Isolating Oscilloscope Channels

Nice to have the good old scopes with differential plug-ins and

Are any of the grounds, grounded?

greg

No the "grounds" are the emitters of the upper leg and lower leg of the H Bridge. Maybe I should just arbitrarily assign "ground" the upper emitter which is also the lower collector invert one of the signals and tweak the scope vertical. I am still waitng for my test fixture to get built so I can snoop these signals.

Thanks for helping me start thinking again.

Ed V.

Get one of the Tek isolated scopes? They're fabulous.

If it's just an edge, transformer couple it.

John

See Tektronix P5200 series of HV diff probes and similar products. That's the right way to do it. Paul Mathews

As John mentioned transformer coupling can achieve that. Mini-Circuits has lots of those. Possibly you could also get away with LAN transformers from an old Ethernet card. But make sure nobody needs that card anymore ;-)

Wont work I.M.O.

The current through the primary - secondary capacitance will screw your measurement over AND if the (on the day they used the right goop for encapsulating) 500 V (or so) isolation level in the cheap(!) Chinese LAN trafo from "Electronics Factory No 5, Huawei" ever goes away then the H bridge - off-mains since the O.P. needs isolation - might then make lots of colours with your scope (before the muffled Boom from the sub-basement fuses cuts the show ;-).

The O.P. should buy the proper differential / isolated probes for the voltage level. They will cost money. But so will the Health and Safety incident, the trip through court, and compensation payments should some worker get injured. The word "production" ups the stakes quite a bit.

The Chinese will print whatever safety labels are requested from the user of the part - it does not follow that any of that actually applies; as f.ex. HP found out the hard way with a run of power supply "Y" capacitors ...

Possibilities:

- Turn off power (to the bridge) and measure directly

- Disconnect gate drives from bridge and measure at ground

- Couple with C, R and transformer (you're looking for the edges, right?) -- with care you can also measure the edge height

- Watch carefully at high voltage (15V of gate drive does kind of get lost in 600V or whatever though)

- Build your own differential probe ;-) Etc.

Tim

-- "Librarians are hiding something." - Steven Colbert Website @

This link looked interesting:

I am going to experiment with these circuits. Unfortunately the suggested transformers are not meant to withstand some of the voltages that might appear if something went wrong. :-( I think spending $1K each for two good active diff probes from Tektronix is going to make the most sense.

I do a pile of consulting for a company that often needs the same thing. They have some fancy large Tektronix 8 channel isolator they use. Me I us an old THS series battery powered Tektronix scope for this. It has two channels, each is isolated from ground and from each other. It's not fancy but it would pretty good for that purpose. They run about $800 used and I think $2500-$3500 new depending on sample rate.

Hawker

On 4/18/2007 9:15 AM, The digits of EdV's hands composed the following:

Just don't use audio mic transformers for microsecond stuff. Also, their breakdown voltage isn't very high, or there is no rating for that at all on some.

You can always roll your own or buy from a reliable source where there are breakdown specs. If you roll your own you could get some Rubadue wire:

Certainly makes sense, especially if you have to do this often or if you must conduct documented test. Saves a lot of ECO writing for test jigs which probably costs more than $1k in EE hours.

I am stuck with the oscilloscope that is in the tester because it used to test many different baord level assemblies. Kind of remeber asking "why don't we have isolated channels"? I believe it was cost. Oh well.

A shame about the microsecond thing. The signal should always be 20 Vpp now that I review the circuit, when it is working correctly. I could protect the transformer with clamp diodes but not much point if the timing would not be preserved.

Thanks, Ed V.

You can get fast transformers at places like Mini-Circuits. I don't know how high LAN transformers go and where they saturate but a datasheet of one should tell. Reason I don't know is because I always keep a pound of #43 ferrites in my lab so I can quickly wind one up if in need. Faster than filling out a Digikey order ;-)

Heh yep. I've had excellent response from even really crappy transformer designs. 100 turns of 30AWG on a high-permeability core (toroid or etcetera) will get you there. Primary then secondary, don't really even need to interleave them. Use *two* layers of masking tape between windings if insulation spec is really high. ;-) My experience gives bandwidth into the 10s of MHz (risetime comparable to the driving circuit: 100-200ns).

I would use a series capacitor to block DC and a (parallel) resistor to dampen oscillations. With small C, you can get away with pretty wimpy windings, since the pulse ends up really narrow. With large C (and L!), you can see the whole square wave with maybe just a little droop along the way.

Tim

-- "Librarians are hiding something." - Steven Colbert Website @

Looks like Ed is only interested in the transition times. My staple for that looks like this: About 6" PTFE wire, 6" of a CAT-5 strand or whatever, twist the two together in 1/2" turns or so, wrap four times through a 1/2" OD toroid of #43 material. I get sub-10nsec transitions through those. The PTFE insulation is only needed where there is a lot of bzzzt between primary and secondary.

Oops. I've done it a few hundred times by now. Or maybe thousands. Do I have to go to confession for that now?

Well, you have to do it right. Transformer coupling can be very helpful especially in production because space it usually cramped and it doesn't exactly help to cram two or more diff-probes in there. That's one reason why there is PTFE cable available where they test and certify every foot of it for dielectric strength.

Anyhow, that's also how we design patient interfaces. Lots of them out there in the field. So if you really don't trust such transformers it would be best to eat healthy, exercise, and absolutely not develop a coronary artery problem because chances are they'll use one of those machines on you. But rest assured, the ones I designed are all defibrillator proof and tested for that. I wouldn't do them any other way.

the

Yep, that sure can happen. Know thy sources ;-)

On 4/19/2007 1:11 PM, The digits of EdV's hands composed the following:

I have often told clients "if you want the work done you need the correct tools" if I can't get the correct tools I can't get the work done. If they need this tested then they need to get the correct tool for the job no two ways around it.

It amazes me how often a client is willing to pay me $5,000 - $25,000 for consulting work, but won't spring an extra $500-$1000 for the specialized tools I need for the job. I understand that basic tools are my responsibility to supply (a standard scope is a basic tool) but since every job is unique there are often a few special tools required (esp ICE tools for whatever processor I am using or HV tools for HV products).

Hawker

From a business point of view this is often the reason: Our consulting services are billed just like materials. They get purchased and then consumed. A tool becomes inventory and must be entered into the amortization tables. Depending on cost and type it has to remain there for several years. Inventory Dollars is a number that needs to be as low as possible so the CFO can leave the next board meeting a happy camper. Payroll Dollars is the other number that needs to be low and that's why US companies like consultants. Which is a good thing :-)))

This is also one reason why businesses often spend more money on long term rentals than we would, beyond the point where an outright purchase would have been a better deal.

Thanks! I'll try this one.

Ed V.