PRP1 - Low cost 2GHz power rail probe

[tszaboo]

I had an other fancy pot in my collection. Brand new.

Vishay PRV6

Edit: the remainder are some low cost pot's not worth testing.

I quickly measured the trimpot that I removed. The 4th Digit on the 6.5 digit DMM was already jumping around.
It kept on doing this even
Measured an ordinary 0805 resistor, only the last digit moves. I also measured the pot that I wish to use in the final design. I noticed some effect, that after moing, it was drifting for more than a minute to the same direction. I set  it up to ~10KOhm, and it would drift to 9981 Ohm I'm not sure if this is a heating up effect of the DMM, or something mechanical. Anyway after the drifting, only the last digit changes again, so it's 100 times less noisy than the trimpot. Unfortunately my 34401A doesn't do ratiometric measurement.

So rather than giving up, I placed the pot in the fridge for about 10 minutes, and after taking it out, connected the DMM across the two end terminals. It went from 11560 Ohm to 11146 Ohm, Unfortunately Alps doesn't specify the tempco of their pots. But I use it in ratiometric mode, and wiper current is minimal, so there shouldn't be any drift and low frequency noise happening.

And here is the interesting part: I grabbed another 10 turns trimpot from the same type. And it didn't have drift and noise.
So I have to assume handling it there is a failure mode when it becomes noisy.

To be not only off topic, I made quite a bit of progress on the probe. Found the issue of the full scale 24V problem, added ESD protection and ferrites to the USB line. In my investigation, the USB outputs of scopes didn't always play nice with the probe, they were injection noise, now it's not measurable on the scope I have.

[temperance]

The pot drift/stability might have something to do with the track lubricant which lifts up the wiper ever so slightly while turning the pot. The RPV6 pot I tested is pretty expensive but worse than the 91-95. But the RPV6 pot is meant as a position sensing pot and probably contains more lubricant.

I'm interested in such a probe. My last design containing a 2.4 GHz RF soc went trough RED testing with a very small margin for spurious emission. (probably caused by noise from the fly back power supply finding it's way into the PLL)

I was thinking this topic would be gaining more attention.

[tszaboo]

Is there a reason why the front and back are made of aluminum? You could use FR4 and no extra manufacturing steps would be required.

These two prototypes were (quite badly) modified by me with a drill and some sawing and filing.
I was also waiting for the supplier to give me a quote for the modifications. It's not too bad, but I think I'll go with FR4 end plates for the first batch. I'll don't want to end up in a situation where the silkscreen is bad, and I have to reorder the housings.

Some more measurements!
This is the USB ripple when the PRP1 proto 2 is connected to the USB port. This is the 5V voltage directly from the USB.
The voltage is measured with the other PRP1!



After manual modifications to the input circuit:


And the output of LDO on the board:


As you can see the PRP1 was capable to measure this voltage ripple.
The actual modifications reduced the ripple from 52mV peak to peak to 3mV. The rest is handled by extra ferrite beads, and the PSRR of the Opamps. The ripple there is on the output of the LDO is barely 600uV, and they have more than 30dB PSRR at these frequencies. The noise is not visible on the output

[tszaboo]

Another day spent well. I've made quite a bit of progress, schematics is complete I think. I was focusing a lot on the coarse-fine adjustment, improving the linearity of the two potentiometers. The fine should have about the range to be able to change a signal in the active range now, so the coarse only needs to get the signal in view.
I also spent time on the mechanical design, pictures are below. Instead of 2 wire assembly and 2 PCBAs it's now 1 PCB. Added an LED indicator. And spent 4 hours staring at potentiometer datasheets trying to figure out how to get the right one, then settling for something close to it.
With this I'm off to BOM setup.

[tszaboo]

I'm ready with everything to order the next prototype.
I'm only waiting for some samples from TI to test out some more parts, just making sure that the probe schematic is optimal.
Also I'm still waiting for some Coax cables, because RG316 with low quality connectors is not a good choice for the connection between the scpe and the probe.


I dare to ask: Who is interested in getting one of these?
Price 250 EUR + Shipping + paypal fees

Contents:
PRP1
SMA-BNC coax cable 0.8m for connecting scope (TBD likely RG142 or LMR240)
SMA-UFL coax cable for connecting DUT
SMA-pigtail coax cable for connecting DUT
SMA-2.54mm square header coax cable for connecting DUT
USB cable
Optimally P57 terminator (OK performance up to 200 MHz)

I'm asking because there isn't a lot of price difference between ordering 2 or a few.

[Calvin]

Hi,

triggered by Your thread I quickly built a probe based on the circuit suggested by Tommi Otsavaara in his thesis "1 GHz Power Rail Probe".
Seemingly he orientated on the Keysight N7020 probe that Sharia kindly dared to open up for us 
The major differences to the Levido DIY-probe seems the higher output resistor of the LF-path of 590R which requires the second inverting stage to have a gain of -11.8, and the increased Offset range of up to 50V
I had some OP270 lying around, which is a low en-noise dual OP27, which in simulation results in very nice low noise, but it´s decompensated and faster brother OP37 would be better suited to the second inverter gain stage.
In the Levido probe where the second inverter is just a buffer with a gain of -1 the OP27 may get problems due to the required higher load currents.
As alternatives TI lists some ´Soundplus audio-OPAs´ like the OPA1655/1656 with lownoise, sufficient GBW and high current outputs.
If higher supply lines are possible the JFET-Input OPA827 appears interesting.
Due to the overall high offset-gain of 118 the Otsavaara probe´s offset-setting becomes critically sensitive and rather requires multiturn-potis even though potis are used for coarse and fine tuning already.
Another point is the lowpass filtering of the LF-path.
The Otsavaara relies on simple RC-Filtering, which in this case leads to a shelving-filter character.
In simulation this adds ~0.3dB of output signal level to the HF-path.
The Levido probe on the other hand utilizes a RLC-filter that follows a LP-filter curve much closer, hence attenuates the LF-path alot better.

@tszaboo
increasing the Offset range from +-24V to +-50V or even more (R&S claims up to 60V) seems easy ... just a bit of shuffling values of the Potis and the offset-input resistor.
Do You plan on offering a probe with higher offset range also?

regards
Calvin

[tszaboo]

Hi,

triggered by Your thread I quickly built a probe based on the circuit suggested by Tommi Otsavaara in his thesis "1 GHz Power Rail Probe".
Seemingly he orientated on the Keysight N7020 probe that Sharia kindly dared to open up for us 
The major differences to the Levido DIY-probe seems the higher output resistor of the LF-path of 590R which requires the second inverting stage to have a gain of -11.8, and the increased Offset range of up to 50V
I had some OP270 lying around, which is a low en-noise dual OP27, which in simulation results in very nice low noise, but it´s decompensated and faster brother OP37 would be better suited to the second inverter gain stage.
In the Levido probe where the second inverter is just a buffer with a gain of -1 the OP27 may get problems due to the required higher load currents.
As alternatives TI lists some ´Soundplus audio-OPAs´ like the OPA1655/1656 with lownoise, sufficient GBW and high current outputs.
If higher supply lines are possible the JFET-Input OPA827 appears interesting.
Due to the overall high offset-gain of 118 the Otsavaara probe´s offset-setting becomes critically sensitive and rather requires multiturn-potis even though potis are used for coarse and fine tuning already.
Another point is the lowpass filtering of the LF-path.
The Otsavaara relies on simple RC-Filtering, which in this case leads to a shelving-filter character.
In simulation this adds ~0.3dB of output signal level to the HF-path.
The Levido probe on the other hand utilizes a RLC-filter that follows a LP-filter curve much closer, hence attenuates the LF-path alot better.

@tszaboo
increasing the Offset range from +-24V to +-50V or even more (R&S claims up to 60V) seems easy ... just a bit of shuffling values of the Potis and the offset-input resistor.
Do You plan on offering a probe with higher offset range also?

regards
Calvin

Hi Calvin. I didn't see this thesis before, it's quite interesting.
The circuits looking like the N7020 probe: There is an expectation on how the probe is supposed to work, if you deviate from it it would work differently. When taking measurements, the probe should be "transparent" meaning that it shouldn't show any behavior which makes you doubt :whether it's the DUT or your test setup is making certain outcomes. Wat I'm trying to say there is only so many ways this circuit can be built, if you deviate from it, it's a different probe.
The thesis is quite ambitious for a BSC. I run into the same issue, a semester is not enough to design a board build it test it, but more importantly to document it. Because you get your grade based on the documentation.
His circuit is a good effort on the analog path. His power supply is very noisy, these DC-DC modules can easily have 100mV ripple on the output. When you measure some high voltage, the potmeter is set to high, you would be measuring all the noise and ripple coming from that DC-DC, amplified. Even though he has a RC filter with an electrolytic cap. I don't think he built the power supply, probably used a bench power supply.
Not to diminish his efforts for a BSC thesis this is very good.

I don't think I would want to offer it with a higher voltage for now. Going for 48V would mean that I need to comply with the Low Voltage Directive, that I'm not willing to do now. If you are interested, I could accidentally  send you a probe that has it's resistor replaced, and it could measure higher voltages. It would also have higher noise levels.

[temperance]

I saw this thesis. What I don't understand is all the fuzz about the coupling capacitor on page 23. The thesis also blames R&S for not plotting such data on page 11.

On the plot shown on page 23 is a reference to a GRM188R72A104KA35 capacitor which is a X7R capacitor. This cap must be an NP0 or C0G type. His supervisors should have caught that.

[Neganur]

On the plot shown on page 23 is a reference to a GRM188R72A104KA35 capacitor which is a X7R capacitor. This cap must be an NP0 or C0G type. His supervisors should have caught that.

It's "just" a B.Eng thesis. The supervisors do not typically pay attention so detailed on that level (sorry, Heikki  ). I would have slapped his wrist for not labeling the y-axis though 

PS: 1206, 100nF, 100V C0G/NP0 from Kemet is pretty darn expensive on Digikey. ~10 EUR for one. Maybe that was a money choice rather than a design choice.

[temperance]

PS: 1206, 100nF, 100V C0G/NP0 from Kemet is pretty darn expensive on Digikey. ~10 EUR for one. Maybe that was a money choice rather than a design choice.

I've seen different prices some time ago on the mouser website. Let's see.

Some 100 V C0G capacitors in 1206 can be had for less than €1. 200...250 V versions in 1206 are available for less than €2. Both in single quantity.

[Neganur]

oh right, I filtered for 1% tolerance - my bad. 2% and 5% are less than one Euro. good point

[tszaboo]

And now something different.
I'm exploring the possibilities of making a browser for the PRP1. This is a DIY solution, example of a commercial part:
https://nl.rs-online.com/web/p/oscilloscope-probe-accessories/1444395

It's made out of a 100MHz low end oscilloscope probe, it's internals were gutted and replaced with rigid coax. And glue.
Not pretty, but I measured <1.5 VSRW up to 2 GHz with it. It's a bit of work to put together, the probe was glued.

[temperance]

https://nl.rs-online.com/web/p/oscilloscope-probe-accessories/1444395

2K6 pour une pièce de résistance... Did they hire an Audiofool to develop this cable?

For me an SMA to open end cable will do. They cost nearly nothing.

[Neganur]

The Keysight SMA-SMA cable for the power rail is similarly expensive. At least with the Lecroy cable, you're also getting a special directional resistive browser head. At the end of the day though, it's the low quantity sold that sets the price.

[Neganur]

I'm ready with everything to order the next prototype.
I'm only waiting for some samples from TI to test out some more parts, just making sure that the probe schematic is optimal.
Also I'm still waiting for some Coax cables, because RG316 with low quality connectors is not a good choice for the connection between the scpe and the probe.


I dare to ask: Who is interested in getting one of these?
Price 250 EUR + Shipping + paypal fees

Contents:
PRP1
SMA-BNC coax cable 0.8m for connecting scope (TBD likely RG142 or LMR240)
SMA-UFL coax cable for connecting DUT
SMA-pigtail coax cable for connecting DUT
SMA-2.54mm square header coax cable for connecting DUT
USB cable
Optimally P57 terminator (OK performance up to 200 MHz)

I'm asking because there isn't a lot of price difference between ordering 2 or a few.

I'd pledge 100 for just the populated PCB, no need for cables or enclosure.

[tszaboo]

I'd pledge 100 for just the populated PCB, no need for cables or enclosure.

I'm not planning on selling this with a discount.
What I can offer is a free unit, in exchange for writing a review here, and verifying the performance of the probe.
I think you have the test equipment to do that?

[Martin72]

What I can offer is a free unit, in exchange for writing a review here, and verifying the performance of the probe.

How “deep” should the review be, in your opinion?

[tszaboo]

What I can offer is a free unit, in exchange for writing a review here, and verifying the performance of the probe.

How “deep” should the review be, in your opinion?

I'm not trying to entrap you with a lengthy homework.
I only want someone competent to verify my claims to the specifications, to avoid making a mistake and disappointing people. RMAs.
Spend however much time on the review you think a free probe is worth. It's a gentlemen's agreement anyway.

[tszaboo]

PCBs, PCBAs and Enclosure ordered. I've requested quotations for the cables.
Next thing is packaging. And instruction manual.

[nctnico]

I'd pledge 100 for just the populated PCB, no need for cables or enclosure.

I'm not planning on selling this with a discount.
What I can offer is a free unit, in exchange for writing a review here, and verifying the performance of the probe.
I think you have the test equipment to do that?

Before anything, I think it is a good idea to figure out what kind of industry standard testing method exists for these kind of probes so specifications / test results can be compared with other probes.

[tszaboo]

I'd pledge 100 for just the populated PCB, no need for cables or enclosure.

I'm not planning on selling this with a discount.
What I can offer is a free unit, in exchange for writing a review here, and verifying the performance of the probe.
I think you have the test equipment to do that?

Before anything, I think it is a good idea to figure out what kind of industry standard testing method exists for these kind of probes so specifications / test results can be compared with other probes.

I agree, I was going to put all that documentation on Github, including my test setups and results. And instruction manual.
There is a performance verification guide in the instruction manual of a Tek probe:
https://download.tek.com/manual/TPR1000-and-TPR4000-User-Manual-077154200.pdf
From page 26. It gives a good baseline on what to measure. I found because of the changing input resistance of the probe (which is the whole point of it) measuring it was quite challenging. The high frequency verification seems to be just sweeping it with a VNA and finding the -3dB point. Which seems to be above 2 GHz in my case, but I still want to sweep the final design with a higher quality VNA before calling it faster. Using a SMA calibration standard on the probe's output. With a scope it's slower anyway, with a 50 Ohm adapter + regular scope input, I think it's limited to about 200 MHz. With a 50 Ohm input scope it should reach the scope's nameplate speed.

[tautech]

Nice, OP image needs resizing....good you've done it.

Dunno what it is but we have one coming soon too, a 4 GHz version.

Wanna know the specs of what you're up against ?

Thanks. I'm eager to find out.
There are many benefits of using a factory built probe, it tells you on the screen what the voltage you are looking at, and with the active probe it's a better interface.
Still, I hope people with low-end scopes or budget will be able to use a probe like this.

Complete specs available now:
https://int.siglent.com/u_file/images/24_09_10/SAP4000P_UserManual_EN01A.pdf

Yours will be a much lower cost version......

[tszaboo]

Nice, OP image needs resizing....good you've done it.

Dunno what it is but we have one coming soon too, a 4 GHz version.

Wanna know the specs of what you're up against ?

Thanks. I'm eager to find out.
There are many benefits of using a factory built probe, it tells you on the screen what the voltage you are looking at, and with the active probe it's a better interface.
Still, I hope people with low-end scopes or budget will be able to use a probe like this.

Complete specs available now:
https://int.siglent.com/u_file/images/24_09_10/SAP4000P_UserManual_EN01A.pdf

Yours will be a much lower cost version......

Looks like a very nice bit of kit.
Any of these probes, especially the "low speed probe browser" would be an excellent accessory for my design.
The high speed browser is an interesting design, I wonder if we can learn something from it's construction.

BTW
"5.3 Pigtail Cable Welding" - you might want to correct this it's a very obvious translation mistake from Mandarin.