Question about wideband SDRs

[mehdi]

Hi
I was thinking why some SDRs (like bladeRF and USRP B200) have lower limits (300MHZ and 70MHZ respectively)
This is the limitation of underlying RF chipset (for example LimeMicro's) and now there are wideband SDRs that support DC-3.8 or DC-6 GHZ
(like the upcoming LimeSDR)
Someone told me supporting lower frequencies introduces some design problems. So my question is whether newer devices  have problems (lower quality/sensitivity/selectivity/etc) in HF/VHF or those problems are fixed and today's technology has addressed them?
Are there any tradeoffs when designing/building wideband SDRs/receivers?
Thanks in advance

[Fank1]

Most likely limited by the local oscillator in the chip.
SDR is nothing more than direct conversion with 2 mixers and a quadrature LO feeding them.
Google YU1LM he has some easy to build designs that work well for the low bands.

[KE5FX]

Wideband (meaning HF-microwave) SDR receivers tend to work by running a local oscillator at several GHz, dividing it down as necessary to perform direct downconversion of the target frequency to baseband.   This technique works well for the most part, but it runs out of steam at low frequencies because the receiver is almost as sensitive at odd harmonics of the LO as it is at the fundamental.  If you were to extend a typical AD9361 or Lime Micro-based SDR's coverage down to the LF/lower-HF region by adding more divider stages, you would need to add increasingly elaborate front-end filtering to avoid interference from strong signals in the lower VHF region such as FM and TV broadcast, pagers, and who knows what else.

The best way to get lower-frequency coverage out of these chipsets is with an upconverter that translates the LF-HF spectrum up to a VHF or UHF IF that the SDR can handle.  That still requires filtering but at least it doesn't need to be bank-switched, the way it would be if you tried to perform direct conversion all the way down to LF.

As RF ADCs get better, it starts to make more sense to switch the antenna directly to the ADC input for low-band coverage.  But this isn't feasible yet for the popular ultra-low cost SDR dongles.

[mehdi]

...
As RF ADCs get better, it starts to make more sense to switch the antenna directly to the ADC input for low-band coverage.  But this isn't feasible yet for the popular ultra-low cost SDR dongles.

So that's why we don't have wide-band direct-conversion SDRs? (To my knowledge, current direct-conversion based systems only support HF and some parts of VHF)
If I understood you correctly, with the mass-production of advanced and high-speed ADCs, we can have HF-Microwave direct-conversion receivers which somehow fix the problem we have now (filtering, harmonics, etc.). Right?

[KE5FX]

...
As RF ADCs get better, it starts to make more sense to switch the antenna directly to the ADC input for low-band coverage.  But this isn't feasible yet for the popular ultra-low cost SDR dongles.

So that's why we don't have wide-band direct-conversion SDRs? (To my knowledge, current direct-conversion based systems only support HF and some parts of VHF)
If I understood you correctly, with the mass-production of advanced and high-speed ADCs, we can have HF-Microwave direct-conversion receivers which somehow fix the problem we have now (filtering, harmonics, etc.). Right?

Yes, that's where things are headed, although slowly.  The best 12/14-bit ADCs are capable of direct conversion up to 1.5 GHz or so in their first Nyquist zone, with no filtering beyond an LPF in the front end.  Unfortunately they cost multiple thousands of dollars per part, they have demanding power requirements, and they can only talk to JESD204B cores that run on higher-speed FPGAs.  So low-cost receivers are going to have to rely on mixing for a while until the mass market catches up.

[C]

The radio's dynamic range is what you need to look at.
Converting the RF to digital is not the problem. The problem is the dynamic range you have to work with.

A radio is doing two things
Increase the amplitude of wanted signals.
Decrease the amplitude of UN wanted signals.

When you have broadband input to a mixer or adc, how do you do this?

If you have a bandwidth of just 3x to the wanted signal you could have a very powerful signal on each side of the wanted signal. You need to limit the gain so that all three signals do not overload the mixer or ADC. At the same time you still need the wanted signal to still be present.

With a radio you do not know what you are missing until some other radio received what you could not but would have if your radio was better.

[rfbroadband]

One key issue is the generation of the LO signal. The first limitation is the tuning range of the VCO, you will have to use a varactor with multiple capacitor banks to maximize the tuning range of the VCO and that of course has its limits (think about a LC tank in a VCO, L can't be tuned and there are limits on what you can do using switchable capacitor banks).

Most of these chips are direct conversion transceivers, thus you need to generate quadrature LO signals (I,Q) over a wide freq. range. The wide freq. range prohibits  the use poly phase filters to generate IQ signals, so typically you follow your PLL output with a muxed divider section at the output. If you need 6GHz you would have to use a 12GHz signal, divide it by 2, to get IQ.

Thus, the closer you want to get to DC the more insane the divide ratio becomes unless you start integrating multiple PLLs and mux them.

The RX, TX signal chain is not that difficult to design for large BW so NF, IP3, IP2, gain requirements are not 'that difficult' (of course that's a 'relative')  to meet.

That's a short, somewhat general explanation.