开云体育

OK, this one is only artificially aligned with this group buy using NanoVNA to actually measure what I expect to be answered in a more "academic" fashion :)

How does MOSFET behaves in the triode region, or BJT transistor in saturation when it comes to the frequency response.

What I'm finding is what to expect in the active regions, but could not find anything when it comes to the "extreme" cases when they are in the "wide open/conductive" state.

Can I use a MOSFET as a replacement for a mechanical relay to switch between my antennas even if MOSFET's max declared small signal frequency is quite lower?

This is not the use case where MOSFET is used for fast switching - I'll turn a MOSFET on or off once a day, and then expect it conducts RF (or not).

Might be different between two - MOSFET is using majority charges, BJT is using both and is acting as fully biased diode.

And, just to justify this topic being asked at this smart group - what would you see if connected to 50ohm dummy load and used NanoVNA to measure S11 :)

Thanks!

Miro,

Even if we assume that you can bias the MOSFET properly in the presence of RF and try to use it as a switch, and even if we assume that it behaves pretty well in the ON state given its pretty low Rds, you will need to check what OFF isolation you can achieve given the device's drain to source capacitance.

Eamon | VE2EGN or AB1NK

On 11/8/22 8:23 AM, Eamon Egan wrote:

Miro,
Even if we assume that you can bias the MOSFET properly in the presence of RF and try to use it as a switch, and even if we assume that it behaves pretty well in the ON state given its pretty low Rds, you will need to check what OFF isolation you can achieve given the device's drain to source capacitance.

That will be the challenge - off isolation. (and the usual issues with biasing, just like using diodes as switches)

You might look at T/R switch literature - It's a similar problem - slow switching, high RF frequencies. You probably aren't as concerned about the gate capacitance or Miller effect, since you're going to drive it pretty hard.




Note this: "PIN diodes are offered by many suppliers of RF components. For example, M/A-COM's MA4AGBLP912 is an AlGaAs PIN diode with just 4 Ω “on” resistance, low capacitance, and an extremely fast 5 nsec switching speed (Figure 8). It can be used up to 40 GHz in a shunt configuration, with an operating bias of +10 mA for the low-loss state, and 0 V for the isolation state, using a simple +5 V TTL gate driver."

So you have a diode switch 40GHz, but the diode itself is a 50-100 MHz kind of part (5ns switching speed)

Figure 13 has S11 and S21 plots, like the ones you're thinking of measuring, eh?

On Tue, Nov 8, 2022 at 10:23 AM, Eamon Egan wrote:

Even if we assume that you can bias the MOSFET properly in the presence of RF
and try to use it as a switch, and even if we assume that it behaves pretty
well in the ON state given its pretty low Rds, you will need to check what OFF
isolation you can achieve given the device's drain to source capacitance.

Fair point - my use case is more forgiving - I'm "exploring" options to replace relay based HF antenna switch, where "OFF isolation" is marginally critical.

But back to "academical" side of the question - is frequency characteristic of MOSFET in triode or BJT in saturation commensurate to the equivalent in active regions, or extends to higher frequencies

On Tue, Nov 8, 2022 at 10:38 AM, Jim Lux wrote:

That will be the challenge - off isolation. (and the usual issues with
biasing, just like using diodes as switches)

Let's say that my use case is the one where "cross channel leakage" and "off isolation" are not overly critical (HF antenna switching).

The biasing - to be honest, in my "thought experiment" I completely ignored that my signal is AC (RF AC), so some creativity will be required or even render my scenario i possible :)

You might look at T/R switch literature - It's a similar problem - slow
switching, high RF frequencies. You probably aren't as concerned about
the gate capacitance or Miller effect, since you're going to drive it
pretty hard.

Yes, switching is slow (minutes, not nano seconds), so gate capacitance is not a big problem here

Note this: "PIN diodes are offered by many suppliers of RF components.
For example, M/A-COM's MA4AGBLP912 is an AlGaAs PIN diode with just 4 Ω
“on” resistance, low capacitance, and an extremely fast 5 nsec switching
speed (Figure 8). It can be used up to 40 GHz in a shunt configuration,
with an operating bias of +10 mA for the low-loss state, and 0 V for the
isolation state, using a simple +5 V TTL gate driver."

In antenna switching 4 ohm is a lot! Also, powers I want to run through are not of a "signal level PIN diodes" (1KW). There are PIN diodes for high power switching, but I was wandering if MOSFET can do the same

So you have a diode switch 40GHz, but the diode itself is a 50-100 MHz
kind of part (5ns switching speed)

That's exact analogy I'm asking about - MOSFET works differently then PIN diode, even BJT saturation is not the same, but at least what you are saying keeps my question alive :)

Figure 13 has S11 and S21 plots, like the ones you're thinking of
measuring, eh?

I knew my question is appropriate for this group - we just added S11 and S21 to discussion :)

On 11/8/22 12:50 PM, Miro, N9LR via groups.io wrote:

On Tue, Nov 8, 2022 at 10:23 AM, Eamon Egan wrote:

Even if we assume that you can bias the MOSFET properly in the presence of RF
and try to use it as a switch, and even if we assume that it behaves pretty
well in the ON state given its pretty low Rds, you will need to check what OFF
isolation you can achieve given the device's drain to source capacitance.

Fair point - my use case is more forgiving - I'm "exploring" options to replace relay based HF antenna switch, where "OFF isolation" is marginally critical.
But back to "academical" side of the question - is frequency characteristic of MOSFET in triode or BJT in saturation commensurate to the equivalent in active regions, or extends to higher frequencies

I've used a lot of MMIC FET switches over the years, but not actually designed them, so I can't speak to "bandwidth of the part" vs "bandwidth of the switch". I'm going to guess, based on the diode notes cited in another post, that the FET would work well beyond it's nominal switching speed.

Since you have a NanoVNA, hook one up and try it. Pick a FET with Vds and Ids bigger than you'll be using with your transmitter or receiver, and just breadboard something.



has a lot of interesting info on MMIC switches.

In particular:

has some modeling examples

And

talks about Switch Figure of Merit, with a note that SwFoM * 10 is the highest frequency it will work at as a switch.
So, FoM = 1/(2*pi*Coff*Ron)

On Tue, Nov 8, 2022 at 03:06 PM, Jim Lux wrote:

Since you have a NanoVNA, hook one up and try it.

Yep, that's the next on my list :)

talks about Switch Figure of Merit, with a note that SwFoM * 10 is the
highest frequency it will work at as a switch.
So, FoM = 1/(2*pi*Coff*Ron)

This is interesting value, will need to read a bit more, appears to represent N channel as RC comprised of Coff and Ron

On 11/8/22 1:14 PM, Miro, N9LR via groups.io wrote:

On Tue, Nov 8, 2022 at 03:06 PM, Jim Lux wrote:

Since you have a NanoVNA, hook one up and try it.

Yep, that's the next on my list :)

talks about Switch Figure of Merit, with a note that SwFoM * 10 is the
highest frequency it will work at as a switch.
So, FoM = 1/(2*pi*Coff*Ron)

This is interesting value, will need to read a bit more, appears to represent N channel as RC comprised of Coff and Ron

It's not a circuit model, just an arbitrary number to allow comparing switches.

The linked articles talk more about the theory.

On Tue, Nov 8, 2022 at 04:06 PM, Jim Lux wrote:

FET would work well beyond it's nominal switching speed.

Because the FET is not really switching things at the RF rate, the gate being DC biased; the channel is just being opened or closed by the gate and looks like a bias-dependent resistor (plus all of its parasitic elements) on the RF side.
Best regards, Don Brant

On 11/8/22 1:31 PM, Donald S Brant Jr wrote:

On Tue, Nov 8, 2022 at 04:06 PM, Jim Lux wrote:

FET would work well beyond it's nominal switching speed.

Because the FET is not really switching things at the RF rate, the gate being DC biased; the channel is just being opened or closed by the gate and looks like a bias-dependent resistor (plus all of its parasitic elements) on the RF side.
Best regards, Don Brant

In fact, that may actually help.. If the gate circuit is slow, that's great. The bane of switches is when the RF signal causes the switch state to change (a classic problem with diode switches, which is why biasing is so important). So you'd look at whether RF voltage can (capacitively) couple to the gate and move it from the desired state.


Pick a FET, use a stack of 9V batteries as appropriate to turn it on and off, and see what happens to S11 and S21.

(I seem to recall people using FETs, or attempting to do so, for spark gap replacement on a Tesla Coil - the switching speed required is slow (~100 Hz) and it's carrying a 100-300 kHz RF current.)

For a Tesla coil, one wants to be able to turn the switch off at exactly the right time.

Miro:

As far as your "academic" question goes, it's clear that any frequency specs of a MOSFET pertain to how fast it can be turned on and off, not how high frequency current they can pass. In fact, for switching type MOSFETs I don't think I ever see a frequency given in the datasheet. The high-frequency switching performance is, I believe, simply limited by the various capacitances.

What you would care about here would be
a) off isolation vs d-s capacitance, what I said earlier
b) channel to ground capacitance, if the device is large - likely not a huge issue at HF
c) lead inductance - again probably not all that significant at HF

For turning the gate on, I'd suggest using a photocoupler with a photovoltaic output like what you can find here . This is because these photocouplers have a photovoltaic output that is isolated from the control (LEDs), and the entire output circuit may may be able to surf on top of the RF; this is a much easier approach than anything that involves a direct connection.

For improving off isolation, you could put multiple identical switches in series. I would start by simply taking the device and measuring its off state, to see how many I might need to chain up.

Lastly - I was assuming MOSFETs weren't used as RF switches until I asked Google. Apparently there are such applications. Have you had a look under the Google search "MOSFET as RF switch"?

Eamon | VE2EGN; AB1NK

On 11/8/22 5:58 PM, Eamon Egan wrote:

Miro:
As far as your "academic" question goes, it's clear that any frequency specs of a MOSFET pertain to how fast it can be turned on and off, not how high frequency current they can pass. In fact, for switching type MOSFETs I don't think I ever see a frequency given in the datasheet. The high-frequency switching performance is, I believe, simply limited by the various capacitances.
What you would care about here would be
a) off isolation vs d-s capacitance, what I said earlier
b) channel to ground capacitance, if the device is large - likely not a huge issue at HF
c) lead inductance - again probably not all that significant at HF
For turning the gate on, I'd suggest using a photocoupler with a photovoltaic output like what you can find here . This is because these photocouplers have a photovoltaic output that is isolated from the control (LEDs), and the entire output circuit may may be able to surf on top of the RF; this is a much easier approach than anything that involves a direct connection.

I don't know of any PV output optos that have 10V kind of outputs, which is what you're going to need for Vgs on a big FET.

An isolated supply is fairly easy - as in an off the shelf DC/DC isolated converter. You could use it as the gate drive directly, since you're not looking for high speed.

For improving off isolation, you could put multiple identical switches in series. I would start by simply taking the device and measuring its off state, to see how many I might need to chain up.

Or a configuration that has a series switch followed by a shunt to ground.

On Tue, Nov 8, 2022 at 04:06 PM, Jim Lux wrote:

Pick a FET, use a stack of 9V batteries as appropriate to turn it on and off,
and see what happens to S11 and S21.

I'd expect MOSFET to behave in a similar way, with benefit that I have some high voltage (>400V) and decent current (>6A) handy.

Another question that came to mind - how will the the AC (RF is not DC, current crosses polarity) work with established channel - as I recall from theory, in that might be possible as there is no PN layer to go against

On Tue, Nov 8, 2022 at 07:58 PM, Eamon Egan wrote:

What you would care about here would be
a) off isolation vs d-s capacitance, what I said earlier

Once I get time to put this on my bench I hope that "off isolation" will not be a big deal given that the load on another switch is 50 ohm antenna

b) channel to ground capacitance, if the device is large - likely not a huge
issue at HF

As power dissipation will be fairly small I will try NOT to ground the case/heat spreader. Also, should not be a big deal on HF

c) lead inductance - again probably not all that significant at HF

Have that same problem with mechanical relays

For turning the gate on, I'd suggest using a photocoupler with a photovoltaic
output like what you can find here
.

Not sure that will work - Vgs needed to turn the MOSFET on is quite high

This is because these photocouplers have a photovoltaic output that is
isolated from the control (LEDs), and the entire output circuit may may be
able to surf on top of the RF; this is a much easier approach than anything
that involves a direct connection.

I'll start with AA bateries, and then if concept works I have isolated DC/DC converters just for this use :)

For improving off isolation, you could put multiple identical switches in
series. I would start by simply taking the device and measuring its off state,
to see how many I might need to chain up.

Let's see if results are OK with just one. Getting them in series will overly complicate the circuit and add more inductance

Lastly - I was assuming MOSFETs weren't used as RF switches until I asked
Google. Apparently there are such applications. Have you had a look under the
Google search "MOSFET as RF switch"?

All that I found is MOSFET as high speed switch - I need slow speed switch for high frequency circuit :)

Since this is an antenna switch, one really doesn't have to worry about timing issues. It would be a very simple straight forward design.

On 11/9/22 6:31 AM, John Cunliffe W7ZQ wrote:

There are actual parts that are designed to do this, it is called a PIN diode.Easy to use in that application, no timing issue to worry about , high reverse isolation because of low capacitance and low losses if biased correctly.Very low distortion and intermodulation.
A simple UM9415 or more modern equivalent can do a far better job than a mosfet for RF switching. I know, I know, PIN diodes are more expensive than simple mosfets but you would not attempt to replace your cars motor with a horse just because motors are rated in horse power lol.

But the FET idea is interesting, and the NanoVNA makes it possible to easily test it.

Not so expensive $10 each for the UM9415 direct from Microchip.

Miro, I would imagine that a MOSFET, with 10V or so on the gate, will behave like a low value resistor in series with a small inductance. Given the small size of the MOSFET chip, at HF the resistance shouldn't be much different from the DC value given in the datasheet (RdsON). And the inductance is that of the terminals and bonding wires. A typical TO-220 MOSFET will have roughly between 7 and 10nH. Larger encapsulations have higher inductance, while small SMDs have lower one.

The problem is how to turn a MOSFET off, for RF. Simply shorting the gate to the source is definitely not good enough. The drain-source capacitance gets very large when there is no voltage on the drain, enough so to be very far from an open circuit, at HF. And even if you could find a MOSFET with low enough capacitance even at zero drain voltage, a large RF voltage across it would turn on the parasitic drain-source diode, which turns on much faster than it turns off, so that any large enough RF voltage across the MOSFET will turn it on and keep it on.

So to use a MOSFET as RF switch, you need to bias its drain to a high voltage to turn the switch off. High enough to be comfortably above the peak RF voltage, so that even at the negative RF peak the MOSFET has enough positive drain-source voltage to keep its capacitance small enough. And even then the capacitance might still be a serious problem!

So I would say that using MOSFETs as RF switches is generally not a very good idea. It can surely be done, but it would require several MOSFETs per switch channel, a high voltage source, supply chokes or resistors and DC-blocking capacitors, and much care would need to be placed on switching slow enough so that no high voltage pulses are coupled into the signal inputs and outputs, and so on. In the end it's more complex, less good and more expensive than other solutions, such as PIN diodes (which have some of the same problems, but to a lesser extent), or even plain old relays.

If you use the NanoVNA to measure a MOSFET used as switch, keep in mind that the signal provided by the NanoVNA is small. It will probably not even begin to make the drain-source diode conduct. But a transmitter signal will!

Manfred

MOSFETs as RF switches are ubiquitous in cellular phones. You can never use just one FET as a switch because the gate-source/drain voltage is limited and will cause breakdown which is easy to realize when you have a high VSWR when the load impedance is high and more than small signal is involved. The typical way it is done in cellular is to stack several FETs and use a high value of resistor in the gate of each FET tied to the control signal. Also you do not bias the FETs with drain or source voltages, they are referenced to ground through the source/load. See the link for some details. SOI FETs are used but discrete FETs are much the same.

Gary
W9TD

On Wed, Nov 9, 2022 at 08:48 AM, Jim Lux wrote:

UM9415

Yes, with 1kW PIN that affordable I question my attempt to use MOSFET with all the inherit problems it comes with

On Wed, Nov 9, 2022 at 10:17 AM, Manfred Mornhinweg wrote:

Good points!

Worth rereading again!