Discussion of subcircuits for homebrew valve testers

[Thatvalveguy]

I would like to dedicate this topic to a collection of ideas and circuit examples that can be used in the building of valve testers.

Over the last few months I have been thinking up circuits for a valve analyser, whose aim is to be more stable (and perhaps more accurate) than any of the rather expensive old production kit.

Take for example the AVO series of testers, they are wonderful if in good nick. however their patented circuits require a thourough understanding of the AC measurement used. Furthermore calibration can only be done with a 10Kohm/volt voltmeter.

The Sussex valve tester is a good compromise between complexity and function, quite frankly the only thing I dislike is the G1 control voltage.

Any type of test equiptment ideally has no drift, and can be calibrated by a DC voltmeter.

If one wishes to measure a lot of valves in quick succession without having to adjust voltages because they have drifted out of their initial setpoint. one needs active regulation and a good voltage reference.

The advantage of having active regulated supplies is that they can be voltage controlled, a nice feature if one would like to add computer control at a later stage.

[kellys_eye]

Suggest you check out this site:

http://www.dos4ever.com/uTracer3/uTracer3_pag0.html

if not for the completed device then at least for the sub-circuits - very clearly detailed - that may prove useful by adoption/modification.

[Bazz4CQJ]

Originally Posted by Thatvalveguy

Over the last few months i have been thinking up circuits for a valve analyser, whose aim is to be more stable

Regarding the accuracy/stability of the old Avo instruments, let’s consider the exercise that was run on this forum last year, in which several owners compared readings of standardised valves produced by David Simpson. My impression was that the great majority of the AVO’s produced results within quite a narrow spread (better than +-5%). I don’t recall how the results from Sussex’s compared.

Obviously, with the solid-state components available today, any new instrument can easily have well-regulated supplies. However, I think the AVO’s show that voltage regulation is not the most important feature in the design of a valve tester and that numerous other factors need careful consideration.

That said, I can well see the appeal of running a tester from a PC, and however good/bad the AVO's were, they are now quite historic.

B

[Thatvalveguy]

True, I must admit my first post was quite incoherent.

I am well aware of the good accuracy of the AVO.

I have built a homebrew tester that has a single +250V regulated anode supply, and a variable negative that has its own dedicated panel meter. another panel meter is switchable between reading triode section 1/2 or anode/screen grid. Sockets are pre wired for the valves I use.

Simplicity at its finest.

The last thing I worked on was designing a grid regulator that can regulate between 0 and -100V, is voltage controlled and short circuit proof (it ended up being a shunt regulator) My idea was to make this a Din module so it can fit into a rack.

Furthermore after a few days of breaking my head I devised a way to measure grid current directly on a 200uA scale. Using the high imput impedance of a LF411 and a HCPL7800 linear optocoupler.

The offset trim of the LF411 is used to also trim the offset of the HCPL7800.

[Diabolical Artificer]

It depends what you want it do, to trace Va/Ia curves or test pass/fail.

As for bang on precision, valve circuits are ish and tranny circuits are the same to a point. As long as a valve is about right it's good enough for most application's. One other thing, a valve datasheet is a fiction, a representation of an ideal valve. I've yet to test a valve that behaved exactly like the datasheet example.

Testing valves is a pain in the bum, re setting anode, screen, bias and heater for every valve tested would drive you nuts, unless you like that sort of thing.

If you want precision, build a computor controlled Tek 570 clone but it's not needed for your average valve testing.

Some ideas here - http://sound.whsites.net/project151.htm here - http://www.valvewizard.co.uk/valvetester.html here - http://www.valvewizard.co.uk/curvetracer.html and lastly owhat one bloke thought about whizz bang testers the Ampitrex AT 1000 - https://www.youtube.com/watch?v=SRkUbYtVawU

Andy.

[ionburn]

Apart from the obvious of testing valves in a known working circuit, my approach has been the totally modular one, in that my tester consists of a range of modules with jump wires for connection. That approach is obviously rather messier than a specifically designed single box tester, but, as my requirements can be very varied, I can rapidly re-connect (or even make another module) for most specific tests. The heart of my system is basically a multiple power supply unit, with external socket and resistance boxes. On thinking about it, it does seem that it would be possible to test transistors with the same unit. I did use it recently for obtaining comparison curves for crystal diodes with valve diodes.

I would agree that the computer controlled unit would be the most useful for most requirements these days.

[Thatvalveguy]

Well one of the things I wanted to do was to at least build something that has all its measurement voltages near ground, computer control is not a necessity, that's why I chose to keep the DAC/ADC module modular as well. Most of the axillary supplies can be made from +-15 volts.

There is something to be said of using the classic 470K-1Meg resistor for a gas/grid emissions test. it works really well, however it never quantifies the amount of gas.

The aim of this was to make a rack that has most of the supplies inside, all the supplies could theoretically be derived from a 24V bus by means of switch mode supplies. An inverting boost controller for 24V to -120V is not that difficult.

Pass fail consists of a few parameters: anode current, screen grid current, control grid current, cut off voltage (E.g. anode current goes below 100uA) mutual conductance or (S) and last but not least current from cathode to heaters. I don't really care about the curves of a tube as I have a Curve tracer.

You can actually test about 80 percent of receiving valves with a fixed 250V supply, for outputs this might even be better then running them at higher voltages as you can test at higher cathode currents.

Modular approach is indeed preferred, I thought about building it to fit the DIN standard of 100 x 160mm

[Argus25]

Originally Posted by Thatvalveguy

Any type of test equipment ideally has no drift, and can be calibrated by a DC voltmeter.

A simple but very insightful remark. The beauty of it is that the DC voltmeter calibration itself is very easy to test these days because of the availability of precision IC voltage references.

I made a semiconductor device to calibrate Hickok valve testers and the device itself has its calibration checked with a meter too. The Hickok testers use an operating principle which is ingenious and as it turns out, do not require a transconductance device or a reference tube for accurate calibration. If you are interested, the calibration tool and information on Hickok valve testers is here:

http://worldphaco.com/uploads/Hickok...d_the_ACS..pdf

[Thatvalveguy]

That looks like a very impressive project! I have never dealt with Hickock units.

The ultimate in simplicity/accuracy should go to the Russian built L3-3 type of testers.

In those machines the Shunt resistor for the dynamic transconductance measurement is included in the feedback network of the regulator supplying the anode. The Ac component of the transconductance (Steepness S value call it whatever you want) is Ac coupled into the bandpass filter producing the reading for the panel meter to display.

The Wien bridge oscillator feeds into a precision resistor network, the divider is tapped at 445mV in calibration mode this voltage is fed directly into the bandpass filter to calibrate the bandpass filter.

The smart bit is that this approach allows calibrating the entire circuit, the accuracy is derived from the ratio between measurement resistor and grid voltage divider resistors, any drift in amplitude and to a certain degree the freqency of the oscillator and bandpass filter is calibrated out. Taking .1% resistors for the divider and .5% resistors for the measurement shunt, the circuit is capable of astonishing accuracy.

The only drawback of the L3-3 itself (I have one of those) is that any drift in the bandpass filter or oscillator itself requires you to recalibrate every 10-30 minutes depending on the stability of your particular unit.
I know from people that have replaced the triode wien bridge oscillator by a lamp stabilized op amp oscillator for much improved drift performance.

Furthermore the Bandpass filter has a very steep frequency curve, the -3dB points are about 10Hz apart if memory serves me well.

For any implementation of this circuit, i need to take a good look at available true RMS converters to produce a 0-2V or 0-4096V output signal. Building a stable bandpass with an opamp is not that difficult. you can achieve about 5 Q factor maximum with one single stage.

Calibration would be manual, as i doubt solid state circuits operating at low power will drift more then 1%

Any solid state implementation would likely use a lot of Reed relays for improved reliability, those can be switched by a microcontroller and a NPN or a switch quite easily.

For those interested in the exact workings and Circuit topologies used in the L3-3 there is a very elaborate write up on the Jacmusic site. (Unfortunately in German)

[Thatvalveguy]

For those Who need a 240VPP buffer, there is the AN18F Figure 9 from linear technology written by the famous Jim Williams.

I have build and verified this circuit with MPSA42/92 for the drivers and 2SA1837/2SC4793 as outputs MJE340/350 will likely work as well. The Toshiba transistor are sadly discontinued. The circuit achieves its rated bandwith without problems, and well compensated the overshoot is rather small.

The opamp really needs to be a FET opamp, or the inputs will latch up.

If you plan on building a curve tracer, this could be the output stage. I suggest putting some parallel diodes over the outputs, and running the opamp from a bipolar supply. Also the circuit will latch up to one of the supply voltages if the high tension rails vary more than 25-30V

There is also the 783 regulator that can be used up to 125V input-output voltage differential.

[Diabolical Artificer]

This video - https://www.youtube.com/watch?v=VTqNi7UexBI came out today. He has some interesting ideas on valve testing especially matching.

A.

[Thatvalveguy]

Well matching is sometimes required, the curve tracer kit from Ronald Dekker is best suited for that. However matching shouldnt be that important, negative feedback takes care of most tube parameter variations.

Testing at the data sheet settings of a tube particularly the 250V specification makes sense because that's what the manufacturer uses for in house QC.

For me personally its much more interesting to be able to measure the vacuum directly, or to be able to determine the cut off voltage or the control grid, because those things tell you something about quality, not about the normal distribution spread of a particular lot.

[Bazz4CQJ]

Originally Posted by Thatvalveguy

For me personally its much more interesting to be able to measure the vacuum directly, or to be able to determine the cut off voltage or the control grid, because those things tell you something about quality, not about the normal distribution spread of a particular lot.

I find your comments interesting...but quite confusing.

One of the things that people generally do when they first get a valve tester is to test valves which are in service in various bits of kit they have. The usual outcome of that exercise is they find that they have equipment which is working very well which actually has some valves which give pretty poor test results. When I went through that phase, what I was told that good designers of valve equipment anticipated that valves would deteriorate and designed to tolerate that.

One of the other things you learn when you first get a tester is how very variable brand new valves are.

So, my point here is that most valves were produced to tolerances that were not too tight and most valve circuits will work with valves which are appreciably off spec. That being so, what is it that leads you to say "measure the vacuum directly, or to be able to determine the cut off voltage or the control grid, because those things tell you something about quality". Generally, people use testers to establish which valves they have are which usable and which need to go for re-cycling. The criteria for that selection is generally based on measurement of Ia and Gm, which are the key measurements on most commercial testers.

Your thread would be helped a lot if you could produce a list of key features you'd like to have. Maybe you want features are are not found on any commercial tester which has ever been produced, but are more some form of one-off experimental rig which meets your particular hobbyist's interest?

B

[Argus25]

Originally Posted by Thatvalveguy

That looks like a very impressive project! I have never dealt with Hickock units.

It is worth looking at the way Hickok does their transconductance test, I described it in the article.

Just one of the brilliant parts of the Hickok tester is the fact that all the voltages throughout the unit are derived from the one precision wound power transformer and remain proportional and the transformer primary is driven by a calibrated line voltage.

Of course, for a home constructed valve tester project, to manufacture such an electromagnetic mil spec masterpiece of a transformer is not practical and out of reach (at least it would cost a small fortune at a transformer winder too) and this leaves home constructors of valve testers out of the loop for deploying Hickok's advanced transconductance testing concept.

The only way to get it is to buy the actual Hickok tester that contains the transformer. These were the testers used by the American Military, typically the TV7 and TV10, so they were generally manufactured to an excellent standard. There were a number of other variants with additional metering. But the ingenious operating principle was invented & patented by Hickok.

Even the meter module in the tester is a mil spec sealed unit. Of late some folks have been making drop in digital replacement meters as the originals are getting difficult to source, but I prefer the original meter.

[Ed_Dinning]

Hi Gents, I believe that Mullard also used this transformer technique on the HSVT, where the mains is standardised before testing.

The Sussex has the facility with 10v mains tappings (Europe model) to keep the voltages close to standards, but is not fully ratiometric as it uses some regulated supplies.

An improvement on the Sussex if size and cost are not constraints would be to use stabilised DC heater supplies.

I have a large number of early valves, often unidentified that appear to work OK. One of my requirements to identify them is to check their ra, so I can see if they are best suited for Det, RF/ AF or Power use.

For audio use I would think a facility to look at hum levels would also be useful, possibly using a computer sound card for analysis.

Ed

[julie_m]

How about a "Peak Atlas Mode" where a computerised valve tester attempts to identify the electrode structure(s) in a valve?

First do some quick continuity tests to identify the heater pins.

Once the correct heater voltage / current rating has been found (by possibly-manual experimentation) then try grounding each non-heater pin in turn, and applying some HT to each other pin in turn via a current sensing and limiting resistor. Only the cathode(s?) will be near enough to the heater to emit any electrons, although any grid will be able to act as an anode (though maybe not for long, at high current .....)

Lastly, for each possible pair of cathode and "anode" in turn, take each other pin in turn to some negative DC voltage. If the emission current is cut off, that means the negative voltage was applied to a grid which is nearer the cathode than whatever is acting as the anode (maybe the real anode, maybe another grid).

Now you know what the structure is, more or less, apart from any internally-connected grids which can't be discerned apart; a "G2+G4" will block emission to a G3 but if G2 is acting as the anode, only G1 will block it; the electrons will have reached G2 and made a bid for freedom before they enter the G3 - G4 space where G3 could stop them.

It could certainly be an impressive party trick .....

Heaters: 4 and 5

1 as K: no emission
2 as K: no emission
3 as K: emission with 2 acting as anode
6 as K: no emission
7 as K: emission with 1, 6, 8 or 9 acting as anode
8 as K: no emission
9 as K: no emission

Emission 3-2: no way to block
Emission 7-1: no way to block
Emission 7-6: no way to block
Emission 7-8: no way to block
Emission 7-9: 8 blocks

3 and 2 form an isolated diode structure.
7 is a common cathode with 1 and 6 as diode anodes, and 8 and 9 as grid and anode of a triode.

[Station X]

Just what I need to test and identify valves with all the writing rubbed off.

Plug in the valve and the tester tells me what it is and whether it's any good.

Setting up all the data files would be fun. Derive them from the OCR version of the AVO valve data manual?

[Thatvalveguy]

First of all, i regard tubes as things that are to be studied. i don't have a huge inventory of old B/W TV's or 1920's radios with tubes to test.


Specifications
First of all, i'd like to be able to measure plate current/screen grid current at DC bias conditions. dynamic Transconductance (like in the sussex) would also be useful as the grid shift method is annoying and prone to errors. Cathode to heater leakage, is not that difficult to implement as well, and allows a rather complete good/bad analysis.

One of the novel things was my idea to measure control grid current, outside of a few laboratory grade machines there are very few that can do this, (One of the reasons could be that sub 50uA movement meters where ridiculously expensive back in the day. receiving tubes typically have sub 1uA grid current. Neccesitating some form of active amplification.

One must realise this functionally is an elaborate way of trying to achieve what a 10 cent resistor can do(Tell you if a tube has grid emissions or not), However some tubes(Particularly Philips and Mullard SQ series) have grid emissions specified in their end of life data, and i'd like to check against those parameters.

Furthermore a resistor grid leakage test will tell you the increase in anode current, but depending on the S at this particular point this will still only give you a theoretical answer to the question: What is the grid emission?

Digital control
For me digital control is optional, that being said, its easy to include the header and some holes for mounting at a later stage on a PCB. The digital add on PCB, can be SMD if size becomes an issue, i prefer through hole for everything else, for the simple reasons that i have a few dozen boxes of metal films to use up.

I have no desire to write elaborate software and protocols for this kind of project, it quite frankly doesn't suit my own needs, nor do i have any desire to commercially exploit the design. for my personal use it would be a bare bones GUI that displays control voltages, allows for range switching and displays measurements.

I'm going to start with one module at a time, that all share the same 100x160mm PCB size and 31 Pol DIN41617 connector. Uniform pin assignment would be nice. i was thinking of assigning pin 15 through 25 to module outputs, with each module featuring a row of 10 jumpers, giving you some leeway on final wiring.

I have yet to fully specify the pinout on the connector, and the supply voltages. For the digital interface i thought it was smart to include a RS485 transceiver IC footprint. (those are about €4 each, auch!) allowing all the different modules to communicate with a central module without running the risk of significant interference.

I have thought about Building the sockets on a PCB based fixture, getting ten PCB's 100x100mm Pcb's done at one of the board houses in China runs you under €18 shipped. This will allow type specific test sockets. This has one major advantage: it allows you to control the parasitics very well, and design in grid stoppers and screen stoppers. You can also just dedicate two pins on a 31P connector to achieve remote sensing from the stabilized heater supply.
Last but not least: its rather foolproof, use the correct socket for the correct tube type and you avoid the risk of damage.

For most audio work 5-6 of these PCB's will be enough to test 80% of what Mullard made, due to the simple fact that one board can test for example ECC81 through 83 E180CC E182CC 12BH7, well you get where this is going.

Personally, i can recognize most valves by the anode structure, and if that fails factory code. Also the P series valves heaters differ in thickness from the E equivalents, which allows me to differentiate between say a PF86 and EF86. Quite frankly, i don't see the need. It is however quite feasible, one should look at the testers used to distinguish bad connections in receiving tubes. I beleive this was a circuit consisting of a string of resistors paralleled by neon bulbs that would light if a connection was unused.

Sussex improvements:
About the possible improvements to the sussex heater supply: There are linear regulators that can work with dropout voltages as low as .4V at several amps, these and perhaps Skottkey rectification could allow you a stabilized supply from the existing transformer. Cost would be ranging from €10 to €20 depending on complexity and whether or not you use a 10 turn potentiometer to set the output voltage.

Hickock testers
The hickock units, well their bridge? approach is a nice idea. However it cannot be calibrated by a voltmeter. Furthermore it doesn't test the tubes near published standardized test voltages and currents. Its good at identifying bad tubes, but its not a labratory grade machine. The average/advanced user could not possibly test a tube from its datasheet. You need a RCA WT100 or a new london 901A for that. and even those suffer from their use of vacuum tube circuit.

To me the Russian L3-3 is the pinnacle of tube analysers any solid state machine that can match its performance would be the new king of the hill so to speak. its dynamic S/transconductance measurement circuit is perhaps the most accurate approach to measure outside of using a true Weston brige. Its drawbacks are many, not limited to rather poor versatility.

The Avo approach is quite interesting, however getting a transformer wound with that many HT taps will be ridiculously expensive. did you know that Metrix made the 310? that uses the same operating principle.

For those that want to do a little construction work, one can always clone the metric U61. Its leakage test is also quite nice, i had that working on a early unit i built.

A Tektronix 500 series O'scope transformer will yield most of the voltages you need. The Chinese panel meters will match or outperform the original meters

[Bazz4CQJ]

Originally Posted by Thatvalveguy

First of all, i regard tubes as things that are to be studied. i don't have a huge inventory of old B/W TV's or 1920's radios with tubes to test.

Ah Ah! I can understand that; a while ago I was looking at some very old technical papers dealing with the development of cathodes and the materials which were progressively adopted for use as cathode coatings (starting around a century ago) and it was really interesting.

Just occasionally, I look at my VCM163 and think that it would be a wonderful project to rip it apart, junk the bad and olde worlde bits and use the good and useful bits together with some with modern components/circuits to start again. However, realistically, that's 'unlikely' to happen.

B

[Thatvalveguy]

Well the VCM163 is a nice bit of kit, (Like all the avo's) very versatile, however some parts are hard to replace, if replacements can be found at all. Any monkey mistake can wreck the panel meter. (This might be different on the VCM163) If i remember correctly the VCM163 uses a bandpass filter as well, that is transformer coupled.

The avo's, you cant really change much except changing the panel meter out for a digital unit, in that way you can display negative numbers as well. so the backing up control works more logically (The 163 doesnt have backing off, like the CT160) DC heating is a pain for indirectly heated tubes, as one part of the heater is going to be elevated.

Tube testing itself is not an accurate buisness, due to cathode interactions. most tubes that have been unused for some time will change their readings in the time following the first test, this is due to the composition of the barium surface layer changing.

Sussex regulated heaters

Whilst writing my last post i suggested using a LDO to supply regulated heaters to a Sussex tester, The IC i was looking for was the LT3803, that IC can supply 3A at a dropout voltage of approximately 300mV and stay in regulation, enough to test a 2A output tube before the foldback current limiting kicks in during warm up.

There is the option of using some special IC so you can use four N channel mosfets as a bridge rectifier and get somewhat higher input voltage for the regulator.