Homebrew spectrum analyser high-impedance probe

[philpem]

I'm not sure whether this belongs here or in Test Equipment, but here goes anyway

A while ago I picked up an old Advantest R3361A spectrum analyser. The old girl works a treat, despite being near enough twenty years old now. Or rather - she works a treat when probing anything which provides a standard 50-ohm output impedance. I'm sure you can see where this is headed

I'm currently debugging a low-distortion receiver for the phase-modulated time signal on BBC Radio 4 (198kHz) and a separate but similar receiver for the Anthorn / MSF time signal on 60kHz. The output impedance of the ferrite loop is high enough that loading it with 50 Ohms is enough to completely kill any signal it picks up. Probing the output of my first-stage RF amp (a common-source JFET amplifier with Zo of ~10k-ohms) works to some extent, but the JFET clearly isn't happy driving it. Probing with an oscilloscope is a complete waste of time - even the amplified signal from the JFET is well below my oscilloscope's noise floor (although admittedly my oscilloscope is a Tek TDS2000B series machine which doesn't have particularly good noise characteristics to start with...)

My question is thus: how would I go about building a high-impedance probe for my spectrum analyser?

I've looked into MMICs (I have a couple of MAR-6 and ERA-2SM chips dotted about) but these appear to be internally terminated for 50 Ohms and will load the 'target' circuit down far too much.
JFETs are another possibility, but the output impedance of such devices is rather low. They might work as a preamp stage, but for a 50-ohm driver? Forget it.

Can anyone suggest anything else I could look into? I'm only after ~150MHz bandwidth at the most - heck, even 20MHz would be a nice start.

Thanks,
Phil.

[olsenn]

Active probes that only need a few hundred MHz bandwidth are easy to make, even with just carefully selected op-amps. If you have the equipment to measure/calibrate its linearity and gain, you should be fine

[pauln]

Active probes that only need a few hundred MHz bandwidth are easy to make, even with just carefully selected op-amps. If you have the equipment to measure/calibrate its linearity and gain, you should be fine

I made an active probe several years back, using a J310 followed by an MMIC, voltage regulator and a handful of other bits all SMD - its not Agilent but it works well .  The original article was from a very early 73 magazine.
Happy to send circuit, photo and description via PM if required.  I may well have a PCB as well.
Let me know.
Paul

[KJDS]

The lazy way of doing it is to just use a length of 50ohm cable with a 470R resistor as the probe. This will reduce the signal seen at the analyzer by (about) 20dB but means the circuit is loaded much less.

The proper way of doing it is with a FET input stage followed by a buffer amp. The main problems with these are stability, probe near a high Q resonance where the JFET is a little marginal and it can oscillate and also the risk of static damage.

[fpliuzzi]

Maybe the "Poor Man's 1GHz Active Probe" project by Elektor magazine would help you out...

http://elektrotanya.com/files/forum/2009/10/e04a036.pdf

This high impedance, active probe is on my to do list (need it for my RF Explorer-3G analyzer).
I have the components already, but I need to make the PCB for it.

Specs listed in article for the DIY probe:
Input-Z: 0.75pf//10Mohm
Output-Z: 50 ohms
Bandwidth: 100KHz - 1.5GHz (+-2.5dB)
Gain: -20dB nominal
Noise figure: 1dB (using SOT143 BF998 MOSFET)

Regards,
Frank

Edit: Updated the broken link.

[philpem]

Wow, that's a lot of replies! Thanks, guys -- I'll answer these in one post if you don't mind 

Active probes that only need a few hundred MHz bandwidth are easy to make, even with just carefully selected op-amps. If you have the equipment to measure/calibrate its linearity and gain, you should be fine

Actually the spectrum analyser has a Normalise function. If you connect the TG output straight through to the input via whatever probe you're using, it figures out the correction factors (I think it works by subtracting to the lowest level, though the manual isn't especially clear about that).

I made an active probe several years back, using a J310 followed by an MMIC, voltage regulator and a handful of other bits all SMD - its not Agilent but it works well .  The original article was from a very early 73 magazine.
Happy to send circuit, photo and description via PM if required.  I may well have a PCB as well.
Let me know.

Wow - that'd be great! Will send a PM shortly. Thanks!

The lazy way of doing it is to just use a length of 50ohm cable with a 470R resistor as the probe. This will reduce the signal seen at the analyzer by (about) 20dB but means the circuit is loaded much less.

The proper way of doing it is with a FET input stage followed by a buffer amp. The main problems with these are stability, probe near a high Q resonance where the JFET is a little marginal and it can oscillate and also the risk of static damage.

The problem is I'm dealing with signals which are quite weak to begin with - losing 20dB would probably push them below the noise floor (or very close to it).

Maybe the "Poor Man's 1GHz Active Probe" project by Elektor magazine would help you out...

www.elektor.com/magazines/2004/october/poor-man-s-1-ghz-active-probe.57219.lynkx

This high impedance, active probe is on my to do list (need it for my RF Explorer-3G analyzer).
I have the components already, but I need to make the PCB for it.

Now that's useful!

I've actually got a bunch of BF998s in my junk box - I bought them ages ago thinking it'd be useful to have some dual-gate MOSFETs. For once, it seems I was right!

The only problem there is that -20dB drop. I wonder if an MMIC (per the 73 Magazine design pauln mentioned) would be enough to bring that within a few dB of unity. Again, junkbox provides - I have MAR-6 and ERA-2SM MMICs; just need to read the specs and pick the best of the two...

[JackOfVA]

A BF998 is a  rather stout device and will drive a  50 ohm spectrum analyzer without too much loss when configured as a source follower. And by "too much" loss, I would expect something between 3 and 6 dB. Since the BF998 has useful gain around 1 GHz or more, parasitic oscillation is always a concern, and yes, a source follower can oscillate as it has power gain.

To check the resonance of a 60 KHz loop antenna, in the absence of a high Z probe, either purchased or home brewed, I would try an inductive pickup. A sensor coil with a turn or two on it, connected to the spectrum analyzer by a meter or two coaxial cable. Keep the inductive pickup a reasonable distance from the 60 KHz tuned antenna and you may be able to pick up enough energy to see the resonance peak.

Spectrum analyzers are designed for linearity and have miserable noise figures - 30 to 40 dB is not untypical. You can improve the sensitivity by a low noise high gain preamplifier. You can also pick up 10 dB of usable gain by setting the minimum SA attenuation to 0 dB instead of the default 10 dB. One point of warning ... if you blow up the front end of your SA, you will unhappy, and the 10 dB attenuator is a safety measure to prevent damage to the  SA input stage, so bypass it with care.  There's a non-obvious key sequence to set input attenuation at 0 dB on the Advantest instruments (I have a pair of R3463s here). It's something like set to manual attenuation, enter 0 dB with the MIN ATTN button and then press the MIN button a second time. Been so long since I've done it that I've forgotten the exact key sequence.

I assume you've also done the usual things like use the minimum possible span and narrowest bandwidth on the SA and also cranked down the video bandwidth. You can also turn on trace averaging to reduce the noise more, but it will slow updates way down.

You can also make a simple voltage probe with a length of coax cable unterminated with a few cm of the center conductor exposed beyond the braid. Probably not going to help with very low level amplifier stages, however.

[Marco]

Why is the 20 dB drop for the poor man's probe necessary BTW? (ie. why not use a couple of parallel pF range capacitors for AC coupling or even just DC coupling.)

[philpem]

A BF998 is a  rather stout device and will drive a  50 ohm spectrum analyzer without too much loss when configured as a source follower. And by "too much" loss, I would expect something between 3 and 6 dB. Since the BF998 has useful gain around 1 GHz or more, parasitic oscillation is always a concern, and yes, a source follower can oscillate as it has power gain.

That's interesting - the Elektor article quotes gain as -20dB (minus 20dB) - i.e. a loss of 20dB.

To check the resonance of a 60 KHz loop antenna, in the absence of a high Z probe, either purchased or home brewed, I would try an inductive pickup. A sensor coil with a turn or two on it, connected to the spectrum analyzer by a meter or two coaxial cable. Keep the inductive pickup a reasonable distance from the 60 KHz tuned antenna and you may be able to pick up enough energy to see the resonance peak.

I'm not trying to find peak resonance on a loop antenna, I'm trying to figure out if my ferrite antenna is picking up a signal, whether that signal is being amplified correctly by the JFET (and following stages) and finally track down any distortion which might be creeping in.

I started out with a 198kHz receiver based on a design from EPE magazine, but the output is clipping quite a bit. I found this out when the PLL locked fine from cold, but if the input signal stopped it'd drift up to Fmax then never recover.

Spectrum analyzers are designed for linearity and have miserable noise figures - 30 to 40 dB is not untypical. You can improve the sensitivity by a low noise high gain preamplifier. You can also pick up 10 dB of usable gain by setting the minimum SA attenuation to 0 dB instead of the default 10 dB. One point of warning ... if you blow up the front end of your SA, you will unhappy, and the 10 dB attenuator is a safety measure to prevent damage to the  SA input stage, so bypass it with care.  There's a non-obvious key sequence to set input attenuation at 0 dB on the Advantest instruments (I have a pair of R3463s here). It's something like set to manual attenuation, enter 0 dB with the MIN ATTN button and then press the MIN button a second time. Been so long since I've done it that I've forgotten the exact key sequence.

My Advantest does something similar - COUPLE, ATTEN,

• [dB]. You can't select 0dB attenuation using the spinner - it has to be directly entered.

I assume you've also done the usual things like use the minimum possible span and narrowest bandwidth on the SA and also cranked down the video bandwidth. You can also turn on trace averaging to reduce the noise more, but it will slow updates way down.

Already done that :)

The problem I have is that the 50R load kills the amplifier. It's a relatively high impedance output (a few Kohms or so) compared to the 50R load the SA presents to it. It just gets loaded down and the SA sees almost nothing. With attenuator disabled I can get a bit of a signal -- just high enough above the noise floor for Zero Span to work (the MSF signal is OOK/100% AM so zero-span with Fc=60kHz shows the transmitted data). Sadly this is nowhere near good enough to see if there are any harmonics being introduced, which this early on would be indicative of bad FET biasing (hardly surprising - this is the first time I've used a JFET in a linear application and I haven't found any good, solid references on JFET -- or indeed MOSFET -- biasing for RF).

Thanks,
Phil.

[Marco]

That's interesting - the Elektor article quotes gain as -20dB (minus 20dB) - i.e. a loss of 20dB.

That's almost entirely because of the decoupling capacitor (and it's effect as a divider with the source follower input capacitance).

[JackOfVA]

I build and sell to the amateur radio and shortwave listener market a voltage probe active antenna. It's basically a high input impedance source follower, with careful attention to maximizing dynamic range and minimizing noise figure.  The input stage is a BF996 dual gate FET, which drives a BFQ19 follower, and the output is a BFQ19/BFQ149 CS follower. Frequency coverage is 10 KHz - 30 MHz where an input filter starts rolling off the coverage to avoid FM broadcast band overload. Without the input filter, it's usable well above 100 MHz. This is, excepting that low frequency response does not extend to 0, an FET scope probe. (It's also a bit physically large to use as a probe.)

In developing the antenna, some early experiments involved driving a 50 ohm load directly with the BF996 and I found it does a credible job, based purely on amplitude. It does not provide the best intermodulation performance, so a more complex circuit is required, but you should not see more than 6 dB loss into 50 ohm load.

The input capacitance of a BF996, plus stray trace capacitance and, in the case of my active antenna, a disconnect  relay, is on the order of 8-10 pF. To prevent the input blocking capacitor from being a limiting factor (voltage divider effect) it's necessary to make it considerably larger - I use a high voltage (630V) 1000 pF C0G capacitor. 

If you don't want to fiddle with a physically small SMD part like the BF996, a TO-92 style J310 is nearly as good and much easier to throw together in a one-off arrangement. For minimum distortion, set the gate bias to provide about 20 mA idle current through the J310. That's a lot more than most people think of when an FET is used, but the "sweet spot" for the J310's intermodulation distortion is in that range.

[philpem]

To be honest, I prefer working with SMD -- it means I don't have to get the Dremel out to drill the PCBs

At the moment I'm reconstructing the CAD files for the Elektor probe, to which I'll add a MAR-6 as a test. It turns out I have almost all the parts I need scattered amongst my spares boxes

Chances are I'll build up the Elektor probe as-is, then add the MAR-6 and see what (if anything) happens. I'd like to do a board shrink-and-split - move the voltage regulator into a metal can mounted on the back of a bulkhead N plug, then run a cable to the Probe Power socket on the SA. That'd mean the probe board was just the probe, meaning I can machine a nice highlighter-pen sized case for it.

I sense a weekend project coming on

I'll let you know how I get on - it should be interesting.

[fpliuzzi]

Hi Phil,
Your idea of adding a MAR-6 to the Elektor probe sounds very interesting. Please keep us updated.

Regards,
Frank

P.S. When I went to purchase the tiny $20 PCB from Elektor to build their "Poor Man's 1GHz Active Probe", they wanted to add an additional $30 to mail that puny thing to me, here in the US. $50 total is unreasonable.

[fpliuzzi]

There seems to some interest in this - I have sent the whole article to others via PM .  Here is the schematic I built and a picture of a prototype.
Originally from an early 73 magazine article.  Works well but is not HP/Agilent standard by long distance.

Thanks for the schematic. It's very helpful to see how the MMIC can be incorporated into my active probe (hopefully built soon).

Regards,
Frank

[KJDS]

How stressed is that 78L05.

7V across it, and at a guess 40mA? A quick check suggests it'll get warm enough to worry me enough to either lower the supply voltage if practical or if you're going to mount it away from the probe then use a 7805 instead.

[branadic]

Here is another FET probe based on TI's JFET OpAmps from DC to 500MHz:

http://www.mikrocontroller.net/topic/188227#new

and another AC FET probe (page 112):

http://www.bartelsos.de/index.php?dl_file=3TNJFF9YR3

[philpem]

How stressed is that 78L05.

7V across it, and at a guess 40mA? A quick check suggests it'll get warm enough to worry me enough to either lower the supply voltage if practical or if you're going to mount it away from the probe then use a 7805 instead.

I'll probably use a normal 7805 anyway - I don't tend to keep 78L05s in stock.
Exceeding voltage regulator specs seems to be Elektor's SOP - I've seen them do it in a few designs. That and using solder pads barely larger than the component pins...

Just patching together an EAGLE library for the Minicircuits MAR-6, then modifying the Elektor design and etching a PCB!

[philpem]

Well, this is what I've got so far -- schematic and PCB attached. Would be interested in any comments, criticisms or suggestions for improvement you might have

Dimensions of the board: around 14 x 50 mm. In old money, a whisker over half an inch wide by two inches long.

My eventual plan is to (hopefully!) build up some form of shielded case for the thing. That's still on the drawing board, though.

[fpliuzzi]

Thanks for posting the schematic and board layout (I'm assuming that you're using Elektor's original bottom layer layout as is). Nice job on the PCB layout. I also like your idea of moving the 5V regulator off of the probe board to minimize the probe's dimensions.

I am happy that I held off building Elektor's original probe board design. I like your implementation much better. I'm sure that you cannot wait to test drive your new probe.

Regards,
Frank

[vk6zgo]

Exceeding voltage regulator specs seems to be Elektor's SOP - I've seen them do it in a few designs. That and using solder pads barely larger than the component pins...

I thought that was just a generic European thing,like through holes barely larger than component leads.
Great for flow soldering,but a nightmare to rework! 

[BravoV]

Poor Man's 500 MHz Active FET Probe -> http://welecw2000a.sourceforge.net/docs/Hardware/Aktiver_Tastkopf_mit_OPA659.pdf

Its written in German though, but with lots of photos especially tons of results of the probe's profiles from the spectrum analyzer, it should be easy to interpret the probe's performance.

[philpem]

Should probably update this...

I eventually did build one of the probes -- the Elektor original. The EAGLE CAD files are here: https://bitbucket.org/philpem/rf_probe

Enjoy.

[Alex Eisenhut]

My favorite is the P6032.

http://w140.com/tekwiki/wiki/P6032

Or you can look into this

http://w140.com/tekwiki/wiki/282

[philpem]

Interesting!

Apparently my spectrum analyser (Advantest R3361A) has a Lemo probe-power socket which matches the pinout of the Tek 492 and 492P connector. The matching probe is apparently the P6201. Sadly they're pretty thin on the ground in working condition; the majority I've seen for sale are either overpriced, broken or in some far-flung locale being sold by someone who'd rather not sell internationally.

I'm sure there's a later replacement for the P6201 but I'm not enough of a Tek guru to know what that would be. And I'll bet there's a HP/Agilent/Keysight equivalent too.

[Conrad Hoffman]

There's a high speed probe on page 96 of this Jim Williams LT app note- http://cds.linear.com/docs/en/application-note/an47fa.pdf