Constant Current and Developments

[BusyBee]

I just thought I would slip this little technical short into the system as something which may be of some interest. It came along as a diversion between working on my electrometer and the 1ж37б valve gain control by DC voltage which I have been looking at.

Attached I show three constant current circuits compared. The results (and diagrams) I show here are, pretty scruffy, first stabs but it was a bit of an aside to what I have been doing recently. I think the results show themselves fairly well without much explanation. Obviously the 1ж37б could be replaced by most any valve with negative bias but it does seem to compare well with the, albeit sub-miniature, traditionally constructed pentode 6ж1б. With direct heating, the Russian rod pentodes would be more easily used with negative bias rather than the usual cathode bias, this being difficult to design with these valves. I use these 6ж1б valves, for many tasks. The 2SK3666 Fet is one I bought a pack of as I did not have any general purpose low power Fets and these are such, and have fairly low capacitances. They are junction Fets. I find that I do swop back and forth a little these days with valves and Fets. The constant currents I set were around 40 / 50 uA, the circuits used being in sink mode.

From these and doing some reading up on Fets I decided to put together a source follower and evaluate it a little. Such a circuit is (in basic form) very simple. I made two of these, one with 1MOhm gate resistor and one with 100MOhm. These were a further offshoot of electrometer work as I had, some time back, tested a similar cathode follower using a 6BS7 valve and achieved good results. After testing the circuits for linearity I thought a good use was on checking grid current of valves which is something highly relevant to work in the high ohmic region with valves. The first, fairly rough, test being that of a 6ж1б. The results I show were achieved using the circuit with 100MOhm gate resistor. This resistor also functions as valve grid resistor to develop grid voltage from the grid current. The results obtained seem to agree well with what I would expect and from what I read in an article on grid current on the National Valve Museum website. In simple circuit form the Fet output at zero input is not zero (apparently would be with matched Fets and resistors) but is stable, only tens of mV, and so the calculations are easily adjusted. Variation in supply voltage varied the zero point, but on test, a well regulated supply was used, plus and minus voltages such that the Fet limits were not exceeded. Obviously there are a number of variables which would have to be clarified, as far as practical, for any sensible determination of grid current in relation to intention to utilise a valve in a particular application (bit of a getout clause but electrometer use has sharpened my mind here lol).

All in all perhaps some useful ideas for further investigation.

Attached:

Constant current circuits circuit diagrams
Constant current circuits results graphs
Fet source follower circuit
Fet source follower negative input voltage graph
Fet source follower positive input voltage graph
6ж1б grid current result graph

(all graphs produced using Gnuplot)


Tracy

   
  CC_Plot.pdf (Size: 17.29 KB / Downloads: 4)    
  Fet_SF_Neg.pdf (Size: 14.23 KB / Downloads: 5)
  Fet_SF_Pos.pdf (Size: 14.34 KB / Downloads: 5)
  6j1b_Grid_Current.pdf (Size: 15.95 KB / Downloads: 5)

[ppppenguin]

Interesting work. For electrometer use I wonder if you would get better DC stability with a long tailed pair, with the CCS in the tail. If you need any 6BS7 I think I have a few in the stores. I think it's electrically similar to the 6BR7 but with the grid brought out to a top cap for ultra low leakage.

[Diabolical Artificer]

You've put a lot of work in there Tracey, nice work. Re 6BS7's I had 5 NOS in a box and tested them for grid leakage when somebody was after one for the same application. I used I think a 1m as Rg1 as a sense resistor, ran the the valve with fixed bias, probably something like -3v on the grid, out of the five I got two that tested as per the datasheet.

Re CCS I've always used a zener (6v2) to fix the base, but in most applications I've had plenty of voltage to play with, not so good if you only have 30v or so. Making Re partly adjustable helps too.

Andy.

[BusyBee]

Thank you for the offer of 6BS7's Geoffrey. I still have my original though and do have three others (I think I bought four at the time) so will be Ok. As implied by Andy's experiences, these valves were originally specially selected for electrometer use. The datasheet gives an example circuit. I have only used the one of mine. which I used in a cathode follower, testing to check voltages across a resistance box made up using ten 10MOhm resistors. The figures came out as may be expected rather than the incorrect values which were given when using my (10MOhm) digital meter. I have had somewhat a fixation on linearity which makes a, somewhat more intrinsically linear, cathode follower more appealing rather than focussing more on stability and reproducibility. In tests I have used one circuit with a Russian  6н16б double triode with a 6ж1б CC tail which was useful experience.

The unit I am working on, which is complete, but needing some fettling (as Andy would say), has working front end undergoing tests has two channels, using an эм6 double / split tetrode electrometer valve feeding an IN126 instrumentation amplifier on one channel and an IN125 instrumentation amplifier (with direct input) on the other. It also provides voltage references. The unit has an Arduino micro which provides switching and timed operation of electrometer valve heater and HT for stability at the moment. Future intention is to link readings to a PC via processed via a 16 bit A to D (already fitted). On both channels output is available directly post instrumentation amplifier before any processing for the micro / PC. I can see that some of the work may well be useful here but I wasn't intending to go through the complete system as it has been and will be very involved plus also not of too much relevance to most vintage radio etc. I am happy if anyone has queries or wishes to liaise if they are doing similar things as I know information / experience is difficult to get (Keithley handbooks and site are very useful). Like any extreme measurement some issues are very expensive - valves are a minor cost; high value precision resistors dictate a bank loan lol - I have accumulated some NOS which should be sufficient for my needs though. These will be less precise but are making me think about ways of checking them, which is good.

Andy - I also use a voltage reference for most of my CSS, usually using a TS431 (~2.5V) reference as I bought a bag (50) some while ago and use them often. The circuit would otherwise probably be similar to yours. Mostly I am using my 60V supply but electrometer valves usually often anode voltages below 10V (at least mine [эм6,эм7]) so these are more transistor levels. It did make it easier with interfacing to the instrumentation amplifier (on +/- 12V). I most always use semiconductor CCS but (as indicated by my results above) valves can be more effective in simple form if voltages are high and currents low. My take on the Fet CCS source follower was that a fixed resistor would besimpler and less noisy / more stable but use of a variable source resistor should perhaps enable output zero setting (the circuit boad is all surface mount and not cramped at one and a half by one  and a half inches).

Tracy

[Diabolical Artificer]

Interesting stuff, I also like marrying tranny circuits and valves but haven't got the patience or nous to program an Arduino. There's something about the anti logic used by computors that winds me up, much like French classes did when I was 11; "Why is the monkey dans la abre"..... "what!!!" "la singe et votre petit small furry shorts???" Sorry, I digress, away with the faeries again, morning coffee kicking in.

If you don't fancy writing a long thread some pics would be nice Tracey so we can ogle your work and make sure the resistors are all lined up nicely : )

Andy.

[BusyBee]

Hi Andy! I'll get a few photos sorted. This thread might be the best place for them. There may be a few technical bits worth sharing from time to time anyway. I spend quite a bit of time testing this and that and plotting graphs. Some of my ideas are very abstract but they do occasionally work lol

Tracy

[BusyBee]

I have now pulled a few pictures together of my progress so far:

The first three show the electrometer valve and some of the first electronic units – The power input board and Arduino. The home-made screening for the electrometer valve is from brass sheet with an Nickel Chrome foil screen (both electrically connected). Hopefully to minimise electric and magnetic field interference.  This screen is designed for direct soldering to the valve base thence to the cathode. It sits away from the valve tube glass.

Number four shows further circuitry – Anode regulator (differential with HT level and differential adjustment for the anodes – a bit overkill but does work). INA125 and INA126 instrumentation amplifier boards.

Number five shows the electrometer tube heater supply regulator. Following an earlier work with low voltage / current supplies I have worked on a constant current supply. More tricky with the 4.5V 75mA but I thought useful as the constant current, rather than voltage, supply will eliminate switch on surge. Stability is the thing with an electrometer and they can take a considerable time to settle if things are less than perfect (hence timed heater / HT switch on).

Number six is the further assembled INA125 and INA126 boards with their op amp offset circuits (for plus and minus ranging for the Arduino A to D). and the protection for Arduino input.

Number seven shows the unit cases. Originally there was only one PSU module but I thought the transformers were at their limit so now it has one case containing the mains transformer plus discrete and electronic smoothing and a second containing the main power regulators; providing outputs of -12V, +12V, +9V and +5V. The power units were separated from the main unit to limit the possibilities of heat and magnetic field interference.

The last three photographs show the completed unit in it's current state. As yet no labels have been produced.

At present I have only done basic tests but the valve circuit seems stable in that there is little drift after switch on settling. A return after zero setting within the hour finds the setting still within 1mV but, as I say, early days. This is measured at the instrumentation amplifier output (The INA125 has a basic gain of 4 and the INA126 of 5). One thing that is a little more troubling is that there is a noise / fluctuation resulting in a random wandering of output of two or three mV. This does not seem time dependant. I have been through the system shutting off various possibilities of interference but may have found the area when I tried linking the g2a input (topcap) with ground directly. This did seem to affect it and shows itself slightly differently now so it is probably an input component,  connection or static error. Further work in other directions (Fet circuit) has shown that a redesign of the input may be good though.

Further down the chain I realised that I had an A to D test program running so I have verified that there is sensible data being transferred to the PC. It's too early for accuracy checks though. I do plan to change the programming to increase the heater to HT switch on time in line with other recommendations / references from 20 seconds to 1 minute (see 6BS7 datasheet, and elsewhere).

The last photograph is one of the two versions of Fet source follower I have made. One has 1MOhm and the other 100MOhm gate resistors. One interesting observation I made earlier in the week was when trying a rough test on a 6BS7 (my previous test example) for grid current. This was noted to be in the picoamp region, but the interesting observation was with regard to light sensitivity. In a reasonably lit small room I switched on the room light (pretty much straight above and consists of a high wattage (maybe 120W equivalent) bulb. This was like operating a switch with grid current instantly changing by 20mV. With my 100MOhm resistor input this equates to 200picoamp. Maybe not a great concern to most uses but a big issue for low level DC voltage or current. With low level measurement obviously the valve needs to be in the dark.

I had another interesting experience yesterday too. I had seen on the internet that someone had used a neon for gate resistor with a Fet for field detection. Their design used a meter in the drain circuit with variable source bias. I built one of my source follower circuits using this idea. Using a 2” square piece of double sided copper board on the input I tested using a voltage source and various input resistors. I was very pleased to find that with just a lash-up I could get figures in the ball park for resistors of 1GOhm, 5GOhm, 10GOhm and 60GOhm, roughly calculation the input resistance of the Fet circuit to around 100GOhm. Obviously not accurate at all but a good step in the right direction and good experience. With a reminder to earlier things mentioned I had put some heat shrink onto the neon as I was aware it was light sensitive. Interestingly there was still leakage as the room light switch test did noticeably affect things. Note 1: Beware of light interference lol. The stability of this circuit also got me things about my valve electrometer input. I am very pleased with these little Fet circuits as they are stable and good for exploration.

[BusyBee]

The two Fet circuits mentioned in the last post. Unless I am missing things, I hit the limit of attachments:

Tracy

[Murphyv310]

Some really nice work there Tracy.

[BusyBee]

Thank you Trevor