Reduce capacitance

[raphaeldavidf]

Hey guys,

The capacitance from the pins SCLK and MXSX are messing up the communication and I'm looking for a way to lower that capacitance.

The capacitance now, measured from GND:
SCLK:   7 nF
MXSX:  7 nF

I need to keep then under 70 pF.

is there a way?

Many thanks!

[T3sl4co1l]

Um, measured how?  Where is 7nF coming from?  Have you given this measurement a basic reality check?

I have no idea what the NLX90316 is, but I'm guessing it's not 7nF.

What is the rest of your circuit, and how is it connecting?  Layout?  I would be much more concerned about signal quality and EMC than the exact magnitude of capacitance on the lines.

Tim

[raphaeldavidf]

Um, measured how?  Where is 7nF coming from?  Have you given this measurement a basic reality check?

I have no idea what the NLX90316 is, but I'm guessing it's not 7nF.

What is the rest of your circuit, and how is it connecting?  Layout?  I would be much more concerned about signal quality and EMC than the exact magnitude of capacitance on the lines.

Tim

Hi Tim, thanks for helping out.

I measured it with a cheap LCR meter that I got from ebay. The 7nF is coming from the IC I suppose.

The MLX90316 is a rotary position sensor an I'm using it with a TXB0108 voltage level translator so it can talk with my pic. The problem is that the TXB can't work with capacitive loads higher than 70 pF. Other stuff, that have capacitance lower than 70 pF, are working ok with the TXB, so I guess that the 7nF might be right.

Attached the circuit that is trying to talk with the MLX and the MLX board layout.

[tggzzz]

I measured it with a cheap LCR meter that I got from ebay. The 7nF is coming from the IC I suppose.

Those usually work by injecting a voltage through a resistor and timing how long it takes to reach a certain voltage. There will be many non-linearities in your measurement that will "confuse" that way of measuring capacitance.

Add up all the known capactitances in your circuit (look at the data sheets) and I would be very surprised if you can get within two orders of magnitude of 7nF. If there is an identifiable  capacitance of that magnitude, tell us what it is and then we can suggest how to reduce it.

In any case, what makes you so sure that capacitance is the cause of your problems?

Are the lines open-collector or open-drain? If so they will need a pull-up resistor. Again, the data sheets will tell you.

[raphaeldavidf]

I measured it with a cheap LCR meter that I got from ebay. The 7nF is coming from the IC I suppose.

Those usually work by injecting a voltage through a resistor and timing how long it takes to reach a certain voltage. There will be many non-linearities in your measurement that will "confuse" that way of measuring capacitance.

Add up all the known capactitances in your circuit (look at the data sheets) and I would be very surprised if you can get within two orders of magnitude of 7nF. If there is an identifiable  capacitance of that magnitude, tell us what it is and then we can suggest how to reduce it.

In any case, what makes you so sure that capacitance is the cause of your problems?

Are the lines open-collector or open-drain? If so they will need a pull-up resistor. Again, the data sheets will tell you.

Hi tggzzz, thanks for the help

I don't have any other way of measuring capacitance 

I looked the datasheet and it doesn't give any capacitance, so the only known capacitance are those two 100nF caps in the schematic. Is there another way to find them out?


I think capacitance is the cause because of this pdf that I found when searching about the problems I'm getting:

With regard to capacitive loads, TXB translators are designed to drive up to 70 pF without issue. If
capacitive loading is larger than 70 pF, the O.S. will time-out after 10 ns and subsequently turn off before
the output voltage is driven fully high or low. From there, the output will continue to rise or fall based on
the RC time constant determined by the 4-k? buffer, load resistance and capacitive loading. To ensure
reliable operation, system designers should keep capacitive loading for the TXB-type devices to 70 pF or
less.

http://www.adafruit.com/datasheets/txb0108appnote.pdf


From the TXB datasheet typical application it doesn't look like I should put any pull-up resistor.

[Marco]

What are you connecting these boards with?

PS. this level shifting IC is expensive and unnecessary for an unidirectional protocol, just use a CD40109 to go up and a resistor divider to go down.

[raphaeldavidf]

What are you connecting these boards with?

ribbon cable with KK connector

[T3sl4co1l]

What is the rest of your circuit, and how is it connecting?  Layout?

[raphaeldavidf]

What is the rest of your circuit, and how is it connecting?  Layout?

Attached the schematics from the TXB board and the layout from the MLX board in the reply, 3º post.

[T3sl4co1l]

What is the rest of your circuit, and how is it connecting?  Layout?

Attached the schematics from the TXB board and the layout from the MLX board in the reply, 3º post.

But those just show off-sheet connectors, not headers, and what kind and how long the cables between..?

Tim

[raphaeldavidf]

What is the rest of your circuit, and how is it connecting?  Layout?

Attached the schematics from the TXB board and the layout from the MLX board in the reply, 3º post.

But those just show off-sheet connectors, not headers, and what kind and how long the cables between..?

Tim

40 cm ribbon cable with KK connector and header, but I already tried with shorter cables, and the capacitance measure was made without the cable(direct on the MLX board header)
If you like I could post some pictures

[Richard Crowley]

Are you using the SPI version of the MLX90316?  That is what it appears from your board layout.

Why do you need any kind of logic level converter at all?
The MLX90316 says that it operates on 5V Vdd and the data line must be pulled up to 5V or 3.3V through a 1K resistor (R2)
It also needs that funky MOSFET transistor to use the single data line as both input and output. (See the data sheet Fig 18 on page 36)

From what we can see in your postings, it appears that your complete interface to the MLX90316 is fundamentally incorrect.
Back to the drawing board and consult the MLX90316 data sheet more carefully.

[T3sl4co1l]

Probably enough cable to be worth adding some source termination resistors, about 50-100 ohms.  More if you don't need the speed.

Use more pins, and ground every other wire (50 mil pitch ribbon, dual row header, one row grounded -- nice and simple).  This gives a transmission line impedance of about 100 ohms, and reasonably low crosstalk.

Mind that SPI will be susceptible to more than modest EMI (i.e., >3V -- enough to violate logic thresholds, despite the extra grounds).  You can help that a little by adding filtering (ferrite beads and extra capacitance), but this directly affects maximum baud (just as the extra resistance does).  If you only need a few hundred kHz, it can help a lot.

Tim