed LNA located 5 km from receiver. Again, no losses.
If we want to have higher dynamics we have to pick LTC5562 which has +15dBm Pi. That's teasing me now :-) Maybe I should use it.
The only caveat - I am unsure how to properly design injection of audio signal for transmission path and acquire audio from second mixer.

What would you recommend instead of HMC544? Relay? My plan was to get rid of relays whenever possible and use latching ones for LPF in transmitter. I know, I still have to switch antennas using relay.

Thank you for your help, Victor! This is incredibly valuable for me.

Thank you for excellent explanation, it is very interesting for me. Nevertheless use fist LNA very near P1dB is risky, it will intermodulate for sure. You can not assume, that any amplifier is perfectly linear up to P1dB and then is hard break in the characteristic. P1dB is more important for TX design, for you is more important IIP3 on the RX.

Yes, I would recommend RF-relay. I know, it is an issue for life time, but I would go this way. I have seen nice and cheap relays from COTO, e.g. 9814-05-00TR or 9814-03-00.

(Don't thank, my English must be horrible).

You have are really good English. It's a second language for me too and we are able to communicate which is totally great! I truly appreciate your help!

I think I found what can give me really good performance.
BGA420, Pi=-6dB, 19dB gain.

Lets see what we are getting here.
From previous calculations we have -14dBm at the input of LNA which is totally fine for this one.
So we are getting +5dBm at the mixer input. Which is right under 6dBm input 1dB compression point for LTC5562.
I plan 8 crystal filter which has about 80+ dB loss outside of passband.

At the same moment we have S1 signal level -111 dBm. After LNA we have - 92dB, and if I will use AD8338, I will get -12 dB at input of the second mixer.
From the other side Pi for AD8338 is about 10dBm per Fig 23 from manual but specifications says 3V p-p which translates to 0dBm because input resistance is 1kΩ. So it translates into -20dBm after band pass filters with active LNA!
I think, this means we can pretty good grip on the dynamic.

Regarding Mixer schematic I am slightly unsure.

I drawn it like this. I have certain concerns about audio injection in TX.
Also second mixer has to do something with audio.

Almost forgot, for relays I found G6K-2F-RF, those quite affordable on aliexpress also G6K esily can be used as pin-to-pin replacement, not fancy but should work.

Hi Philipp.

Short calculation:

BGA420: OIP3=13dBm => IIP3=13dBm-19dB=-6dBm

Two interferers, power per each Pim= -30dBm (as an example)

IM3 product at LNA = 3*Pim - 2*IIP3 = 3*(-20dBm)-2*(-6dBm)=-78dBm, which higher than -111dBm wanted signal. I don't know, what you want to achieve, but this simple calculation shows, that the RX-LNA can intermodulate. 

I am working now with LTC5562, it is 7GHz mixer. It seems to me like "with a cannon on a sparrows" for the applications at 30MHz. 

If I understand the schematic, you want to use mixer for both RX and TX modes. For the AM in the TX mode you can use also  PIN-diode attenuator (or AGC circuit) instead of common RF mixer. 

Hi Viktor, so, how can I find "a perfect" LNA? :-)

I know that this mixer looks like overkill but it has a perfect dynamic range.

For TX I am trying to achieve single side band modulation. Double balanced mixer should suppress carrier signal but leave DSB, then one of the sides will be filtered by the narrow band quartz crystal filter. Depending on the band there will be either upper side band or lower side band.
I don't have plans to transmit AM/FM but it would be cool. Right now it is SSB or CW.

Are you doing any ham-radio stuff, Viktor?