Someone just shoot me..

[itsmoked]

I have a problem and am looking for some insight.

I need to monitor plasma in a chamber. My customer has a probe that has a high source impedance, (it's plasma. DUH!) I'm required to provide a simple divider network and was directed to use 100k between the probe and the AMP input which is a J-FET with insane input resistance and with attenuation turned on another 100k to ground. With no attenuation, e.g. the bottom resistor not in-circuit everything is just peachy-keen.

The problem is when 50% attenuation is desired. Switch in that bottom 100k divider resistor and now you have a 200k load piled onto the probe. The result is a fraction of what's expected. A little pencil work shows the probe has about 1 mega-ohm of source impedance!

They don't understand this concept. I made them a little test jig to demonstrate this. Here's the circuit; Coax in to a 100k resistor to a coax out. Coax out to a little N.O. push-button to another 100k to ground. This mimics my device with the AMP but without the AMP. They can hook a scope to the output the probe to the input and see an unloaded probe then press the button and see a 200k loaded probe.

I do this board on my own time donating the parts. Today they come back with, "before we bother to do this test we want two more boards; one with 100k RF resistors and one with 1M RF resistors. They want RF resistors so there is "less noise".

Please correct my thinking here;
'RF resistors' are not about "less noise" they provide no less noise, they provide less self inductance. Correct?
'RF resistors' are typically 1k or less and not 100k or 1M. Correct?
Does anyone know where one could procure 100k and 1M RF resistors? Is it right next to the shelf covered with golden geese or is it the shelf with five leaf clovers? Maybe I'm just looking in the wrong place.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[ScottyUK]

Hi Keith,

Years ago I worked at a plant making resistors and hybrids. I was an engineer in the hybrid division, but occasionally spent time over in resistors division. The RF products were designed with very low self-inductance as you already noted. 'RF' was GHz rather than MHz in that product line because the standard surface mount types were pretty good up to the MHz range.

Electrical noise is largely Johnson noise which is a function of temperature and resistance, so lower resistances are generally preferred for ultra high bandwidth applications where there isn't the luxury of cooling the resistor. That will probably explain why you're not finding higher resistance values in the RF product ranges.

Can't you / they attenuate the signal after the high impedance buffer amp at the front end?

 

[Comcokid]

SMT Thin-film resistors are typically used in 100's MHz to low GHZ range RF, the issue being skin-effect which affects the apparent resistance. But as ScottyUK pointed out, standard surface mount (thick film) resistors are good well into the 10's of MHz range.

But, your issue may be the probe setup as you've tried to point out. Plasma (~~1Meg source impedance) to coax (50 Ohm impedance, short run which will look more like a capacitor [divider] to gnd) to op amp (very high impedance), and 100K to ground for a divider. Not exactly standard RF practice from a matched impedance and attenuator prospective. You will only be able to get meaningful/predictable voltage division after the op amp. Dividing the signal before the opamp with all of the unknown elements of parasitics and source impedance will be trial-and-error value selection.

 

[Skogsgurra]

What voltage?

Use the FET as a source follower and forget the load. You will miss the G-S voltage drop, but that is usually quite constant. It will give you better result than any voltage divider (if you do not go for giga-ohms). But that is another problem.

Or do you already use a source follower? Not quite clear from your question.

Gunnar Englund

http://www.gke.org

--------------------------------------
Half full - Half empty? I don't mind. It's what in it that counts.

 

[itsmoked]

Thanks gents.

Thanks Scotty and Comcokid for completely confirming my understanding.
And, YES! I'll be changing the design to a buffer AMP and the attenuation after it. The current design is theirs.

The probe signal can (but rarely) get up to 20V which makes the front end buffer a bit of a pain given the available power supplies, but worth it due to all the issues you mention Comcokid.


Skogs not a follower an J-FET front-ended OPAMP. Source can hit 20V transiently. Available power supply is 12Vdc. Currently using a DIP 12V to isolated +/- 15V. Supporting a 20V input will need something else.


Keith Cress
kcress -

http://www.flaminsystems.com

 

[VEBill]

Is this to monitor a voltage. or a high frequency signal?

Your description implies some similarities with RF design.

Lessons from RF:

If you want less noise, then the very first stage should be the (Low Noise) amplifier. If you put an attenuator stage (or any loss) first, then those dB of loss are added right onto the noise figure. (Then again, if you have too much signal, then perhaps the noise isn't the big problem after all.)

With a high dynamic range signal environment, it might be worth running the amplifier (<- the preamplifier) with fairly high voltage rails, and signal handling capabilities, to keep it linear with lots of headroom. (This approach also supports moving the attenuation further down the signal chain.)

It seems like you're on the right track.

 

[itsmoked]

Hi VE1BLL! This is a voltage monitoring thing. The logging is exceptionally low rate, something like a couple a second max, but a secondary function is just about the rate of change and so this buffer needs to pass thru fast stuff, though still it's only low kHz stuff. Yes, I'm with you on the buffer amp with high voltage rails +/- 24V. That's where I'm headed now. I'll just prototype a buffer board that goes inline with the existing stuff, which will land the current front-of-the-amp attenuation into back-of-the-buffer attenuation.

Hopefully it will cut thru all the meetings..

Keith Cress
kcress -

http://www.flaminsystems.com