Tube Tester/Matcher Talk

[R.G.]

All the discussion on selecting tubes diverted my streams of designs back over to the idea of testing tubes. These ideas are not fully fleshed out circuits and very far from a buildit kit, but they're the necessary first considerations.

Conceptually a tester/matcher is simple. You apply a plate voltage, a screen voltage with pentodes, and a grid voltage. You set the grid (and perhaps screen) voltage to a desired value, and measure the plate (and perhaps screen) current. Record the voltages and currents. Change the grid and perhaps screen voltages to different values and repeat. When you've exhausted the state space (or perhaps your patience), shut it down, replace the tube, and repeat.

All of the necessary DC values for the tube can be derived from this data. The only problem is that doing this with manual setting of everything is a grand PITA. Curve testers were one way to cope with this. They have a CRT display a large part of the possible data automatically, but older analog CTs make writing down the data another PITA. I gave a tube-based and a solid state curve tracer away for this reason - I hated trying to derive any useful data from them.

Computers changed this. If you can have the computer set the operating conditions, measure the results, and record them, all while you're having a nice cuppa tea, your nether regions experience far less pain. There are devices on the market IIRC that do just this; I dimly recall one DIY sort of thing.

Measuring and recording voltages and currents with a computer is pretty trivial; it's ASMOP (A Simple Matter Of Programming), and it's always good weather for programming.

Aside from whatever special precision limits you want to put on your data gathering, the only difficulties with doing automated testing/matching on tubes is the design of the circuits to provide the right voltages to the tubes under test (TUTS... ) and read the currents. There are ADC and DAC modules available for ~ US$2 each on ebay to read voltages and currents by computers. The difficulties are almost entirely in the circuits to provide the voltages to the tubes.

A tester needs DC supplies. The simplest way to do this is to cannibalize an old tube amp for its high voltage power supply. This will give you a source for the applied voltage and heater supplies. (there's an asterisk implied here; anyone see what it is?) From there, what you need to make up is a way to have a computer control the plate (and if needed, screen) voltage to the TUT, step the grid voltage in useful steps under program control, and read the currents.

The plate voltage (and possibly screen) is the first nut to crack, and what I've been thinking about. Ideally, you could use a computer DAC module to set the plate voltage in steps, and then tinker with grid and screen. So the first hurdle is converting a low voltage DAC (or other digital output) to a specified plate voltage. And that's what I've been looking at. Ideally, one could just buy a 500V output range opamp and drive it with a DAC, but that seems to be missing from ebay. So I'm off looking at MOSFET DC regulators and how to make them programmable.

[pompeiisneaks]

We discussed this a long while back and you sent me a board I might be able to use to do this. At that time I didn't have have the experience with SMD but since have bought a microscope and played around and even fixed a few smd devices. SO... I'm ready to try again if you can come up with the schematic and I can use that board.

~Phil

[Phil_S]

R.G. I don't begin to measure up to your level of knowledge and experience. I'm an accountant, not an engineer. My dabbling is hobby level, so please give me some latitude here and please be gentle!

It seems to me that, on the one hand you argue that the matter is fairly easily simplified by using an old amp. On the other hand, you seem to suggest anyone who can do the programming can employ a computer to do the dirty work. It is easy to agree with all of this. Of course, I don't have the programming chops or the equipment to do the latter. And please forgive me for oversimplifying.

Not so long ago, I had some bone pile parts, an itch to do something, and managed to built a crude rig to test tubes. I happen to have a modest pile of 6AQ5's so I built it around that with a 7 pin socket and Va about 250V, figuring there is nothing inherently wrong with building a one trick pony. Then I added an octal socket in parallel, figuring 6AQ5 is the same as 6V6. I was trying, in my own way, to figure if I had any matched pairs, even if crudely matched, and that I might get lucky with a matched quad (no quads, over 50 tubes.) I used a pot in the bias circuit that allowed me to get readings on each tube at 3 different bias voltages. (Easy-peasy, at max, mid, min on the pot sweep.) Of course, I couldn't hold screen or plate voltage constant, but it didn't vary all that much. Nevertheless, I was able to get three sets of readings for plate, screen, and cathode, with 3 known bias voltages. From this I figured, it's not so hard to compute the plate current and the transconductance at the three data points and attempt to understand the operating parameters of each tube. In the end, I wasn't sure I'd accomplished anything, but I had some fun along the way.

What, if anything, is fundamentally right or wrong with this approach? If this really does work, it sure is the cheapest tube tester one can make. If you are willing to comment, I'd be very interested and I think it's a fair game topic for the forum that isn't so far from your opening post.

[R.G.]

We discussed this a long while back and you sent me a board I might be able to use to do this. At that time I didn't have have the experience with SMD but since have bought a microscope and played around and even fixed a few smd devices. SO... I'm ready to try again if you can come up with the schematic and I can use that board.

I'll go take a look and see if I can find it.

It seems to me that, on the one hand you argue that the matter is fairly easily simplified by using an old amp. On the other hand, you seem to suggest anyone who can do the programming can employ a computer to do the dirty work. It is easy to agree with all of this. Of course, I don't have the programming chops or the equipment to do the latter. And please forgive me for oversimplifying.

Not so long ago, I had some bone pile parts, an itch to do something, and managed to built a crude rig to test tubes. [...]

What, if anything, is fundamentally right or wrong with this approach? If this really does work, it sure is the cheapest tube tester one can make. If you are willing to comment, I'd be very interested and I think it's a fair game topic for the forum that isn't so far from your opening post.

There is nothing at all wrong with that. My advancing laziness just makes me look for the solutions that let the machines do the work for me. Well, OK, the only thing wrong is that it is limited in the number of tubes that can be tested, and my mental path is on testing, characterizing and matching/documenting right now.

I can see I was not clear. Using an old amp gives you the heavy mechanical stuff - power supply, chassis, and filament supply. Programming gives you a way to automate the process by having a computer take the data for you and arrange it into a usable form. What goes in the middle is a way for the computer to set the voltages and read the resulting currents and such, replacing what you did manually on your simple tester.

What I had in mind was a way to build a few things to bridge the gap between the heavy mechanical stuff and the computer. I think this amounts to:
> a plate supply that can be set to a specific voltage by computer action
> a screen supply (if needed) that can be set by a computer
> a grid voltage settable by a computer
> a plate or cathode (and maybe screen) current reader that a computer can read.
These are the bridges for a very simple tube characterizer. There are other fancifications, like directly measuring AC gain and noise that could be added, but I'm most focused on the matching aspect at the moment.

Most particularly, I've been looking at a computer driven plate supply. That is conceptually easy, but is turning out to have some complications. I'll post more when I get a handle on that.

[pompeiisneaks]

Could you have something like several transistors that source multiple different zeners for differing drops?

~Phil

[pompeiisneaks]

I was not very clear.... something like arduino/raspberry pi or through darlington array that chooses where to drop the ground? or a computer of some kind.

~Phil

[R.G.]

A raspberry pi would be great. My approach would be to give the pi a d to a converter output with a range of, say 0 to 3Vdc, and run that into a DC amplifier to get 0 to perhaps 400V, amplifying the DAC output up by a factor of 15. This would be the plate (and screen) supply. Another DAC would set the grid voltage for 0 to -60 or so.

The point in this is to set up the input and output devices so the pi could set the tube supply and input to literally any reasonable value and make the readings automagically without having to do manual settings. It also offers many small steps on the excitation voltages, so you don't need to cope with the several-zeners or switching things in and out.

I'm not explaining this well. I'll do a couple of drawings tomorrow.

I just spend a couple of hours on the circuit simulator trying to make a DC amplifier with a gain of 15 and an output range of 0-400v, without much luck. Classical voltage regulators handle the voltage and currents easily, but mostly don't have much range. Opamp circuits easily have the dynamic range on the outputs, but don't get to the high voltages. Opamp boosters (so far) haven't been stable because of the added parts needed for the high voltages. I still haven't found the right mix. In my mind, a reasonably accurate DC amplifier with a high voltage output is the crux of the problem.

[roberto]

Hi,

if you are not in a hurry, I can make the software for Arduino, or at least contribute to that part.
For the DAC, we can have a VVR for each connection of the tube, and target the voltage via a PID on the analog read.
That means that every connection of the tube to test will have a VVR to set the voltage, a voltage reading (scaled down to 0-5V) to set it, plus a current reading to plot the curves.

This way we can have grid, screen, anode and cathode current reading.

What would be great then, is having someone who can implement part of the software to chomp the data automatically and give out the LTSpice model of the tube.


Post Scriptum
R.G. we are waiting your next palindromic post


Post Post Scriptum:
http://www.valvewizard.co.uk/analyser.html

[dorrisant]

I'm pulling up a chair...

[pompeiisneaks]

Hi,

if you are not in a hurry, I can make the software for Arduino, or at least contribute to that part.
For the DAC, we can have a VVR for each connection of the tube, and target the voltage via a PID on the analog read.
That means that every connection of the tube to test will have a VVR to set the voltage, a voltage reading (scaled down to 0-5V) to set it, plus a current reading to plot the curves.

This way we can have grid, screen, anode and cathode current reading.

What would be great then, is having someone who can implement part of the software to chomp the data automatically and give out the LTSpice model of the tube.


Post Scriptum
R.G. we are waiting your next palindromic post


Post Post Scriptum:
http://www.valvewizard.co.uk/analyser.html

I can usually do either arduino coding or python for the raspberry pi. I wrote my own custom software in python for the raspberry pi that controls my brewhouse in my home brewery... Uses a pid controller library to do a ton of the algorithmic work, I just had to tune the PID.

~Phil

[roberto]

I don't do Python, just Arduino.
I did some tests on cheap LCR to find its characteristics, and that structure could be used for this equipment as well.

I copy it here, in case of need, and I attach one photo of the setup I used:

Code: Select all

#include <Wire.h>

//to manage the LCR-0202s
#include <Adafruit_PWMServoDriver.h>
Adafruit_PWMServoDriver pwm = Adafruit_PWMServoDriver(0x40);

//just to show results
#include <LiquidCrystal_I2C.h>
LiquidCrystal_I2C lcd(0x27, 20, 4);

//analog readings in voltage dividers
int a_read[] = {0, 0, 0, 0};

//LCR-0202 initial resistance values
float LCR_ohm[] = {0, 0, 0, 0};

//define "top" resistors in voltage dividers
unsigned long topR_ohm[] = {100000, 100000, 1000, 1000};

void setup() {
  //PWM startup
  pwm.begin();
  pwm.setPWMFreq(1526); //max pwm frequency to avoi wavering of led

  //LCD startup
  lcd.init();
  lcd.backlight();
  lcd.clear();

  // Serial startup
  Serial.begin(9600);
}

void loop() {
    for (int i = 0; i < 4096; i = i + 4) {
      for (int led_num = 0; led_num < 5; led_num++) {
        pwm.setPWM(led_num, 0, i);
      }
      read_values();
      show_values(i);
    }
    }

void read_values() {
  delay(200);
 for (int gpio = 0; gpio < 4; gpio++) {
    LCR_ohm[gpio] = 0;
    for (int num_read = 0; num_read < 1024; num_read++) {
      a_read[gpio] = analogRead(gpio);
       LCR_ohm[gpio] = LCR_ohm[gpio] + ((a_read[gpio] * topR_ohm[gpio]) / (1023.00 - a_read[gpio]));
    }
    LCR_ohm[gpio] = LCR_ohm[gpio] / 1024;
  }
}

void show_values(int i) {  
  lcd.clear();
  lcd.print("LCR0202 Test");

  lcd.setCursor(13, 0);
  lcd.print("PWM");
  lcd.print(i);
  Serial.print(i);
  Serial.print(", ");

  lcd.setCursor(0, 1);
  lcd.print(a_read[0]);
  Serial.print(a_read[0]);
  Serial.print(", ");

  lcd.setCursor(5, 1);
  lcd.print(a_read[1]);
  Serial.print(a_read[1]);
  Serial.print(", ");

  lcd.setCursor(10, 1);
  lcd.print(a_read[2]);
  Serial.print(a_read[2]);
  Serial.print(", ");

  lcd.setCursor(15, 1);
  lcd.print(a_read[3]);
  Serial.println(a_read[3]);

  lcd.setCursor(0, 2);
  lcd.print(LCR_ohm[0]);
  Serial.print(LCR_ohm[0]);
  Serial.print(", ");

  lcd.setCursor(10, 2);
  lcd.print(LCR_ohm[1]);
  Serial.print(LCR_ohm[1]);
  Serial.print(", ");

  lcd.setCursor(0, 3);
  lcd.print(LCR_ohm[2]);
  Serial.print(LCR_ohm[2]);
  Serial.print(", ");

  lcd.setCursor(10, 3);
  lcd.print(LCR_ohm[3]);
  Serial.println(LCR_ohm[3]);
}

[R.G.]

I got some time to play with the trains (i.e. the circuit simulator) today and hacked out a high voltage amplifier.

The design requirement is to take in zero to 3V, and have this drive the output to 0-400V. This lets the computer use a DAC to output 0 to 3.3V out of a D to A Converter (DAC) and then have this amplified up to ~400V. The usually resolution is one part in either 1024 or 4096, those being the 10 bit and 12 bit versions of common DACs. So you can set plate or screen voltage by computer to the nearest 0.4V or 0.1V out of 400V. The computer can output "255" and have the 10 bit amp put out 102V.

I'll do a bit more to get an inverting one for grid voltages in the next few days. Same deal, computer says "0512" and the output voltage goes to -25, or whatever.

With those available, the instrumentation can begin.

One can buy a 4-channel, 16bit A to D Converter (ADC) on ebay for about $2. It takes some tinkering with resistor dividers to scale down plate, screen, and grid voltages for measurement. Measuring the voltages you set means that you can calculate out any impreciseness of the DAC-> amplifier chain. These modules speak I2C and are a natural fit for Arduino, Raspberry Pi, and PICs if you like those.

Current measurement is fairly easy. There are I2C current modules based on Hall effect with I2C output available as well for a few bucks each, again for the Arduino/Pi market.

With this set of stuff in hand, you can set voltages by programming, and read the results by computer.

There are a few extensions that might be useful for tube characterization. A relay (solid state or otherwise) can switch an signal into the grid (although dithering the grid voltage by a few bits might be just as good) and you can easily construct a plate signal converter to tease out the AC signal caused by the grid signal and convert to AC gain, readable with the same ADC.

On making measurements:
I had in mind programming the machine to mostly work the tubes in cutoff at very low plate voltage. To make a measurement, the plate would be stepped to the desired measurement value (and screen, if desired), let settle for a few milliseconds, then the grid stepped to the desired point, let settle for a few millliseconds, then the readings made; then put the grid and plate back at the resting values. This sampling approach will probably let you handle power tubes without excessive heat generated in the tube or the amplifiers feeding them. This is an old variant of how tube testers did semiconductor and power device tests without overloading the tester. I'm guessing that a reading might take 10-20mS per reading. It's a little bit of messing with programming, but it saves a lot of excess heat and worrying how to get rid of it. The tubes are at least ultrasonic, even RF capable, so they will happily slew to a new position, then go back to resting.

Once the mess works, the ASMOP begins.

[R.G.]

Got a better high voltage amplifier designed.

Had to go back to audio power amplifier design style. It's basically an audio power amp front end driving an asymmetrical output stage. It uses the now-standard diffamp/voltage gain/current source voltage gains stage. This makes the DC accuracy less than you could get with opamps, but with far fewer issues with stability. The front end up to the VA transistor is powered by +/- 70Vdc. I was trying to make the plate and screen supplies work with a single power supply (0-450Vdc) but I realized that the power supply has to generate about -70V for a grid supply anyway, so I just used the minus supply to make the whole plate thing easier. The VA transistor is loaded by a current source floated up at +450Vdc, so the output is asymmetrical, minus 60(or so) up to +400V at clipping. Not good for audio, but plenty good for setting the output to +300 or +400 with a DAC. The plate and screen amplifiers don't need to go negative anyway. The plate and screen amps will run with a closed loop gain of about 150, so a 0-3V DAC signal can drive them to 400V and over. There's not a lot of bandwidth left over for low AC distortion, but then that's not needed. It's just a DC amplifier and we can wait till it slews to the right DC point.

The grid amplifier does need to go negative, so it can run entirely from +/-70. Same front end ought to work, but the output only needs to swing from 0 to -70ish, so the high voltage side of things is much simplified.

Mouser has a 36VA transformer for $19 that generates 120Vct at 300ma, should be good for the "low voltage" part of the drive amplifiers.

There's a parts threshold at about 500Vdc for plates and screens. That's the highest voltage rating for a P-channel power FET to do the current source load for the plate and screen amplifier VA transistors that's easily available. Above 500V and you need to start using multiple devices and voltage sharing - as well as wasting a large amount of power in the CCS for the VA transistor.

More later, but this gets the design into the raw feasibility range. It's a truism that power amplifiers are primarily a fancy DC power supply with some other insignificant stuff tacked on to make it do the actual job. A power amp design is properly done by designing the power supply and power amplifier concurrently to avoid running into not being able to buy the power supply parts at a decent price once the fancy, high-spec power amplifier is done.

[R.G.]

Here are some other bits and pieces to go into the thing.

> digital to analog converter, 12 bit resolution, I2C input interface, based on MCP4725; $1 to $3 each on ebay and amazon
https://www.ebay.com/itm/1PCS-MCP4725-I ... Sw-HRfUPKd

> 4 channel A to D converter, 16 bit resolution, I2C interface board, based on TI ADS1115; $1 to $3 each on ebay and amazon
https://www.ebay.com/itm/ADS1115-4-CHAN ... Swk5Fd-fHb

> power transformer for high voltage amplifiers; 120v to 120Vct, 36VA; Mouser 530-ST-7-120; Bel Signal ST-7-120, $18.35.

Notice that the fancy electronics are cheap, but the power parts are expensive.

[martin manning]

RG, have you examined previous designs such as: S Bench's rec.audio.tubes (RAT) tube tester (the grandaddy DIY all manual Ia, Gm tester), R Dekker's u-Tracer, and C Chang's e-Tracer (microprocessor controlled mS pulse tracers)? Dekker's site has a ton of info on the development of the u-Tracer, and I've "blogged" extensively about my experience with one in this TAG thread: https://ampgarage.com/forum/viewtopic.php?f=6&t=21903 As posted earlier by Phil, Blencowe's site has something very close to where you are headed: http://www.valvewizard.co.uk/analyser.html