I use two transformers, one for HV where i use a bridge rectifier to make the Plate drive signal. I also use a small 6.3 volt transformer that i use for the grid step generator, I pass the output thru the phase adjust and then to the schmitt trigger. I dont use a rectifier for the step generator. If you use your transformer just for the step generator then you can bypass the bridge rectifier and you should then get the 200hz without modifying the divider on the PLL.
An externally hosted image should be here but it was not working when we last tested it.
Feeding the AC directly to the LM311 will require more hacking on the PCB than changing the PLL frequency division from 4 to 2 only one connection to change from O4 to O2 I guess.
Regarding the RC network will add it to play around with the phase adjusting.
Oh man, this is a project that is coming slowly and demanded a lot of effort so far
Can't see the end of this....
Cheers,
Ale
Regarding the RC network will add it to play around with the phase adjusting.
Oh man, this is a project that is coming slowly and demanded a lot of effort so far
Can't see the end of this....
Cheers,
Ale
...
Oh man, this is a project that is coming slowly and demanded a lot of effort so far
Can't see the end of this....
You are doing it the hard way. My curve tracer is using (or will use) software to drive a set of DACs to make the wave forms. The I use ADC to measure. The best part is that today's $5 microprocessors have several ADC and DACs built-in. They typically have only 10-bit resolution but that is more than enough. With the uP the parts count is way down. So with a uP the wave forms are nearly trivial to produce, I still need high voltage supplies and that is where the work goes.
But the down side is that I will not get to see the result on my scope, it will get pushed unto a computer and plotted there.
But there is room for innovation now. How best to display the curves on a computer screen. One now can use 3d color graphics and even animation whereas the old tube data books only had black ink. I'm sure this is where most of the work will need to go.
Chris,
Which microprocessor do you plan on using? I wanted to take a similar approach but need to find a processor that can do a bunch of measurements synchronously(or asynchronously really fast). Otherwise there will be an ADC board with a CPLD or something to make measurements synchronous. I wanted to measure currents from all electrodes and produce some really nice curve data and just plot in Excel.
Which microprocessor do you plan on using? I wanted to take a similar approach but need to find a processor that can do a bunch of measurements synchronously(or asynchronously really fast). Otherwise there will be an ADC board with a CPLD or something to make measurements synchronous. I wanted to measure currents from all electrodes and produce some really nice curve data and just plot in Excel.
The key is that because you are NOT driving a real time display like the analog scope, you can go as slow as you like. With the uP you do not need to retrace a curve at 60Hz, 1Hz is fine. You don't need the "really fast" part. Put yes you do need "synchronous". technically there is no reason you can't set a voltage on say the grid, then wait 10 minutes to measure the current, as long as the voltage you set does not change. I will not wait 10 minutes but I do consider all the voltages to be "DC"
You can set the voltages, wait for them to "settle" and measure each current one at a time, store to a file, then step to the next voltage. If the whole process takes 10 seconds no big deal.. I can see doing 6,000 data points per tube and doing about 100 per second, could take one minute. I don't know the real rates yet.
A scope needs to be re-freshed before the phosphor fades but computer memory does not need the re-fresh.
There are cheaper uPs but I like the Arduino. Mostly because it is easy and because it will be easy for others to also use and program it. If it is easy then maybe people will do it. Main criteria is if people new to uP can use it. Arduino sees to be dead-easy. Even if it does add $30 to the project.
Current state is just the very beginning with breadboard experiments to prove some techniques. one for example is how to pass analog data through an opto-coupler and how linear is the result. First try failed. Next try showed it could be done. Next will use paorts I have on order. Lots of things like this to check out..
Last edited:
One potential problem I can see is that grids tend to draw a lot of current at low plate voltages, control grid if you are doing positive curves, screen grid no matter what. If you go slowly this could be very hard on the grids.
With a fast sweep you can draw the curves out past the tubes max dissipation and just give the tube rest periods to keep the average down. As you slow down the sweep, you may be limited in that regard.
With a fast sweep you can draw the curves out past the tubes max dissipation and just give the tube rest periods to keep the average down. As you slow down the sweep, you may be limited in that regard.
I view the microprocessor option as a tool to sweep the traces rapidly so you minimize power dissipation. Most tubes have frequency response above a MHz so there is no reason (that I see) not to sweep as fast as you can within the limitations of your sweep circuit.
I've got dozens of 89C51s so Ive used them in the past, but have an Arduino Uno on order to play with.
Also, with a microprocessor you can perform the sweeps in a non-sequential order. Lowest dissipation, then highest dissipation, next lowest dissipation, next highest dissipation.....
This minimizes the peaks and pushes dissipation towards average.
I've got dozens of 89C51s so Ive used them in the past, but have an Arduino Uno on order to play with.
Also, with a microprocessor you can perform the sweeps in a non-sequential order. Lowest dissipation, then highest dissipation, next lowest dissipation, next highest dissipation.....
This minimizes the peaks and pushes dissipation towards average.
Last edited:
Working slower allows you the benefit of adjusting settling time, averaging measurements before you dump them to the host, removing outliers, and loop compensation is easier.
I'm designing a system that uses pulse measurement techniques ;-) for each point e.g. I'm using a D/A converter that has an synchronous load function such that each measurement point is a brief pulse from a specified baseline condition. In this way, the dissipation can be limited by the duty cycle of the measurement pulses.
For example, each plate curve would consist of a sequence of higher and higher plate voltage pulses until some limiting condition is met e.g. I*V. For each measurement point, all electrode voltages will be pulsed together, held for a settling and measurement time, and then returned to the baseline voltages.
An arduino has good A/D capability so I am using an arduino plus a separate AD7808 D/A chip having the required global load/enable function and up to 8 10 bit outputs..
I find that 10 bit resolution is sufficient as long as there is range scaling for each electrode e.g. plate current measurement range selection of say 1mA, 10mA, 100mA, and 1A to cover a range of tubes and operating points.
I'm providing a separate power supply for each electrode (there are only 3 after all...) such that current can simply be measured by means of a shunt in the return path. Using pulse measurement, these supplies can be relatively small in terms of continuous power and have big storage caps to provide the short term current needed for the measurement pulse duration.
I'm working on a simple arduino "shield" for the AD7808 D/A converter and a simple programmable regulator to drive each electrode. A rudimentary tester/tracer would consist of an arduino board, the D/A shield, and 3 regulator boards for g1, g2, and plate. The regulator board will also carry the rectifier and filter so will only need 3 small toroids (or a toroid with 3 sec coils...) and 3 regulator boards for the electrode supplies. The programmable regulator circuit is opamp+MOSFET based, similar to the one posted earlier in this thread.
Range selection is through a rotary switch for each electrode. The arduino code would iterate through the set of points comprising the plate curves and output CSV lines for the plate curve data which could then be plotted using any of a number of apps e.g. ploticus or MS excel.
Also the tracer could be fitted with a realtime display consisting of a bitmapped LCD panel and another arduino to draw the curves. Uno boards are about $30 each and there are LCD display shields.
I want to do this as an open hardware and open software project.
PS also thinking of using it as a physical modeling box for use with spice
Last edited:
Yes you are right, there is no reason to go slower then you have to. My point was that you don't need to go really fast.
Another thing a micro processor can do is consult a database of tube specs. so if you are testing a 6v6 tube the tester can know the max plat voltage and so on and not go over the limits. The old style testers had their "database" stored on a paper roll or chart. A new design tester would have the tables stored internally.
Now I have a question. I will also post this in the solid state forum but some one here may know. I need a MOSFET for the pass regulator for the variable plate supply. I will construct this tester to use a fixed voltage bench B+ power supply but I'll build a pass regulator using a power mosfet. So the question is "which mosfet?" I ant one that is very common that anyone can find, low cost. and over spec'd so it will be hard to fail even if some one tried to test a shorted tube. Figure that we want to be able to test power tubes like a KT88 Are there any 800 volt .5 amp mosfets. I had also thought about using a vacuum tube as the pass regulated but
the mosfet's gate will be controlled by the Arduino's analog output. But via an opto-coupler and with a 7 volt senior diode as a backup safety device to shunt high voltage to ground. I really don't want any of that 800 volts B+ getting into the USB cable even after multiple component failures and a resulting small fire. If people suggest some part numbers I'll order a couple of each and try them.
Sorry I did not mean to hi-jack this thread. I really do like the scope based tracer, it's way-cool.
I've mentioned this at least once, before, somewhere. But I don't think that anyone else realized what it could mean.
Using a computer and software, with A/D converters measuring all voltages and currents of interest (which the computer would save), could free us from needing to use synchronized waveforms at all.
To make the point, maybe even noise, if large-enough in amplitude, could be used for both the high-current path drive and the control drive. The point is that the computer could measure and record for a short time and then sort and reassemble all of the measurements by voltages and currents in any way we wanted and still produce the same familiar-looking curves (except that it could show the steps for ANY values we wanted to see).
i.e. The time-sequencing of the stimuli would be unrelated to the ability to construct the desired plots.
I'm not suggesting using noise instead of waveforms, necessarily (unless it would be advantageous). Just making the point that digital sampling, storage, processing, and display-management would open a whole new universe of possibilities, and not just for the use of asynchronous or random test signals.
Imagine also collecting temperature data and plotting the curves versus temperature, or any other variable that can be measured and digitized. Long testing time periods could also be used, and/or multiple devices, and statistical plots could be created. The possibilities are probably endless. Matching or selection would be trivial to automate, for example.
Maybe a good "curve tracer" could consist of only a set of high-impedance probes and differential probes and current probes, with A/D converters, and "curve tracer" software on a computer.
It could, for example, be connected to a real device in a real application circuit. The circuit could be used for the real application, while being measured. Having collected data for the desired operating scenarios, the "curve tracer" software could assemble and present the data in whatever sets of curves one wanted to see. Or it could provide real-time plot displays like an old-fashioned curve tracer, too.
Of course, you could still also use it with a dedicated "test fixture", when desired, which would look like most of the rest of a standard curve tracer, although it might be even nicer if the test signals and settings and other functions could also be controlled by the software.
Also, imagine the possibilities for the automatic generation of Spice models.
Sorry to blather on, again. I'm sure it's all been done at places like Agilent. But even for a "typical" analog curve tracer, like what is being discussed, it would be quite simple to ADD ON the digital sampling capability (while not losing or changing any of the original capabilities), to be able to record any one of the voltages and any one of the currents of interest (or all of them at once), as numbers in a file on a standard PC. After that part was done, there would be infinite options, starting with just using off-the-shelf software to make the usual plots (even just using MS Excel, or freeware).
To do it really cheaply and easily, maybe something like a soundcard could be used as the input. Then it could be done without even knowing anything about A/D converters. (But I'm not sure if they can have their AC coupling capacitors removed to enable DC coupling or not. Might have to buy a cheap A/D board or a USB A/D box instead, or (gasp!) use a parallel or serial port.)
EDIT: Ah, cool, I see that ChrisA and others DO "get it". An Arduino or other dedicated hardware would probably be the best way to go (maybe by far; I haven't done any of the math). But I was thinking more about the fact that almost everybody already has a PC, and maybe also some old PCs that they don't even use anymore, and thought maybe there would be some basically-already-available way to get the data into the PC, at least for the case of only two data streams at a time. But wow that WOULD be very limiting, now that I think about it. I went into a Radio Shack for the first time in quite a while, last week, and they seem to have reversed course, again, at least by a bit! They actually have Arduino and Basic Stamp kits, and lots of protoboards, and some other stuff I haven't seen there for a long time! The manager said something about a new CEO or something, taking them back more toward their roots...
Cheers,
Tom
Using a computer and software, with A/D converters measuring all voltages and currents of interest (which the computer would save), could free us from needing to use synchronized waveforms at all.
To make the point, maybe even noise, if large-enough in amplitude, could be used for both the high-current path drive and the control drive. The point is that the computer could measure and record for a short time and then sort and reassemble all of the measurements by voltages and currents in any way we wanted and still produce the same familiar-looking curves (except that it could show the steps for ANY values we wanted to see).
i.e. The time-sequencing of the stimuli would be unrelated to the ability to construct the desired plots.
I'm not suggesting using noise instead of waveforms, necessarily (unless it would be advantageous). Just making the point that digital sampling, storage, processing, and display-management would open a whole new universe of possibilities, and not just for the use of asynchronous or random test signals.
Imagine also collecting temperature data and plotting the curves versus temperature, or any other variable that can be measured and digitized. Long testing time periods could also be used, and/or multiple devices, and statistical plots could be created. The possibilities are probably endless. Matching or selection would be trivial to automate, for example.
Maybe a good "curve tracer" could consist of only a set of high-impedance probes and differential probes and current probes, with A/D converters, and "curve tracer" software on a computer.
It could, for example, be connected to a real device in a real application circuit. The circuit could be used for the real application, while being measured. Having collected data for the desired operating scenarios, the "curve tracer" software could assemble and present the data in whatever sets of curves one wanted to see. Or it could provide real-time plot displays like an old-fashioned curve tracer, too.
Of course, you could still also use it with a dedicated "test fixture", when desired, which would look like most of the rest of a standard curve tracer, although it might be even nicer if the test signals and settings and other functions could also be controlled by the software.
Also, imagine the possibilities for the automatic generation of Spice models.
Sorry to blather on, again. I'm sure it's all been done at places like Agilent. But even for a "typical" analog curve tracer, like what is being discussed, it would be quite simple to ADD ON the digital sampling capability (while not losing or changing any of the original capabilities), to be able to record any one of the voltages and any one of the currents of interest (or all of them at once), as numbers in a file on a standard PC. After that part was done, there would be infinite options, starting with just using off-the-shelf software to make the usual plots (even just using MS Excel, or freeware).
To do it really cheaply and easily, maybe something like a soundcard could be used as the input. Then it could be done without even knowing anything about A/D converters. (But I'm not sure if they can have their AC coupling capacitors removed to enable DC coupling or not. Might have to buy a cheap A/D board or a USB A/D box instead, or (gasp!) use a parallel or serial port.)
EDIT: Ah, cool, I see that ChrisA and others DO "get it". An Arduino or other dedicated hardware would probably be the best way to go (maybe by far; I haven't done any of the math). But I was thinking more about the fact that almost everybody already has a PC, and maybe also some old PCs that they don't even use anymore, and thought maybe there would be some basically-already-available way to get the data into the PC, at least for the case of only two data streams at a time. But wow that WOULD be very limiting, now that I think about it. I went into a Radio Shack for the first time in quite a while, last week, and they seem to have reversed course, again, at least by a bit! They actually have Arduino and Basic Stamp kits, and lots of protoboards, and some other stuff I haven't seen there for a long time! The manager said something about a new CEO or something, taking them back more toward their roots...
Cheers,
Tom
Last edited:
I hope the Arduino is going to open a whole new world up to kids with scientific/engineering interests.
Since ICs took over from discrete logic there has been a continuing recession of available 'toys' so to speak, for kids to play with.
I hope this will lead to a new field of educational tools to stimulate their interest in the world of electronics.
Back to the topic of interest, I find the Arduino solves a number of issues, such as getting data into the computer (via USB), Readily available HW, Unified F/W development package, it is quite inexpensive, etc.
ChariA, I have been using the IXHF9N80 which is rated at 800V, 9A continuous, 180W, 36A peak pulse.
Even with this I have been blowing up devices in a pass regulator.
The issue is you need a lot of capacitance following the regulator to meet the peak current demand (unless you want to make it quite complex and achieve low z0 with a lot of capacitance prior to the regulator). That high capacitance means you now have to add current limit on the output of the regulator to protect it from power on surges. I finally got it stable in the curve tracer I built with the Lampizator module.
If you want some samples of the 9N80, PM me your address and I'll get them in the mail.
Since ICs took over from discrete logic there has been a continuing recession of available 'toys' so to speak, for kids to play with.
I hope this will lead to a new field of educational tools to stimulate their interest in the world of electronics.
Back to the topic of interest, I find the Arduino solves a number of issues, such as getting data into the computer (via USB), Readily available HW, Unified F/W development package, it is quite inexpensive, etc.
ChariA, I have been using the IXHF9N80 which is rated at 800V, 9A continuous, 180W, 36A peak pulse.
Even with this I have been blowing up devices in a pass regulator.
The issue is you need a lot of capacitance following the regulator to meet the peak current demand (unless you want to make it quite complex and achieve low z0 with a lot of capacitance prior to the regulator). That high capacitance means you now have to add current limit on the output of the regulator to protect it from power on surges. I finally got it stable in the curve tracer I built with the Lampizator module.
If you want some samples of the 9N80, PM me your address and I'll get them in the mail.
I do think arduino is the way to go. I have been building and programming arduino projects for a few months now and it's hands down the easiest to work with ever. BTW, I've used a lot of different uControllers in the past and even designed bit slice processors back in the old silicon age.
Arduino doesn't have analog outputs, only PWM, so an external D/A seems like a good idea.
For Spice, I think the way to go is physical modeling, where Spice provides the inputs to drive the tester, the HW tester measures the tube currents etc., and provides the outputs back to spice. Otherwise there is still a crapload of curve fitting and trend extraction to do.
A micro-programmable tube tracer/tester would be able to do so much more than anything out there now. Imagine being able to plot the exact load line for your tube at your op point. Matching at your op point. actually measuring grid currents for A2 and screen drive designs. Fiddling with g3 voltage... It would be super easy to create a custom measurement regime for your own needs.
Using arduino, open hardware, and open source would allow many people to contribute in their own area of expertise.
As for the pass regulator, I would only ever put the storage "behind" the regulator. Also make sure that DUT and fixture capacitance doesn't cause reverse voltage to be exceeded when rapidly reducing output voltage. Gate protection diodes are essential.
Arduino doesn't have analog outputs, only PWM, so an external D/A seems like a good idea.
For Spice, I think the way to go is physical modeling, where Spice provides the inputs to drive the tester, the HW tester measures the tube currents etc., and provides the outputs back to spice. Otherwise there is still a crapload of curve fitting and trend extraction to do.
A micro-programmable tube tracer/tester would be able to do so much more than anything out there now. Imagine being able to plot the exact load line for your tube at your op point. Matching at your op point. actually measuring grid currents for A2 and screen drive designs. Fiddling with g3 voltage... It would be super easy to create a custom measurement regime for your own needs.
Using arduino, open hardware, and open source would allow many people to contribute in their own area of expertise.
As for the pass regulator, I would only ever put the storage "behind" the regulator. Also make sure that DUT and fixture capacitance doesn't cause reverse voltage to be exceeded when rapidly reducing output voltage. Gate protection diodes are essential.
I thought of the sound card. But I think many of them are AC coupled and can't hold a DC level (maybe) but the real problem is that I want at a minimum 6 analog inputs. I want to measure the voltage and current on every pin. Think of a pentode.
And then there is the problem than a few people will not want to risk pugging something that runs on 500 volts into a computer.
The big task for a tester like this will be software. So it is best to use something that is VERY day to program so more people can contribute. Also you wan to be multi-platform. Arudino is easy and it's development system runs on Windows, Mac and Linux equally well. As a test I wrote something that would read the voltage from a pot (wired as a voltage divider) and use PWM to have a LED brightness follow the pot while also displaying the pot's voltage on an LCD. It took all of about 20 minutes and that was the first evening I have the Arduino. So it is a very quick learning curve. OK I do software at work 25+ years at it not on Arduinos or AVRs
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.