Little wobler idea, very wideband mixer design

[Yansi]

Hello!
me and my RF dumbness again! 

Been thinking about building a simple wobler with rather high frequency span (up to 4GHz or so), however it turned out to be not so simple. Here is my idea:


To conserve simplicity, a single wideband synthesizer is used as LO. Let's say it can do 30-4000MHz. But I want the wobler to go from "DC" (say 0.1MHz) up to whatever it can. So a mixer is added to down-convert the range by the LO min frequency.
Then I need a fixed IF to be able to filter and log detect on, so another mixer is used on the receive port to convert the LO to fixed IF equal to the XO (= LO min) shift  frequency.  30MHz is pretty nice for a log detector with preceding filter.

Turns out this topology it seems to have some nasty drawbacks:

1) TX port contains two products 2*XO apart (LO +- XO). Not usable as a standalone RF gen, cannot be filtered out easily if at all.

2) Both mixers have to be ultrawideband:  MIXer 1 needs an RF port capable of almost DC (but full RF) and two other ports working on low (tens MHz) frequencies.
MIXer 2 needs also a port working from DC to the max RF.
I think this may be a showstopper as I do not know if such mixers could be ever designed to work good enough.

Usually MMIC mixers are designed so that the IF port works from DC up to few GHz at most (but really not 4GHz), while the RF and LO ports are somewhat limited at the GHz range, not being able to go anywhere near tens MHz.

Do mixers exist or are there reasonable implementations of mixer for the above requirements? 
Or is there any other simple approach for such instrument, that would not have such crazy requirements for the mixers?

Thank you for any insight or ideas,
Y.

//EDIT: Just noticed a fatal flaw: Both mixing products at the TX port will mix back to the output IF. That is unfortunately not usable, at all. I need to come up with a different solution.

[rfeecs]

You could have your XO at 4000MHz and mix down instead of up, filter out the upper sideband.
But then with your block diagram you would have a 4000MHz IF.

For the mixers, the RF/LO bandwidth is usually limited by the bandwidth of the baluns.
If you have a fully balanced RF, LO and IF without passive baluns, then you don't have the lower frequency limit.  So yes, those mixers are possible.  I don't know off hand a part that you can buy off the shelf, however.

[hendorog]

For the mixers, the RF/LO bandwidth is usually limited by the bandwidth of the baluns.
If you have a fully balanced RF, LO and IF without passive baluns, then you don't have the lower frequency limit.  So yes, those mixers are possible.  I don't know off hand a part that you can buy off the shelf, however.

Something like this perhaps:
https://www.digikey.co.nz/product-detail/en/peregrine-semiconductor/4140-52/1046-1021-1-ND/2614525
http://www.psemi.com/pdf/datasheets/pe4140ds.pdf

[Yansi]

That is an interesting part number. Never seen such.

However the ideas above should be probably scrapped, due to the fatal problem with the mixing products being to close to each other.

[T3sl4co1l]

You can strip one sideband (sum or difference) using quadrature mixing.  But that's hard enough for SSB (usually solved with DSP!), let alone over two decades of bandwidth.  It's also a nulling method, so you'll always have some residual.  Getting better than -40dB on the undesired product is a challenge.

Tim