The "Sussex" Homebrew Valve Tester.

[crackle]

Hi everyone
I am still trying desperately to get to the bottom of why my Sussex Valve tester has problems testing EL34 valves.
I have done some more tests today in the hope that somebody can come up with the cure to the problem.

I tried a 150k resistor in the grid circuit, this made no difference to the readings of scope.

I then checked the heater volts on no load (7v) and under load EL34, (6.46v) I then used a resistor to reduce the voltage to 6.2v and the gm went up from 3.2 to 4. and A1 current from 77mA to 72mA.

This got me thinking and I did some more experiments with the screen volts reducing this to 175 to 200 increases the gm to about 6
Increase the grid –volts to 16.4 and the gm goes from 2.1 to 8.2 the A1 current reduces from 86mA to 54mA.

I took some photos of the scope during these and a few other tests.

The scope was connected to anode and cathode using a x10 probe, the scope settings should be visible in photo.

Tests on a fairly new EL34 (Groove Tube)
Photo 1
Grid set –13.5 anode current 86mA GM 2.1

Photo 2
Grid set –16.4 anode current 55mA GM 8.2(optimum, any further increase or decrease in grid reduced GM)

Following tests using a resistor as the anode load.

Photo 3 same as photo 3, but changed the time base
6kohms anode current 42mA

Photo 4
3kohms anode current 81mA

Photo 5 same as photo 5, except I had turned function to leakage, LED was now glowing.
3kohms anode current 0

Where the hell does that sawtooth waveform get generated..
thanks

Mike

[crackle]

Originally Posted by mole42uk

Looking at your 'scope images, I'm reminded of some similarity between those and my Sussex when I was having similar problems with EL34 valves.

The symptoms are similar, the gm reading is always low and variable, the Anode current tends to vary a lot but is always much lower than it should be. In fact, the anode supply voltage will be seen to vary by a hundred volts or so in a cyclic manner. In my experience this was caused by spurious oscillation in the EL34 valve, cured by soldering a 150k resistor to pin 5 of the octal valveholder and connecting the grid wire to that.

I suggested, a few posts ago, that you might try putting your hand around an EL34 under test and see what, if any, difference the hand makes to the readings on the Ia and gm meters. I might have missed the answer, but that was, for me, the clue to the final solution of this very perplexing problem! Caution, though, because if the valve heater has been switched on for a while, the valve will be uncomfortably hot to touch.

Neither of these suggestions made any difference to the way the EL34 performed in the tester.
It would seem that a few have suffered similar problems to me, and some are still waiting for a resolution.
The next thing I am going to try is to buy some IFR830 mosfets as alternatives to the BUZ80, as suggested by topcap in an earlier post.
Mike

[mole42uk]

I'm sorry that the hand didn't make any difference!

Can you put a DVM or a 'scope on both the anode and screen supplies when testing an EL34?

I found that the screen supply was stable, but the anode voltage was hunting up and down with weird 'scope traces. I thought the current limit cut-off was activating until I reduced the current sense resistor to allow 120mA anode current. That made no difference.

[crackle]

Originally Posted by mole42uk

I'm sorry that the hand didn't make any difference!

Apart from making me move my hand by fast.

Previous tests have shown that with an EL34 the voltage on both the anode and screen was stable at the desired setting whilst testing.
When testing the anode current rises slowly and the GM goes quickly up to 6 or 8 then drops back to 2 or 3 as the current settles down to 70 to 80 mA.

However I have experienced what you describe as "hunting" with a couple of EL84 valves.
When switched to test, the anode current and GM built up to a reasonable setting and then after a minute or so the GM went up to over 24 and the current was going up and down. I took this a the valves were faulty, as they did look quite "tired". Newer looking EL84 valves seem to test OK, with a GM in the region of 5 to 8 mA/V
I have wondered if the slightly high voltage on the heaters (6.6v) was causing the valve to get too excited.

[mole42uk]

For the EL34 can I just check that you are setting anode and screen voltages to 250, and the grid bias to -13.5v?

I would then expect to see Ia of about 75mA and gm of 11 according to the AVO tables. Most of the new valves I've tested, ranging from new Mullard to some less reputable ones, have managed a gm of at least 11 with the grid bias at -13.5v

I have never left a valve under test for more than a minute. They all have come up to temperature and stabilised after, say 10 or 15 seconds and I usually make a mental count to five once the Ia has settled before noting the readings.

I am playing around with heater voltages on my Sussex, mainly because I wanted to use DC current-limited heaters to see what difference that made. A different heater voltage doesn't change the slope, it just moves it on the x axis of the chart. I would expect your noted gm to change but not by the amounts you are seeing.

[vidjoman]

I can't give any answers to the problems experienced by crackle. There is no definitive way to construct the tester.

My Sussex is only part constructed - but I have read every post and comment and realise that every Sussex will be built differently. Different layout - different wiring - use of ferrite beads or not - use of screened wire - proximity to other components etc., etc.
It will be very difficult to give any definitive answers to any questions raised as there are so many variables.

I'm sure we all know about hum problems in audio amplifiers due to different earthing arrangements - whether the chassis is used as an earth path - whether all earthy points go to a single point etc.

No doubt everyone will try to help but I can see the same problem with different outcomes due to the different construction techniques used.

High gain output valves seem to be the biggest problem at the moment, yet some people have theirs tester working well. Perhaps they can explain about the way theirs has been constructed - types of wire, layout, etc. and that may give pointers to the problems experienced by crackle.

[crackle]

Originally Posted by mole42uk

For the EL34 can I just check that you are setting anode and screen voltages to 250, and the grid bias to -13.5v?

I would then expect to see Ia of about 75mA and gm of 11 according to the AVO tables. Most of the new valves I've tested, ranging from new Mullard to some less reputable ones, have managed a gm of at least 11 with the grid bias at -13.5v

I agree, that is what the book says, and it is how I had been testing them, but as per my last few posts I have just recently been experimenting with different values for bias, heater, anode and screen volts. It all points to if the anode current is reduced the GM goes up.
What is the reason for this. Is it just self distortion or oscillation within the valve, or is it the mosfet regulated supply causing a "dirty" HT when they are passing more current or reaching their current regulation limit.

I would love to understand how the semiconductors in the circuit design worked and where the sawtooth waveform is generated, how does the semiconductor function in the "pre-regulation" circuit, could this generate a sawtooth signal.?
What is the pulse waveform seen on the traces taken from the anode when testing an EL34. Could these cause high gain valves to not perform to spec.

If you go back to post 1081 you will see in photo 1 that there is multiple sine waves at I believe 50 Hz, yet in photo 4, although this is approximately the same current (in a resistor), the waveform is entirely different.

Thanks
Mike

[mole42uk]

Originally Posted by crackle

I would love to understand how the semiconductors in the circuit design worked and where the sawtooth waveform is generated, how does the semiconductor function in the "pre-regulation" circuit, could this generate a sawtooth signal.?

The MPSA92 in the 'pre-regulator' circuit is in fact the reference voltage regulator. It provides a very accurate reference for the divider chains that are switched by the anode and screen voltage switches. The switch-selected voltage is applied to the gates of the IRF830 power MosFETs to set the output voltage for the anode or screen.

There are two types of current limiting. One is the maximum current that the transformer can supply. The other is provided by the MPSA42 transistors on the gates. Their base voltage is held below the turn-on point by the current sensing resistors, R6 and R10. If the current across those resistors gets high enough, the voltage drop will turn on the MPSA42 which will turn off the gate of the MosFET.

I can see a situation where the current limit could cycle - the valve draws too much current, so the limiter steps in, then the anode voltage reduces so the valve draws less current, so the limiter switches off again, and so on. This could produce the sawtooth you are seeing, but because you don't see it when using a resistor as a load, the sawtooth has to be generated by some combination of the valve and the power supply. Maybe try a small capacitor across the current sense resistor to damp it?

[crackle]

Hi
When you say small are you thinking 10uf electrolytic or .01uf poly. I suppose the larger the better the damping and less attenuation of any radiation of unwanted signals.
I guess nearly all the anode current is going through this resistor and also all around the wiring to the electrode switch. Which is close to the same bundle as the rest of the electrodes. Perhaps this could be a good place to look to decouple R6 and separate out the wiring more.
The grid wire is screened and wires direct from the relay to the grid electrode socket on my patch panel.

Mike

[crackle]

The new mosfet devices arrived today, I fitted them and it made no difference.
I also tried the capacitor idea on R6 I used a 0.1uf and that made no difference.
I checked the wiring to the electrode switch and the current limit switches and it is not bundled with the HT or screen.
Mike

[crackle]

Hi Guys
Thanks very much for your help, I have now also asked for help on the Vrat forum to see if they have any fresh ideas. If I find a solution I will post it back here.
I will pop back daily in case any new ideas turns up here.

Thanks
Mike

[Top Cap]

Hi Mike, looking at the photos of your Sussex in posts 1032 & 1033, I am thinking that your problem may be due to the layout. The leads to the Mosfets look very long, well as best as I can see on the photos, it suggests to me that this could contribute to oscillation problems. It could be a combination of this and the veroboard layout too, lots of capacitance between strips, it reminds me of my younger days when I tried to make a superhet radio using Veroboard. I spent hours trying to stop the I.F. strip oscillating and eventually gave up. You may have to do some careful lead dressing or suppression using ferrites or something. I once had oscillation with an EL34 and I wrapped a couple of turns of 24 Gauge tinned copper wire around the envelope, not too tight so I could slide the loop up and down the valve glass envelope. I attached to the loop a length of wire, say 2 to 3 feet, this acted as a kind of antenna but its effect was to provide a kind of loading to the oscillation which then ceased with adjustment of the loops position. Sounds crazy but that worked for me but this was pure instability around the EL34. In the case of your Sussex, I would be prepared to think that the problem you have could be a bit of both, i.e. EL34 related plus wiring to the Mosfet's. Because of the unique layout of your Sussex, it makes it very difficult to analyse exactly where the problem could be. My only advice that I can offer is to perhaps treat this problem like one would in solving an interference issue. Try to rule out certain parts of the circuit by maybe re-running the wiring differently or adding suppression devices in the hope of finding the method by which the oscillation is able to manifest itself.

[crackle]

hi Top Cap
Some good points there.
Regarding the veroboard layout for the Pre-regulation and current limit boards, I removed alternate strips as I thought they were too close for the voltages involved.
I will be working on the tester tomorrow trying out a few ideas given to me on another forum. I will check back here daily, and if anything improves I will let you know on here.
Thanks
Mike

[mole42uk]

Thanks to TopCap, I revisited the photos of your Sussex and have to emphasise his comment that the leads to the MosFETs are way too long! As a first step, make those as short as you can. The gate of a MosFET is very high impedance and can pick up all sorts of stray signals from attached lengths of wire (think aerial) especially when you have an unstable beast like an EL34 running at it's optimum setting to radiate spurious oscillation at surprisingly high power!

The best place for the MosFETs is right on the PCB (I know you don't have one, but the principle is the same) and they don't need a heatsink of anything like the size you have - those devices are dissipating only a few watts most of the time. Mine have a heatsink about a quarter the size of yours, and the temperature probe I used on that showed only a few degrees above ambient when I was running a 6V6 under maximum current for a while!

Richard

[crackle]

Hi Richard and Les
I have shortened the mosfet leads by 6 inches, and this seems to have cleaned up the scope trace a little, but this also makes it very difficult to remove the circuit board for maintenance. So I am going to look to see if there is a way to mount the circuit board on the heat sink, or mount both on a paxolin board, then I can reduce the leads to about an inch..
When I was planning the construction of the Sussex I read that some had installed cooling fans, so I assumed if I used a large heat sink it would be sufficient with out a fan. The one I am using is out of a Dell PC cpu. the outer 2 fins were bent down and round to form feet to mount it on.

Thanks
Mike

[mole42uk]

I don't have a fan either.

[mole42uk]

I'm interested that the guy in Vrat seems to be re-designing the power supply.

While I accept that there may well be some useful improvements to be made to the Sussex power supply, it is worth bearing in mind that there are several Sussex testers in the world, working perfectly with the standard power supply. In other words, the design as it currently stands works well enough.

It is my experience of many years of electronics development work, that if a device doesn't work in the same way as the prototype did, it is preferable to find out why it doesn't before committing to a re-design. If you change something fundamental then you really are on your own because no-one else has a piece of equipment the same as yours.

In the present case, you are getting good help from several members of is forum who have already successfully built the Sussex. That, IMHO, will be the easiest route to sorting your Sussex tester.

Sorry if this seems like a rant - it isn't, it's just an observation which I hope will help to get your tester running sweetly.

Richard

[crackle]

Hi Richard
I have just completed isolating the octal valve socket and re-wiring it direct to the screen, anode, cathode, and bias supplies.
The traces on the scope seem less messy to me but still seem to be incorrect.

I have taken 3 more photos,
the 1st shows the anode trace at -13.5v bias.
2nd shows the anode trace at -14v bias
3rd shows the anode trace with -15.5v bias.

If i continue to lower the bias voltage to -20 and beyond then the "spikes" seen in photo 3 just reduce in size and eventually disappear at the same time the GM value reduces also.
So it is almost as if the spikes are what the meter is reading to get its GM value.

The lower the bias the lower the current as expected, but at -15.5 to -16v the GM gives the highest reading of 6.8 (almost believable)

Would it be possible for you or somebody else to test an EL84 valve on their Sussex tester and compare the traces to the ones I have just obtained.

Thanks
Mike

[Omegaman]

Hi Mike,
I've just this afternoon built a 3rd Sussex and it was exhibiting the same problem as yours with the gm reading rising and then dropping back to a low figure and the anode current increasing.
The first two I built did not do this and behaved themselves.

Throwing a handful of ferrite beads into the equation did not help.
Un-looming all my lovely laced wiring loom and separating all the cables (it now looks like an explosion in a spaghetti factory!) did not help either.

BUT!

Removing the protection diode across the 10 ohm gm measurement resistor did. I didn't have this diode in the first 2, but included it as Les had suggested it may help protect the meter. My first one blew up and destroyed the meter when a 6L6 flashed over, but the second was fine.

Have you incorporated the diode in your build?

The readings are now rock steady with an EL34.
It worked fine with everything except an EL34 before removing the diode.

I have not worried about the noisy waveforms in any of my builds as it doesn't seem to affect operation.

Hope this may help you.

[crackle]

Hi Howard
That's very interesting, yes I incorporated the protection diode.
I will remove it tomorrow and try things again.

"(it now looks like an explosion in a spaghetti factory!)"
That's what mine looked like before and after messing around rewiring it.

Thanks for posting.
Mike