To keep things simple, I will ask the question first and give the background later.
The question is: What leakage resistance would you expect between the pads of a 1206 size part on a fiberglass PCB. Assume that the board does not have solder mask and has been thoroughly cleaned using isopropyl alcohol and is at 50% relative humidity.

The cap does not appear to be the limiting factor since I have seen spec's for a ceramic SMD cap having as much as 10,000 Meg ohms of resistance.

The background:
I am considering building a version of the Atansoff Berry Computer (ABC) using CMOS logic instead of over 300 vacuum tubes and thyratrons (or even discrete MOS-FET's).
https://en.wikipedia.org/wiki/Atanasoff–Berry_computer

What intrigues me right now is the memory system that was used. It was mechanical and stored the bits as charges on a bank of capacitors. The caps were selected by a motor-driven rotary switch.
https://en.wikipedia.org/wiki/Regenerative_capacitor_memory

I would like to do my initial testing at slow speed -- ideally without a motor. To do this, I need to store the charge on the caps for a very long time. It would be nice if this time was on the order of tens of seconds. This would allow me to manually enter the bits using switches and view the data using an LED. For the initial tests I will use a 10 position switch but eventually I would make a PCB switch. The PCB switch would have something like 50 caps and contacts plus a motor driven wiper mechanism. Lots more work needed here...

 

I've never measured it, but for a clean board I'd guess 10's of Megohms minimum. PC board fab shops probably have data on this.

ak

 

Reed switch logic?

 

Hola RichardO

IIRC Bob Pease wrote a couple of columns about leakage on caps and PCBs. There is a salad of those somewhere on the Web.

 

IIRC Bob Pease wrote a couple of columns about leakage on caps and PCBs. There is a salad of those somewhere on the Web.

Thanks! Pease, of course. I think that I will start by looking in his book.

 

View attachment 92208 Reed switch logic?

I haven't decided if I will use 74Cxx or CD4000 series logic. It might have to be 4000 series instead.

 

Here is a first cut at what the test circuit might look like.

 

I finally have the test circuit working. There have been a few minor changes but the concept is the same. One change is that I used a 74HC00 instead of the 74C00 because that is the part I had in stock. This change required operation at about 5 volts so I can't run the circuit from a 9-volt battery.
A change that is not reflected in the schematic is the rotary switch. The rotary switch I used has 11 electrical positions and 12 mechanical positions. The 12th position is the common contact and is a convenient position for the switch when measuring data storage time.

The (ceramic) memory caps were soldered directly to the rotary switch. This way, the leakage currents are all in the switch/cap assembly and therefore are not dependent on PCB wiring.

I put alternating 1's and 0's into the memory and the pattern is stored for over an hour! This is with the dry air here in Denver but no special cleaning of solder flux from the switch and memory caps.


A few comments on making a version of the memory using a motor...

My version of the mechanical memory would have SMD caps mounted on a stationary PCB with switch contacts part of the PCB. The motor would drive a custom wiper to select the memory caps.

Using modern (40 year old, actually) "high speed" parts should allow running the motor at as high as 3600 RPM.
I looked in a Grainger catalog and highest speed sychronous motor I found is 600 RPM. This is ten times faster than the ABC's 60 RPM.

Effects of high RPM:

Balancing of the moving parts is harder. (Balancing the ABC must have been _really_ hard).

Faster wear which causes fine conductive powder that will "shorts" across and between the caps.
(I suspect that this is why the ABC has the caps inside a cylinder with only the contacts on the outside. Easier cleaning is traded off for having to spin the much higher mass of the caps and cylinder).

Contact bounce is a greater percentage of contact time.

Capacitor charging time is shorter.

 

I haven't decided if I will use 74Cxx or CD4000 series logic. It might have to be 4000 series instead.

My picture above was an old "4000 series Reed switch Logic Module" from Control Data corporation. I'm talking really old - early 1960s. Nothing to do with Cd4000 series. It is a big block about 4" x 4" x 0.5". I only posted it because you referenced mechanical. I figured a reed switch was boarder line mechanical.

 

My picture above was an old "4000 series Reed switch Logic Module" from Control Data corporation. I'm talking really old - early 1960s. Nothing to do with Cd4000 series. It is a big block about 4" x 4" x 0.5". I only posted it because you referenced mechanical. I figured a reed switch was boarder line mechanical.

Yep. I realized that. My reply was a bit too subtle it appears -- CD4000 verses 4000...

Thanks for the history of the module!

I once had some reed relays that had several small reed switches. The relays also had bias magnets built in to make some of the reeds latching. I took them apart for the switches. Probably should have saved some of the relays for historical reasons.

 

Yep. I realized that. My reply was a bit too subtle it appears -- CD4000 verses 4000...

Thanks for the history of the module!

I once had some reed relays that had several small reed switches. The relays also had bias magnets built in to make some of the reeds latching. I took them apart for the switches. Probably should have saved some of the relays for historical reasons.

I did save them for historical reasons and now I just have a box full of junk if anyone but me looks at it. I can see someone throw the boxes in the garage at the estate sale after I'm in the dirt - "50 cents each" or "whole box $5"

 

I did save them for historical reasons and now I just have a box full of junk if anyone but me looks at it. I can see someone throw the boxes in the garage at the estate sale after I'm in the dirt - "50 cents each" or "whole box $5"

I have seen my father's library (5.000++ books) vanishing in less than 12 hours) I managed to catch some 10 of them... so I could be not surprised of that happening to all what I will leave behind. In fact, I do not care much, really.

 

I have seen my father's library (5.000++ books) vanishing in less than 12 hours) I managed to catch some 10 of them... so I could be not surprised of that happening to all what I will leave behind. In fact, I do not care much, really.

My post above was not complete. The last thing I want to become us a burden on my wife and kids. I also don't want to be a burden after I'm gone by leaving a house full of shit. I'm always looking to move some of my crap out to good homes.