Basic accuracy of the uTarcer3P at low voltages and high currents

[Never Mind]

I have tested the uTracer3P for it's accuracy at low voltages and high currents.

I used a 250 ohm precision resistor power resistor in all testing and the latest version of uTracerJS.

uTracerJS has two additional calibration values that allow hardware limitations in the uTracer design to be corrected for improved accuracy at voltages below 100 volts and high currents.

These additional calibration fields are...

Vadc offset

Rextra losses

The results for a uTracer3P using the current version of uTracerJS software and no addition calibration from the standard calibration fields are quite good.

Errors measured less than 10% below 100V into a 250 ohm resistor and on average were about 5% off compared to measurements made during the standard calibration process for the uTracer3P.

This is likely good enough for most uTracer3+ users and in sharp contrast to a uTracer6 were peak uncorrected errors were found to exceed 100% at low voltages and high currents.

There is however a opportunity to further improve uTracer3+ accuracy at low voltage if the new calibration fields are used with the uTracer3+. 

Maximum errors can be reduced to about 5% and average errors reduced to about 1% using the correct values for Vadc offset and Rextra losses.

See the attached charts below.

In my testing I used Vadc offset = -0.1 and Rextra losses = 8 for best results with the uTracer3+.

See the attached Excel sheets for the detailed data.

[Martin Manning]

Yesterday I did some experimenting with the new cal features which is captured in the attached .pdf. I used a 5AR4 rectifier rather than a resistor. I found a Rextra value of 1.5 ohms to produce the best match to a DC measurement of the rectifier Ia at Vak. 

Unfortunately the tracer seems to have some damage, which I believe will require replacing the transistors in the HV switches.

[Never Mind]

Martin

Sorry to hear you have had a failure. 

From your data it seems you are using a modified uTracer3 for higher current levels of 600mA. Is that correct?

I have not done any testing on a 600mA uTracer3 as I do not have one however the extra calibration fields should be even more important on your unit at the higher currents it supports than a stock uTracer3.

In my testing I found selecting double acquisition resulted in larger nonlinear measured errors I did not understand and so I did not use the feature.

As a result I have little experience with this setting.

Your method of using a actual tube where you carefully measured it on a bench supply at 20V 280mA will work but only provides a single data point. 

That would be a 71.43 ohm resistor at a Va of 20 volts, Ia 280mA.

That will work but limits your ability to find the values for both calibration variables. 

The two values interact. 

The effect of "Vadc offset" on the result of Va/Ia increases as Va is lowered

The effect of "Rextra losses"  on the result of Va/Ia decreases as Va is raised.

A number of data points is needed on the Va / Ia curve to find appropriate values for both "Vadc offset" and "Rextra losses".

You could plot a larger number of data points on your 5AR4 to get a correct curve over a wide range of Va over Ia and that would allow the correct calibration of both Vadc offset and Rextra losses to be found.

This does seem tricky to get 100% accurate however.

I chose to use precession resistors in my calibration work so I would not have to worry about tube drift during testing spanning a few weeks time complicating my results.

Be interesting to compare the results you got from your reference tube to measuring a power resistor that is close to 71.43ohms like a 68 or 75 ohm resistor.

Then calculating out the value of Va/Ia for each point to show the error in measuring the resistance.

A advantage of using a resistor is you can plot the measured resistance and any error from a low Va value up to your measurement systems maximum current.

The excel sheets I posted show this method and plots the resulting errors.

I found errors in the uTracer below 100V were very dependent on the levels of Va as well as Ia. 

By using a resistor you can watch the errors change as Va sweeps up in value.

By using  a series of resistor values and repeating the Va sweep you can check if the correction is still valid over a range of Va and Ia with different resistor values.

In my uTracer6 testing I used resistor values from 10 ohms up to 10K ohms to check that calibration held over a wide range.

Using a test resistor that resulted in maximum test current (Ia) occurring at about Va=50V I found resulted in a best values of correction to be calculated by the excel sheet for "Vadc offset" and "Rextra losses".

It would be of interest to know what you find failed in your uTracer3

Take care

Bob

[Ihor]

Hi Martin, 

Thanks for sharing the tests and pity to hear that your uT is not functioning properly at this moment. I was looking at the plots but I was not sure what the most important conclusion would be. Probably indeed it is necessary to do multiple points and with at least a couple f resistors. I assume your 5VDC heater for the direct current measurements was exactly the same as for the measurements with the uT(?). The "double acquisition" uses a very simple rule, so probably it helps to understand the exact behavior. On the first attempt uT takes a measurement at a given voltage Va, and measures what was the actual Va_meas after the measurement, after that, during the second attempt, it applies Va + (Va - Va_meas) to compensate for the drop. Within the linear region and low currents it works reasonably good. 

About the last slide in the pdf, where it says that uTJS hangs on HV LED and Windows software works but too slow, I assume uTJS just stops taking measurements and quits at some point, there is a delay of 10 or 15 sec hardcoded, to wait for a measurement, and if the caps are not or cannot be charged (I think it is the reason then and the place to look for the breakage), it will stop, while window software probably waits until forever and in thin case actually gets lucky:) 

Ihor

[Ihor]

By the way, this might be relevant to the discussion of diode testing but also voltage drops and corrections. Ronald actually mentions a kind of VoltageOffset in his log book of the initial uT3 development, but it is not in the software for uT3+ and further. Here is the source and the screenshot:

"Diodes: AZ1, EZ81, EAA91" https://www.dos4ever.com/uTracer3/uTracer3_pag6.html

 

[Never Mind]

For basic accuracy the uTracer3 does much better in stock form than the uTracer6 below 100V.

My results show the uTracer3 errors to be a maximum of about 10% and a average of about 5% below 100V. 

My results are a bit worse than Ronald's lab notes however I agree that a stock uTracer3 is "good enough" for most uses even without full calibration correction.

It does seem a lost opportunity not to provide the extra calibration ability to get better accuracy with uTracer3 hardware given a bit more software and calibration effort.

Then came the uTracer6. 

Errors that could be ignored most of the time on a uTracer3 become very large errors with a uTracer6 where being off by 40% to 90% is not uncommon with Va < 60V at high currents.

For me the errors rendered the utracer6 of no value in a lab setting. It made nice curves but checking the accuracy told a story of pretty pictures with limited data integrity.

Be especially careful using Va=Vs with Ronald's software as there is a bug that reports the wrong values for screen voltage in this mode.

See my post on "https://www.diyaudio.com/community/threads/utracer6-gui-measurement-accuracy-issue.409428/"

With a uTracer6 and Ronald's software data for Vs can be off by 100%. Ouch.

At the time I thought the bug may only affect the uTracer6 but I now know it affects all utracer users and high current uTracer versions far more.

The bug has been there for years it seems as I recently spotted a old post from a uTracer3 user noting that the loading on the anode port affects the voltage results on the screen port.

At the time there was no explanation for the effect and it was decided to ignore it as unimportant. It is a small error most of the time on a stock uTracer3.

[Martin Manning]

Bob wrote:

From your data it seems you are using a modified uTracer3 for higher current levels of 600mA. Is that correct?

I have not done any testing on a 600mA uTracer3 as I do not have one however the extra calibration fields should be even more important on your unit at the higher currents it supports than a stock uTracer3.

Thanks for the reply, and you are correct. I have been using the high current modification for a couple of years, and with 6-ohm current sensing resistances I have  ~600 mA capability . On the first page of the pdf above you can see that adding the 0.1 Vadc and 8 ohm Rextra has almost no effect on the current calibration at 200V and 20 mA. On the second page, it appears that without using those extra cal inputs I was missing the current by 10 mA out of 280, or 3.6%. The 280 mA at 20V was measured with a DMM and using the same heater supply (a B&K 3A bench supply) I use for tracing. The 5AR4 is inserted in the uT3's octal socket, using the same heater cabling. I could have measured the rectifier current at a lower Vak to get a second reference point, but your point about drift over time is certainly valid, even over the time of the calibration. A long warm up time is needed for the tube and the tracer to stabilize. In previous testing I found that about 40 minutes is required for the tracer itself to settle in. 

I haven't started looking for the hardware problem, but it looks like a partial failure in both channels. 

[Martin Manning]

Ihor wrote:

About the last slide in the pdf, where it says that uTJS hangs on HV LED and Windows software works but too slow, I assume uTJS just stops taking measurements and quits at some point, there is a delay of 10 or 15 sec hardcoded, to wait for a measurement, and if the caps are not or cannot be charged (I think it is the reason then and the place to look for the breakage), it will stop, while window software probably waits until forever and in thin case actually gets lucky:)

Exactly. The HV-On hang-up is difficult to clear too, I have to shut the tracer off, disconnect everything, and cycle the power switch to stop the HV LED coming on with the main power. I've wondered if it would be a good idea to add a reset button for the PIC for such cases.

[Ihor]

Exactly. The HV-On hang-up is difficult to clear too, I have to shut the tracer off, disconnect everything, and cycle the power switch to stop the HV LED coming on with the main power. I've wondered if it would be a good idea to add a reset button for the PIC for such cases.

I had the similar behavior many times, so reset button for the PIC is not really necessary. Usually when it hangs like that, you have to power off and on utracer and just connenct with uTracerJS. When uTracerJS starts, the first thing it does is it sends a special "300000...." command which forces utracer to discharge the caps. You just have to wait 5-10 sec and it is done. 

Ihor.

[Never Mind]

Ihor

"add a reset button for the PIC for such cases"

Hmmm, maybe one driven from DTR? 

Something to do with all that spare time...LOL

I wonder the exact state of the hardware when the  PIC HV-On hang-up happens?

Does all the other power supply code keep running? If there is a strobe that normally runs but stops in the hangup then a hardware watchdog can recover the condition.

Wondering about a small hardware watchdog IC reset by transitions in hardware that may freeze during the hang. 

Then the hardware watchdog timer would reset the PIC.

I have used the DS1232 in a large number of designs for this.

Take Care

Bob

[Ihor]

As far as I remember, from following Ronald's blogs, all the algorithms and routines described there and also looking at the assembler code, available for his early versions, there are no real watchdogs there and hanging is only possible when the caps cannot reach the set voltage (the routine will just try and try), and in all other cases it can be due to some hardware glitches due to electricity jumps and so on, which are very rare so should not be taken into account probably. During the measurement pulse, all the interrupts are turned off, so nothing can disturb the flow of the code, but it also means that if the hang happens there, it is just a hardware (one in a million chance) issue. The HV LED just indicates if there is a high voltage present (where the threshold for indication is also not so high, 30 or 60V I believe). So in summary, for myself I decided to reset the power manually and that's it and that' enough. The only useful case to have the automatic reset, what Bob once mentioned, is when the control software (for example uTracerJS) is running on a remote system, next to the uTracer and the user is using the webinterface remotely, but sill I think when you test the tubes you will be around the whole setup so things can be reset. In that case it is more about automatic resetting/restarting of the software which will crash in thin case. 

[Never Mind]

Ihor

I agree for a hobby piece of equipment the hanging is not a issue.

I run my uTracer off a current limited power supply and just restart everything. Then all is good.

I found since going to a real FTDI USB to RS-232 9 pin connector adapter hangs now rarely happen.

Bob

[Bill van Dijk]

Just curious,  I've never had a lockup on mine yet. Since the filament power is regulated by the PIC, what happens there? Does the output go to zero or full power? Perhaps it depends on the exact state of the FET when the lockup occurs?
Bill v

--
You received this message because you are subscribed to the Google Groups "uTracer" group.
To unsubscribe from this group and stop receiving emails from it, send an email to utracer+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/utracer/153988f2-75ea-491f-bed0-fd83f1fafa0an%40googlegroups.com.

[Never Mind]

Bill

I have never used the internal filament supply on the uTracer3 as I found in testing it not very accurate.

Measure a tube with a very accurate 6.3 volt DC external supply measured close to the tube socket and then remeasure it using the utracer3 internal power supply.

I saw a 10% to 20% drop in emissions between the external and internal power supply.

As I was too picky to accept the loss in accuracy for data to create spice models I just moved on and used a bench supply for the filament.

Never seen a report of tube destruction from the internal power supply so the risk seems low.

Bob

[Ihor]

Just curious,  I've never had a lockup on mine yet. Since the filament power is regulated by the PIC, what happens there? Does the output go to zero or full power? Perhaps it depends on the exact state of the FET when the lockup occurs?
Bill v

The internal PWM filament voltage, if set, is present all the time, except the moment (milliseconds) when the measurement is performed. So the PWM is switched off at that moment, and then reactivated. If something hangs (almost surely outside the time for the measurement) the voltage is still there I believe, because PIC uses separate independent PWM register which does not care about the main loop in the code. Again, I think the "hang" that we observe is not like halting of the CPU or a segmentation fault that crashes the main loop and goes nowhere. The PIC just stays in some infinite loop (trying to charge the caps or convert the measurements from the ADCs) which interferes with its responsiveness, but the clock, PWM, computations are still running. Also, the voltage level for the filament is never adjusted because there is no feedback or any sensing circuit, so it is just set and that's it.      

[Davo]

Would a simple double pole push button with resistor & as short as possible wires work also?

Or would it add unwanted capacitance, oscillations, or interference?

That HV bug is very annoying though

Op 23 sep 2024 om 15:38 heeft Ihor <iste...@gmail.com> het volgende geschreven:



Just curious,  I've never had a lockup on mine yet. Since the filament power is regulated by the PIC, what happens there? Does the output go to zero or full power? Perhaps it depends on the exact state of the FET when the lockup occurs?
Bill v

The internal PWM filament voltage, if set, is present all the time, except the moment (milliseconds) when the measurement is performed. So the PWM is switched off at that moment, and then reactivated. If something hangs (almost surely outside the time for the measurement) the voltage is still there I believe, because PIC uses separate independent PWM register which does not care about the main loop in the code. Again, I think the "hang" that we observe is not like halting of the CPU or a segmentation fault that crashes the main loop and goes nowhere. The PIC just stays in some infinite loop (trying to charge the caps or convert the measurements from the ADCs) which interferes with its responsiveness, but the clock, PWM, computations are still running. Also, the voltage level for the filament is never adjusted because there is no feedback or any sensing circuit, so it is just set and that's it.      

--

You received this message because you are subscribed to the Google Groups "uTracer" group.
To unsubscribe from this group and stop receiving emails from it, send an email to utracer+u...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/utracer/c7741cc3-0ac7-4df1-9b22-4b648394bb17n%40googlegroups.com.

[D.A.R Achterberg]

To discharge both HV caps

Op 23 sep 2024 om 15:49 heeft Davo <djda...@gmail.com> het volgende geschreven:



To view this discussion on the web visit https://groups.google.com/d/msgid/utracer/347C024B-D20E-496D-A265-EB26D8A311FA%40gmail.com.

[Bill van Dijk]

Thanks.
Bill v.

To view this discussion on the web visit https://groups.google.com/d/msgid/utracer/c6f6e5b7-5add-45cb-9691-c5f7f2b32dddn%40googlegroups.com.

[Bill van Dijk]

Ah, yes, if the PIC simply gets stuck in some internal loop, it would not affect the PWM. That would also make it very difficult to implement an external watchdog.
Bill v.

--

You received this message because you are subscribed to the Google Groups "uTracer" group.
To unsubscribe from this group and stop receiving emails from it, send an email to utracer+u...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/utracer/c7741cc3-0ac7-4df1-9b22-4b648394bb17n%40googlegroups.com.

[Martin Manning]

Investigating the situation with my uT3+ (600 mA) I found it passes all hardware checks, but those are at low current. 

Next I ran a known 6L6GC at 400V Vg2 as a test case for high currents. Results there are interesting, as the Dekker GUI can do it, uTJS succeeds with Double Acquisition off, but fails with it turned on. Terminal log is attached below. 

At this point I'm questioning the assumption of a hardware failure. 

The low-current hardware checks also look different using uTJS with Multiple Acquisition turned on and off. 

The Dekker GUI results look like what I have seen before, but there is an unexplained kink at the low end of the current amp cal run.

Here is the effect of 1.5 Rextra and 0 Rextra,  Dbl Acc on:

Here is the effect of 1.5 Rextra and 0 Rextra,  Dbl Acc off (note kink at the low end of the traces):

[Martin Manning]

Ignore the first screen shot in the post above, and note that the current amp cal traces are using 2x 10k resistors, and the traces should be centered around 20 mA at 200V.

[Never Mind]

For interest I retested my stock uTracer3P with a 121 ohm load on the anode port using DA OFF and DA ON.

Both modes worked fine on my stock uTracer3.

In contrast to past experience with my my uTarace6 I found results improved in accuracy with Double Acq turned on.

Interesting and I will have to further investigate.

Averaging was left at Auto

Voltage correction was ON

Here is the attached data.

Top data is with Double Acq turned OFF

Bottom data is with Double Acq turned ON

Note that with DA turned ON data errors are smaller and the measured values more consistent.

Note I used in testing for best results.

V adc offset = -0.2V

V rextra losses = 5 ohms

The corrected accuracy of a stock uTracer3 is really very good in these test results.

I am impressed.

[Never Mind]

Martin

If I can make a suggestion in tracking accuracy issues using a reference test resistor.

Measure a wide range of test voltages and so test currents and the resistance value can be calculated over a wide range of test voltages and currents.

For example in your testing measure the 10K resistor from 2 volts to 210 volts using 30 or more points.

Then paste the resulting uTracer V/I data into a spread sheet to calculate the resistance values for each v/I data point.

This can uncover any nonlinear accuracy results in the uTracer over a range of test voltages and currents that will not be seen with single measurements or a narrow ranges of test voltages.

The resulting resistance values can provide clues as to what or where the accuracy issue may be.

Good luck

Bob

[Ihor]

Hi Martin, 

I checked your log where uTracerJS hangs with the Double acquisition and it is apparently because for the low voltages unexpected things can happen. I should add some code to monitor such thing. The Double acquisition applies the set voltage and then expects the drop in the measured value and then increases that set voltage a bit. But as you can see in the screenshot, you set 2V but measure not a low voltage but higher, 4V, so uTracerJS tries to reduce it even more, and then problem is that if one tries to set any voltage <2V, uTracer will hang (it will go to that infinite loop trying to charge caps which is not possible, this is also described on forum here). So that logic with the double acquisition should be active only on higher voltages or only in the direction of increasing the set voltage. 

About the kink, two questions would be it is really reproducible (and not a "noise" of ADC at that low "current" level for your high-current-mod uTracer3+) and if you lower the voltage that you start with to 180 or 170V, is that kink always the first point? In the case of Double acquisition measured curves might potentially be smoother because the HIV caps are not charged or discharged to the new voltages for the new point but just checked and slightly adjusted because it is almost the same measurement point, but it is probably visible only when the step along the Va voltage is large.  

[Never Mind]

If "V adc offset" is set to a much higher than correct value (above 1 or so volts on a uTracer3) it could result in a returned Va voltage higher than set.

On my utracer3 I have found "V adc offset" should be -0.1 to -0.2V.

I wonder what the cal file values used are?

Bob

[Never Mind]

Ihor

I did a test setting V adc offset to 2 volts in a Utracer3 and set double Aqu. ON.

The log does show measured Va higher than the set Va.

However on my uTracer3 the "Reacquiring with adjusted voltages!" occurred and then testing continued anyway.

Not sure what this may mean but thought it may be of interest.

Bob

[Ihor]

Hi Bob, 

If I have time I will probably create a hardware simulation for that. From your log I see that actually I already have that problematic case handled in the code. As you can see in the screenshot, setting Va to 2V returns higher voltage, and then uTracerJS tries to lower it but in the "Retaking" step that voltage should be less than 2V but it is properly kept at the allowed minimum (of 2V). That also happens for the next measurement as well, and only for the 3rd point the set voltage moves on. 

Because of the different calibration values for you and Martin, those physical 2V can be converted to different HEX values, and of the HEX value for the Va or Vs is lower than 002A then uTracer will hang. You can also send such command directly from the terminal and it will happen. In all your cases I saw values of 0030HEX and above, but also with Martin's log it did not look wrong. I can try to put the same calibration values as Martin has and see what kind of numbers are generated, but I am pretty sure it is something related to that. For a simple test, starting from 4 or 5V should not cause any problems. 

[Martin Manning]

Ihor wrote:

I checked your log where uTracerJS hangs with the Double acquisition and it is apparently because for the low voltages unexpected things can happen. I should add some code to monitor such thing. The Double acquisition applies the set voltage and then expects the drop in the measured value and then increases that set voltage a bit. But as you can see in the screenshot, you set 2V but measure not a low voltage but higher, 4V, so uTracerJS tries to reduce it even more, and then problem is that if one tries to set any voltage <2V, uTracer will hang (it will go to that infinite loop trying to charge caps which is not possible, this is also described on forum here). So that logic with the double acquisition should be active only on higher voltages or only in the direction of increasing the set voltage. 

About the kink, two questions would be it is really reproducible (and not a "noise" of ADC at that low "current" level for your high-current-mod uTracer3+) and if you lower the voltage that you start with to 180 or 170V, is that kink always the first point? In the case of Double acquisition measured curves might potentially be smoother because the HIV caps are not charged or discharged to the new voltages for the new point but just checked and slightly adjusted because it is almost the same measurement point, but it is probably visible only when the step along the Va voltage is large.

Thanks for looking into that. The kink is reproducible and always on the first point in the current cal test. I see a similar artifact on other traces too. 

Bob wrote:

If I can make a suggestion in tracking accuracy issues using a reference test resistor.

Measure a wide range of test voltages and so test currents and the resistance value can be calculated over a wide range of test voltages and currents.

For example in your testing measure the 10K resistor from 2 volts to 210 volts using 30 or more points. 

Then paste the resulting uTracer V/I data into a spread sheet to calculate the resistance values for each v/I data point. 

This can uncover any nonlinear accuracy results in the uTracer over a range of test voltages and currents that will not be seen with single measurements or a narrow ranges of test voltages.

The current amp cal test is focused on 200V for that single-point calibration. I've done as you suggest re calculating resistances previously for various purposes. For the present issue, I traced a 50 ohm resistor from 2 to 30V, and for linearity I looked for the trace to pass through 200 mA at 10V and 400 mA at 20V. It's pretty close, but the traces are not as smooth as they should be. 

Based on the above behaviors, I'm back to the conclusion that there is some hardware damage, and I'm thinking I will just replace the transistors in the HV switches. 

[Never Mind]

Re:"I've done as you suggest re calculating resistances previously for various purposes. For the present issue, I traced a 50 ohm resistor from 2 to 30V, and for linearity I looked for the trace to pass through 200 mA at 10V and 400 mA at 20V. It's pretty close, but the traces are not as smooth as they should be."

Would you consider to share the used setup file, the cal file and the resulting measured data files from your reference resistors?

I would find the setup, cal files and your data files sweeping the resistors interesting to compare my setup and data.

Take care

Bob

[Never Mind]

Re:"About the kink, two questions would be it is really reproducible"

Ronald noted kinks from the auto range circuit as certain current levels are crossed.

See

[Martin Manning]

Bob wrote:

Would you consider to share the used setup file, the cal file and the resulting measured data files from your reference resistors?

I would find the setup, cal files and your data files sweeping the resistors interesting to compare my setup and data.

Sure. I can run some data, but It wouldn't be representative given the issues I'm having. I would like to replace the transistors in both HV switches and see if that clears things up. I might as well change the boost converter FET's and diodes to the current part numbers too.

Attached here is my cal procedure, which explains the setup.

[Never Mind]

Martin

Thanks for your reply and hopefully you can sort out your uTracer issue without too much trouble.

When you get your hardware working again it would be interesting to see what results a 600mA uTracer3 produces when measuring lower value reference resistors in the 50 to 85 ohm range.

I was impressed with how accurate the measurements of my stock uTracer3 were using a 120 ohm test load. They are far better in accuracy than my uTracer6 using a 50 ohm load.

As I have only a stock utracer3 and a uTracer6 to test with it would be interesting to see data from a uTracer3 600mA unit.

I am sorry I was not clear when I asked for your cal file. My mistake.

What I was hoping to see was the "_eeprom.csv"  file that utracerJS uses to hold all calibration values used in testing.

I am especially interested if you have set any values for "Vadc offset" or "Rextra losses" and if so what they may have been.

Thanks and take care

[Martin Manning]

Bob wrote:

What I was hoping to see was the "_eeprom.csv"  file that utracerJS uses to hold all calibration values used in testing.

I am especially interested if you have set any values for "Vadc offset" or "Rextra losses" and if so what they may have been.

eprom file is attached below. I have not changed the calibration inputs other than to experiment with Vadc and Rextra.

The 600 mA modification is current sensing resistors (R45 and R20) and the hardware current limit resistors (R38 and R24), plus the "battery" caps (C19 and C14). 

[Martin Manning]

PS Ihor,  I assume the values for "Rbot" and "Rtop" (R32 and R33 in the above circuit snip) are reset when using uT JS for a uT3+ as opposed to a uT6?

[Davo]

Hey Martin,

Where did you get that 600mA mod from?

I thought it was only changing the current sense resistors…

Op 28 sep 2024 om 12:26 heeft Martin Manning <mman...@fuse.net> het volgende geschreven:

PS Ihor,  I assume the values for "Rbot" and "Rtop" (R32 and R33 in the above circuit snip) are reset when using uT JS for a uT3+ as opposed to a uT6?

--

You received this message because you are subscribed to the Google Groups "uTracer" group.
To unsubscribe from this group and stop receiving emails from it, send an email to utracer+u...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/utracer/f22d6cda-2faf-4aa4-87e9-18582be6a381n%40googlegroups.com.

[Martin Manning]

See here: https://www.dos4ever.com/uTracerNotebook/Notebook.html#Current1

[Ihor]

Hi Martin, 

Those two resistors are hardcoded for each version of uTracer that one can choose in the Calibration tab. So they are fixed and properly used every time the measurement is done. If those would be some other values then I could either introduce one more “version”  there or just put then in the eeprom.csv file so they could be set from outside. 

[Martin Manning]

Thanks, I was pretty sure you would have that covered ;^)

Yesterday I changed the three transistors in both HV switches and I am still getting strange results:

Still have the kink at the left end (first point) with Dbl Acc off. 

Still getting the voltage droop with Dbl Acc on, third plot is with scales adjusted to show the entire trace. 

The I vs. V looks right for 2x 10k.

At low voltage with 50 ohm resistor, the screen channel is showing current that is much too low, and much higher current on the unused channel (Anode).

[Never Mind]

Martin

Martin wrote "At low voltage with 50 ohm resistor, the screen channel is showing current that is much too low, and much higher current on the unused channel (Anode)."

The screen data was taken with Va,Vg with Vs constant set to 5V.

As the screen is set to a single fixed 5 volts it is very hard to see what issues if any may be present in the screen circuits.

Keeping in mind that uTracers loose accuracy rapidly below 10 volts especially if the extra correction factors are not applied in the config file the screen port results may be normal.

 Given the anode sweep shows ~ 80mA at ~ 5 volts it's possible the screen port may be working correctly.

If a full screen sweep from 2V to ~ 35V to is done to get close to a full 600mA with your 50 ohm test resistor the UTD data file created can then be used to calculate levels of error over a large range of current and voltage.

This is a good way to confirm that a port is working fully in my experience. 

The anode voltage sweep at 50 ohms seems to be showing about the correct 50 ohm resistance although it would be easier to confirm exact error levels with the UTD file.

A sweep to the full current capability is also useful to confirm the port is working correctly up to the full 600mA current level.

It does however seem possible the anode port is working correctly.

A  UTD data file of a full current sweep for both anode and then screen would allow the calculation of exact values for Vadc offeset and Rloss extra for your unit.

From of a 50 ohm sweep of 2V to ~ 35V getting close to 600mA if you send me the UTD file and the eeprom config used to create the data I am happy to find correct values for Vadc offeset and Rloss extra for your utracer3.

Take care

Bob

[Martin Manning]

Bob wrote:

The screen data was taken with Va,Vg with Vs constant set to 5V.

As the screen is set to a single fixed 5 volts it is very hard to see what issues if any may be present in the screen circuits.

Keeping in mind that uTracers loose accuracy rapidly below 10 volts especially if the extra correction factors are not applied in the config file the screen port results may be normal.

Thanks for looking. Yep, my bad, I just moved the resistor's banana plug to the screen terminal thinking I was tracing using Va=Vs. Corrected results are shown below, Double Acquisition on and off, and look much better!

Bob wrote:

A  UTD data file of a full current sweep for both anode and then screen would allow the calculation of exact values for Vadc offeset and Rloss extra for your unit.

From of a 50 ohm sweep of 2V to ~ 35V getting close to 600mA if you send me the UTD file and the eeprom config used to create the data I am happy to find correct values for Vadc offeset and Rloss extra for your utracer3.

Great! I will run through the calibrations with Vadc and Rloss =0, and then run traces of the 50 ohm exactly as above and using Va=Vs. Should Double Acquisition be on or off? 

I am thinking I might change the current sense resistors to 6.8 ohm 1% 50 ppm (need to order those), and the hardware current limiting resistors to 1 ohm 1% to get rid of the parallel resistors used for the 600 mA modification. That should result in a 529 mA compliance limit per R Dekker's formula, and in that configuration I will call it a 500 mA uTracer.

[Martin Manning]

[Never Mind]

Martian

Martin wrote "I will run through the calibrations with Vadc and Rloss =0, and then run traces of the 50 ohm exactly as above and using Va=Vs. Should Double Acquisition be on or off"

I would create UTD files with the double aqu. turned on and then turned off. With both we can determine what mode provides to most accurate results.

It is important to getting a good estimation of the correct values for V adc offset and R extra to acquired data as close to the maximum current the system can supply.

If you use the 50 ohm resistor try to get as close to 600mA as possible. Set it all up do a test sweep and then truing compliance off run a finial sweep to get the data file helps.

Normally compliance must be turned off for the full current ability of a utracer to be used at low voltages due to the voltage droop in the storage capacitors during the test pulse.

This is because the compliance current measurement is made at the start of the test pulse before the storage capacitors have drooped in voltage and then the finial measured result for voltage and current is made at the end of the test pulse after the storage capacitor voltage has dropped when both voltage and current can be much lower in value.

The result is compliance will always limit the test current to well under maximum values at lower test voltages.

In my experience the most accurate values for V adc offset and R extra can be found when test data reaches the maximum test current of the system at a test result voltage of about 50 volts.
Best results come from adjusting the voltage sweep range to make sure test pulses of a high enough voltage are used to get to the maximum test current, select a test resistor value for your system based on its maximum current ability at 50V and then turn compliance off for the last sweep to collect the data file.
Redoing the sweeps over several test resistor values is also useful in confirming the calibration values are valid over a wider range of test conditions.

uTracerJS has a current limit function "Ia max", "Is max" that limits the test current based on the test current after the capacitors have drooped in voltage. This can be very useful at test voltages below 100V.

This allows testing up to the full system current ability while still providing some protection for over current events when compliance is turned off. 

On a uTracer3 the switch not robust enough to survive a dead short with compliance turned off ( it can even will fail sometimes with compliance on!) but modest overloads will now shut down measurements pulses based on "Ia max" or  "Is max" even with compliance turned off.

Looking forward to seeing data files when you have time to collect them.

Bob

BTW

I am considering creating a small PCB to hold a alternate switch circuit design for utracer3 users. 

This is because the utaracer3 switch is run very very hard in Ronald's design and so is vulnerable to damage from mishaps. 

The switch can become damaged when compliance is shut off the get full current measurements at lower voltages if a short should occur in testing.

Ironically Ronald recommends you never turn compliance off and then admits he often does so to get measured results.

So I am considering making this small add on PCB for the uTracer3 that will feature a blow out proof switch design even if full voltage measurements are made into a dead short with compliance turned off on a 600mA uTracer3.

It would also improve accuracy for high current  lower voltage testing.

[Martin Manning]

Bob wrote:

If you use the 50 ohm resistor try to get as close to 600mA as possible. Set it all up do a test sweep and then truing compliance off run a finial sweep to get the data file helps.

...

In my experience the most accurate values for V adc offset and R extra can be found when test data reaches the maximum test current of the system at a test result voltage of about 50 volts.

When I converted to the 600 mA version I made some runs with compliance off to see exactly where the hardware limit started rolling so I could be sure of getting to 600 mA unimpeded.

I'm going to go ahead with my plan to change the current sensing and hardware limiting resistors to get a "500 mA" device. I've ordered the parts and included some 100 ohm resistors for the test runs so I can hit 500 mA at 50V. I think I will also change the boost converter FET's to the current SPA07N60, which have better specs than the original IRF840's. It'll be a few days before I get the parts, swap them in, and make the runs.

Bob wrote:

I am considering making this small add on PCB for the uTracer3 that will feature a blow out proof switch design...

I'm very interested in seeing what this is about from a design standpoint, but I have very little space left in my uTracer enclosure, so implementing it might not be possible in my case.

[Martin Manning]

Made some good progress...

Replaced Boost converter FET's and diodes, current sense and hardware limit resistors in both channels. 

Re-ran the standard calibrations. 

Double Acquisition ON still shifts the traces down for the 2x 10k current amp cal when tracing over a small voltage range.

Double acquisition OFF makes a nice trace when spanning 150V (1st screen capture).

Remarkably, the tracer no longer "chirps" when acquiring data. I suppose that must be due the new SPA07N60 FET's and/or the new diodes (UF4007). Also noticed that the resistors on the reservoir cap drains are discolored, so they  have been hot; more so on the screen channel (2nd screen capture). 

Traced 2x 100R resistors to approach 50V with and without Double Acquisition (3rd, 4th, and 5th screen captures). 

Adjusted Rextra to line up the high current end, Rextra = 5 ohms.

The low end looks good at 5V and 50 mA with Vadc offset = 0. 

Tested the hardware limit using 2x 100R. Both channels clear 500 mA (6th screen capture). 

These are 1% wire wound, and measure 100.53 and 100.56 ohms using my B&K 879B LCR with Kelvin clips.

Done for the day now, but I'll work on getting .utd files for the 100R resistor traces. 

[Ihor]

Hi Martin,

Nice to see that fixing some hardware issues resulted in better acquisition. About the Double acquisition results, I see that it is supposed to measure 10 points but there are only 8 visible. It looks like the last two are cut, so it is not a shift probably but they are not measured. DA in this case should have worked as expected. Were there some compliance issues? Or going above the Imax thresholds? In the log it should be visible if some points were skipped. 

[Never Mind]

Martin

A word of caution on the uTracer3 switch as I would not want to see you newly repaired uTracer3 becoming damaged again.

Ronald's utracer3 design has little margin in the switch at high currents.

The safe area of the KSA1156 is only Vce=25V @  Ice=650mA

See attached picture of switch safe area.

To make the danger larger Ronald's design only limits current in the switch and does not protect the switch based on the safe area (Vce * Ice)

This means when the hardware current limiter is active the switch is in great danger of overload and failure as the current stays high and the Vce keeps rising on each measurement until the switch fails from safe area overload.

In your data the last one or two measurements when the current limiter circuit was active the switch may have been operating past it's rated safe area of only 650Ma @ 25V.

Personally I never design past the data sheet safe area values for a BJT due to the high risk of failure although some designers do.

A solution is to always set in uTracerJS Ia and Is max. current value to no more than the system's maximum current measurement capability and most importantly to less than the hardware current limiter set point. 

In your system's case that would be the  system maximum current measurement capability of about 500mA. 

This way uTracerJS will stop measurement pluses when current levels reach the maximum the system can handle before the hardware current limiter becomes active or the current measurement AD reaches numerical saturation.

This will prevent the switch from being damaged by the test voltage continuing to increase on each pulse when the hardware current limiter is active and so the measured current no longer is rising.

Bob

[Martin Manning]

Bob, 

Ok, so I may have fried my switches again. I need to go back and revisit the calibrations to see if anything has changed. I did try again to trace a 6L6GC using uT JS, and it failed, resulting in a HV LED on hang-up. I tried using R Dekker's GUI and it worked, with the curves looking good for this tube, which I have used as a reference since building the tracer. 

On the KSA1156, the "Pulse" see operating area is for a square wave? Maybe it can survive the duty cycle in the uT3+? I would choose 400 mA at 400V as the design point for this tracer, with 500 mA being absolute maximum. 

Ihor, 

I will get a capture of the log for that DA on case.

[Never Mind]

Martin said " Ok, so I may have fried my switches again"

Possible but maybe not as you were running close to the edge of safe area.

With the margin a manufacture builds into semiconductors for production tolerance your switches may still be fine.

Martin said "On the KSA1156, the "Pulse" see operating area is for a square wave? Maybe it can survive the duty cycle in the uT3+?"

Safe area is a issue with the switch based the voltage across the switch AND the current through the switch.

So with the switch fully off (no current) the full 400V rating applies and there is no safe area issue.

With the switch fully on (very small voltage across switch) the full 800mA for 1mS pulse rating applies.

Safe area rating is not based on a repeating pulses like a square wave.
Safe area is based on a single pulse of the length shown on the data sheet of 1mS, 100uS or 10uS.
The single pulse will heat up the transistor junction very quickly and with time it will cool off and be ready for the next pulse.
There can be a lot of heat in the single pulse event. For example a pulse of 400mA with 400V across the transistor for 10uS is 400V * .4A  = 160 watts. OUCH that is hot.
The data sheet does not tell us how long to wait between pulses so I have assumed by the time the next 1.25mS test pulse comes around the transistor junction has cooled off.

Safe area becomes the main danger when the switch is not fully on or fully off. 

That happens when the hardware current limiter comes on during over current testing.

When the hardware current limiter comes on during a pulse it partly turns the switch off to limit the current to a maximum current value. 

This allows the voltage across the switch to rise while there is still a lot of current through the switch and you will very quickly exceed the safe area for the switch used in a uTracer3.

With compliance turned on the PIC will shut down the switch very quickly on over current and normally will save the switch in a stock utracer3.

The data sheet shows a maximum of 400mA at 400V for 10uS. This allows compliance and the PIC time to shut down the switch on a stock uTracer3.

With compliance off a dead short will often destroy a uTracer3 switch as it lacks the safe area margin to survive even a single pulse event into a short.

A high current uTracer3 runs at a current higher than 400mA.

On a high current uTracer3 by the data sheet there is risk to the switch into a over current event even when compliance is turned on above ~ 250V test voltage.

With compliance turned off you have to pay very close attention to the maximum test current.

Take care what  Ia and Is is set to making sure pulses stop when the maximum system current is reached before the hardware current limiter becomes active.

The limited safe area available to survive a over current event is why I redesigned my uTracer6 switch after the switches failed during use.

The new switch and driver design I used for my uTracer6 by the data sheet is able to survive a full 1 amp pulse at the full 1KV test voltage even with compliance turned off into a dead short.

This is also my goal for the replacement switch design for my uTracer3. 

To also allow any uTracer3 to survive a full voltage and current pulse into a dead short even with compliance turned off.

[Martin Manning]

Ihor, 

Here are runs with and without DA, first 2x 10k and then 2x 100R where some of the specified points do not appear in the plot.

The I vs. V looks the same as in the plots above, so its seems the hardware is functioning as it was after I replaced the transistors in the switches. 

[Ihor]

Hi Martin, 

Thanks for the test. Are you using any compensation in terms of Vadc or that R_extra_losses because it looks strange to me that the adjusted voltage is lower then the set one, even for the very first simple case. So it is either a bug in the software that was introduced with all those new parameters, or those compensation values are messing things up. About skipping some points I will check the logs more carefully a bit later.

Ihor

[Martin Manning]

Hi Ihor, here is the cal:

[Martin Manning]

Thanks for the detailed response. This is interesting stuff.

Bob Wrote:

Safe area is an issue with the switch based the voltage across the switch AND the current through the switch.

So with the switch fully off (no current) the full 400V rating applies and there is no safe area issue.

With the switch fully on (very small voltage across switch) the full 800mA for 1mS pulse rating applies.

Yes, understood.

Bob Wrote:

Safe area rating is not based on a repeating pulses like a square wave. 
Safe area is based on a single pulse of the length shown on the data sheet of 1mS, 100uS or 10uS.
The single pulse will heat up the transistor junction very quickly and with time it will cool off and be ready for the next pulse.
There can be a lot of heat in the single pulse event. For example a pulse of 400mA with 400V across the transistor for 10uS is 400V * .4A  = 160 watts. OUCH that is hot.
The data sheet does not tell us how long to wait between pulses so I have assumed by the time the next 1.25mS test pulse comes around the transistor junction has cooled off.

Safe area becomes the main danger when the switch is not fully on or fully off. 

That happens when the hardware current limiter comes on during over current testing.

When the hardware current limiter comes on during a pulse it partly turns the switch off to limit the current to a maximum current value. 

This allows the voltage across the switch to rise while there is still a lot of current through the switch and you will very quickly exceed the safe area for the switch used in a uTracer3.

 

In my case I turned compliance off and let the hardware limit hold the current with a 100.5R load. I believe the HV switches were subjected to the following conditions, which looks like I was just inside the safe zone on an energy basis, except that each data point is the result of multiple current pulses, and Tj will be rising with each additional pulse.

Bob Wrote:

With compliance turned on the PIC will shut down the switch very quickly on over current and normally will save the switch in a stock utracer3.

The data sheet shows a maximum of 400mA at 400V for 10uS. This allows compliance and the PIC time to shut down the switch on a stock uTracer3.

With compliance off a dead short will often destroy a uTracer3 switch as it lacks the safe area margin to survive even a single pulse event into a short.

For perspective, 400V, 400mA, for 10us is 0.0016J, 1/10 of the 1ms safe limit. If the compliance is turned off, and the pulse is 1.25ms, the energy is 0.2J, 12.5x the 1ms safe limit.

Bob Wrote:

A high current uTracer3 runs at a current higher than 400mA.

On a high current uTracer3 by the data sheet there is risk to the switch into a over current event even when compliance is turned on above ~ 250V test voltage.

With my current 529mA compliance, the maximum time at 400V to deliver 0.016J would be about 60us. 

[Ihor]

Hi Martin, 

I did some simulations of that voltage drop during the acquisition and for me it actually did the good job, so uTracerJS was increasing the voltage during the Double acquisition as it should. At the same time, I also make a small change tot he code, so please download and try the latest version. Before that compensation that Bob proposed (with Vadc and Rlooss) was active anly if both values were non-zero, and in your case only Rloss is non-zero so it was no properly used. 

[Martin Manning]

Hi Ihor, here is a repeat of the 2x 100R traces using the updated uT JS. I did not change the values for Vadc (0) or Rextra (5) from what you see above. Looks like the traces pass through the expected Ia/Va points.

[Ihor]

Yes, I see also that the compliance in the end actually stops uT from acquiring the last points, I was always looking where are the points for the last Va specified in the range :) 

I think for such test it is also better to use for example a range of Va from 5 to 60 with 12 points, then (or 5 to 50 with 10 points) so then all the measured voltages should fall on the Va axis in the steps of 5V, which is easy to align with the grid.   

[Martin Manning]

Here it is again with tweaks to Vadc and Rextra to line up the I-V trace. Looking good. The points don''t align with the grid, though. I'm also trying to reset the x axis to manual, 0 to 45 by 5 and save it, but it keeps resetting it to 5 to 45.

[Martin Manning]

Hi Ihor, 

What's happening here? The DA "off" curve looks correct, DA "on" does not. DA "on" also fails and hangs up if the starting anode voltage is less than ~10. I think that's because a negative voltage is requested? Maybe the minimum voltage target for the second pass needs a minimum value of 2?

[Ihor]

Martin

"Here it is again with tweaks to Vadc and Rextra to line up the I-V trace. Looking good. The points don''t align with the grid, though. I'm also trying to reset the x axis to manual, 0 to 45 by 5 and save it, but it keeps resetting it to 5 to 45."

I will check later if I can reproduce this behavior. I do not see the bottom part in those screenshots, so I am not completely sure what you set and what changes, basically the steps to reproduce the problem. 

About the different curves, indeed it is strange. I would try to measure some other tube on a more reasonable scale lets say up to a hundred of mA, and also put those parameters Vadc and Rlosses to zero, and then see is the double acquisition works as it should. I would not expect any problems with the code, and typically two curves should be the same just with shifted points (still along the same more or less curve). Right now the curves are quite different which is also just the response of uTracer to double acquisition, so it is a reaction of hardware to two pulses at the same (more or less) set point, I would say, because decoding and encoding of the values and messages is pretty robust and should not affect that. You can also add a Delay (0.5-1sec) just to test, if taking some breaks in between would help, but I am not sure if that delay in fact is also works with double acquisition pulses. 

Another option for comparison is to show two curves in one plot, so you can acquire one set of curves, then save them in the Clipboard, then acquire the second set of curves, and then in the dropdow box of Y2 axix jsut select the first set from the Clipboard.       

[Martin Manning]

Ihor, here is an EL84. First trace is with DA off, and it looks good. Second curve Y2 axis is DA "on" and it shows the dip and data points shifted to lower voltage that I saw with the 6L6GC. The third curve is DA "on" with Vadc and Rextra both set to zero.

I tried making plots using the clipboard to compare the Vadc and Rex corrections in and out, but when I changed the cal values the stored curves were lost. The second plot shows that case, and it looks just like the dashed curves in the first plot, so no effect from the corrections.

[Ihor]

Hi Martin, 

Indeed, if you go to another tab and come back the curves in the clipboard will be lost. I was planning to implement it long time ago but always postponed. I just updated the version of uTracerJS so now it will keep the curves from the clipboard after refreshing the main page or switching to another tab, even if you restart uTracerJS they will still be there. They are stored in the browser now, so only if you close the browser tab they will be cleared. But even with the old version I was just opening the Calibration tab in a new browser tab, so the main one always kept the Clipboard curves. 

Also when you get those hangs, it would be interesting to see the log, basically if it happened in the beginning of the curve or somewhere in the middle. 

About the different in curves, I still suspect that the way of doing that double acquisition affects the results. Probably the compensation of the Vs in this case affect the hight of those curves. Probably you can send me the logs for both cases, to see what exactly the voltages Va and Vs were during the DA. And for a moment I would indeed keep Vadc and Rlosses set to zero.  

[Never Mind]

Martin said

"For perspective, 400V, 400mA, for 10us is 0.0016J, 1/10 of the 1ms safe limit. If the compliance is turned off, and the pulse is 1.25ms, the energy is 0.2J, 12.5x the 1ms safe limit."

I am do not understand what you are saying here. Perhaps my last post could use some improvement.

The charts on the KSA1156 data sheet give us current/voltage/time only.

Safe area is a complex process in a BJT and a lot of research went into the issue in the 1960 as power transistor failure was common even when devices were used within the know dissipation limits(V*I). 

The result of this research was the idea of safe area. 

A BJT's safe area tends to be nonlinear in relation to voltage, current and time. It is unclear to me that extrapolation based on Joules of one part to another of the safe area chart can be valid.

The KSA1156 is a Epitaxial type device. They are fast but more prove to secondary breakdown than older diffusion types.

A  safe area issue with a BJT is that at higher voltages across the junction the current tends to become focused on small areas of the die creating hot spots.

This is called Secondary breakdown. 

It tends to surprise designers just how just how little current many BJT can handle at higher voltages.

For example the KSA1156 data sheet shows at only 10mA at 400V for 1mS or 4 watts power. 

For 1mS at 200mA and 90V the KSA1156 is rated for 18Watts.

Because of Secondary breakdown effects the data for a KSA1156 at 400V shows only 10mA for 1mS. 

10mA at 400V is not much current compared to the 600mA current limit in the switch during a full measurement pulse of 1.25mS with compliance turned off.

Martin said

"With my current 529mA compliance, the maximum time at 400V to deliver 0.016J would be about 60us."

By the data sheet at 10uS you get 400mA at 400V.

The data sheet shows for 529mA the 10uS limit is only about 330V. 

Any amount of time at 529mA and 400V is outside the safe area of the data sheet and a no go zone.

Some designers will extrapolate from the data sheet to "extend" the ability of a device. You do so at your risk and it is a practice I have avoided.

The manufactures perspective is that if you operate outside the data sheet you are on your own if you get device failures.

In summery 

For myself I have seen uTracer switch failures.

Looking at the uTracer designs I can see failures are to be expected if you operate at the highest current levels.

Now that calibration compensation for switch losses can be set in uTracerJS allowing changes to the switch design I am in the process of making my utracers more robust and accurate at the highest current levels.

Many users may find my changes unneeded.

[Never Mind]

Martin said "First trace is with DA off, and it looks good. Second curve Y2 axis is DA "on" and it shows the dip and data points shifted to lower"

I too noted odd accuracy issues using the DA mode when I was investigating high current low voltage accuracy of utracers.

I simply stopped using it.

It will be interesting to see what is discovered as the root cause of the accuracy issues.

Do I understand correctly this is done on the the GUI side?

If so what is the sequence of commands vs normal acquisition?

If multiple measurements that are identical in set voltage values are quickly sent (simulating the DA process except keeping the set point unchanged) it would be interesting to see if the returned results are consistent from first to second, third, etc sample. May give a clue as to any hardware/firmware issue.

I will set up a 6BQ5 on my stock uTracer3 and try DA to see if I see the same issue here.

Bob

[Ihor]

Hi Martin, 

To dissect the problem with those curves, I suggest to set those Vadc and Rloss to zero and then run 2 times the measurements with DA, to see if they are reproducible and then set the Delay to 1 or 2 sec and also run the acquisition with DA to see how those curves will look like. This would clarify if the utracer hardware has some problems with setting up the set-points. For comparison you could also take one curve without DA and Delay, as a baseline. Also now you can visualise everything in one plot without problems. If that clarifies something, then we can see how Vadc and Rloss influence the measurements.  

So far in all my tests with the simulator I do not see such inconsistencies, meaning that most likely there is no problem with the code or transformations between the real V/I values and HEX bumbers obtained from uTracer. 

  

[Never Mind]

Ihor

Thinking on this last night an old issue with the uTracer3+ came to mind.

This old utracer3+ issue caused a unexpected loss in anode current in measurements.

This was a disturbance in the PWM signal used for the grid circuit voltage supplied by the PIC during a measurement pulse.

The disturbance is caused by ground bounce inside the PIC due to the high currents used to turn on the two switch opto couplers.

Changes to the PWM filter time constant reduced this issue but with a side effect of a longer setting time after each measurement pulse.

At 10mS there is still a lot of grid error voltage and at 15mS still some voltage error on the grid.

With the longer time constants in the filter Spice shows the settling time back to zero is now close to 30mS with some undershoot as this is a under dampened filter.

A repetition rate of pulses faster than about 15mS will cause these errors to build up in the filter and cause the grid voltage to "drift" negative with time.

Any repetition rate faster than about 30mS will cause the grid voltage to "wander" with time by small amounts due to the filter time response.

So I wondered if during DA mode there could be samples take at a rapider rate and so affecting the grid filter output voltage.

[Never Mind]

Screen shots of grid voltage changes

[Never Mind]

Forgot to post circuit I used in the sim

[Martin Manning]

I was the one bugging Ronald about the grid voltage sag back in 2015. He suggested increasing the cap sizes in the SK LPF but maintaining the ratio. The plot below from my sim shows a series of curves for different pulse widths for 220n/320n vs 220n/220n, which is critically damped. The under damped case comes up a bit faster, but overshoots a bit (yellow vs gray). Either way ~15ms is needed for the voltage to stabilize.

[Martin Manning]

Bob, thanks for the additional discussion on the switch transistor failure mode. I was attempting to make some sense out of the limit curves by looking at the amount of energy jammed into the transistor during the pulse. Later I calculated that for the "dissipation limited" corner points in the safe area curves and found that it is not linear, more like an exponential function of power, suggesting the limiting factor is conducting heat away from the die, with the added complexity of the secondary breakdown. Secondary breakdown reminds me of connecting two diodes in parallel and expecting the current capability to double. Am I on the right track?

[Ihor]

I do not think DA happens at the rate of tens of ms, it should be slower but indeed it can be due to the settling of the grid voltage as well, so would be interesting to figure it out. Again adding that Delay in utJS menu would clarify that.

utracerJS also has that option called Parallel acquisition which I added at some point to speed up the acquisition ( which it typically does, 2 to 6 times) by acquiring the curves not one by one for the same grid voltage but all the points for all Vg but at the same Va, and then going to another Va and scanning all required Vgs and so on, so the caps are not charged and discharged for every curve but actually only one time for all the curve. In any case, so that approach at that time was also producing slightly different results, also I believe noticed by Martin at that time, and then I also asked Ronald and he also suspected settling of Vg to be the problem. It is even mentioned somewhere on the forum here. 

[Never Mind]

I set up my utracer3+ to scope the output of U4, the LP filter from the PWM.

The "bump" that happens during a measurement pulse due to the PIC ground bounce is easily seen.

So is a surprising amount of random high frequency noise on the LP filter output pin but that is a second issue.

I noted the  "bump" decayed very much in line with my spice sims taking nearly 30mS to completely decay after a measurement pulse.

I never saw any repeating measuring pulse with DA on or off before the filter had completely decayed.

This suggests the grid voltage "bump" is not the issue here with DA.

The amount of random high frequency noise at the filter output I saw exceeded 100mV. This is with the scope ground connected very close to the low pass circuit.

The scope ground connection point is very important as the 2 layer board used for a uTracer3+ has a lot of noise across the grounds of the PCB for a data acquisition system.

Sitting idle here is easily 100mV or more of random noise ground bounce between the PIC and the grid LP filter circuit. Not much for a micro processor system but a lot for a analog data  acquisition system

It is of course very difficult to see what if any effect the high system ground bounce noise floor is having on measurements.

It did made me really wonder what the grounds are doing during a high current pulse measurement.

[Ihor]

Quite some time ago we already had discussions about the accuracy of the grid voltage and I posted some measurements in this thread as well:

https://groups.google.com/g/utracer/c/6U_E6eS3_0o/m/US9vaSQtAwAJ

As you can see from the pictures, during the measurement pulses, the grid voltage (on a small scale) is jsut a mess and all over the place :) 

[Never Mind]

I ran a older 6P14P (~6BQ5) on my stock utracer3+.

I did two runs.

One with VC on and a second run with VC and DA on.

The look very similar. 

The only difference I noted was a small droop in anode current  in the first sweep at -0.1V grid of the VC and DA curve.

This droop is very small compared to what Martin is seeing but it is in a similar position and so makes me wonder.

Of interest I did the VC and DA curve several times and the small droop seemed to be consistent.

I find it interesting that the grid circuit becomes the most sensitive to noise and ground bounce from the PIC as the grid voltage approaches zero volts.

It is only the -0.1 grid curve that shows the difference on my utracer3+ between DA on and off.

 

I used the values below in the cal file as they provide the best accuracy on my stock uTracer3+.

V adc offeset = -0.2V

Rextra losses = 5

I did have a bit of trouble figuring out how to display two curves at the same time so please forgive my screen shots if this is odd way to get there .

[Never Mind]

Ihor

A minor quibble.

I would find more characters available in the file names and also displayed in the selected fields handy.

My habit is to make my file names tell a story of how the data came to be and so tend on the long side.

For example file "6P14P-pen-Vs250vcda"  is a 6p14p tube in pentode mode Vs=250V VC and DA are both on.

I would add the grid voltages as well if there was room. 

I find this habit makes it much easier later in time to quickly find a setup or data file based on how the measurement was done.

Long ago Unix allowed very long file names then Linux and later windows. Old habits die hard.

Take care

Bob

[Ihor]

Hi Bob, 

The plots with and without DA actually look almost perfect. The small dip is really tiny and it is probably not even reproducible. Even the fact that all the points on the DA curves are slightly above the non-DA curves (barely visible) can be explained with that compensation of the Vs voltage as well (that DA does). So in Martin's case it should be something the with hardware/tube or the fact that the test was also done for Vg voltages that are close to 0, when the grid current starts running is present and the Vg circuit of uTracer cannot deal with it.

About the 28 or 32 characters, that restriction was introduced in ESP32 version of uTracerJS because the file system of that microcontroller has that strict limitation. On the PC version of uTracerJS that limit just stayed (was inherited) without any practical reason. I would have to double check if there are any limitations in the code that would cut the names to be short, but it all can be removed:)

Ihor

[Martin Manning]

Ihor wrote:

in Martin's case it should be something the with hardware/tube or the fact that the test was also done for Vg voltages that are close to 0, when the grid current starts running is present and the Vg circuit of uTracer cannot deal with it.

I could run a trace of Vg1 = 0 with the Low/+Vg loupe engaged to see if this is an issue.

Ihor wrote:
To dissect the problem with those curves, I suggest to set those Vadc and Rloss to zero and then run 2 times the measurements with DA, to see if they are reproducible and then set the Delay to 1 or 2 sec and also run the acquisition with DA to see how those curves will look like. This would clarify if the utracer hardware has some problems with setting up the set-points. For comparison you could also take one curve without DA and Delay, as a baseline. Also now you can visualise everything in one plot without problems. If that clarifies something, then we can see how Vadc and Rloss influence the measurements. 

When I get some time I will make a plan to do these runs and capture the log.

[Never Mind]

Martin

A suggestion.

A quick way to completely eliminate the utracer grid circuit as a possible issue.

Simply disconnect the tube's grid from the output terminal on the utracer and tie it to the cathode terminal.

This insures the grid will be at zero volts.

Then run the two curves one with DA on and DA off.

Any influence of the utracer's grid circuit will no longer be in the results.

[Martin Manning]

Doh! Of course.

[Ihor]

I updated the version of uTracerJS on the website, so now it works with long filenames as well, and everything works after some simple tests but if there are some inconsistencies or problems just let me know. 

[Martin Manning]

Thanks Ihor, I will download that version. 

Re the LPF behavior here is a link to the 2015 discussion on Amp Garage where I was looking to get accurate Vg1 close to zero volts for 12AX7 (as exhibited by g1 shorted to k), save for the last 50mV or so that the grid voltage circuit design can't overcome. https://ampgarage.com/forum/viewtopic.php?p=353226#p353226 I wondered at the time whether the solution of increasing the settling time would go unnoticed in all cases, but it doesn't look like it is an issue here.

Re the current limiting in the uT3, it seems what is needed is some method to prevent the BJT switch Vce from ever exceeding 20V (actually going below -20V). That would result in safe operation for any rapid succession of measurement pulses, up to and including DC at 500mA. As it is, with the HV switch performing the current limiting function it's protecting the DUT rather than itself.

[Never Mind]

Martin said"

Re the current limiting in the uT3, it seems what is needed is some method to prevent the BJT switch Vce from ever exceeding 20V (actually
going below -20V). That would result in safe operation for any rapid succession of measurement pulses, up to and including DC at 500mA. As it is, with the HV switch performing the current limiting function it's protecting the DUT rather than itself."

That would work at the expense of added circuit complexity and the tendency of "safe area" limiting circuits to "latch off" after a over current trip event causes the VCE to rise.
It would also lead to reduced test current ability at lower voltages due to voltage droop in the
storage capacitors during the test pulse.

In my opinion the BJT switch design used on the uTracer3 was obsolete from the very start of utracer project and is best today left behind for modern components and circuits.

The IRF840 would have made a much better switch in the uTracer3 at the project's start.
This old device was tough as nails with a safe area of 2 amps at 500V for 1mS.
It would have easily handled a full 1 amp pulse on the utracer3+ with margin to spare making full protection possible with only a current limiter and eliminated the need for the compliance concept.
The bonus would have been a reliable 600mA utracer from the start.
I designed the IRF840 into a number of high voltage, high power switching products when it came out. It was a great device for it's time.

Now modern MOSFET are even tougher.
The MOSFET I used in my utracer6 switch redesign can take 2.2A at 1000V for 1mS making compliance no longer needed for switch safety.
The MOSFET I used in my uTracer3 redesign PCB can take 4 amps at 400V for 1mS again making compliance unneeded.

You can see Ronald started in a more modern direction with the utracer6 by moving to a MOSFET but he selected such a small part it is still very vulnerable to damage on over current events due to a very limited safe area.

Today MOSFETs with large safe areas are available from several vendors and the cost keeps dropping.
The MOSFET switch I used in the uTracer3 redesign is about 2USD in single piece buys.

The key issue I see with compliance, it is measured at the start of the test pulse before the test current has settled to the finial value.
This results is false compliance tripping in the utracer design. On the utracer6 this reduces the maximum test current to under 500mA below 100v test voltage with compliance turn on.
Pretty frustrating on a 1 amp test system.

The Ia Is maximum now available in uTracerJS measure the test current at the pulse end where it is the final stable value.
This allows the full system test current to be reliability used at any test voltage and uTracerJS will then shut off the pulses after a over current pulse event.
However the switch needs enough safe area for this to be safely used.
On a uTracer6 that is up to about 200V for the stock switch.
With the uTracer3 there is not really a usable safe test voltage (below 20V) for the stock switch if compliance is turned off.

BTW

 I have the PCB is layed out for my updated uTracer3 switch.

It is small enough to fit between the storage capacitors and plugs into the 4n25 opto coupler socket.

Is intended to allow a safe 600mA (or possible 1Amp) utracer3 even with compliance turned off.

I still have to build and test.

[Martin Manning]

Downloaded the new uTracer JS. Loaded the saved EL84 .uts. Went to the cal tab to turn off the corrections, then returned to the acquire tab... setup was gone.

Test sequence:

•Set Vadc and Rloss to zero

•First curve with DA “off” as a baseline

•Second and third curves with DA “on”, to see if they are reproducible 

  - The dip appears and is reproducible

•Fourth curve with DA “on” and g1 shorted to k 

  - The dip remains and the curve rises a bit due to true Vg1=0

• One curve with normal g1 connection, DA, and Delay set to 1 sec

   - Error out

Log attached.

[Ihor]

Hi Matrtin, 

Thanks for sharing the results, pitty that they do not clarify the issue....

About storing/reset of the the config: it is a designed behavior in fact. So the page (after each refresh of going to another tab and back) shows what config was ran/used last, and then information is taken from the "server" that runs the measurements. Whey you choose the config on the webpage and does not "apply" it by running measurements, it is like a temporary values that you fill in on the page for parameters, you can choose another one to see the values and so on. Only if you press Measure, those values and the config become valid and "last used" and will be displayed after the actual refresh of the page. If those are preserved, as you would imaging or thought before, then one cannot retrieve/see the config which  one just used to measure the tube. So, basically the webpage is somethign that jsut displays the info from the server, all the changes and values are temporary and become stored/fixed only when the Measure button pressed, because only then they are sent tot he "server" and made "actual". So it is jus the login of the program. The easiest way is to open Calibration tab in a separate tab in the browser. 

About the delay, strange to see that uTracer lost the connection, so I am not sure if it was again because of the DA and low voltages or something else. Probably it is a good idea to measure not from 2V but from 5 or 10V, just to see if that's the problems.

Another thing I noticed is that you use the log-scale for Va for pentodes, It should not affect anything but you can try to disable it once as well for the DA/nonDA test and have the points uniformly sampled. You probably do not have to scan the whole range but just the one with the problem.      

I also noticed that the behavior of the curves is a bit similar to this case described by Ronald, but it is a different tube, and the g3 in EL84 is connected to cathode so it cannot be the case.   

https://www.dos4ever.com/uTracer3/uTracer3_pag6.html#ECL80

Now that I am thinking about that Delay, and saw your logs and checked once again what it does (because I implemented it long time ago), I think one should be careful with that "Delay". It was implemented to help with the calibration, to keep the Vg Va and so on at some given values for some (specified) time (to measure the voltages with DMM and calibrate), BUT without a tube. What happens there is for example with a delay of 2 sec, there are Va Vs and Vg applied to the tube and kept for 2 sec, and only after that 1ms measurement pulse is followed and the current is sampled/measured, but in fact it runs there for 2 sec before that. So for some high voltage high current some components might overheat. But this mode works exactly as the Ronald's software.  

  

[Ihor]

I am still confused why Martin gets strange behavior for DA vs nonDA modes, while Bob (for the same tube EL84 / 6P14) gets perfect results. I overlayed two plots that Martin shared and another strange thing is that in DA mode all the measured Va voltages are lower compared to the non-DA case, but they should be higher(!) because the compensation should work. It also means that Vs in DA mode is most likely also lower than it should be, so the curves are "correctly" going down (as we see in the plot). But all that is just an indication of some problems, either due to something with the circuit of the uTracer or the tube (weak, gassy or so on, .. but I thought Martin also tried another type of tubes(?)).

   

[Martin Manning]

Tried a pair of 1k to get in the range of the dip. DA "on" has no dip, DA "off" does. Weird.

[Ihor]

Hi Martin, 

I analysed your log with EL84 once again and I think I see the problem, it is with your power supply. I put in the table to cases, the one without DA and then one with. In the first half you see the Va/Vs set and measured and you also see the power supply voltage for each measurement, which drops substantially(!) to 13V from example. First of all it should not happen, second of all Vsupply is involved in the computation of measured Va and Vs, so that's why you also suddenly get Vs higher than 250V, which is not in fact the measured one but the computed one with reduced Vsupply. So the conclusion here is that your nonDA curved are also wrong (both with utracerJS and the windows software) because power source sinks to much while charging caps I think. 

Now the second part of the table shows DA and only the Va, also set, measured, then set(using the compensation) and measured after that. As you can see from the Vsupply values, then second measurement in DA routine does not sink the power supply so much, (probably becasue the caps are already nicely pre-charged around the same point). But here, the "computational overshoot happens: because the first DA measurement is exactly the same as the one in non-DA (also clearly visible from the table), the computed difference is underestimated (too small), then it is applied to the second measurement within the DA routine and then the measured values of Va are way lower just because they were not properly compensated (they should have been taken with higher "set" vlotages, which were underestimated int he fist step). The Vsupply in the second step is surprisingly better, but still going from 20V to 16V looks like a big gap, the powersupply should perform better.

  

[Martin Manning]

Ihor, thanks again for looking at this. As it turns out (looks sheepish), my bench supply's current limit was set at 500mA. It does not display the current limit, and the only way to see it is to short the output. I must have had it set it there for some other purpose. 

With the supply limiter set at 1A, the EL34 trace looks fine with and without DA, and DA "on" shows slightly higher Va and Ia for the data points, just as it should. The additional cal factors are active here, with Vadc = 0.5 and Rextra = 5.5. 

The digital display on the bench supply hits 600mA in the course of making the EL34 Vg1=0 trace. 

Sorry for the diversion!

[Ihor]

One more comment, I now updated the utracerJS version so it deals with the powersupply issues a bit more consistently, but it is still not the best solution. In the ideal case when the power supply does not sink (it looks like Bob has no issues with that), all the previous results are good. Now, that Vsupp voltage (which is about 19-20V) is used to form the measurements commands for utracer (convert set Va and Va into HEX-codes for utracer) and also to decompose the received HEX codes into measured Va and Vs.  uTracerJS creates all the measurement HEX commands in the beginning using the ideal Vsupp measured once without any load. At the same time, id decomposes them using the measured Vsupp. The small issues was already present with the DA and weak power supplies which Martin's experiments highlighted is that in DA mode the second measurement with the compensated voltage is created on the fly (the message string) so it was using wrong (lower) Vsupp. That's why that second set of colums in the lower part of the table was suspicious. Now I changed the code that every command in the whole process is based on the Vsupp of those ideal 19--20V measured in the beginning. I wonder how the results for EL84 with and without DA would look like. I would assume definitely more consistent. 

Now, that way of measuring is still not perfect. I think in the best case one needs to take DA-kind of measurements in any case, so one to just sample the set Va Vs point but ignore the measured Ia an Is and only measuring the Vsupp at that point. The use that Vsupp to form a new (more accurate) command with the set Va and Vs, then measure the Ia Is and Vsupp, now the third time to create an even better command where the Va and Vs are boosted a bit to compensate for the voltage drop. So, probably the TrippleAcquisition is the way to go :) or otherwise the powersupply should be just fixed :) 

Another option, probably it is interesting to add some monitoring curves that collect Vsupp during the measurements (I can dump them in the debug file or so) in order to have some idea how trustable those acquired curves are :)  

[Ihor]

I also just checked your lates log and your Vsupp is very nicely within 20 and 20.2V, so all my approach described above is a bit overkill :) Also compensating new commands with lower Vsupp is not ideal, because all the calibration values that are involved as well, were obtained for perfect Vsupp, so the final result will also not be optimal. Then we stay where we are:) 

[Martin Manning]

My bench supply can do 3A, so it's only a matter of setting it high enough that it doesn't lose voltage. I don't think the laptop bricks that many people are using will have any issue. 

Here is a 6L6GC for a test with higher current with the same uT JS version (haven't downloaded your update yet). The log shows it holds Vsupp with 1A limit. 1.5A might be better for a 500mA DUT. Note that according to the bench supply display my uT3+ idles at 250mA.

[Ihor]

Nice! Those curves look exactly as they should be. When the powersupply is stable, then actually in the previous version of uTracerJS things are going correctly with DA, because the measurement before returns the current Ia and Is which are used to compute the compensation to Va and Vs and there is also the measured Vsupp which is then used to form the next command for that point on the curve. So the final command uses all the actual/realistic (just probed) values to do a one more acquisition of the same point. 

[Never Mind]

Ihor Said " Also compensating new commands with lower Vsupp is not ideal, because all the calibration values that are involved as well, were obtained for perfect Vsupp, so the final result will also not be optimal. Then we stay where we are:) "

A main purpose of measuring Vsupply is that the cathode sits at Vsupply and so Vsupply needs to be subtracted from any voltage measurement of the anode or screen.

Except the cathode does NOT sit at Vsupply under static or dynamic conditions.

Under static conditions as the cathode is coupled to Vsupply through a diode and then pulled to ground by a 1k bleed resistor. The cathode sits at about 0.65V below Vsupply.

Under dynamic conditions the cathode is coupled to ground by a 1000uF capacitor C44. 

If Vsupply suddenly droops then the cathode tends to only slowly droop resulting in a significant error between Vsupply and the actual cathode voltage

The measurement systems can not detect this error and so a stable power supply is required for best utracer accuracy.

During the measurement current pulse the cathode voltage rises due to the charging of C44.

Again the measurement systems can not detect this error however it can be calculated from the test current and corrected fairly well in the new calibration values.

The delta between the cathode voltage and Vsupply during a measurement pulse I have characterized on a uTracer6 as a significant source of error in the reported Va and Vs numbers at lower voltages.

Ideally I believe the Vsupply pickup point should the cathode connection side of D40. That way you would be measuring the actual cathode voltage during a measurement pulse.

This is not how the uTracer hardware and software was is designed. It would seem to me to be problematic to change the design for many uses that do not wish to hack the hardware and would create a lot of compatibility issues.

V adc and R extra used with a a stable power supply does for the most part eliminate the cathode voltage error issue without hardware or further software changes.

My vote is leave the code as is and accept that for best accuracy you need a voltage stable power supply.

[Never Mind]

BYW

On my uTracers I added a local power supply bypass capacitor of 4,700uF right at the power entry point to further lower the dynamic power supply impedance during measurement pulses.

[Martin Manning]

One more, at low current doing the current amp cal with 2x 10k resistors. Recall I was seeing a left/down shift when using DA, which is now gone. That was still present after I changed out the transistors. Maybe even at the few tens of mA in this trace it was bumping the 500mA supply limit?

Interesting and informative discussion, Bob. The uT continues to be an excellent learning experience. 

[Martin Manning]

Bob Wrote:

On my uTracers I added a local power supply bypass capacitor of 4,700uF right at the power entry point to further lower the dynamic power supply impedance during measurement pulses.

Could just increase the size of C23 (1000u at the input to the 5V Reg) to 4700 or 6800, no?

[Never Mind]

Martin wrote:" Could just increase the size of C23 (1000u at the input to the 5V Reg) to 4700 or 6800, no"

Yes that should be fine also.

I just tacked it onto the input terminals for the 19.5V connection to the external power supply on the back side of the PCB as it was quick, easy and there was space in the case for it to sit.

Martian wrote:"Interesting and informative discussion, Bob. The uT continues to be an excellent learning experience"

The same is true for myself.

I find Ronald's utracer designs exceptionally creative and a interesting study of intelligent design compromises made in the goal of using a few low cost components that will allow achieving good levels of performance.

Yes I can disagree with some of Ronald's design choices and wish for better performance in some key areas with hindsight providing me a unfair advantage in any review of his work.

However as a long experienced designer I find myself continually impressed with Ronald's design creativity, excellence in documentation and generosity to the wider community. 

Inspiring is the word that comes to mind.

[Never Mind]

Martin wrote:" Maybe even at the few tens of mA in this trace it was bumping the 500mA supply limit? "

The boost converters draw a large current from the 19.5V rail in charging the storage capacitors up to each test voltage level before each test pulse.

That makes the 19.5V rail current in charging the storage capacitors somewhat independent of the test pulse current.

A factor in the 19.5 rail current draw is size of each voltage step in the test series as that sets the "step" voltage that the storage capacious must be charged by for each test pulse.

I set my power supply current limit to 1.75A and at that level have not seen any droop. 

The power supply current level setting is a compromise as I have seen my utracer6 enter the "crash and burn" mode were the PIC hangs with both of the large MOSFET that charge the inductors stuck on.

This results in current draw that is pretty well only limited by the power supply resulting in the MOSFETs and inductors becoming very hot in under a minute.

[Martin Manning]

In another thread, Bob wrote:

2) The very low current measuring accuracy of a utracer6 is about 3.8 times lower than that of the utracer3+. So very low current measurements are never as accurate as on a utracer3

This was the main reason I decided not to get a uT6, provided the uT3 could be modified to measure 400 mA or so without losing its ability to measure currents in the 1 mA and below range.

I ran a couple more traces using 1% 100k and 1M resistors to see how close the 529 mA uT3 could come to the expected results for these components. The 100k's look pretty good using the standard cal using 10k's. At very low current, the 1M result is a little high, and the error has a slope. Bringing that in at the high end, and then running the 100k's again shows that you can't have both, the low end looks good, but the high end is low. 

The low end of current measurements are a 12AX7/ECC83, where the typical 250V Va and -2V bias results in 1.2 mA, with the extreme being measuring uA of grid current for the same tube using the Low/+Vg loupe.

[Martin Manning]

A couple of questions around how bw=eat to use the new features added to the uTracer JS interface.

First, calibration.

I believe that the original current amp calibration using 2x 10k resistors is still valid, but now there is a second calibration at low voltage/high current using Vadc and Rextra. My approach to that is to use 2x 100-ohm 1% resistors and use Vadc to bring the I-V relationship in at low current, say 5V-50mA, and use Rextra to bring a high-current point in, per Bob's suggestion something approaching the max current capability of the tracer, say 40V-400mA in my case. The easiest way to observe the results is to use the Ia/Is vs. Vak plot and look for the traces to cross at the correct intersections.

Second, how best to use the new Ia and Is current limits on the form?

My understanding is that these are additional to the existing Compliance limit, and similar in function, except that separate limits for Ia and Is are available, and any particular limit can be specified. I think there may be a difference in the way the exceedance is projected and or prevented, but I can't seem to find the discussion on that.

[Never Mind]

Martin

A document on how V adc and Rextra were designed to be used would be a good idea.

I will see if I can motivate myself to create one in the next few days.

For now here is the short form.

1)

V adc was created to allow correction of two voltage measurement offset errors present in the utracer.

1.1) The first error is caused by actual ADC offset errors in the the anode and screen voltage measurment hardware on the uTracer6.

For some reason there is between 2.5 and 3.1 volts of offset errors in the anode and screen ADC circuits in the uTracer6.

These errors can be observed by running a "ping" command in the debug menu.

  On a utracer the returned ping value for the anode and screen voltage should be zero volts (the measured Anode/screen voltage - the measured power supply voltage).

However on a uTracer6 the ping returned value for the anode and screen voltage is between 2 and 3 volts negative.

This causes reported anode and screen voltages to be reported as low by 2 to 3 volts on a calibrated uTracer6.

The error is not really a big issue above 100 volts as the percentage error is modest and becomes smaller as the measure anode/screen voltage increases.

It does however become a larger and larger error below 50 volts reaching 150% error at the lowest test voltage voltage of 2 volts. Ouch!

For some reason the ADC offset errors on the uTracer3 ADC are very low. On my uTracer3 it is very close to zero volts.

1.2) The second source of ADC offset error is the cathode on a uTracer is NOT connect directly to Vsupply but connected through a diode causing the cathode to sit about 0.65 volts below ground.

This has the effect of causing a positive offset voltage of about 0.65V in all anode and screen ADC measurements as the utracer uses Vsupply as the correction voltage for anode and screen measurements when it should be using the actual cathode voltage.

1.3) The uTracer3 uses a darlinton pair for the switch that has a fixed loss of about 0.65 volts. This loss is not measured by the hardware nor was it corrected by the GUI software.

I expect this was do to the fact that the offset error in #2 and #3 are opposite in direction and nearly cancel each other out. 

However the diode and the switch are not at the same current or matched devices and so the cancellation is not complete on the uTracer3

On the uTracer6 the switch is a MOSFET so the diode in the cathode circuit is not canceled at all. 

For some reason the GUi did not take this into account causing utracer6 low voltage accuracy to be low.

2) Rextra

There are many resistive losses in the anode and screen circuits in both the uTracer3 and uTracer6 that past GUI have not corrected for.

As these are resistive losses the voltage loss and so errors in anode and screen voltage measurements are current dependent.

These resistive losses causes little error above 100V test voltages and at current below 100mA.

As a result for most stock uTracer3+ users they are not a issue unless you measure rectifier diodes.

For uTracer3+ with the high current mod they become a significant error at higher test currents below 100 volts.

For the uTracer6 they become very large errors below 100V and above about 250mA test current.

3)

How to set Vadc offset and Rextra.

3.1)

Start with it Vadc offse and Rextra set to zero volts

Use a precession resistor to load the anode port that produces about the full test current of the utracer system at about 50 volts on the anode port.

So for a utracer6 that is about 50 ohms and for the uTracer3+ it is about 250 ohms. 

However the exact value is not important as long as you measure the know the exact value and using as low as 120 ohms on a stock utracer3+ seemed to also work fine.

3.2) Set to Va=VS with 25 to 30 points and do a voltage sweep stating at 2 volts and sweep up to just enough test voltage to reach the maximum current you utracer can deliver.

Work you way us to the highest voltage to get to the maximum test current carefully as running a utracer where the current limiting circuits cut in is not totally safe due to the limited ability of the switches used in the uTracer design to take overloads.

When you reach the desired maximum current save the data file created.

3.3)

The easy was to get the correct calibration values now is to use the excel sheet I posted.

Sorry it was created for myself originally and so is a "bit" cluttered.

Bold is data you enter and other non-bold fields are calculated.

Take the data file just created and paste it into the data area on the left side of the excel sheet.

Then add your test resistance value to the sheet in the box at the upper left under "Rtest".

3.4)

The sheet will then calculate the measured resistance and plot the errors.

At this point you will see how accurate at low voltage and high currents your utracer is with no additional correction. 

If you are a stock utracer3+ user you may be happy with the reported errors and simply move on.

3.5)

Continue below to find the cal values.

Adjust Rextra first to bring the resistance down to the test value you used and produce a lowest overall error.

Concentrate on the right side of the  graph at the higher voltages.

Second tweak Vadc error value to lower the errors below about 20V and to try to straighten as much as possible the resistance error curve on the life side of the graph.

Some back and forth is required

For example 

On my stock uTracer3 the Vadcerror needed was only -0.1 volts and the Rextra was about 5 ohms

On  my stock uTracer 6 the Vadcerror needed was 3 volts and the Rextra was about 9.5 ohms

On  my stock uTracer 6 with the switch update the Vadcerror needed was 3 volts and the Rextra was about 3.2 ohms.

3.6

Finally take the values you found in the excel sheet and enter then into uTracerJS's cal section and you are good to go.

Hope this is helpful

[Ihor]

Indeed, it is a good idea to have a concise information about the purpose and action of those two extra parameters, which I could also link to in uTracerJS. I can also format that text as a webpage and include it as the Help, or part of the calibration procedure. 

About the limits set by compliance, Ia max or Is max, the discussion is also not in one place. Basically uTracerJS stops acquiring the current curve and stitches to the next one when ANY of 3 criteria are violated (actually there are two more, when the powers Pa max and Ps max exceeded). The error about the compliance is returned by the device, so it is easy to abort the measurement. In all the other cases, uTracerJS monitors Ia, Is, ia * Va and Is * Vs, and if those are about the specified thresholds, it continues with the next curve. One One should be careful with the size of the the steps, for example in Va. For large steps (so a couple of points on the curve) the jump in Ia might be quite substantial, from one measurement to the next one (for example 5ma, 10ma and then 100ma), and utracerJS does have to make that higher measurement in order to stop. With small steps there is no danger as utracerJS will be slowly "approaching" that threshold level.