RF direct conversion and input impedance in PCB

[arlen]

Hi everyone.

I`m a little bit confused about the input circuit requirements for a RF signals (RF direct conversion).

When i use an oscilloscope to VISUALIZE signal, I use the scope with 1Mohm and 10 OHm for signal with a low frequency and high amplitude, and  for signal with higher frequency and low amplitude i use scope with 50 Ohm to avoid some signal distortion. But how about to DIGITIZE RF signal?.

So far I've been using ADC keeping in mind that the input impedance of ADC have to be much more higher than output impedance of the signal source in order to capture a very close value to the signal that I want to digitize. But what about digitizing RF signals?.

I've read that one consideration is the sampling frequency and another is the Full-Power Bandwidth. But what about PCB requirements? ,input circuits or elements  before ADC like  filters or amplifiers have to be according to 50 Ohms of impedance for RF signal?Once the signal onto input pins of the ADC alwas is with an 50 Ohms impedance (for single mode) or 100 Ohms differential impedance?

I started to wonder about it since I saw the Red Pitaya Fast_analog_inputs_sch (It is attached), where in the RF input specification says that it has input impedance (1Mohm // 10pF), but in output impedance has 50 Ohm.

I hope someone could help me to understand this .

Thanks in advance

arlen

[Howardlong]

The RedPitaya output impedance is next to nothing irrespective of what the documentation might say. You can measure it by generating a signal, say 1MHz, and attaching a scope (even the Red Pitaya's) in Hi-Z and a decade resistance box in parallel. Adjust the resistance box until the voltage is half what it would be without the resistance box. The resistance box value will now tell you the output impedance.

For the input impedance to be 50 ohm you can use a 50 ohm through terminator, or even solder a 50 ohm resistor directly across the SMA connector.

You don't say what frequencies you're interested in. For HF which is largely what the RedPitaya is all about in SDR terms (or as an IF), things are going to be reasonably forgiving within reason on PCB layout.

At Hf I am sure it will work even without properly terminating, it just won't be quite as good as it could be.

[T3sl4co1l]

Read the ADC datasheet. They typically provide example circuits for AC and RF coupling.

If you look inside the 'filter' block, you'll probably find what you need towards the ADC side of that subcircuit.

Protection, depends.  Back-to-back diodes may be a good idea.  But you will reduce the dynamic range in the process (overload characteristics around large signals, lots of IM3).

Tim

[Howardlong]

Tim

There's a whole bunch of documented and undocumented stuff in front of the ADC. I may have misinterpreted, but I interpreted it that the OP was going to use the Hi-Z buffered input provided (1Mohm || 10pF)

[T3sl4co1l]

I think what he's talking about is, he wants to do RF into the ADC, but hasn't seen a relevant example.  He's listed examples that he has seen, which are wideband and heavily nonreciprocal* in design, so aren't really ideal for dynamic range / SNR.

As for what the ADC input looks like, it should be a lossy capacitor.  You will need to provide balanced input (usually from a transformer / balun at some point), with a fixed common mode DC bias (usually available on an ADC pin), and at high frequencies (where the capacitor's loss dominates), it will terminate okay, but at low frequencies (where the capacitor dominates), you'll need to provide an external termination resistor.  So, probably, you'd use an R+L across the ADC IN_P/IN_N pins, where R is the termination and L is selected to complement the ADC input's RC loss part (thus, keeping input impedance flat over a wide bandwidth).

*Nonreciprocal design is any building-block method where there is "no" loading effect, of inputs, upon outputs.  The typical examples being op-amp circuits (an ideal op-amp has a voltage source output, so an unlimited number of loads can be connected without attenuating the output signal), and digital logic (an output pin driver is capable of driving up to N input pins, with little or no increase in propagation delay).

Nonreciprocal design is typically undesirable for RF circuits, because amplifiers are used to implement the nonreciprocity, at the expense of noise and bandwidth.  You typically do not use op-amps, because they have too much gain, the gain varies strongly with frequency (i.e., you're near fT, where the op-amp looks like an integrator), and the noise is higher than an MMIC or couple-transistor LNA.

Tim

[Howardlong]

On re-reading I think you are right.

The Red Pitaya may not be the best example for an ADC driver for RF as its inputs are primarily designed to be used with a Hi-Z scope probe rather than 50 ohms.

[arlen]

Hi
Thank you very much for your replays , Your answers helped me a lot to understand and showed me a line to continue with the matter,
I know there's a lot of information, but if someone has a book which could help me to begin in this topic it would be great

I have a question about the next text :

Nonreciprocal design is typically undesirable for RF circuits, because amplifiers are used to implement the nonreciprocity, at the expense of noise and bandwidth.  You typically do not use op-amps, because they have too much gain, the gain varies strongly with frequency (i.e., you're near fT, where the op-amp looks like an integrator), and the noise is higher than an MMIC or couple-transistor LNA.
Tim

I don't understand the part "the noise is higher than an MMIC or couple-transistor LNA", pls could you explain it in more detail ? In What applications MMIC (Monolithic microwave integrated circuit) should be used?

Thanks in advance

arlen

[T3sl4co1l]

Just that the noise figure and bandwidth (but not necessarily flatness) will be better for those, instead of an op-amp.

Tim

[dmg]

Input impedance of ADC's that can do tens to hunreds of Megahertz is typically in the range of several kOhm in parallel with low figure picofarads (say 5 pF or so). So it's high at low frequencies and goes low as frequency increases.

To properly drive these kind of ADC you basically exit your controlled-impedance RF section, typically (but now always) at 50 Ohm single ended or 100 Ohm differential. Then you place your main anti-alias filter and after that your ADC driver amplifier. ADC driver amplifiers are fast amplifiers, usually current feedback ones if you need high bandwidth. It's very very important to match the input of the amplifier to the output impedance of the RF section so that the anti alias filter is between a fixed and known impedance environment. The input impedance of the amplifier is typically fixed with the feedback resistor networks and careful layout is required to avoid parasitic capacitance, more so if you're dealing with current feedback amplifiers because they can oscillate.

After your ADC amplifier you stop caring about power transfer and matching and you start caring about the voltage of the waveform which is what the ADC's gonna digitize, so instead of aiming for power matching you aim for voltage matching, i.e. having the lowest possible output impedance at the ADC amplifier to drive the ADC with negligible voltage attenuation. Most ADC amplifiers have output impedances lower than 20-25 ohm. You can't use lossy transmission lines from the ADC amplifier to the ADC like oscilloscopes do to avoid reflections and stuff so you have to keep these traces electrically short and usually a low Q lowpass filter in between helps a lot. That's even more important if you're gonna sample past the first nyquist zone.

[uncle_bob]

I don't understand the part "the noise is higher than an MMIC or couple-transistor LNA", pls could you explain it in more detail ? In What applications MMIC (Monolithic microwave integrated circuit) should be used?

Thanks in advance

arlen

Hi

Op amps are wonderful things when you need to have a response that includes DC in your signal path. They have a number of limitations as well. You *can* (but don't always) design an "RF  only" amplifier to get around many of these limitations:

1) Bandwidth that includes multiple GHz, easy with an RF part, tough with an op amp with feedback.

2) Good noise figure (sub 1 db) out to many GHz ... essentially impossible with a feedback op-amp.

3) Third order IMD intercept points that are into the 30 or 40 dbm range. Again tough with an op-amp.

That's hardly a full list, but it at least is a start. It's not to say op amps are junk, just that you can do a purpose built circuit that does better. Since ADC's *can* have very good performance in some areas, you want a good amp in front of them if you use one.

===

Most ADC's switch a capacitor onto the input line to "sample" the voltage. At the time of the switch, this cap looks like a short circuit. Since the switch has some resistance (10 ohms maybe) it's not a dead short. Once the sample time is done, the switch opens and you have a high impedance (normally capacitive). If you are sampling at a hundred MHz, the input goes from short to open a lot of times a second. The net result is a really weird input impedance for the ADC. The only way most outfits address it is to spec it with a certain "matching network" attached. That network often is the dominant contributor to the load on the amplifier rather than the ADC it's self. If you mess with the circuit, the ADC does not have the proper "low impedance source" when the switch closes. That messes up the sampling process. The net result is a degraded performance level on the ADC ....

Bob

[dmills]

Have a look at the schematics of the Hermes direct sampling SDR from TAPR, readily available on  line and it has a fairly typical attenuator/filter/driver architecture for a band limited 50 ohm input using parts for which datasheets are available.

Regards, Dan.