High Voltage Differential probe design for review

[Jay_Diddy_B]

Hi group,

I took my Saphire SI 9000 probe apart for some photographs.

When you open the case you find two screened boxes.


Under the first box there is a power section. This is used to generate +/- voltage from the 6v battery:



Under the other box is the input section and amplifier:









Regards,

Jay_Diddy_B

[Jay_Diddy_B]

Hi,

I was intrigued by the poor performance of these probes at high frequencies. So I took a closer look.



I measured the response using an RF signal generator and an oscilloscope. 



First the gain at 1MHz:



This is correct it is showing a 500x reduction in signal amplitude.

At 44 MHz I found a resonance:



The gain is much higher than it should be, 23.96mV out with 320mV input. That is divide by 13.3 instead of divide by 500.

I noticed that gain was sensitive to the arrangement of the leads.

A quick LTspice model shows that the 7pF capacitance of the probes is resonating with a few uH of inductance.





I am sure that this could be engineered better to dampen this resonance.

Regards,

Jay_Diddy_B

[tautech]

Hi group,

I took my Saphire SI 9000 probe apart for some photographs.

Scrubbed IC's 

[Jay_Diddy_B]

Scrubbed IC's 

I don't know why they bother.

It has a Harris logo, it is DIP, it is a high speed multiple op-amps that has been around for 10-15 years so that should narrow it down to a handful of parts.

Jay_Diddy_B

[T3sl4co1l]

Try ~500 ohms in series with the "probe" lead.

Note that, even if the differential and common mode sensing is constructed with zero capacitance, you still have the loading effect of the freaking meter (give or take) of wire, which itself becomes resonant at 75MHz or so, with a characteristic impedance of maybe 100-300 ohms differential mode, and 50-600 ohms common mode, depending on how the cables are routed together and what they're laying on.

Tim

[Jay_Diddy_B]

Tim,

I had similar thoughts myself. The 500 Ohm series resistor will damp the resonance.


A similar result is predicted by SPICE:



I found the calibration procedure for the SI9000 probe here:

http://www.sapphire.com.tw/si9001cp.htm


I hope these measurements help the original poster with their design.

Regards,

Jay_Diddy_B

[cjk2]

I want to thank you all for your hard work and investigation on this project. I have ordered my boards and parts today. Hopefully some time in the next month I will be able to build and test this thing!

[janaf]

Just popping in on this thread, you might want to look at fully differential op-amps for the first stage, like THS4521 / THS4541. TI have a bunch of differentials in the portfolio.

[AdShea]

Problem with most those fully differential amps (TI or others) is that the input impedance is pretty low (100s of ohms) and the circuit topologies needed are low input impedance which screws with all the work in making a nice input attenuator.

For reference, the P5200 used a MAX436 to do the differential to single-ended conversion (and a bit of gain) but those are long gone.

[janaf]

You are right.
I looked up the datasheet; not hundreds but around 100K; 72K for the THS452X, 110K for THS4541

Problem with most those fully differential amps (TI or others) is that the input impedance is pretty low (100s of ohms)

[AdShea]

Yeah, I was looking at some really fast video reciever ones that had ~700 ohm differential input impedance.  Still have to wire them up in a configuration that has even less input impedance (similar to inverting opamp circuit) if you want to get good noise performance.

[Mechatrommer]

You are right.
I looked up the datasheet; not hundreds but around 100K; 72K for the THS452X, 110K for THS4541

Problem with most those fully differential amps (TI or others) is that the input impedance is pretty low (100s of ohms)

wrong opamp for the job? maybe LMH6503...

[AdShea]

Better, though the bias current might be an issue (11uA typical) also the output-referred offset voltages are in the ~100mV range and change with gain setting and common mode voltage.  It's also a bit on the slow side for aping the P5200, but getting closer...

I'm almost thinking that transconductance amp-based frontends are the way to go though.