Measurement errors with x10 probe on large negative signals

[MatNieuw]

Hi,
The input impedance of the DSO150 becomes lower if the voltage at the BNC connector is lower than -5V. This is because the TL084 for the sensitive channel is outside its spec. For a low-impedance source, this is not a problem: see the image with a 70 Vpp from a low-impedance source. This image looks the same on an analog scope.
However, if you use a x10 probe with a 9M impedance, the effect is very noticeable, see the pictures of the 220 Vpp signal. On the DSO150 the lower part of the signal is incorrect. The same probe was used for both scopes. Basically, to ensure correct display of the signal, with a x10 probe you shouldn´t measure signals where the lowest voltage is below -50V.

I think the best way to fix this would be some isolation device between the 510k and 5.1M resistor, driven by an unused U2 output. Such as a real signal relay (but that´s big and requires a drive transistor probably). Or maybe a optical solid state relay like a Panasonic AQV216S, but I have no idea what that does to millivolt signals, or which frequency range it handles, and it has a noticeable capacity between input and output, so high-frequency signals or steep edges might filter through.

[update 2017-10-04: Put ¨negative¨ in the Subject title, fix typo]
[update 2017-10-08: Relay type is AQV216S (thanks MNTech); x10 probe has of course 9M impedance, not 10.]

[MNTech]

Hello MatNieuw,

You have done a very good job of describing the problem and suggesting a solution.
There is a typo, the opto-isolator part number should be AQV216S.
As you say there could be problems with using a MOSFET switch including capacitance and leakage current.

There are inexpensive low current reed relays that could be operated by U2. Their price is less than the opto-isolator.
Coto Technology 9007-05-40 5V 5ma
Coto Technology 9007-05-00 5V 10ma
Littelfuse Inc. HE721A0510 5V 10ma

One end of R1 could be lifted and run through a relay. C2 would still load the 10:1 probe at higher frequencies in a manner similar to R1. C2 could also be run through the relay but that may lower the input capacitance below the range of a standard 10:1 probe's adjustment.

[MatNieuw]

I thought of another possible fix.
- Use a bistable single coil subminiature relay with DPDT (2 changeover contacts), like the Axicom IM43GR. A bistable relay only uses power when switching, so this avoids increased power consumption (around 15 mA extra).
- Control this by a PIC12F1501, this 8 pin chip has 6 I/O pins. Each coil connection is driven by two output pins in parallel (4 pins total), to avoid too much voltage drop. Its 2 input pins are connected to:
a) Signal SENSEL3, or a currently unused output of the 74HC4053: that output should switch between 0 and +AV.
b) The center contact of one of the relay´s switchover contacts. The two other parts of that contact are connected to 0 and +AV.
The PIC should be programmed that is its two inputs are not the same, it gives a minimum 5 ms pulse of the correct polarity (one side high, other side low) to the relay so that after switching over, its two inputs get the same polarity. In this way, the relay follows the SENSEL3 signal. It could go to sleep and check every 20 ms or so. If the inputs are the same, remove the drive from the relay outputs.
- The other relay switchover contact is used to switch U1D´s + input between ground and the R1/C2/R2/C3 join. During switching it will shortly be unconnected, but the output cannot go outside +AV or -AV, as now, so that should not be a problem.
- The PIC is very small, needs a decoupling C, and then the relay, so not much extra real-estate is used. But it is a microcontroller with internal clock, and might increase noise (I don´t think so, but it has to be checked).

I think with two pins in parallel, the PIC12F1501 will drive this relay which requires minimum 3.75 V at 15 mA. With a single pin not. But this has to be tried, and I have not done so so far. The PIC and relay are cheap, the PIC programmer board less so (for private persons).

It is possible that with this relay, there is a slight increase in signal quality if the ¨high" channel is selected. Because an overdriven ¨low¨ channel will flip between close to +AV and -AV, a 9V swing, and the 74HC4053 does have some crosstalk between channels. But I have not measured if this really is a problem. With a ¨low¨ channel input to ground, its output is a quiet 0 volts.

if you have the relay, then a future analog board design might as well use a single TL084 op amp and switch its input between the two input attenuators. Then the 74HC4053 can go too.

[MNTech]

Hello MatNieuw,

I think your solutions could be made to work but would add more to the product cost than JYE Tech would be willing to accept.

A answer to your particular need could be to make your own scope probe.
The homemade probe would be a voltage divider. 1 megohm over 10.5K would be reasonably close to 100:1 with the 1 megohm input resistance of the scope. 200 volts input would give 2 volts out to the scope which would be set to the 1V/div or 0.5V/div range. The signal would begin to fall off around 50 KHz but I suspect that would be more than enough for your use.

It could be made as a box between a standard 1:1 probe and the scope. Check the voltage rating of the 1 meg resistor you select, it probably needs to be 1/4 watt or higher, or you could put several in series to get a higher enough rating.

[MatNieuw]

Hello MNTech,

thank you for your replies. Indeed, cost is an issue, the low cost is one of the attractions of this very nice little scope (which I wish I had obtained earlier).
The reed relay solution looks the cheapest, I'm a little worried about the internal resistance of the 74HC4053: it can be 100 ohms, which would be added to the relay's coil resistance. But there are 2 unused channels, so you run them in parallel. The 5 mA one looks acceptable in extra power. Too bad these relays are in DIP form, larger than e.g. the Axicom IM43GR, or for a single contact, the IM23GR (10 mA). What I miss with the reed relays is their contact details which seem important with low level signals. See https://www.phoenixcontact.com/assets/d ... _en_00.pdf . The Axicom ones do specify this.
For me, to modify the scope internally, when buying a single relay, shipping is a large part of the cost, so relay costs are a little less important.

I thought about making a x100 probe, or maybe buy one.

[MNTech]

Hello NatNieuw,

It would take a lot of modification to properly correct the scope's poor input circuit. Just adding the relay wouldn't result in the right input capacitance. Also it would add to the current consumption which would reduce battery life.

Here is a link to a DIY 100:1 probe.
http://how-to.wikia.com/wiki/How_to_mak ... cope_probe

If you do have a 10:1 probe you could make a 10:1 voltage divider to go between it and the scope for 100:1 total attenuation.
A 909K series input resistor (R1) and 113k (R2) in parallel with the scope input are standard values that would give the 10:1 division.
Approximately 2pF in parallel with R1 would give some frequency compensation. If you want to be more exact make the cap adjustable and tune it while using a 1:1 probe.
Then adjust the 10:1 probe's frequency compensation. If it doesn't tune add a cap across the input of your new 10:1 divider. (about 10 to 20pF ??)

[hvorfordetda]

Thanks for bringing up this problem.

I have done some very rudimentary testing with a home made optocoupler consisting of one 5mm white LED and one GL5549 LDR (cost almost nothing on ebay) since I already had them.

Instead of changing the input network as suggested, I just lifted pin 12 on U1D from the PCB. Connect the LDR in series with pin 12 and let the LED be lit when SENSEL3 is high. I did not connect to SENSEL3, I used an external power supply for the LED to keep it simple sisce it was just for testing.

The LDR is more than 10Mohm (mine was more than 40M) when dark and about 500 ohm when exposed to light from the LED. This appear to give the necessary isolation and as far as I could see, it worked OK and behaved like the non-modified circuit for low level signals.

I tested up to 200 VRMS 50Hz and 10:1 probe without any noticeable defects in the displayed trace when pin 12 was isolated by the LDR. My pictures are 100 VRMS 50Hz.

I think this is a possible solution to the problem.Will need more time to "polish" the optocoupler and add some circuitry for LED-control.

[hvorfordetda]

I have modified my scope now, and it appears to be functioning OK.

I use SENSEL3 for switching the LED on/off. To avoid mixing too much into the existing circuit, I choose to use a separete transistor as a switch for the LED and only use one of the spare switches in U2 to control that transistor.

A LDR is most sensitive to green light, and since I had some small 0603 SMD type green LEDs I connected three of them in series. This gives us a forward voltage of about 7.5V which is perfect since the supply voltage is about 8V. If you choose to use a different LED, you must re-calculate the 180 ohm resitor. I used a BC546B NPN transitor, but anything NPN small signal should be OK.

All details are described in the attached pictures, but it is wise to start with lifting pin 12 on U1D and pin 10 on U2.

I used some very thin wires for the new connections. They are made up by twisting 3-4 strands form thin flexible wires and adding solder to their entire lenght.

The LDR was affixed to U1 by some rapid curing clear epoxy. I used a small droplet of epoxy on the LEDs as well to make sure they did not short themselves. When the LDR had cured, I added a new blob of epoxy to its light sensitve side, and placed the LED assembly into that blob.

I have tested the function of this mod with 290 VRMS 50Hz, and the waveform is perfect. I have also tested the modified and the unmodified circuit with both sine and squre wave signals up to 200 kHz with different voltage levels, and I can not see any erros introduced by my mod. When I did this testing, I found that the scope has some difficulties calculating the correct frequency. It appears to be a combination of frequency, timebase and signal level. This is not related ty the mod, but it is clearly visible in some of the pictures of the waveform.

Edit: I can also see that Vpp has a maximum value of 65 before it starts over at 0 again. I guess this has somthing to do with binary resolution in the ADC ?

When everything is tested OK, it is wise to add some black paint to the LDR / LED assembly to keep the ambient light away. It is also smart to add some glue to the wires here and there to make sure they stay in place.

(Click on image for larger version)

[hvorfordetda]

Two more images

[MatNieuw]

Hi hvorfordetda,

You have found a very clever solution to the problem, my compliments. I will implement it as well.
Many thanks too for the detailed photos and instructions.

Best regards, Mat