Lecroy HVFO "High Voltage Fiber Optically-isolated Probe" Modulation Scheme

[ezalys]

Howdy folks!

I've been looking into these fiber isolated probes that Lecroy makes (http://teledynelecroy.com/probes/probeseries.aspx?mseries=537), and I'm curious how they're sending the signal over the optical fiber. I was thinking maybe you could make a probe with a sigma-delta converter in it and then send this digital signal optically. The manual/datasheet claim however that the signal is carried as FM (note "frequency modulating optical transmitter"). I'm curious how you might design a voltage to frequency converter with high linearity and 60 MHz of modulation bandwidth!

Any ideas, folks?

Best,
Evan

[Marco]

I don't really see the point of FM, sure you avoid the drift of the laser/photodiode response but there are other ways to do that and you're limiting your bandwidth. Maybe Tektronix got some patents on IsoVu (even though AM modulating a laser to transmit an analogue signal to an oscilloscope is patently bloody obvious) and they are doing FM to avoid it?

If the drift of the amplification with temperature and age of laser/photodiode was severe I'd just split up the signal into LF and HF and send them with two wavelengths, include a low frequency sine with the HF one so you can automatically detect the amplification and compensate. Send the LF part digitally.

[texaspyro]

Maybe Tektronix got some patents on IsoVu (even though AM modulating a laser to transmit an analogue signal to an oscilloscope is patently bloody obvious) and they are doing FM to avoid it?

Bloody obvious NEVER stops the USPTO from issuing a patent.  I'd say maybe, if you are generous, 1 in 10,000 patents is worth of a patent.

I was sending 150 Mhz analog waveforms over 1+ km of fiber well over 30 years ago.

[Kleinstein]

Analog AM modulation should be good enough for a scope. A scope is usually not expected to be very precise at DC and photo-diodes are not that temperature dependent in the sensitivity - so it might need an optical feedback at the sender side. But this old style - there are already opto-couplers made for this.

US patents are sometimes kind of jokes - but expensive ones

[JohnG]

The Tektronix IsoVu has 1 GHz bandwidth, several kV of isolation, and about 80 dB CMRR even near 1 GHz. There is a hell of a lot of work between this and the obvious use of AM modulation. They get >$10k for these probes and they are selling.

If you don't need this, then don't buy the probe. On the other hand, if you have ever needed to know what is going on with a high side gate drive looking at a gate signal for GaN or SiC FETs sitting on a 500V-1000V square wave with 200 V/ns transitions, then you might realize that it is not so simple. I don't know of any other probe that comes close.

I don't work for Tektronix, and don't even have a Tek scope these days. But I have seen their probe in action. I have never seen any other probe that can do the above measurement and produce consistently usable results.

John

[Marco]

There is a hell of a lot of work between this and the obvious use of AM modulation. They get >$10k for these probes and they are selling.

If the patents are broad enough to deny anyone else putting in work to make AM work then they are far too broad ... we have a case in point here of someone doing it thirty years ago. It's not uncommon in astronomy either. Nothing terribly complex about it, DC needs to be handled and amplification can drift but that's just a little engineering away from fixing.

As for why oscilloscope probes have been mostly retarded for decade and didn't take this patently obvious route before, inertia. We could all be using 200 MHz active optically coupled probes for 20$ a piece and never again have to worry about blowing up the scope due to grounding, but instead we have to screw around with antiquated technology which hasn't made sense in ages ... the human condition sucks sometimes.

[JohnG]

I don't see any evidence that the patents are too broad or forbidding anyone from doing an AM carrier-based optical probe, only speculation that this is the case.

I do think it is not trivial to do this with AM, and it might be easier with FM. And, a high-end scope company does need to be concerned with DC drift and gain accuracy for their probes.

BTW, I have actually looked at some related patents. The interesting part is how they do the modulation. Here are a couple that might be of interest:

https://patents.google.com/patent/US6603891B2/en
https://patents.google.com/patent/US7310455B2/en

If you are really interested, take a look at the references for these.

John

[ezalys]

The tek probes for sure use an electro-optic modulator. Maybe the LeCroy ones do too? If the LeCroy one doesn't, then I just wonder how it's possible to get 60 MHz of extremely linear modulation bandwidth out of a UHF band oscillator

[KE5FX]

The tek probes for sure use an electro-optic modulator. Maybe the LeCroy ones do too? If the LeCroy one doesn't, then I just wonder how it's possible to get 60 MHz of extremely linear modulation bandwidth out of a UHF band oscillator

The Tek probe is an interesting piece of hardware, all right.  (I didn't look at LeCroy's.)  It incorporates four separate lasers and a bundle of five optical connections:



One laser is used for power transmission, presumably on the order of 1W or less, one for transmission of the signal back to the controller box, and they don't say much about the remaining three.  They apparently constitute an elaborate feedback loop which is needed because the signal amplifier isn't actually part of the probe head.

I'm having trouble figuring out what all that stuff is for.  If I were building this, I'd just use rechargeable batteries in the probe head, with power-over-fiber available as an option for the 1% of customers who need more than a few hours' operating time.  A $10 off-the-shelf diff amp like an LMH5401 should be fine for signal conditioning with a suitable high-Z front end duct-taped to it.  At that point, a single fiber connection back to the controller would be enough to get the job done.  Plenty of ways to put 1 GHz of analog bandwidth through a single-mode fiber.  What am I missing?  Why all the wacky electro-optical hardware and Rube Goldberg feedback loops? 

[ezalys]

Flat response, drift, nonlinearity, these sorts of things. I suspect getting an analog signal over a fiber is easy but doing it to the standard of a scope probe is likely very very hard.

[PartialDischarge]

Anyone knows the price of these Lecroy probes??

[PartialDischarge]

The Tek one has a laser diode transmitting some power to the probe head, then the probe head recovers this power to electric power. The power then powers low frequency ADC and DC bias circuitry, while the HF part goes directly to a crystal that modulated another beam directly.
Since majority of information is directly modulated, the laser must be very pure in terms of modes and bandwidth, so a single mode fiber is used for HF modulated signal.
Multimode fibers are used for power and probably DC~LF signal.

*: Heard from a Tek sales a few months ago at a conference.

What the Tek sales forgot to say is that there has to be a feedback loop scheme to compensate for Tx power levels variation in the lasers variations due to heating and time drift. Transmitting analog information over an optical link with precision is not an straightforward task, I guess thats why they have so many fiber optics

[ezalys]

I'm relatively convinced that the lecroy probe and tek probe work differently inside. They're completely different sizes, the lecroy probe is a tenth the price, and the lecroy probe has a tenth the bandwidth.

[KE5FX]

What the Tek sales forgot to say is that there has to be a feedback loop scheme to compensate for Tx power levels variation in the lasers variations due to heating and time drift. Transmitting analog information over an optical link with precision is not an straightforward task, I guess thats why they have so many fiber optics

Let's see... you need 8 bits of resolution for an oscilloscope application, or about 48 dB of dynamic range.  You'd like to have 10 or 12 bits, of course, but users who need common-mode specs measured in kilovolts can't be too picky.

So the most straightforward way to build a 1 GHz differential probe might be to use a 3 GHz 8-bit ADC at the probe head and send the data back to the controller in digital form, with an FPGA to mux it over three 10GBASE-SR fibers.  This seems kludgy, but it's a commodity kludge, much cheaper and less R&D-intensive.  A 3 GHz 8-bit ADC costs less than $200.  For $2K you can upgrade to 12 bits at 3.6 GSPS, but now you need another fiber, I guess.

The biggest drawback of a fast ADC at the probe head is probably the power consumption.  That may be the #1 factor that made Tek choose the EO approach.  A probe that runs on batteries is never going to make everyone happy... but then neither will a $20K+ price tag.

[PartialDischarge]

blueskull, I remember you mentioned sometime ago you were doing something with POF, do you know what kind of photodiode(or whatever) would be needed to generate around 200-500uA of current with the light of a laser?

[raviteza88]

Infact, these 2 probes have me baffling the past 2 months and I have been diving deep into the world of isolated measurement, ADCs, Optical, transformer isolation etc.

Infact pure analog differential amplifiers if designed properly can achieve such high CMRR. A company called CICResearch http://www.cic-research.com/ makes them already. If you look at their datasheets, u will notice they make HV Diff probes upto 30KV and their DP04 series has 120 MHz bandwidth and 8kV RMS voltage ratings. Of course this voltage rating doesn't mean shit if your probes CMRR degrades over frequency, which is a common problem with normal diff probes. However CICreasearch somehow tackles this problem and has a CMRR of -155dB @ 100 Hz and -110dB @ 10 MHz. So one can assume that at around 100MHz it has a very good CMRR although this info is not in datasheet.

http://www.cic-research.com/datasheets/DP04_Datasheet_06-B.pdf

I dived down the rabbit hole of analog opamps world and found that basically one has to compensate for Common Mode impedance at higher frequencies, which is the main culprit for voltage derating, low CMRR etc. ( one can now understand that these are all the same). A differential amplifier with improved CMRR circuits is shown in this book Analog Electronics with Op-amps Page 54,55. Further guard rings etc can be employed in PCB layout which can further improve CMRR.

However an isolated front-end/oscilloscope if designed properly with enough bandwidth can do the same the ISOVu does. I think the ISOVu is really overrated and may have other intentions. Anyone heard about Photonic Sampling in ADC's? Its just a theory until now. Probably they they somehow adapted this tech and use it get such a high bandwidth for the power electronics measurements industry which is now into SIC and GaN Mosfets with new challenges of its own. Infact Photonic sampling research is in 100s of GHz range. So if properly done, 1 GHz should be easier. Probably Tektronix wants to learn about photonic sampling with this probe and learn to develop even higher bandwidth probes.

What I really don't understand is the amount of data conversions taking place in this probe, even in the Lecroy HVFO one. Analog to (Digital probably) to Optical to (Digital probably) to Analog to Digital in Oscilloscope. Looks like a waste of scope hardware. If only they could eliminate oscilloscope front-ends all together and feed directly into the FPGA/ASIC.

Anyways on side notes, there is a kiwi company called Cleverscope. They have released their isolated usb oscilloscope CS448 with 4 isolated front ends upto 1kV and achieve a similay CMRR of 100dB @ 1MHz. Looks interesting too but costs 10k a pop, but with 4 channels though, so that's a slight relief. However 1 kV is a little less. I work with 4kV @ 2 MHz, so wont be of much use.

https://cleverscope.com/

All isolated oscilloscopes achieve high CMRR due to inherent high common mode impedance which alleviates all these measurement problems. A differential probe if designed properly can also achieve higher CMRR  even in the 100s of MHz range.

[JohnG]

Infact, these 2 probes have me baffling the past 2 months and I have been diving deep into the world of isolated measurement, ADCs, Optical, transformer isolation etc.

Infact pure analog differential amplifiers if designed properly can achieve such high CMRR. A company called CICResearch http://www.cic-research.com/ makes them already. If you look at their datasheets, u will notice they make HV Diff probes upto 30KV and their DP04 series has 120 MHz bandwidth and 8kV RMS voltage ratings. Of course this voltage rating doesn't mean shit if your probes CMRR degrades over frequency, which is a common problem with normal diff probes. However CICreasearch somehow tackles this problem and has a CMRR of -155dB @ 100 Hz and -110dB @ 10 MHz. So one can assume that at around 100MHz it has a very good CMRR although this info is not in datasheet.

http://www.cic-research.com/datasheets/DP04_Datasheet_06-B.pdf

Please note that the data sheet specs common mode rejection (CMR), not common mode rejection ratio (CMRR). They are not the same. The probe gain is -60 dB, so the CMRR is 50 dB. Respectable, but not stellar by any means.

John

[raviteza88]

Infact, these 2 probes have me baffling the past 2 months and I have been diving deep into the world of isolated measurement, ADCs, Optical, transformer isolation etc.

Infact pure analog differential amplifiers if designed properly can achieve such high CMRR. A company called CICResearch http://www.cic-research.com/ makes them already. If you look at their datasheets, u will notice they make HV Diff probes upto 30KV and their DP04 series has 120 MHz bandwidth and 8kV RMS voltage ratings. Of course this voltage rating doesn't mean shit if your probes CMRR degrades over frequency, which is a common problem with normal diff probes. However CICreasearch somehow tackles this problem and has a CMRR of -155dB @ 100 Hz and -110dB @ 10 MHz. So one can assume that at around 100MHz it has a very good CMRR although this info is not in datasheet.

http://www.cic-research.com/datasheets/DP04_Datasheet_06-B.pdf

Please note that the data sheet specs common mode rejection (CMR), not common mode rejection ratio (CMRR). They are not the same. The probe gain is -60 dB, so the CMRR is 50 dB. Respectable, but not stellar by any means.

John

I think CMR and CMRR are used interchangeably by many companies. But the text book definition of CMRR is the ratio of differential  gain to common mode gain Ad/Ac, which means this is just a number. And the textbook definition of Common mode Rejection in decibels is CMR =20 Log(CMRR) .

Also nowhere in the datasheet is written it has a gain of -60dB. So I can't agree with you that this probe has only 50 dB CMRR (or CMR to be more specific)

[JohnG]

I think CMR and CMRR are used interchangeably by many companies. But the text book definition of CMRR is the ratio of differential  gain to common mode gain Ad/Ac, which means this is just a number. And the textbook definition of Common mode Rejection in decibels is CMR =20 Log(CMRR) .

Also nowhere in the datasheet is written it has a gain of -60dB. So I can't agree with you that this probe has only 50 dB CMRR (or CMR to be more specific)

On the data sheet, the probe attenuation factor is 1000:1, or a gain of -60 dB. This is this the differential gain Ad.

I have only rarely seen CMR and CMRR used interchangeably in practice, and it is incorrect. Also, I have several textbooks, and none of them defines CMR in the manner you describe. Since CMRR is a unitless ratio, it is enough to use dB to indicate the use of decibels, otherwise just the ratio is used.

John

[raviteza88]

On the data sheet, the probe attenuation factor is 1000:1, or a gain of -60 dB. This is this the differential gain Ad.

[/quote]

The probe attenuation factor is not related to the differential probe's gain. The gain is set by the feedback resistors or whatever front end this probe has. The attenuation factor simply means that the signal arriving at the op-amp's inputs is attenuated by 1000 times. If you calculate in decibels, surely u would arrive at -60 dB Gain for a 1000:1 ratio. However, this has nothing to do with the op-amp's gain which can be independently set. Remember we are amplifying the differential signal and rejecting the common signal. Once you have done the necessary signal conditioning, one can again attenuate the signal to get the desired input to output scaling.

http://www.analog.com/media/en/training-seminars/design-handbooks/Op-Amp-Applications/Section1.pdf

As far as the other CMRR defintion goes, this is just one of the sources i found, it defines CMRR as Ac/Ad. As per TI its again different and CMRR is Ad/Ac, however both define CMR as 20LogCMRR.

http://www.cypress.com/file/58181/download

 

[raviteza88]

Heres the TIs defintion.

[JohnG]

You are the clearly the expert. I have nothing to add that will make any difference.

[HalFET]

The Tektronix IsoVu has 1 GHz bandwidth, several kV of isolation, and about 80 dB CMRR even near 1 GHz. There is a hell of a lot of work between this and the obvious use of AM modulation. They get >$10k for these probes and they are selling.

If you don't need this, then don't buy the probe. On the other hand, if you have ever needed to know what is going on with a high side gate drive looking at a gate signal for GaN or SiC FETs sitting on a 500V-1000V square wave with 200 V/ns transitions, then you might realize that it is not so simple. I don't know of any other probe that comes close.

I don't work for Tektronix, and don't even have a Tek scope these days. But I have seen their probe in action. I have never seen any other probe that can do the above measurement and produce consistently usable results.

John

1 GHz sounds an awful lot as if they're using those very standard Avago/Broadcom analog -> fiber modules. We looked into using them to get signals out of a MRI scanner, and these things are world beaters; quite frankly the only reason you don't see them all around the place is the rather cheerful pricing they apply (50-80 EUR each).
https://www.broadcom.com/products/fiber-optic-modules-components/industrial/industrial-control-general-purpose/1300nm/afbr-1310z
https://www.broadcom.com/products/fiber-optic-modules-components/industrial/industrial-control-general-purpose/1300nm/afbr-2310z

[Cerebus]

1 GHz sounds an awful lot as if they're using those very standard Avago/Broadcom analog -> fiber modules. We looked into using them to get signals out of a MRI scanner, and these things are world beaters; quite frankly the only reason you don't see them all around the place is the rather cheerful pricing they apply (50-80 EUR each).
https://www.broadcom.com/products/fiber-optic-modules-components/industrial/industrial-control-general-purpose/1300nm/afbr-1310z
https://www.broadcom.com/products/fiber-optic-modules-components/industrial/industrial-control-general-purpose/1300nm/afbr-2310z

Uh, uh. Quoted passband ripple for those is 6dB ptp and there's a low frequency cut-off of 50MHz. For simple AM that's not going to cut the mustard; at the very least you'd have to bandshift your signal and 6dB of ripple isn't going to make any scope look good, or even usable.

[PartialDischarge]

Uh, uh. Quoted passband ripple for those is 6dB ptp and there's a low frequency cut-off of 50MHz. For simple AM that's not going to cut the mustard; at the very least you'd have to bandshift your signal and 6dB of ripple isn't going to make any scope look good, or even usable.

Indeed it is advertised as usable from 200MHz to 5.5GHz, and to see how poorly spec'd the datasheet is in terms of graphs over frequency given the price.

[HalFET]

The actual ripple is better than advertised though.
Are we sure it's strictly AM modulated? Or do they use feedback fiber to compensate? And I'd be hard pressed to believe they rolled their own modules for this, $23000 wouldn't suffice for that.