DRSSTC OCD detector... / Projects / Forums

Forums

4hv.org :: Forums :: Projects

« Previous topic | Next topic »

DRSSTC OCD detector...

1 2 next

Move Thread

LAN_403

Nicko

Tue May 04 2010, 10:46AM

Registered Member #1334 Joined: Tue Feb 19 2008, 04:37PM
Location: Nr. London, UK
Posts: 615

Just been doing some playing around for a high frequency DRSSTC.

One of the things that has struck me is that most OCD detectors are a bit suspect in terms of linearity & frequency response. Some very common designs out there are also very sensitive to capacitive loads or impedance mis-match. These issues make such designs difficult to set up reliably and makes their behaviour somewhat unpredictable.

I've been messing around and have tested the design below to 1.5MHz with only a 2% loss in linearity between 50mV and 5V AC input. Its very stable and shows no signs of oscillation or noise susceptibility. The behaviour is completely predictable, allowing accurate current limiting to be set. Note that even though this is a single rail design, the input CT signal from the CT burden resistor can swing between +/- 5V, i.e. the input signal is ground referenced.

It is however a "work in process" as there are some features I wish to add. Whilst the circuit may look complex, its actually quite simple! Its part of the V2 of my u(DR)SSTC project which has a thread elsewhere in this forum...

Here's a clip from an LTspice simulation with 1Mhz +/-4V input with the OCD level set at 3.8V. The delay between hitting the current limit and the OCF signal being asserted is about 120nS - in practice on the built circuit I see closer to 100nS. Green is the input signal, blue the output from the precision rectifier (about 30nS behind the input), red is the voltage on C2 and cyan the logic output. Note that most of the overall delay between input and output is due to the schmitt buffer only switching at about 1.4V, so V(C2) has a long way to fall before OCF is asserted. By using a different buffer or changing its hysteresis, the end-to-end propagation can be reduced to approx. 65nS. Its surprising the amount of time even a small cap can take to drain via a diode...

Just changed the output Schmitt input logic gate for another comparator with a hysteresis of 3.6<->4.0V, and this is the result (cyan is CT output, blue is the voltage on C2, green is output and red the current limit):

i.e. end-to-end propagation drops to 65nS. Probably way more than fast enough.

Cheers
Edit: Misprint fixed - C1 should be 10pF & R6 should be 68K
Edit: Added comments to the schematic to explain operation
Edit: Replaced Schmitt buffer with fast comparator.
Edit: Removed additional hysteresis from IC2 - its own internal hysteresis is enough already!

lightlinked

Tue May 04 2010, 11:50AM

Registered Member #2087 Joined: Tue Apr 21 2009, 08:32AM
Location:
Posts: 115

i am also interested in the idea of using a precision rectifier to do OCD and maybe measuring bridge current on a panel meter with a peak follower, but I am wondering if we could take it further and also use the same signal for zero cross.

Wolfram

Tue May 04 2010, 02:05PM

Registered Member #33 Joined: Sat Feb 04 2006, 01:31PM
Location: Norway
Posts: 971

I really like the fact that it is single supply. Are the particular chips used selected because they are available in LTSpice, or is there a different reason you chose those particular chips? I suppose a TL3116 could well replace the comparator for a third of the price, and hopefully a cheaper alternative could be found for the Op-amp too, as both chips are quite expensive compared to the rest of the parts on a typical DRSSTC controller board.

I will definitely implement this in the next version of my DRSSTC driver board.

A.M.

Finn Hammer

Tue May 04 2010, 06:11PM

Registered Member #205 Joined: Sat Feb 18 2006, 11:59AM
Location: SkÃ¸rping, Denmark
Posts: 741

Nicko wrote ...

Just been doing some playing around for a high frequency DRSSTC.

One of the things that has struck me is that most OCD detectors are a bit suspect in terms of linearity & frequency response. Some very common designs out there are also very sensitive to capacitive loads or impedance mis-match. These issues make such designs difficult to set up reliably and makes their behaviour somewhat unpredictable.

Cheers

The "Predikter" was, I believe, first to use a precision rectifier in the OCD circuit, and a very elegant one, even if I say so myself, with only a single op amp in line. I am not aware of any tendency to unwarranted oscillation, capacitively induced impedance mismatch or set up problems either.

In fact, I have taken the output from it as the gospel truth.
Correct me if I am wrong.

Cheers, Finn Hammer

Nicko

Tue May 04 2010, 08:42PM

Registered Member #1334 Joined: Tue Feb 19 2008, 04:37PM
Location: Nr. London, UK
Posts: 615

Finn Hammer wrote ...

The "Predikter" was, I believe, first to use a precision rectifier in the OCD circuit, and a very elegant one, even if I say so myself, with only a single op amp in line. I am not aware of any tendency to unwarranted oscillation, capacitively induced impedance mismatch or set up problems either.

In fact, I have taken the output from it as the gospel truth.
Correct me if I am wrong.

Hi,

I wasn't aware there was a land-grab going for who published what first in 4HV - I'm just interested in the performance of various precision rectifiers. The design you use in Prediktor was, I believe, first published in the RCA datasheet for the CA3130 in 1974 (not a misprint) and subsequently repeated in the CA3140 and others, so its hardly original (36 years ago). I'm also not sure I mentioned oscillation or set-up problems - impedance variations, non-linearity and consequential unpredictability (as it calculating max current for the OCD) in some designs, yes. I certainly didn't single out your design at all, or even mention precision rectifier OCDs as being the culprits!

I agree the design used in Predikter is simple, but like all designs, tends to be a compromise. Of course it works, but it requires split supplies, has an input impedance that is 2K for negative cycles and up to 13K for positive cycles, and requires a pot to adjust to get the positive & negative peaks the same. In some cases, though I'm not sure about the specific op amp you use, it is insensitive below about 100mV.

My requirements were for a single supply and consistent linearity from low mV to 5V and up to 1MHz - pretty much rail-to-rail - the second op-amp provides an "ideal diode" to clamp the non-inverting input to 0V on the negative half-cycle - note that there is still only one op-amp in-line. I'm also not claiming originality for the design - I quoted where it came from originally (TI App Note SBOA068). I just think it might be an interesting approach, especially if you don't want to fuss with dedicated +/- supplies for the analogue parts of the design - excellent 5V (and lower) single rail devices are now available, and simplifies (IMHO) the overall design - if you already have 5V (or 3V) for the logic, why confuse things when you don't need to - just use that.

With both op-amps in a SOT-23 case and still only a single diode and three resistors, its not a space issue either - it also doesn't use a negative regulator or a trimpot to level the output, i.e. overall, a lower component count. It does have different input impedance for positive (pretty much infinite) and negative (about 5k) half cycles, but allows for a ground-referenced signal of low impedance, i.e. it is ideally suited to a low value burden resistor as its source (10ohms = 0.2% difference).

Complexity is a semantic point!

Cheers

Nicko

Tue May 04 2010, 08:58PM

Registered Member #1334 Joined: Tue Feb 19 2008, 04:37PM
Location: Nr. London, UK
Posts: 615

Anders M. wrote ...

I really like the fact that it is single supply. Are the particular chips used selected because they are available in LTSpice, or is there a different reason you chose those particular chips? I suppose a TL3116 could well replace the comparator for a third of the price, and hopefully a cheaper alternative could be found for the Op-amp too, as both chips are quite expensive compared to the rest of the parts on a typical DRSSTC controller board.

I will definitely implement this in the next version of my DRSSTC driver board.

Ok, I admit it! I chose those two because I could simulate the design pretty accurately in LTspice and I can get free samples from them! However, I've also used a similar Maxim part, the MAX4413 (USD 1.05) with some success.

I like single supply solutions - personally, I fail to see why in this configuration you need dual rails when perfectly good single supply solutions are now available that are pretty noise immune so are completely happy using the logic supply.

There are some astonishing chips and some new, novel, designs out there that need looking at!

Cheers
PS. Did you get the capacitors I sent?

Steve Conner

Wed May 05 2010, 09:23AM

Registered Member #30 Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706

Well, if we're landgrabbing, I had a precision rectifier first! :P

I actually use a window comparator for the OCD. It operates on the unrectified signal, using one comparator to detect the positive peaks, and another for the negative peaks. If you like, I'm rectifying after the comparators. The precision rectifier is for the meter.

I thought I invented this rectifier circuit, but it turned out to be similar to one designed by Douglas Self in the 1970s, for an audio compressor.

And it all runs off a single supply using 1970s technology, no cutting edge ICs needed.

Dalus

Wed May 05 2010, 10:18AM

Registered Member #639 Joined: Wed Apr 11 2007, 09:09PM
Location: The Netherlands, Herkenbosch
Posts: 512

Why the hell do you guys have to argue. Almost every basic circuit has been built before. The only thing you can excel in now is being able to combine these basic circuits into a project efficiently. Just cite the sources if you use parts from other peoples circuits.

on-topic: Has anyone compared the basic diode comparator circuit with this one? I found that even a CT connected to my scope differed about 200A from the value set by the OCD circuit.

Though this isn't much of a problem in very large coils where those few 100A are peanuts.

Nicko

Wed May 05 2010, 11:34AM

Registered Member #1334 Joined: Tue Feb 19 2008, 04:37PM
Location: Nr. London, UK
Posts: 615

Dalus wrote ...

Why the hell do you guys have to argue. Almost every basic circuit has been built before. The only thing you can excel in now is being able to combine these basic circuits into a project efficiently. Just cite the sources if you use parts from other peoples circuits.

Heh. I wasn't aware that either Steve or I were arguing

- both of us cited prior art and neither claimed originality. I've been doing fairly "bleeding-edge" tech stuff for about 35 years, and I agree, there is very little new under the sun. How its packaged and cross-pollination from other fields is what changes - We all continually see "new" ideas in both software & hardware that are small variations or re-arrangements on what has gone before.

Dalus wrote ...
on-topic: Has anyone compared the basic diode comparator circuit with this one? I found that even a CT connected to my scope differed about 200A from the value set by the OCD circuit.

Though this isn't much of a problem in very large coils where those few 100A are peanuts.

The reason I stated looking at this is that my main interest is in smaller coils - big coils simply worry me as I don't know enough about power electronics, but I hopefully do know enough to recognise my SOA (Safe Operating Area) - when I did my degree, it was mostly low-voltage - all those grads, divs & curls of scalar & vector fields were nasty stuff...

I'm still not completely convinced that a precision rectifier is necessary, even for small DRSSTCs, so long as you get your CT right. If the burden is after the bridge and your turns ratio is high enough, the forward drop of the bridge may be insignificant. I simply don't know yet.

A big point here is also the linearity of the CT - the frequency of a given TCs is fairly constant, but the current through the primary obviously varies wildly. I have some Pearson CTs that are wonderfully linear, but also very expensive. My understanding is that once set up & calibrated, a given CT will be fine so long as it doesn't saturate.

Because I like small coils, I like to keep the component count down if I can, hence simple is generally good. Personally, I'd rather have one unregulated 12 or 15V supply for the GDT driver, and a clean 5V supply for everything else (driven from the same 12VDC input). As the 5V demand is low (just a few mA), a simple small 100mA linear regulator generates that with little dissipation - that should be the only regulator needed.

I just don't see the need for the analogue section to have dedicated dual rails with their associated regulators & smoothing etc. when there are components and configurations that allow the use of the existing logic supply. If required, a simple LCR (or similar) filter can isolate the analogue use of the rail as is done in most uP ADC supplies.

Cheers

Finn Hammer

Wed May 05 2010, 06:07PM

Registered Member #205 Joined: Sat Feb 18 2006, 11:59AM
Location: SkÃ¸rping, Denmark
Posts: 741

Nicko wrote ...

Hi,

I wasn't aware there was a land-grab going for who published what first in 4HV
Cheers

Nick, C'mon!

Landgrabbing is the name of the game, everybody likes to appear smart, sophisticated, intelligent, creative and whatnot. Don't everybody wish they had a bright idea?

That someone got offended by it is not even new,.

I guess it is my privilege, as a chiropodist, to step on someones corn. Well, I didn't intend to cause any harm.

Cheers, Finn Hammer

1 2 next

Moderator(s): Chris Russell, Noelle, Alex, Tesladownunder, Dave Marshall, Dave Billington, Bjørn, Steve Conner, Wolfram, Kizmo, Mads Barnkob