Hi Forum,

My names Chris and I'm a senior in high-school and at the moment I am working on my senior project. For my project I plan to create a pair of electrostatic speakers.

These speakers require a high voltage dc bias which I am trying to satisfy with a crockoft-walton generator. I created a PCB to do this, but it is acting up.

To create my circuit I unashamedly based it on a fully functional one by Jazzman and slightly modified it. The only component changes I have made are doubling the the output resistor(R3) from 11M to 20M and reducing the capacitance of the caps from 5 uf to 4.7uf and choosing polyprop caps.

For safety and guidance, I am working with my physics teacher and my mentors at Microsoft, but they are stumped and don't really have too much time to look into it so I was wondering if you guys could help me out.

The Problem

At different tapping points my circuit should be putting out -2,000-4,000V DC instead my circuit is putting out around -250V DC.

The circuit seems to work for the first node. The voltage across C6 is -590V DC up. However after this my circuit acts like a reverse voltage multiplier.

My Testing:

• The diodes are facing the right way
• The transformer is putting out the proper voltage
• The routing is correct

Many Thanks in advance

-Chris

 

Are your caps really 4.7uF, or did you mean 4.7nF? The ones in your parts list are 5nF.
I don't think you can buy 5uF 1kV polypropylene caps.

 

I simulated the circuit. With no load, I got exactly +2kV out. With a 10Meg load to GND, I got about +350V, with about 270V p-p 60Hz ripple.
You can't measure the DC output with a normal multimeter, which typically has 10Megohm input resistance. I guess most meters won't take that much voltage anyhow.
You will need a high voltage meter, unless you are already using one.
I don't understand why your output voltage is negative.

EDIT: Even a 100Megohm load drops the voltage to about 1.1kV. See Electrostatic voltmeter.

 

You have a 'leak' somewhere in the high voltage section. If you can get a meggar from someone, it would help you locate the leaky component.

I would personally start at the capacitors and replace all of them with a different type. CERAMIC high voltage caps are called for and you should try them BEFORE considering any thing else as faulty.

 

You have a 'leak' somewhere in the high voltage section. If you can get a meggar from someone, it would help you locate the leaky component.

I would personally start at the capacitors and replace all of them with a different type. CERAMIC high voltage caps are called for and you should try them BEFORE considering any thing else as faulty.

Or, he's measuring with a meter with finite input resistance, as I said in my post. That is a major leak. The components could be perfectly good.

 

Thanks you guys for the quick replies.

First off Ron nice catch, I am using 4.7nf caps not 4.7uf caps

Let me tell you a little about how I've been taking my measurements. 2-4KV would blow up the multimeter I am using (Fluke 289 I believe).
So my mentor and I put together 10 10M resistors across the reference point (the net that goes from the bias supply to the audio transformers) to the net leading from R3.

I have been consistently measuring 74V DC across each 10M resistor and 250ish V DC across all 10...?

Kermit, I'm pretty sure I can get my hands on a meggar so I'll try that when I go to Microsoft again. I bounce your idea of replacing the polyprop with ceramics off my mentors and see what they thinks.

-Chris

 

Thanks you guys for the quick replies.

First off Ron nice catch, I am using 4.7nf caps not 4.7uf caps

Let me tell you a little about how I've been taking my measurements. 2-4KV would blow up the multimeter I am using (Fluke 289 I believe).
So my mentor and I put together 10 10M resistors across the reference point (the net that goes from the bias supply to the audio transformers) to the net leading from R3.

I have been consistently measuring 74V DC across each 10M resistor and 250ish V DC across all 10...?

Kermit, I'm pretty sure I can get my hands on a meggar so I'll try that when I go to Microsoft again. I bounce your idea of replacing the polyprop with ceramics off my mentors and see what they thinks.

-Chris

I went to the Fluke web site but could not find the input resistance when measuring volts.

Did you read my post about loading?
If the meter's resistance is 10Megohms, and you are measuring 74V across 10Megs out of a 100Meg load, the actual output voltage is 1026V. This is consistent with the simulated results (1091V) I got with a 95Meg load, which is your load if you have 10x10Megs with a 10Meg meter across one of them.
I think you have a measurement problem.

EDIT: I found the input resistance spec. It is 10Megohms.
EDIT: If you are concerned about the 20Meg resistor on your output, keep in mind that it is not connected to GND, so it is not a DC load on your multiplier.

 

Thanks Ron I get it now .

I think there is one more quirky thing that is going on with my circuit. I created tapping points for the circuit each separated by a stage for added versatility: J5-J7.

For some reason the voltages decrease as I move to higher voltage junctions.

For J6, measuring one resistor of the 100M resistor bridge, I get 66V DC and across the entire thing I get 244V DC.


For J7, measuring one resistor of the 100M resistor bridge, I get 58V DC and across the entire thing I get 242V DC.

Any thoughts?

Here is my schematic:

 

I tested the speakers out with my physics teacher. The circuit did provide enough bias voltage to get some sound out of my proof of concept.

However I am still wondering if anyone had any insight on my last post.

 

I tested the speakers out with my physics teacher. The circuit did provide enough bias voltage to get some sound out of my proof of concept.

However I am still wondering if anyone had any insight on my last post.

You really don't appreciate the loading issue.
Do you have a simulator? I used LTspice, which is free, and very good.

First, let me be sure you understand that, when you measure across one of the 10 Meg resistors in your 100 Meg string, and you are using a multimeter with 10 Meg input resistance, and you have an additional 20 Megs in series, your actual voltage will be
Vact=Vmeas*115/5=23*Vmeas.
The 115 Meg load is moved to each node before the simulation is run.
The only load at "no load" is a 1e12 resistor from node E to GND.

 

Show us a photo of the high voltage construction.

You might have mounted all these components onto board of unsuitable material or board with low insulation resistance.

Are you really using 1N4001?

 

Isn't the 1kV PIV variant of the 1N400* family the 1N4007? 1N4001 seems to be only 50V. http://www.fairchildsemi.com/ds/1N%2F1N4001.pdf