High voltage capacitive pick up and input protection circuit

Thread Starter

patpin

Joined Sep 15, 2012
401
hi p,
The single diode is upside down, refer to my earlier sim image. [other thread]
Eric

PS:

See my post on your other thread regarding Cap values, you show all in Farads!!
Thanks !!.... First experience with LT Spice. Though C was defaulted to pF... i'll correct and add the scope probe 10:1 which I forgot to.

Whats the LT Spice equivalent of a 10:1 ratio probe?
Which TVS would you advise?
 
Last edited by a moderator:

Thread Starter

patpin

Joined Sep 15, 2012
401
I'll give it a try.. correct my failures please.
I have a ratio of 1000:1 on the left diagram which is what I wanted.
Remains the selection of the TVS's...
 

Attachments

Last edited:

MrAl

Joined Jun 17, 2014
8,063
Hi there Eric, and patpin and others,

A couple notes...

1. I prefer not to use Neon bulbs for anything because their characteristics can change over time and abuse.
2. If you used diodes, use anti parallel diodes to be sure negative transients are also clamped. Kickback from coils can go to both polarity extremes.
3. Not sure what that 100k resistor R3 is doing for the circuit. What is that for?

After a friend had a power strip fail because of an open thermal fuse, i meant to test the TVS but did not get to it yet, and i wanted to get a couple new ones for testing rather than use an old one that may or may not still conform to the specs. Which reminds me, you may want to build in some sort of test so that you can test the ability of the circuit to clamp properly before actually connecting it to anything sensitive. A test phase would ensure everything is working properly. Say a push button or something that disconnects the scope and applies a high voltage pulse to the input and measures the output to ensure that it is low enough. Components age and change characteristics, but a quick test beforehand could ensure nothing expensive gets blown out. In fact, that might be the most important point in the design :)
 

Thread Starter

patpin

Joined Sep 15, 2012
401
Hi there Eric, and patpin and others,

A couple notes...

1. I prefer not to use Neon bulbs for anything because their characteristics can change over time and abuse.
2. If you used diodes, use anti parallel diodes to be sure negative transients are also clamped. Kickback from coils can go to both polarity extremes.
3. Not sure what that 100k resistor R3 is doing for the circuit. What is that for?

After a friend had a power strip fail because of an open thermal fuse, i meant to test the TVS but did not get to it yet, and i wanted to get a couple new ones for testing rather than use an old one that may or may not still conform to the specs. Which reminds me, you may want to build in some sort of test so that you can test the ability of the circuit to clamp properly before actually connecting it to anything sensitive. A test phase would ensure everything is working properly. Say a push button or something that disconnects the scope and applies a high voltage pulse to the input and measures the output to ensure that it is low enough. Components age and change characteristics, but a quick test beforehand could ensure nothing expensive gets blown out. In fact, that might be the most important point in the design :)
at 1: OK I can omit that and replace it with a TVS which is apparently more robust. I read that only with MOVS the leakage current augments after a serie of HV events, but not TVS's.
at 2: which diodes would you take and where: instead of D1-2-3?
at 3: Eric considers this the equivalent of a spark plug
Indeed a good test whithout the scope would be perfect.... but how to evaluate the circuit then. Simply measure some currents through the clamping circuits or temperatures and see what degrades after eg 100 pulses of >5KV?
I'm thinking of a Vbr 5.6V TVS with Ppp of 600W like: the 3V3 in http://www.farnell.com/datasheets/1600353.pdf
Could one put two of those back to back? Is there an LT Spice equivalent?
 
Last edited:

Thread Starter

patpin

Joined Sep 15, 2012
401
I' introduced the 2 TVS's (unidir.) after creating a symbol in spice and can compare with the diodes.
I guess I also could leave the diodes in place and add the tvs's?

PS. V004=Vout; V003=between R1 and TVS (cant find how to label a trace..)
PS2. I temporarily let the neon in place so I can see if there is a firing goiing on (for testing)
 

Attachments

Last edited:

ericgibbs

Joined Jan 29, 2010
12,214
3. Not sure what that 100k resistor R3 is doing for the circuit. What is that for?
Morning Al,
The 100k is just to keep me and LTSpice happy.
I agree about the neons, I would not use them in this circuit.
My original circuit does have anti parallel diodes.

A pre-self test of any HV circuit should always be carried out BEFORE connecting to sensitive measuring equipment, especially in the VERY electrical noisy environment of a vehicle engine bay.

E

#patpin.
I would suggest more realistic cap values in your LTS sims, 1pf is close to nothing.;)
You have a disembodied GND in the lower end of C2

E
 

Thread Starter

patpin

Joined Sep 15, 2012
401
Morning Al,
I would suggest more realistic cap values in your LTS sims, 1pf is close to nothing.;)
You have a disembodied GND in the lower end of C2

E
Hello Eric, good morning too.
Thanks for the tip on the C2; The value of the 1pF was an estimation of the developper of the PU probe described in http://www.lotuselan.net/forums/download/file.php?id=9946
Do you think that a split metal tube of 6mm around a sparkplug cable has a larger capacitance? Guess we have to account for the shielded cable of 1.2 meter too...
Can anyone give a tip on how to be able to see the label, that I add in a diagram, in the graph.
 
Last edited:

Thread Starter

patpin

Joined Sep 15, 2012
401
I've been looking at the shape of the output (on top graph). There is a considerable deformation of the shape of the pulse. With 10pF or 100pF instead of 1pF even more.
Do you think the shape of the HV pulse will still be 'interpretable"? Or is there something we can do about it?high volt prot 03vraag.jpg
 

MrAl

Joined Jun 17, 2014
8,063
Hello again,

I must have missed seeing D1 as that is connected anti parallel, and that is good.

Capacitors do not behave as short circuits when connected to other capacitors, they only behave that way when connected to non capacitors.
For the example here we seem to have a 1pf connected to a 1000pf which creates a capacitive voltage divider. Since 1pf is 1000 times smaller the theoretical output would be 1/1000 of the input. So for 1000v peak in we'd see 1v peak out. That is for idea capacitors. For real capacitors, the ratio also depends partly on the ESR and any inductance, especially when working with caps as low as 1pf, which brings up the question, are you sure the probe is really 1pf? That sounds too low. Did you provide the physical construction of the sleeve? We can go by that as there are many formulas on the web for concentric circular cable (coax section) capacitance.

When we measure voltage or current for an unknown source the best method is to set the meter to the highest range, take a measurement, then adjust to a lower range when we know for sure the amplitude will be acceptable for the meter set at a lower range. We could probably do the same thing here. Create an resistive attenuator that maybe allows steps of 1/1000, 1/100, 1/10, and 1. Starting at 1/1000 take a reading and then work your way down. It would be good if you could generate a high voltage pulse for the test that way everything could be tested first.

Another idea is to use a buffer between the circuit and the scope. If something goes wrong, the buffer burns out but not the scope. An op amp would probably do it, with the standard input protection anti parallel diodes.

Also, because some expensive equipment might be involved, i would not feel comfortable about this unless i included a warning that the final consultation should be with an expert in high voltage circuits that has a lot of practical experience in high voltage circuits. Sometimes things that are particular to a given application type are found out by trial and error and the only people who know about these things are the people who have to make a living at it. They may be written down somewhere or may never be written down ... they may be on the web or never put on the web, so no one knows except them. For example, in the power conversion field transformer core flux ratcheting was not talked about much anywhere except in the lab, but could easily blow out the driver transistors on the very first run. Someone who did not know about that could spend days or weeks trying to figure out why the transistors kept blowing out.
 
Last edited:

Thread Starter

patpin

Joined Sep 15, 2012
401
Thanks for the info Mr AI! Not every thing can be calculated as you said.
I think that putting a voltage follower after the capacitive attenuator 100:1 as 'suicidal' chip is a good idea. Maybe it could resolve partially the deformation of the shape of the pulse to but w'll have to do some field work here... or can this also be resolved on LTS. And which OpAmp would U prefer?
 
Last edited:

Thread Starter

patpin

Joined Sep 15, 2012
401
Hi,

LM358 is cheap, but what bandwidth do you expect?
The firing orrurs every two revolutions of the crankshaft and sequentially. A few examples:
RPM 800 (idle)
cylinders 12
firing freq 80 per sec

RPM 3000 (max RPM for test session)
cylinders 12
firing freq 300 per sec
This 300Hz is the highest freq of the large oscillation. many smaller (but important) oscillations follow the large one as here:
upload_2016-3-5_23-15-16.png
 

Thread Starter

patpin

Joined Sep 15, 2012
401
Hi,

The LM358 should handle that, especially at low gain like 1.
Hello, I tried with one of the comparators already present in LTSpice and the shape is much better than in #26.

I attach the graphs and diagram (much trial and error!) too. The top trace is the output. The bottom one the orig pulse of 10K.
I stll find that the part after the peak is to negative. Is there a solution?
 

Attachments

Last edited:

ericgibbs

Joined Jan 29, 2010
12,214
Morning patpin,
The circuit is far to complex for the simple task that is required, to sync a Scope trace.

I suggested earlier that you could sync on the injector current low going leading edge as you are only interested in the trailing rising edge of the current waveform.

As a side note regarding your sims, N004 ident on the sim plots is meaningless to us as we do not know which circuit Node N004 is...?

Use key F4 to create a unique label for a node being plotted, Press F4, type in say Vout and attach the label to the node that you are plotting, it will then appear on the plotting window.

E
 

Thread Starter

patpin

Joined Sep 15, 2012
401
Morning patpin,
The circuit is far to complex for the simple task that is required, to sync a Scope trace.

I suggested earlier that you could sync on the injector current low going leading edge as you are only interested in the trailing rising edge of the current waveform.

As a side note regarding your sims, N004 ident on the sim plots is meaningless to us as we do not know which circuit Node N004 is...?

Use key F4 to create a unique label for a node being plotted, Press F4, type in say Vout and attach the label to the node that you are plotting, it will then appear on the plotting window.

E
Hello Eric, This is not for the TRIG input of the scope but to study the wave of each firing for each cylinder at RPM 800 and 3000. The shape of the wave after attenuation/ampli is very important. This morning I tried this . In the mean time we have a sacrifial cheap IC in case of firing to the probe.
For the trigger I 'll do as you said, but there also I would like to protect it for the same reason. What do you think of it? (upper trace is output; lower is input; thanks for tip on labeling. I ll try it nxt time)
 

Attachments

Last edited:

Thread Starter

patpin

Joined Sep 15, 2012
401
Hello, have been optimizing the PWL to imitate the real pickup signal.
I now realise that the signal is decoupled. The platform of the V at the sparkhigh volt prot 1.jpg is very important. Dont see it, neighter at output, neighter at input (node V005) of Opamp . How can I resolve that?
 

Attachments

Last edited:
Top