What we need for testing HV tube supplies

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
The issue came up in the discussion of the LT3080 as a high voltage floating regulator.

There are four basic tests that you have to perform on all power supplies. Power supply rejection ratio (PSRR), noise, impedance and stability. (From impedance you can examine stability.)

Easily done with the tools at hand for opamps and ss power amplifiers. It gets really tricky to do so inexpensively for high voltage power supplies.

Take PSRR -- you could use a transformer to modulate the voltage input of a high voltage supply -- if you can find one with bandwidth of 10 Hz to 10 MHz that can sustain 100mA or more. Cost $300 and up.

Build a modulated DC power supply -- like the Kepco BOP -- the bandwidth of the Kepco is only 20kHz. Maybe a VDMOS amplifier using ham radio techniques? (The beloved F5 amplifier I built has bandwidth to around 1MHz with IRF devices) There's are several designs in ARRL handbooks which could be modified.

Well, I am open to ideas.
 
I don't think anything is insurmountable or even costly: take the PSRR, if you feed a floating supply from the ground side using a wideband buffer or amplifier, you can easily arrive at decent results provided the capacitance of the DC supply wrt. ambient is reasonable.
Here is an outline, the LT1210 only has a 66MHz GBW, but if you accept lower output currents (or if you shop around), wider bandwidths are possible
 

Attachments

  • Jack.png
    Jack.png
    5.3 KB · Views: 408
Here is the way I would do it. Modern VNA's have a usable range >100dB, meaning the setup will be able to resolve ~0.5mΩ.
If that's not sufficient, a wideband buffer followed by a 5Ω resistor could be inserted in port 1. This would better the floor by 20dB
 

Attachments

  • Zmeasure.png
    Zmeasure.png
    11.3 KB · Views: 359
In practise, does PSRR measurement need to extend to 10MHz? The LT3080 datasheet extends to 1Mhz, and shows a consistent loss of PSRR above 100kHz.

Is that extent based on rectification diode turn-off resonance parasitic voltage levels that could be present at the input to such a HV DC regulator? Or is the regulator being driven by a mains switchmode supply that has a HV DC output, and the concern is that the switchmode output has content that extends to 10MHz?
 
Last edited:
PSRR to1MHz would be satisfactory. For measuring Zout of the power supply, perhaps (I think) a decade more as the gremlins seem to manifest themselves any from 10s of kHz to 10MHz.

If you had a sound-card analyzer, you would not be able to see the Jung-Didden super-regulator in oscillation mode with the AD797 error amplifier. With the remarkable AD797 you can get the Zout to under 10 micro-Ohms, but... With the AD797 I was quite able to see New York's AM station WABC on the HP3577 spec analyzer.

As suggested by one of the European engineers from ADI, a simple a.m. radio can be your best sensor for power supply instability.
 
Here is the way I would do it. Modern VNA's have a usable range >100dB, meaning the setup will be able to resolve ~0.5mΩ.
If that's not sufficient, a wideband buffer followed by a 5Ω resistor could be inserted in port 1. This would better the floor by 20dB

Looking at Semtech and Littlefuse TVS on Digikey's site. (I've used small incandescent bulbs as current suppressors. They do have DC resistance which contributes to noise.)

TVS have pretty significant capacitance, 100's to 1,000's of pF's, or not in this application?
 
Please tell me more about instability.

This is what I am thinking about stability.

The power supplies that we are speaking of will be used to power tube amplifiers.

What is the issue with connecting the known load, the amplifier, to the power supply under test and using a function generator to sweep, pulse or whatever at the input of the amplifier?

While sweeping pulsing or whatever at the amplifier input connect the test probes to the HV output of the regulator under test. Trigger the bode plot from the function generator.

If the power supply + audio amplifier assembly is stable under test conditions it will also be stable functioning as an audio amplifier.

DT
 
TVS have pretty significant capacitance, 100's to 1,000's of pF's, or not in this application?
The ones intended for the protection of high speed signal lines have a much lower capacitance, typically <1pF, sometimes much smaller.

Some examples:
Low capacitance ESD protection for high-speed interfaces | Nexperia

https://www.google.com/url?sa=t&rct...ort_Form.pdf&usg=AOvVaw3_eCXTsyqAGuGWEU5Qp6ZA

https://www.infineon.com/dgdl/Infin...N.pdf?fileId=db3a30433e9d5d11013e9d6619a20002
 
Please tell me more about instability.

The power supplies that we are speaking of will be used to power tube amplifiers.

What is the issue with connecting the known load, the amplifier, to the power supply under test and using a function generator to sweep, pulse or whatever at the input of the amplifier?

You don't generally regulate the output stage of a tube amplifier although you can regulate the screens if not running ultra-linear.

It's recommended to test for stability with low perturbation current. Omicron-Lab's uses the Pico-Test current injector which provides 25mA -- pretty consistent with what you would expect from the input stage and phase-splitter. From the standpoint of consistency and repeatibility you would want the same kind of standards that a journal would require in the physical sciences.

I think that many of the regulators for tube circuits don't meaningfully contribute to performance, or may hurt...
 
I have the impression that the potential number of regulators used for tube output stages may be under estimated.

Or perhaps I am the exception using regulators as the rule rather than the exception. My complaint about tube amplifiers is the noise, particularly line frequency/harmonics and IM products. Even using a C-L-C pre-filter and LM350 Maita type regulator does a much improved job over a C-L-C filter alone.

In the recent couple of years I have been using audio analyzers to see the influence a voltage regulator has.

There is one brand of regulator kit that does a particularly bad job. This is a linear regulator that uses an IXCP10M45 current source. This regulator creates a huge broad noise hump that centers near 3K Hz. The hump does not flatten out until there is better than 50VDC delta across the regulator.

To see instability and distortion past the traditional 20K Hz BW the recent crop of audio analyzers has much broader BW. The Keysight U8903B has a 1.5M BW option. The Audio Precision APx555 comes standard with 1M BW.

I agree that the 25 ish ma injectors will have limited utility for testing other than low current HV regulators. Perhaps we could repurpose a single end output transformer.

DT
 
There is one brand of regulator kit that does a particularly bad job. This is a linear regulator that uses an IXCP10M45 current source. This regulator creates a huge broad noise hump that centers near 3K Hz. The hump does not flatten out until there is better than 50VDC delta across the regulator.
DT

Sorry, I'm not seeing it -- this is Pete Millett's "Engineers Amplifier" which uses the IXCP10M45.
 

Attachments

  • EngineersAmplifier.png
    EngineersAmplifier.png
    35.6 KB · Views: 206
My hat is tipped to Pete and the IXCP10M45.

The Engineer’s Amplifier uses the IXCP10M45 in the tail of the phase splitter.

Jackinnj, the FFT noise plot of the Engineer’s Amplifier shows the entire family of power line frequency and harmonics.

Memory failed me. The HV regulator that I mentioned having the noise hump centered at 3K was centered at 1,300hZ. The regulator name is PS-1, the performance is much improved with 50+ volts across the regulator.

Jackinnj, please share with us the stability test procedure.

Do you plot phase or impedance to determine if instability is likely?
Do see instability on a wide band FFT?
Do see instability on a Bode Plot?
Do you do something to trigger the instability?
Or does it just happen?

DT
 

Attachments

  • PS! 200VDC input.png
    PS! 200VDC input.png
    40.9 KB · Views: 209
  • PS1 250VDC input.png
    PS1 250VDC input.png
    41.9 KB · Views: 180
Jackinnj, please share with us the stability test procedure.
https://www.omicron-lab.com/fileadm...ote_Traditional_NoninvasiveStability_V3.0.pdf

You don't need the Omicron instrument to perform the measurement, but it
is a breeze to operate. Red Pitaya anyone?

The non-invasive method described in the application note uses measurement of phase-angle to calculate Q and this yields phase margin. The non-invasive method also demands a wideband modulated current source. Interesting Red Pitaya calculates "Q" on the fly.

Try out the non-invasive method with LTSpice and Linear Tech's examples of reference regulator circuits. I think that it can be quickly adopted for higher voltage circuits.
 
Power Distribution Network analyses

jacjinnj,

Thank you.

Steve Sandler and his affiliated businesses are all at the center of this Power Distribution Network analyses industry. To look at the many application notes and videos someone might think this is about independent tests and test equipment. If you look past the spotlight on the merits of a specific state of the art test tool we can see the importance PDN analysis. If there is one most common culprit, it is instability. Not so much a single regulator being at fault but multiple items in the network not being compatible with each other. A voltage regulator may have a little funky PSRR and the attached load may have an equally funky reverse transfer. In terms of stability the PSRR and Reverse Transfer may not play well together. Then again a bypass capacitor ESR may just need a little tuning.

The problem with applying this Power Distribution Network analyses to HV is that most all the tools are built for increasingly low voltage and high efficiency devices. These tools are for cruise missiles and cell phones.

I thought about spending some of the toy budget. I do think that an active differential probe may be a good addition to the audio analyzer and oscilloscope. The oscilloscope does do Keysight’s version of Bode plots.

For the time being I will use the connected amplifier as the load and the amplifier input as the injection point.

DT
 
The problem with applying this Power Distribution Network analyses to HV is that most all the tools are built for increasingly low voltage and high efficiency devices. These tools are for cruise missiles and cell phones.

DT

Means that the demands of test equipment today may not be up to the standards of 5G! Keeps Keysight and Tektronix (through Fortive) in business.

I have a "sacrificial" low noise amplifier I use ahead of the HP3577a. Uses a small incandescent bulb (about 20 ohms) as a current limiter with clamp diodes. Like the Tektronix differential amplifier, the input cap is "charged" and shorted to ground to drain the DC. I haven't done any HV stuff with the Bode-100.

3577a only goes down to 5 Hz -- this might not be low enough to look at power supply issues which trigger motor-boating.
 
Not inexpensive

Just back from playing in the cold Nevada desert.

I took a look at the Bode 100 stuff and found that the Picotest injection transformers are designed and built to 600V / CATII specifications. The Picotest injection transformers are 600V / CATII certified. In their (Picotest) literature they discuss using the Bode 100 and the injection transformers to test 400 plus volt circuits. This is right up the alley of injecting voltage into the control loop of a regulated vacuum tube power supply.

Add a couple of differential probes and we are in the business of using the Bode 100 to test 400volt plus DUT’s. It makes me feel somewhat better seeing the manufacturer making such recommendations.

Also I like the software that is bundled with the analyzer.

https://www.picotest.com/downloads/DIFFPROBE/080305_Data_Sheet_Diff_Probe_PMA_ext.pdf

Not inexpensive; $525 for the injection transformer and $500 times 2 for the differential probes. These are just accessories for the $5600 Bode 100 analyzer. Something to put in the next equipment budget?

DT
 
DT: Dispensing with the injection transformer -- it's only useful if you can break the feedback loop for the error amplifier since it is good for just a few mA.

If you look at the two-MOSFET HV power supply in "Art of Electronics" == the first device can be modulated with an A.C. source. Thus you can supply a current to the output of the HV supply and sweep impedance -- Phase angle, by the Erickson-Maksimovic approximation is 50.36 * Q^-0.907 -- accurate to within 5%
 
Hello,

The 3rd edition of "The Art of Electronics" is open here on my lap. Looking at Figure 9.110, this must be the high voltage regulator you are speaking of. The source input voltage to the inverting input of the error amplifier op-amp (LT1236-5) is 5 volts. Inject a little AC at the input and we are in the DC + AC test business.

This power supply / amplifier is looking a lot like a Do-It-Yourself version of a Kepco BOP kind of thing. If picotest made it, it would come with a shiny logo and a certified/tested data sheet. . I like it.

This looks a lot like the voltage controlled resistor kind of thing mentioned in a previous post. The feedback at the error op-amp non-inverting input gives a good head start towards a good PSRR. (This is negative feedback as the MOSFET is inverting.)

The 20ma max output at 330 ohms may be a little problematic.

DT
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.