Modifying the Mullard High Speed Valve Tester to read gm

[Electronpusher0]

It seems, reading the forum posts, that the main disadvantage of the MHSVT is that while it performs tests using DC conditions to the Mullard spec it is, essentially, a go/no-go tester.
It would be far more useful if it could also measure mutual conductance, gm.
I am therefore looking at adding a gm feature to the MHSVT. There are a few basic assumptions I am making
1. Nothing is to be done to the tester that prevents it being restored to its original condition. That precludes drilling any holes!
2. Use of modern electronics including digital panel meters is OK, the object is to increase its usefulness not retain the technology (however see rule 1)
3. Any additional circuitry, readouts and controls are to be in an outside housing, painted silver Hammerite with a black front panel, preferably sloping to match the tester.
4. Ideally no additional cards should have to be made, gm should be capable of being measured with the standard cards.

For an understanding of how to measure gm I looked at the manual for the “Sussex” tester.
To measure gm a 100mV rms signal is applied to grid.
The Anode current is monitored by measuring the a.c. voltage across a 10R resistor. This gives 10mV per mA of Anode current. With 100mV of grid drive the meter reads mA/V directly, 1mV equates to 1mA/V. (the meter used in the Sussex reads 200mV ac)

In the MHSVT there is a resistor, R89 (100R), always in the Anode circuit.
The proposal is to change this for a 1% component and measure the ac volts across it while injecting the 100mV ac signal on the grid. In this scenario 10mV equates to 1mA/V, a 200mV meter would read 0 to 20mA/V!, with a resolution of 0.1 mA/V
I would need to add a 10K resistor in the grid circuit so as to prevent the existing grid bias supply loading the 100mV 1kHz (ish) sine wave inserted via an isolating capacitor.

Before starting to make any modifications I decided to hook up a “proof of concept” adaption, this was done by making an adapter to plug into the Octal base with a B9A base on the top. This adapter was wired for an EL84 as I have a quantity of these, old and new, to test.
An additional 100ohm resistor was wired in series with the Anode lead in the adapter with a wire brought out from each end of the resistor for the ac mV meter.
Likewise a 10K resistor and 0.1uF coupling capacitor were wired into the grid lead, wires from the capacitor and Cathode were also brought out for the external oscillator.
The schematic and photo of the adaptor is attached.

I have tested this set up using several valves but the results reported below were for a NOS device.
I first tested the valve without injecting any signal to make sure that I still get the same results as I do when using the normal B9A socket (I do).
I then used an external signal generator at 1kHz set to 100mV as read on my digital multimeter, this was injected into the grid. The Anode ac current was monitored by reading the voltage across the 100 ohm resistor using the same multimeter.

The tester was set to position 6, used to check emission, this tests the valve at the Mullard published spec:
Characteristics (from Mullard Data sheet)
Pentode Connection
Va 250V
Vg2 250V
Ia 48mA
Ig2 5mA
Vg1 -7.3V
gm 11.3 mA/V

I was very gratified to find that when the 100mV signal was injected the meter read 112mV representing a gm of 11.2, precisely to spec!
I have tested several other valves and got a range of values from 9.3 to 11.6
I am satisfied that this proves the concept will work and am proceeding to build the final version.

For the final version I plan to remove the front drop down flap by unbolting the hinges, (I will put the flap aside carefully in case I need to put it back to original), there is enough space to get cables past the mains selector / presets panel to the inside of the tester.
I plan to mount a sloping front box in place of the drop down panel using lift off hinges and a single screw using the existing captive nut used to hold the flap. This allows the box to be put aside to access the mains selector and presets if needed.
For the oscillator I decided to use a Wein Bridge oscillator using diode stabilisation, this gives 1 to 2% thd but that is more than good enough. The output of the oscillator is about 1V pk-pk. This is attenuated by a pot and buffered before being output to the grid.
Also for the final version I also plan to include a mains indicator digital panel meter. I have already marked the voltage settings on the mains selector switch (using peel off Dymo of course) and intend setting the mains switch by reading the actual mains voltage rather than using the mains card.
I will post further to report progress, however it may be several weeks as the meter has to come from China.

[David Simpson]

gm is just a STATIC function, i.e a DC Function of a valve. Then we have mu which is the DC Amplification Factor of a valve. mu = ra(the internal resistance of the valve) x gm. Both must follow Ohms Law & Barkhausen's Law. Then we have Gain "G" which is a DYNAMIC Function i.e. a Signal Function & must comply with the Manufacturer's spec for Ra - which is the Anode Load Resistance. Hence G = V out/V in = Anode signal / Grid Signal = (Ra x mu)/(ra + Ra). Note - as Ra/(ra + Ra) is always a fraction then G will always be less than mu.
I apologise if I need to go to Specsavers, or need some maths lessons, but gm is NOT G.
The low value anode resistors in AVO's, Mullards, etc. are NOT the true Anode Load as specified by valve manufacturers. Those diddy resistors a just for (how shall I put it) bleeding off a few mV to the meter & its bridge circuitry. The meters only being able to take a few tens or hundreds of microamps.
Incidentally, someone in another thread post has just recently mentioned that VCM163's have a 14KHz signal put on to the grid for gm measurement purposes ? ? ?

Regards, David

[Ed_Dinning]

Great piece of work, but may need a little refinement to work across a range of valves. I look forward to the finished version
My HSVT is due an overhaul, so I will probably add this feature at the same time.

Thanks, Ed

[David Simpson]

Right enough, Electronpusher0, I'll echo Ed's encouragement to forge ahead & develop gm circuitry.
I've just remembered that my Post27 of the Thread "Mullard HSVT 7600/4 - started on 27th Feb 2016, proposes a method that might help you. From my experiences back then, there are a number of really experienced & helpful HSVT enthusiasts out there. Thanks again guys.

Regards, David

PS How about a Christian name ? We're all just Dave's, Ed's, Frank's, etc out here

[Electronpusher0]

Thanks for the encouragement gentlemen. The names Peter.

I have started the main project, starting with the mods to the tester.

To be able to test the gm on the maximum number of valve types a couple of mods have to be made to the tester itself. These are fairly minor and easily reversible.
The first involves a resistor, R89 (100R), which is connected in the Anode circuit upstream of the Gate switches that select the valve pins.
This needs to be changed to a tighter tolerance component and a couple of wires attached, one to each end. The ac mV meter will be attached to these wires. It should be borne in mind that these will have HT volts on then during testing so the insulation has be chosen appropriately. I used double insulated 2 core mains wire for this. I also put a ferrite on each wire. I also used 4 resistors in series/ parallel to give the same value but increase the power and voltage rating, not needed in normal conditions but possibly under fault conditions. The original resistor was a 4W component.
The second mod is to cut a link on the main selector switch and replace it with a 10K resistor, a co-ax cable is then attached to the switch to inject the 1kHz signal.
The switch segment is SAda (/4 version) / SW3 (/3 version).
In position 6, (marked 8 on the schematic but equates to position 6 as you rotate the switch), this segment connects the grid bias volts to the grid, it is upstream of the gate switches that select the valve pins. This connection is only used during emission testing and the 10K resistor has no effect on this test.
I attach an annotated schematic, the areas for modification are circled in red.

The link between the tags marked 6 and 8 on the schematic (position 4 and 6) is cut and the 10K resistor connected in its place. In my tester there was a green wire attached to tag 8, this had to be moved to tag 6
The Screen of the co-ax is connected to position 2 (marked 4 on schematic) of the same switch segment which is 0V.
During testing the Cathode of the valve is held at 0V except when push button 3 (PB3) is pressed.
I added a small ferrite to the inner of the co-ax to minimise the possibility of parasitic oscillation

Additionally the mains needs to be brought out to provide power to the meter and oscillator, the switched Live can be picked up from the fuse holder, the switched Neutral can be picked up from the wiper of the mains selector switch, SW12 and the Earth from a convenient point on the chassis. Of course you could also provide a separate mains supply to the external circuitry but then both units would have to be switched on/off.

I attach some photos to help.

In my tester I have removed the front drop down flap by unbolting the hinges, (I have put the flap aside carefully in case I need to put it back to original).
I plan to mount a sloping front box in place of the drop down panel using lift off hinges and a single screw using the existing captive nut used to hold the flap. This allows the box to be put aside to access the mains selector and presets if needed.
I have a right angled chassis made to replace the door and mount the box.

Now to construct the oscillator.

Peter

[Electronpusher0]

I forgot to add that the selector switch needs to be removed from the front panel to access the correct segment.
I attach updated photos with added notes.

Peter

[David Simpson]

Peter, I hope there are some constipated mathematicians out there - who can work it out with a pencil - - in regards to the known formulas for valves.
As far as I can work out, if you leave out the manufacturer's recommended Anode Load RL(usually several Kohms) from the equations, then a hundred ohms or two is insignificant, and your left with gm = approx. the square root of Gain G - your signal gain. Maybe somehow the AC sinewave operating circuitry & your modern DMM, are performing some sort of calculus action. Because delta Ia/delta vg(gm) is a dy/dx function,(i.e. the overall graph is exponential = an x squared function), and the output signal voltage divided by the input signal isn't - its just simple arithmetic.
Come on - you Mullard HSVT buffs - get involved.

Regards, David

[PJL]

gm is the increase in anode current for a small increase in grid voltage normally quoted in mA/V. It is not a constant as it is dependent on the valve operating conditions (if the valve is cut-off gm=0) so it is generally quoted at recommended operating conditions.

The Mullard tester uses a regulated HT and grid supply set for typical operating conditions with minimal anode load so the proposed solution here is measuring gm pretty accurately. Voltage gain is a function of gm, Ra and the anode load but we are measuring the current (voltage across a small value resistor) and not voltage. Ra is another factor that is not a constant as it is the change in anode current for a small change in anode voltage.

I have never sat down to look in detail at how the AVO AC testers work but I assume the results largely reflect the performance at the peak of the AC signal.

[Electronpusher0]

I took the method of measuring gm from that used in the Sussex, I also consulted Mike McCarty's document "Gm vs mu"
vrps.org/documents/Gm%20vs%20mu.pdf
He states that
"Gm is the ratio between the plate current change and the grid to cathode voltage change with plate to cathode voltage held constant. If we call Ip the plate current, and Vg the grid voltage, then
Gm = dIp / dVg
where you may read the "d" as "small change in". Another way to read it is as "The small AC plate current resulting from a small AC grid voltage, with the plate voltage held constant, or short circuited for AC" (like by bypassing the tube with a huge capacitor)."

In the MSVT the Anode is held at a constant voltage by a regulated HT supply. The small resistor used to measure current is negligible. The choice of 100mV signal on the grid is to meet the requirement for the AC grid voltage to be small.

Peter

[Dekatron]

There might be a measurable effect from the 100 Ohm resistor when you measure low Ra-valves like the 6080 (ECC230) just like AVO states in their Valve Data Manuals.

You will have to check that to make sure if there is an effect or not.

/Martin

[ms660]

Originally Posted by David Simpson

As far as I can work out, if you leave out the manufacturer's recommended Anode Load RL(usually several Kohms) from the equations, then a hundred ohms or two is insignificant, and your left with gm = approx. the square root of Gain G - your signal gain.

Regards, David

I don't understand the above, can you explain?

As I understand it the voltage drop across the 100 ohm resistor that's due to the change in grid voltage (100mV) will be minimal, even for a valve with a high mutual conductance, and as such, that voltage drop (reduction in anode voltage) in practical terms will have almost no effect on the mutual conductance.

Lawrence.

[David Simpson]

Peter & Guys,
I applaud any development projects which enhance British valve testers & VCM's. I wish more folk would get involved. When I was first given a HSVT to repair a couple of years back, I was initially put off by its sheer size & weight. But once I got stuck in, I was very impressed with its workmanship & performance. Mullard's, like AVO's & Taylor's, are nigh-on of pensionable age now in the 21st century. With TLC, long may they continue to be cherished in the vintage fraternity.
Back to valves - the manufacturers, since the 1920's, have been producing graphs of families of curves for gm & Ia/Va & so on in their data books, and sometimes accompanying the valves in their boxes. All based on their factory DC operating test rigs, therefore obeying the accepted mathematical formulae. Right enough, circuit design folk have used these curves to add on ac signals at various points to examine classes of amplification & research distortion, and so on. Signal gain G can only relate directly to gm when certain circuitry conditions are met. So any designs by keen folk which can do this are most welcome.

Regards, David

[PJL]

For some valves the screen grid is taken off a potential divider without decoupling which will also have a small impact on the gm measurement.

[David Simpson]

Peter, I've attached copies of BBC Technical Instructions from 1939 for their VT/4 DC Test Panel, relating to principals of operation. Back then, and for many years afterwards, they were at the pinnacle of electronic engineering. Their frequent valve testing procedures are 2nd to none.
Hopefully their info is of some help to you.

Regards, David

[David Simpson]

Missed page 3.Sorry. Dave

[Electronpusher0]

Thanks Dave, I will read the instructions with interest.
I have constructed the oscillator and it produces 1080Hz and can be set to 100mV rms. I will draw up the circuit properly and post.

Peter

[David Simpson]

Hello again Peter, well - I've done 10 press-ups for my poor formula transpositions - Right enough, if valve book's recommended anode load Rl is removed from the equations, then gm = gain G, not the sq.root of G. But only for basic testing operations. In circuit reality, the ratio of ra to Rl can be significantly different. Sometimes 5 : 1, 10 : 1, even 20 : 1. i.e in valve data books some valves are quoted as having an int. resistance ra of 50Kohms & a desired anode load Rl of 5Kohms, and so on.
Another thought, care will need to be taken with high-slope valves, 807's, KT66's, and so on, - to avoid any parasitic oscillations due to your additional circuitry. Using some anti-parasitic ferrite beads on your leads might be a good idea, just in case.
You've certainly raised a very interesting thread. Hopefully the HSVT buffs out there will take it on board.

Regards, David

[Electronpusher0]

Hello David,
I have read the BBC technical instructions, quite difficult as the copy quality is not the best, but it is very interesting.
They define gm as "the change in Anode current measured in milliamperes per volt change of grid potential, ie. gm=dIa/dEg"
They also state that a simple method of measuring gm is to "change the Grid voltage by 1 volt and notice the change of Anode current produced" however they state that this has difficulties as a 1V change is too great for modern valves and because of the difficulty in holding the Anode volts fixed (at DC)
They describe "Another method for measuring mutual conductance based on the application to the grid of a 50c/s voltage". In their circuit they apply a 200mV 50Hz signal to the grid and develop a voltage representing the Anode current across a 100ohm resistor in the Anode circuit while holding the HT fixed (to ac).
In order to measure the ac current they compare the ac voltage across the 100 ohm resistor against a reference 50 Hz signal using something called a "vibrating galvanometer" and nulling the reading on the galvo using a "slide wire", the slide wire is graduated in readings of gm. I guess that in 1939 they did not have an easy way to read mV ac voltages directly.

re your comment about ferrites, I totally agree and have noted the need for ferrites in my post, if you look at the photos posted of how I attached the take off wires you will see ferrites fitted to the leads.

Peter

[David Simpson]

I guess that the Vibrating Galvanometer was the most accurate way of metering null conditions in their balancing circuitry back then. Even AVO were using 444uA fsd meters in their VCM's into the early 50's, before moving on to the delicate 30uA in the Mk 3's, 4's & CT160. Also, more reliable & accurate 270 deg. potentiometers were available by then, used in balancing the bridge networks. Rather than insert a centre (null) reading meter in their VCM's, AVO just used a +ve reading ordinary mc meter. So do Mullards & Taylors. -ve of null being of no use, but its still there.
Certainly, this phenomenon of signal derived gm for testing purposes, is of great interest. Compared to the rigid mathematical delta Ia/ delta Vg.

Regards, David

[ms660]

Another method was to null using audio detection, eg: 'phones or loudspeaker:

http://www.r-type.org/articles/art-118.htm

Here's the full VT/4 document including full schematic:

http://www.bbceng.info/ti/early/VT_4...st%20Panel.pdf

In the full schematic the 100 ohm resistance referred to concerning fig.1 is composed of the two 50 ohm resistors in series in the anode circuit, also note that the circuit arrangement in the full schematic for the slide wire/galvo/HT feed etc is different than that shown in the explanation circuit, the function and operation remains the same.

Lawrence.