Inexpenive 50 Ohm Feed Through Terminations

[graybeard]

In this video I review why you need to use a feed through termination and show an example. 
Then I do review of a low-cost (2 for $10) feed-through termination and compare it to ones from Tektronix and HP including measurements with a network analyzer. 
The results are interesting and not what you might expect.  The low-cost ones and expensive name brand ones all perform the same when connected to an oscilloscope input.

Chris

[n3mmr]

You connect the f-t termination to the scope input and then connect the attenuator to the termination?

I e source->cable->attenuator->f-t_termination->scope_input?

[TimFox]

That is correct.  When the 50-ohm terminator is on the attenuator output, the impedance looking into the other end of the attenuator will be 50 ohms.

[RoGeorge]

Thank you for showing the Smith chart for when the terminator is also directly connected to an oscilloscope input.   

[David Hess]

Most modern oscilloscopes only have one high impedance attenuator stage (1) which explains the difference between 100mV/div and 200mV/div and why there are basically only two sets of curves.

Is it possible that at 100mV/div and lower, the signal level from the VNA is triggering the oscilloscope's shunt input protection?  If there was really that big a difference, it should show up as a change in probe compensation between 100mV/div and 200mV/div and if that happens, the oscilloscope is broken.

I have two of the 5 watt Tektronix feedthrough terminations that you mention.  Actually one was sold by Probe Master and is marked as such so either Probe Master or someone else was OEMing them.  They are lower performance than the smaller Tektronix feedthrough terminations but up to 300 MHz, perform just as well.

(1) This also means that they operate their high impedance input buffer over a much larger signal range which is detrimental to performance.  Older oscilloscopes with 3 or even more high impedance input attenuators perform better.

[vk6zgo]

Just for chuckles, have you tried the 20dB attenuator alone, without the termination?

I have played around with these, & found that 20dB attenuators approximate a pretty good termination by themselves.

I didn't have the good equipment you have, just a Spec An with a tracking generator, a (salvaged) Seimens directional coupler, plus a limited number of other bits & pieces, which I had to characterise as well as I could, prior to the job on hand.

I found, with this gear, (at around 400MHz), that a 20dB attenuator with its output open circuit, delivered a return loss of 40dB, compared to a plain open circuit cable.
This seems anti-intuitive, with the impedance mismatch, but was completely repeatable.

I would often use two such attenuators in cascade to give 80dB return loss.

Due to the limitations of my equipment, it would be interesting to see if there are problems with this which would show up with a VNA.

[TimFox]

An unterminated 20 dB attenuator will show a return loss of 40 dB, with a complete reflection at the other end but 20 dB attenuation each way.

[Kleinstein]

Is there a way to build a terminator that is actually better than a normal one, when connected to the scope. By first guess would be something with some inductance / longer wire in series to the resistors.  So something that one would normally call a crappy build termination. Changing one of those cheap Chinese ones would not be such a big loss.

[vk6zgo]

An unterminated 20 dB attenuator will show a return loss of 40 dB, with a complete reflection at the other end but 20 dB attenuation each way.

I do believe that is exactly what I said happened under test.
For most practical purposes in RF work, a return loss of 40dB is "near as dammit" to a correct termination.

One would assume that the return loss of a 20dB attenuator in cascade with a 50 Ohm termination connected to an oscilloscope with an input impedance made up of 1MOhm resistance in parallel with 16.5pf would be 40dB better than with the termination alone.
The OP seems to get a best figure of around 40dB

[Kleinstein]

The 40 dB extra return loss is for a perfect attenuator. The one used seems to be not that good. It could also be the cable with not a perfect 50 ohms that can make a termination / attenuattor look bad.

[langwadt]

An unterminated 20 dB attenuator will show a return loss of 40 dB, with a complete reflection at the other end but 20 dB attenuation each way.

I do believe that is exactly what I said happened under test.
For most practical purposes in RF work, a return loss of 40dB is "near as dammit" to a correct termination.

One would assume that the return loss of a 20dB attenuator in cascade with a 50 Ohm termination connected to an oscilloscope with an input impedance made up of 1MOhm resistance in parallel with 16.5pf would be 40dB better than with the termination alone.
The OP seems to get a best figure of around 40dB

an unterminated  PI attenuator (61.11R,247.5R,61.11R) is a ~51R load, which is ~40dB return loss

[tautech]

The Tek pads Chris used are listed in the attached pdf

[bob91343]

My main criticism is that I can't understand all of the discussion.  I guess my ears are getting old.

Bottom line, since most of my work is below 100 MHz I guess I can stop worrying about termination quality.

I have a home brew 50 Ohm termination that I use sometimes with a T connector and it works reasonably well also.

[ejeffrey]

Then I do review a low-cost (2 for $10) feed-through termination and compare it to ones from Tektronix and HP including measurements with a network analyzer. 
The results are interesting and not what you might expect.  The low-cost ones and expensive name brand ones all perform the same when connected to an oscilloscope input.

This isn't that surprising to me.  I don't doubt that there are plenty of ways to screw this up but it isn't like they are magic.  The performance of a feedthrough terminator and its advantage over a tee is mostly due to geometry.

[graybeard]

Is it possible that at 100mV/div and lower, the signal level from the VNA is triggering the oscilloscope's shunt input protection?  If there was really that big a difference, it should show up as a change in probe compensation between 100mV/div and 200mV/div and if that happens, the oscilloscope is broken.

No the input shunt protection was not activated.  The VNA 10 MHz waveform only exceeded the screen limits on 10mV/div and and more sensitive settings.   There was no change in the measured input impedance from 500µV/div to 100mV/div.  There was one tiny change above 200mV/div I noted in the video, there was an associated relay click when it happened.

Older oscilloscopes with 3 or even more high impedance input attenuators perform better.

This is not always true.  I tested the the inputs to my 200 MHz 7A26 plug-in on my Tektronix 7904A 500MHz CRO and it was worse than the Siglent SDS1104X-E since the 7A26 input capacitance is 22pF vs. 15pF for the SDS1104X-E.   The 7A26 measured input impedance with the feed-through termination varied a bit on almost every input attenuator setting.

I omitted the measurements on 7A26 (200 MHz 1MOhm||22pf) and my 7A29 (1GHz 50Ohm) from the video to keep the run-time reasonable.

Chris

[David Hess]

Just for chuckles, have you tried the 20dB attenuator alone, without the termination?

I have played around with these, & found that 20dB attenuators approximate a pretty good termination by themselves.

I have done that when I did not have a feedthrough termination handy and it worked fine.

Good feedthrough terminators however place their resistive element as close to the male connector or even make it part of the male connector to minimize the length of the unterminated transmission line so in theory they could perform better.  BNC connectors with a built in 50 ohm resistor disc as part of their construction used to be available.

Older oscilloscopes with 3 or even more high impedance input attenuators perform better.

This is not always true.  I tested the the inputs to my 200 MHz 7A26 plug-in on my Tektronix 7904A 500MHz CRO and it was worse than the Siglent SDS1104X-E since the 7A26 input capacitance is 22pF vs. 15pF for the SDS1104X-E.   The 7A26 measured input impedance with the feed-through termination varied a bit on almost every input attenuator setting.

I meant they perform better because their high impedance buffer can operate over a smaller signal range.

The 7A26 has a higher input capacitance so will present a worse match but it should be more consistent between attenuation settings unless the compensation for the attenuators is out of calibration.  It would make for boring measurements though.

[thm_w]

Here is a photo of one of the cheap "plastic" terminators you said were bad. In reality they are all metal, and the plastic is simply a cover.

There are two 1210 resistors inside which can take 0.5W each, so the 1W rating should be accurate. I don't have equipment to test the frequency response though.

The PCB inside might be useful for making custom filters or attenuators.

[graybeard]

Here is a photo of one of the cheap "plastic" terminators you said were bad. In reality they are all metal, and the plastic is simply a cover.

There are two 1210 resistors inside which can take 0.5W each, so the 1W rating should be accurate. I don't have equipment to test the frequency response though.

The PCB inside might be useful for making custom filters or attenuators.

It appears to be a repurposed housing for a low frequency oscilloscope probe compensation network.   My guess is holes in the shielding are to facilitate adjustments. 

Given the extended ground path for the resistors I would not expect it to work well.  My guess is it will work OK up to the high tens of MHz, but look inductive above.  The specification of DC to 1GHz is BS.

In my experience the plastic outer pieces of BNC connectors do not last.

Since you have it you should use it.   It will probably be fine below 100 MHz.  However I would not buy one.

[bitseeker]

Here is a photo of one of the cheap "plastic" terminators you said were bad. In reality they are all metal, and the plastic is simply a cover.

There are two 1210 resistors inside which can take 0.5W each, so the 1W rating should be accurate. I don't have equipment to test the frequency response though.

The PCB inside might be useful for making custom filters or attenuators.

Thanks for the peek inside. As graybeard said, it does look like a re-purposed scope probe compensator.

[vk6zgo]

Here is a photo of one of the cheap "plastic" terminators you said were bad. In reality they are all metal, and the plastic is simply a cover.

There are two 1210 resistors inside which can take 0.5W each, so the 1W rating should be accurate. I don't have equipment to test the frequency response though.

The PCB inside might be useful for making custom filters or attenuators.

As an "even greyer beard", you had me going for a while with "1210"!
I thought "120 Ohms--- what the heck?", until I realised that it was the size of the resistor package

The older Tektronics 75 Ohm feedthrough terminations used two leaded 150 Ohm resistors in parallel.
These terminations had a habit of becoming unscrewed after years of use, & detaching one or more resistors.
We routinely soldered them back.

The externally physically identical 50 Ohm ones never seemed to have this problem, (probably because they  weren't used as much  as the others), so I wasn't able to check if they used the same type of internal parts.

They both used a brown plastic "sheath" with the impedance value & Tek part number, slid over what looked like an assembly made up of standard  BNC fittings.

We also had a larger 50 Ohm one, which had a higher wattage rating & had a bigger plastic body, with no obvious way of dismantling it.

[joeqsmith]

Experiments with a damaged part.

https://www.eevblog.com/forum/testgear/pasternack-pe6008-50-thru-term/msg947602/#msg947602

[graybeard]

Experiments with a damaged part.

https://www.eevblog.com/forum/testgear/pasternack-pe6008-50-thru-term/msg947602/#msg947602

Joe,

I believe you were measuring the termination itself.   You did not measure it on the input of a scope.  I suspect if you try that again with the termination on a scope or meter input there well be very little difference for the various configurations.

Making a calibration termination (not a feed through) is far more critical.  I have made SMA cals for my VNA using two 100 ohms 0805s based on this article.

Chris

[joeqsmith]

Experiments with a damaged part.

https://www.eevblog.com/forum/testgear/pasternack-pe6008-50-thru-term/msg947602/#msg947602

Joe,

I believe you were measuring the termination itself.   You did not measure it on the input of a scope.  I suspect if you try that again with the termination on a scope or meter input there well be very little difference for the various configurations.

Making a calibration termination (not a feed through) is far more critical.  I have made SMA cals for my VNA using two 100 ohms 0805s based on this article.

Chris

I agree that with a meter you wouldn't likely see any difference, outside of the resistance tolerance of the parts.   All of the meters I have, run at DCish.     

I had ran some test using 0805s vs the 1206 and also saw some improvement.  This is why the cal standard for my old 8754A is using two 100 ohm 0805's for the load.   I should have saved that data.     

My scopes have 50ohm support.  There is no way to disable it on my higher BW unit.   I made a video a while back for someone who had asked me about measuring the speed of light.  In that video, I show many tests, some optical.  One test was looking at a  square wave using my function generator and my slower scope.   I did this to show the contrast to the sub ns edges we would be looking at for the other experiments.   Granted, when you are not looking at transmission lines,  the termination will be less important, just like with a meter.

[ocw]

I had ran some test using 0805s vs the 1206 and also saw some improvement.  This is why the cal standard for my old 8754A is using two 100 ohm 0805's for the load...

I have had better luck using Vishay FC0603E50R0BST1 RF resistors.  They 50 ohm 1/8 watt 0.1% 0603 resistors with a factory rating of under 1.2/1 VSWR up to 30 GHz for just $1.61 (Qty 1 Mouser).  See:
https://www.mouser.com/datasheet/2/427/fcseries-947854.pdf
Those have worked better for me than using two 100 ohm 0.01% regular resistors.  I prefer staying away from the similar FC0603E50R0BTBST1 resistors since they only have top sided soldering.

[joeqsmith]

I had ran some test using 0805s vs the 1206 and also saw some improvement.  This is why the cal standard for my old 8754A is using two 100 ohm 0805's for the load...

I have had better luck using Vishay FC0603E50R0BST1 RF resistors.  They 50 ohm 1/8 watt 0.1% 0603 resistors with a factory rating of under 1.2/1 VSWR up to 30 GHz for just $1.61 (Qty 1 Mouser).  See:
https://www.mouser.com/datasheet/2/427/fcseries-947854.pdf
Those have worked better for me than using two 100 ohm 0.01% regular resistors.  I prefer staying away from the similar FC0603E50R0BTBST1 resistors since they only have top sided soldering.

I am very interested in seeing your test data if you don't mind.   

[ocw]

I am very interested in seeing your test data

With my prior 2 x 100 ohm on a SMA connector load being trashed, I thought that I would make a new one using two APC0603T100RZ 100 ohm 0.01% resistors.  It came out better than my prior one--I was impressed!  The attachment shows my Anritsu MT8222A evaluation of it.  I decided to use that instead of my Agilent E7495A or HP 8753C.  I used a Suhner 18 GHz calibration load rather than one of my four other choices. I compared those and thought that the Suhner produced the most accurate results.

[joeqsmith]

Are you performing a calibration using the load only?   Does the load and reference have the same connector?  If not, how did you compensate for the difference? 

I wonder, did you happen to save plots for the other resistor combinations you tried?   

I double checked to see if I had saved any of the data I collected back then but came up short. 

[ocw]

I have 7/16" DIN, SMA and N connector calibration kits.  I obviously used the SMA version this time.  I have wondered about your calibration kit.

... the cal standard for my old 8754A is using two 100 ohm 0805's for the load.

A calibration standard will always be measured by the same instrument as being perfect regardless of how good or bad it is.

I provided a link to the manufacturer's RF specifications for the resistor which I suggested.  Where are similar specifications for what you used?

While you are making it difficult, I am trying to be helpful not confrontational.

[joeqsmith]

I have 7/16" DIN, SMA and N connector calibration kits.  I obviously used the SMA version this time.  I have wondered about your calibration kit.

... the cal standard for my old 8754A is using two 100 ohm 0805's for the load.

A calibration standard will always be measured by the same instrument as being perfect regardless of how good or bad it is.

Obviously. 

I provided a link to the manufacturer's RF specifications for the resistor which I suggested.  Where are similar specifications for what you used?

They don't normally provide s-parameters for standard resistors like I have used.  As I mentioned in my previous post, I couldn't find the data I collected for the standards I made. 

While you are making it difficult, I am trying to be helpful not confrontational.

Sorry you feel I am being difficult and confrontational.  Had I known asking a few questions would upset you,  I wouldn't have posted them.  Lesson learned.

[tim_]

I encountered the same issue of wanting to connect a 50 Ohm amplifier to the input of my oscilloscope (Rigol MSO5000).

I bought one of the "cheap plastic P57" through terminators. It turns out it really does not work that well up to 1 Ghz, see S11 measurement of the open terminator below.

What is a bit more concerning to me is that the 50 Ohm terminator does not really help to get the RF signal into the scope input. See below. One can also see that the gain selection of the channel has a clear impact.

Any ideas how to solve this? From the video above it seems that the easiest solution would be to add an attenuator? Of course that is not really preferrable.

[David Hess]

A feedthrough termination is not really suited to operation above 200 MHz where the input capacitance and stub length of the high impedance input become significant.  High impedance probes include high frequency compensation to handle this which is missing in a feedthrough termination.

At higher frequencies the change in input characteristics with input attenuator setting is insignificant because it is above the bandwidth of the oscilloscope.

[joeqsmith]

Experiments with a damaged part.

https://www.eevblog.com/forum/testgear/pasternack-pe6008-50-thru-term/msg947602/#msg947602

Joe,

I believe you were measuring the termination itself.   You did not measure it on the input of a scope.  I suspect if you try that again with the termination on a scope or meter input there well be very little difference for the various configurations.
...
Chris

Is four years considered a necropost?  Revisiting my modified terminators with a SPICE twist.