Yes but when you are squirting 10MHz through the device, have you measured what comes out of the other inputs / outputs?
I presume you mean in the demux direction. No, I haven't but I will.
I'm getting around 340 mV of the signal. But you'd expect that wouldn't you? It is getting cross-talk when effectively that channel is high-impedance.
I just tried it the other way around. Pumping in a different frequency into a non-selected line didn't seem to affect the output. (eg. 10 MHz into pin 14 (IO-1) when IO-0 was selected).
It is bidirectional so it dosn't matter.
When I was making my RFID sequencer I used these to try and switch the output from lots of pickup coils to the one decoder. I found that signals were appearing on the other coils through the multiplexer. So much so that it was not usable. This was with all the inputs terminated as well. I did try two different manafacaturers devices that claimed to be low cross talk but they were equally as bad?
Grumpy_Mike:
It is bidirectional so it dosn't matter.
I think it does matter because you have two different situations.
8 --> 1 means you have one output, and you are hoping it accurately reflects the selected input.
1 --> 8 means you have 8 outputs, but only one is active, the others are, presumably, floating.
I think he means it doesn't matter which way round you use it to test crosstalk. No matter which end you inject a signal, it should only come out of one other pin - i.e., either the selected output (for 1->8 ) or the common output (for 8->1) - all the other outputs should be completely silent. Further more, signals injected into any other pin, regardless of the direction you are using it, should not affect the signal passing through between the two pins that are actively connected.
My vote is for "It doesn't matter".
Because there is no output and there is no direction. One of the 8 is connected to the common pin. Perhaps the circuit around the chip uses it as input or output, but the chip is not handling it as input or output. It only makes a connection to one of the 8.
Matters Only: that the switched signals are within 20 DB or 100X of each other and do not exceed Vee or Vcc (Vdd) by more than .6 V (Vee is the analog ground).
RFID is a possible issue here as it is a pickup coil resonant at 13.56 MHz and the additional capacitance represented by the I/O connections (20 ~ 50 PF) guestimated... would be more than enough to detune the pickup coil enough to make the device numb... Or with a free running oscillator, change the operating frequency. (same thing) long cables would certainly be an issue as well for added capacitance
I've used the 4051BE's and the AE's for RF switching but the added capacitance of the gates was accounted for in the design. With the 4051, 2, 3 and 4016/66/4566 series IC's, switching any tuned circuit isn't a great idea because of the added capacitance of the gates and possible Vcc coupling issues which pretty much leaves out anything RF except the output of a buffer amp to the input of another or similar non tuned RF switching.
Bob
RFID is a possible issue here as it is a pickup coil resonant at 13.56 MHz
Yes but it was a 125KHz RFID system I was making.
switching any tuned circuit isn't a great idea because of ........
Yes I know, I wasn't doing that. I was taking the rectified and smoothed output of a tuned coil.
Signals were coming through inputs that were not the one currently being addressed. These were the same size as the signal from the addressed input.
I think that you didn't have Vee connected to analog ground (it goes to the most negative potential used) Vee must be connected either to ground or analog ground in the case of both because Vee is the switch ground and there is a digital ground for return of control signals. both muist be grounded at a minimum.
I've used a lot of those chips.. they make great glue logic but the only time I've ever had troubles was if I either tried to exceed the 20 DB differential between inputs that I found worked.. experimentally or if the analog ground wasn't used properly...
Bob
If you look carefully in the different datasheets, you'll see the different parts have
different values of stray capacitance to ground, and that effects the transfer ratio
and phase shifts of the signals at high frequencies. Looking in the box, I see I have
the following, all in DIP16,
CD4052BE (TI)
MC14052BCP (Mot)
HCF4052BE (ST)
CD74HC4052E (Harris)
NJU4052BD (JRC)
It's been 10 years now, but I ended up using the last part, since as I recall, it's performance
blew away all the others.
I'm not sure I'm measuring this correctly, and my test setup probably has a lot of stray capacitance.
However re-testing today indicates different amounts depending on frequency. I hope I have the formula for dB of v1 compared to v2 correct?
db = 20 * log10 (v1 / v2)
The tests were with output 0 selected (pin 13) and measuring output 1 (pin 14).
Freq Hz Input V Stray mV dB
Pk-Pk Pk-Pk
100 5.12 13 -52
1K 5.12 79 -36
1.5K 5.04 144 -31
1M 5.08 144 -31
10M 3.12 130 -28
20M 2.00 148 -16
You can see the effect here:
Channels 1 and 2 have different scales, it isn't that dramatic! Channel 2 is the "non-selected" output.
Moving the output being tested just one further over (to pin 15) shows quite a drop in the crosstalk.
Erdin:
Nick, could you write in the tutorial something about the switchable resistor. That would make clear how it works. Perhaps something like this:If a line is selected, the selected line has a small resistance to the common output/input pin (about 80 ohms). If a line is not selected it has a large resistance (too large to measure) to the common output/input pin. The resistance is created with fets.
So if I understand this correctly, each line has two FETs side by side, operating in opposite directions. The gate is connected to the decoded "enable" lines, so at a given time one FET is saturated (on) and the others are off. The "on" FET thus presents a small resistance (Rds(on)) to its signal (and since there are two side-by-side this low resistance works both ways). The "off" FETs present a large resistance.
Thus the signal (coming from either end) follows the path of small resistance. The cross-talk could be accounted for by the signal crossing over to nearby FETs by capacitance.
Have I got it right?
That's basically the right idea. One of the pass-channel MOSFETs is n-channel and the
other is p-channel, so they turn on and pass signals for different polarities, eg the
n-channel passes for Vgs > 0 and Vds > 0, and vice versa for p-channel.
Your dB formula is correct for voltage ratios, whereas the multiplier is 10 instead of
20 for power ratios.
When you say
I'm not sure I'm measuring this correctly
you're apparently measuring feedthrough with the channel turned off, so your values
pretty well match the datasheets.
The stray capacitance I was talking about is the inherent capacitance in MOSFETs
measured between source and gate, and between drain and gate. For these devices
this appears to be on the order of 15-25 pF, and you can see, in combination with the
signal source resistance and the channel resistance, this will act like a low-pass filter,
and both attenuate the signal and produce a phase lag.
Nick,
You might also like to take a look at the 74HC4067, a bidirectional 16 i/o analog multiplexer. It also isn't fussy in which direction the data flows. Of course, it can also be used for multiplexing digital signals. I'm using two to detect which of 32 inputs is pulled to ground from just 6 Arduino pins.
Henry_Best:
Nick,You might also like to take a look at the 74HC4067, a bidirectional 16 i/o analog multiplexer. It also isn't fussy in which direction the data flows. Of course, it can also be used for multiplexing digital signals. I'm using two to detect which of 32 inputs is pulled to ground from just 6 Arduino pins.
It's on my "to be purchased" list. 
One of my projects in the pipeline is something very similar to that, not all of the parts have arrived yet.
One thing that bothers me, is that it is hard to tell how accurate the chip is in terms of voltage.
Looking at your oscilloscope plots there, you have the time delay to consider and the frequency response
to consider, if that is important to you. In my application, it isn't particularly important. But looking
at the difference between the yellow trace and the blue trace, at the lower voltage levels, the difference
seems quite marked.
But looking at the difference between the yellow trace and the blue trace, at the lower voltage levels, the difference seems quite marked.
Look at replies #17 and #20 for channel throughput. The traces in reply #23 are
apparently measuring channel crosstalk.
It should be pointed out that the levels are measured on unterminated outputs. Open circuit measurements of this nature really don't apply to "the real world". If both input and outputs are terminated in a resistance minimally equal to the on channel resistance and the generator impedance in series then what is measured will depend on the test jig more than the device. At the point of equality the impedances are said to match, power transfer will be maximum and strays minimized.
Basic Electricity 101, power transfer.
Bob
oric_dan:
Look at replies #17 and #20 for channel throughput. The traces in reply #23 are apparently measuring channel crosstalk.
True and that was at 10 MHz. I found a lower drop between input and output at lower frequencies, and in the case of using this device to multiplex, say, a light sensor, you are effectively dealing with DC.
Docedison:
It should be pointed out that the levels are measured on unterminated outputs. Open circuit measurements of this nature really don't apply to "the real world".
Yes, that was the setup, and certainly applying loads reduced the cross-talk. However I was assuming you might "load" it with an analog pin on the Arduino, which itself would be high impedance.