Isolated or Differential for floating measurements?

[MustardMan]

I'm a bit reluctant to post this (after reading the 'newbies sticky') as there are a plethora of threads that involve discussions on Oscilloscope Isolation and Differential Probes. However, pushing forward...

I thought I knew what a differential probe was, and what it's limitations are, and that was why isolated input scopes were available. Many discussions though seem to imply (some in fact recommend) the use of a differential probe for making measurements with large (greater than 200v) floating references. I will abstain from the word "ground".

Many threads also ask for requirements from the OP. Often the OP complies, occasionally not. Here are mine...
- Ability to measure small voltages (50mV to 5v) with a common-mode offset of 300v to 600v.
- Ability to also measure large voltages (300v to 600v) with the same common-mode offset.
- Bandwidth of 500kHz minimum, 10MHz nice.

One application might be a solar inverter with multiple 400-600v DC solar input strings and a 240vAC output (at least one floating rail, 'high' currents, and very very lethal).

However, a better example of a circuit to probe would be a switch-mode power supply with an isolated secondary circuit. In such a circuit I might want to probe the BE drive voltage and simultaneously look at the CE voltage while also probing at least one point on the secondary side. BE is going to be pretty small, while CE is going to be pretty big.


I was under the impression that a differential probe could not do this - as soon as I want to look at a small signal (eg: the intricacies of the BE drive signal) the common-mode ability of the device smashes down to the order of tens of volts. To do this I thought my only option was an isolated input scope. But some threads I've read here seem to imply otherwise. And it seems manufacturers go out of their way to hide this sort of detail.

I started looking at 'isolator attachments' for scopes, and got disillusioned pretty quick.
- Elector designed an isolated adapter (single channel) 60kHz that needed a jumper to select 250mv/2.5v/25v input. I could maybe live with 60kHz for the price, but I've only seen one for sale on ebay, once, ever. Schematic available, but I've not found a layout. And it can't do high voltages.
- Micsig do a 2 channel isolated adapter 1MHz with 200:1. The attenuation factor makes it useless for small signals.
- And as for the fiber isolated probes from the likes of LeCroy and Tektronix. Big dollars - in the case of the Tek devices, the entry model (one channel) was more than the Cleverscope with four channels! (admittedly the isolation was better, but I really don't need tens of thousands of volts isolation!!)

I looked at differential probes, but I must still be missing something in how they operate...
- A multitude of manufacturers with prices to suit both hobbyist and government buyers!
- Circuit Cellar did a fairly good write-up/design of a differential probe. Again although they show the schematic, I've not been able to find the layout.
- MicSig have a small range, of which I thought the DP10007 could fit my scenario (10x)
- Siglent also have a differential probe (DPB4080), but the fact they also have the ISFE (isolation adapter) says that there MUST be a difference, and specific use cases for both. [What are they???]

So then I started looking at scopes with isolated inputs...
- MicSig, Tektronix, Cleverscope, Metrix (and no doubt more if I looked harder).
- Dave did a good teardown of the MicSig 300 series. Hopefully the build quality has improved since then.
- It would seem that for a "lower cost" option (under 5K), two channels is all that you can get.
- I really dislike the prevalence of buttons where knobs should be. It seems even the better devices (and higher end manufacturers) have all gone 'full button'!
- I was considering a second-hand Tek TPS2000 series (discontinued for some time), but the waveform update rate is horrible! At least it had knobs!
- The Tek THS3000 looks good with 4 channels, but has sadly gone 'full button'.
- I could even spring for a cleverscope (4 channels) at 10K (US dollahs), but needing a laptop/PC to run it puts me way off. Not only do you now have no knobs, but something else on a bench where things are already falling off the sides and back due to lack of room.
- I did really like that for all these I could ramp the sensitivity from often under 50mV/div to often over 50v/div. Not something I could do with either Isolating adapters or Differential Probes.

Regarding Differential Probes: I thought (take the MicSig DP10007 at 10x) the rating of max differential of 70v meant that the max common-mode offset was also 70v. But the faceplate of the Siglent says that at 10x the max differential is 80v with a common-mode of 800v. But it also seems to say that at 100x the max differential increases to 800v, but the common-mode stays the same at 800v? Does anyone have any pointers on how to use a differential probe for small voltages with large offsets? And how that differs from an isolated channel? An internet search does not return anything useful, although that might be the particular search terms I use - rephrase the question and the answer appears...


I did buy a MicSig STO1104 only a couple of months back to replace my now very old Tek 222, and I must say it is awesome! I like that it has knobs where knobs should be. The knobs (and buttons) are seamlessly integrated with the touchscreen - I honestly thought I might use the touchscreen once in a blue moon, but I use it a LOT. However, when I need to fine adjust a setting, there is nothing as good as a knob!
Serial decode, triggering options (including triggering serial), cursors, four channels, battery operated, truly portable, big colour screen... The only thing I miss (really miss) about my Tek222 is the isolated channels.

Cheers, MM.

[nctnico]

Unfortunately there seems to be a gap in the low-cost market where it comes to isolated probes which support small signals riding on high voltages. I have been thinking about designing something but it isn't an easy problem to solve. There are all-in-one chips like the Analog devices ADUM4190 but that is 400kHz only.

[bdunham7]

The Tektronix TPS series (still made) is one option.  I have one, and while its specs as a scope are pretty lame compared to any new model, it still seems perfectly adequate for standard sort of viewing signals.  How many waveforms per second do you need or want?  Just be aware that it isn't magic as far as CMRR goes--it still can have issues with a small signal riding on a huge, noisy base.

I know you mentioned knobs, but I think the go-to instrument for this, especially if you use it in the field at all, would be the Fluke 190-series scopemeters.  Tek (the THS you mentioned) and Rohde & Schwarz have their own versions of handheld scopes as well. 

[MustardMan]

@nctnico
Yeah, there does seem a pretty big gap. I got my Tek222 scope for $50 in a pallet of surplus crap, threw the rest of the pallet out and kept just the scope. It is highly doubtful the sale house even knew it was there. I've used it for probably two decades as a portable and for isolated measurements. I have no idea what the things would have cost new.
I don't have a problem paying for good quality, but I don't like paying exorbitant prices simply for a name. For example, without doing sufficient research I bought a Tek MDO3104 - the serial decode "module" (the software code) cost more than double the MicSig scope I got with serial decoding as standard! Added to which the menus are all over the place and ridiculously difficult to use... having to go back-and-forth from one knob to the menu button to another menu button to another knob. And for the price I would expect a touch screen - but no. Then Tek has their proprietary probes. After owning one it is certainly NOT a scope I would ever recommend.

@bdunham7
I didn't know the TPS was still made!
From the User Manual (page A1 of the TPS2000 series, 071-1441-02 manual) 180 waveforms per sec. You'd want your trigger set right if you were looking for a one-in-a-million runt! That is terrible! Even the MicSig MS300 is 1000 times better! It is possible it is a typo, but Tek are usually pretty good with their specs.

That 4 channel Fluke looks pretty good (I skipped past the 2 channel versions - once upon a time 2ch was adequate, but those days are gone). I find it really interesting indeed that the US Fluke page claims:

The Tektronix THS3000 Handheld Oscilloscope has been discontinued, but customers can find a nearly identical replacement with the Fluke 190 ScopeMeter® Test Tool.

I don't like the Aus R&S dealer (at least the one I've dealt with) at all. I made an enquiry about a VNA last year and found him pushy and reluctant to give you even the slightest snippet of information unless you hand over your entire family history (and that of your grandmother too). Unfortunately that is one of the traits that turns me away from what might otherwise be good products. Needless to say he got no sale.

Cheers,

[bdunham7]

From the User Manual (page A1 of the TPS2000 series, 071-1441-02 manual) 180 waveforms per sec. You'd want your trigger set right if you were looking for a one-in-a-million runt! That is terrible! Even the MicSig MS300 is 1000 times better! It is possible it is a typo, but Tek are usually pretty good with their specs.

That 4 channel Fluke looks pretty good (I skipped past the 2 channel versions - once upon a time 2ch was adequate, but those days are gone). I find it really interesting indeed that the US Fluke page claims:

The Tektronix THS3000 Handheld Oscilloscope has been discontinued, but customers can find a nearly identical replacement with the Fluke 190 ScopeMeter® Test Tool.

Well, the TPS doesn't  have any fancy trigger options either, runt pulse triggering would be light-years beyond it.  So probably not your cup of tea.  It was really made for power analysis and so forth.

Fluke and Tek are both owned by Fortive.  Some Tek products are or were made by Fluke, but I don't know about the other way around.  User blueskull promised us a review of his new 190 series, but I haven't seen it yet. 

[AVGresponding]

A new 190 series will set you back two or three kilobucks, why not just get a couple of used 90 series?
Probably wouldn't cost more than $1k for a pair of really good ones.

[David Hess]

The conflict for a differential probe is that the input dividers which attenuate the common mode signal also attenuate the differential signal, so high voltage operation precludes high sensitivity.  The differential signal cannot just be amplified later because the attenuation also multiplies the input noise.  If the amplifier's input range is 1 volt, which is typical, then making measurements at 1000 volts requires x1000 attenuation, which turns 1mV/div at the oscilloscope into 1V/div.  Better differential probes use input amplifiers with a wider input range so less attenuation is required but this is only a x10 improvement.

Isolated probes do not have the problem of attenuating their input signal but providing galvanic isolation at 10s of MHz is not easy.  In the past, optical isolation was used for the low frequency part and transformer isolation was used for the high frequency part but modern designs, like that found in DSOs with isolated inputs, use digital isolation by moving the digitizer and signal conditioning to the isolated side.  A fully documented example of the analog method including schematics can be found with the old Tektronix A6902 isolation amplifier.

Another difference is that differential probes have symmetrical high impedance inputs, while isolated probes have one high impedance input and a relatively low impedance common connection.  The low impedance comes from the capacitance to ground which can be 10s to 100s of picofarads so this connection to the probe must go to a low impedance point of the circuit which can tolerate this capacitance.

As far as performance, other than the difference in sensitivity and noise, differential probes have limited common mode rejection at high frequencies compared to isolated probes because of mismatch between their input attenuators, mismatch in source impedance, and limited amplifier common mode rejection.  Isolated probes have essentially unlimited common mode rejection at high frequencies so are more suited to making sensitive measurements at high voltages and frequencies.  The common mode rejection of an excellent differential probe might drop to 200:1 at 20 MHz so a 400 volt common mode swing would produce a measurement error of 2 volts on a timescale of 17.5 nanoseconds.  This is why differential probe bandwidth above 20 MHz is of questionable use for off-line switching measurements.

[MustardMan]

@David Hess
Thanks for that explanation, it makes things considerably clearer.

Funny how the Siglent ISFE is a x200 device then! That divide ratio is what one would expect from a Differential Probe design. I wonder if it can be cracked open and hacked to be something more reasonable? Perhaps I'll buy one just to see... it's not as if I haven't wasted more money on less interesting pursuits!

Cheers,

[profanum429]

We have a 500MHz four channel R&S RTH1004 at work and for a handheld scope it is pretty damn nifty. Light years (to be expected probably though) ahead of my 199C at home and the THS3000 at work also. It definitely wasn't the cheapest (around 8500 for the COM4US package that has all of the software packs unlocked) but it definitely works nice. Only one knob though, buttons everywhere else but the build quality is very nice and the display is fantastic when I'm used to the older handheld scopes. It's quick also, it feels like owning an RTB2004 but in a portable form factor.

[David Hess]

Funny how the Siglent ISFE is a x200 device then! That divide ratio is what one would expect from a Differential Probe design. I wonder if it can be cracked open and hacked to be something more reasonable? Perhaps I'll buy one just to see... it's not as if I haven't wasted more money on less interesting pursuits!

The Siglent ISFE isolates two channels and is not inherently differential.  It looks like they made the attenuation x200 so that it could measure a 600 volt input with a 3 volt output, which implies that internally it operates on +5 volts from USB and generates a -5 volt supply internally.  An attenuation of x100 would have been a little too low to measure the rectified output of a 240 volt AC source, 340 volts DC or 400 volts DC if power factor correction is used, which would commonly be found in an off-line switching power supply.

The bandwidth of 1 MHz allows only optical isolation to be used with is much simpler than combining optical and transformer isolation but the associated 350 nanosecond rise time is too slow for many applications.

Altering it to have a much lower attenuation may be possible.

[MustardMan]

@David Hess
Yeah, I suppose that is going to be a problem for *any* externally attaching adapter (having to select a range on the device). And that applies to Isolating adapters as well as Differential Probe adapters. I suppose the likes of Tektronix can get around that because of the data pins on the front of their scopes. It's a pity they are so locked up with patents and proprietary data formats, and their prices pretty much say 'government buyers only'.

As an aside, an unsubstantiated rumour I heard was that the truth behind their change from Tekprobe-BNC to TekVPI was simply because the patents on the older version ran out, and it had nothing to do with functionality or bandwidth (which is their claim).

I guess the market for Isolated inputs is so small that Siglent putting range selection on their ISFE (as Siglent & many other manufacturers do with Differential Probes) is simply not worth it.

@profanum429
That price is truly a surprise! R&S have a reputation to uphold for high prices, and you paid 8.5K (assume USD?) for full bandwidth and all options! Crikey... did the sales guy miss a decimal point?!

[profanum429]

@profanum429
That price is truly a surprise! R&S have a reputation to uphold for high prices, and you paid 8.5K (assume USD?) for full bandwidth and all options! Crikey... did the sales guy miss a decimal point?!

Yep, USD https://www.newark.com/rohde-schwarz/rth-com4us/handheld-digital-oscilloscope/dp/99AC0520

Not sure if it's only available in the US or if other countries can get it; unfortunate if it's US only

[threephase]

Seems to be available in the UK from Farnell, but only up to 250 MHz bandwidth at the moment.

https://uk.farnell.com/c/test-measurement/oscilloscopes?brand=rohde-schwarz&product-range=rth-series

Nice instrument though

[MustardMan]

Having just checked, these are available in Australia through "Element14" (they should have kept Farnell - at least a search for 'farnell' does not bring up lots of spurious crap).

500MHz, 4 channel, Serial decode : AUD 17,400 (inc. GST). Doable after taking a (very) deep breath, although I'd want to make pretty sure it had everything.

I haven't found any others apart from those I listed in the table above, although I have found errors (my fault) in the features for some of them.

[profanum429]

Oof, that's a bit more money there, not sure if we'd have bought it at that price point. It'd be tempting to get a cheap isolated scope and use a bench one for decoding. For our stuff the packages were definitely a nice to have and not totally necessary in the device but it works well enough that I can use it as a bench scope without much compromise.

[MustardMan]

That was one of my thoughts too - ie: use an alternate scope for serial decode. Most (maybe all?) of the MicSig battery scopes have serial decoding as standard (no ridiculously priced "options" to add), so they can be used to decode a serial bus, although I'd still be very reluctant to hook it up to a mains-live circuit.

I've updated the comparison list... I found another 4 channel by Metrix. It is on their website, just not obvious - I discovered it through the Aus Element14 site. Interesting it claims 12 bits resolution. I wonder if that is genuine or "interpolated"? EDIT: Genuine apparently.

Updates in blue, corrected errors in red. Prices are first ones I found.

FURTHER EDIT: Metrix do a benchtop scope range (isolated channels) as well. They seem to have combined the worst of everything - not a single knob in sight (and a stylus for touch!). Sigh.

The MicSig seems the best value for money - especially if you are keen on isolated serial decode. Pity it's only a two channel device.