I keep opening my posts with something like “this is an interesting _____” and so it is. An instrument that is designed to measure from 10 μV to 1 V and 10 pA to 3 mA has little if any practical value to me, yet it is of interest. I had never given the issue of measuring such low electrical values any real thought and the design of this instrument is completely new to me.
I must preface the following attempt to describe the operation of this instrument by saying (as I have said before) that I am an M.E. and asked my friend Kenneth Kuhn http://www.kennethkuhn.com/hpmuseum/ to straighten me out which he kindly did and any errors are due to my failure to correctly transcribe his explanation:
HP in their manual have this to say, “If the 425A is considered a black box with two input and two output terminals, it is a DC amplifier”. If the amplifier were DC coupled, any drift or DC offsets in the amplifier would swamp the extremely small signal. The system uses lamps acting on photoresistors through a chopper wheel to modulate the incoming DC signal so as to present an AC signal to the AC coupled amplifier, the signal is then demodulated to restore the DC.
The output of the modulator is a rough squarewave having an amplitude proportional to the applied DC signal. The chop or modulation frequency is 5/6 of the line frequency to ensure that line noise is not passed, (50 Hz is not monotonically related to 60Hz.) The resulting AC signal is amplified by an amplifier that has a 50 Hz reject T filter in the negative feedback path thereby causing it to respond extremely strongly to the 50 Hz signal and not much else. This greatly amplified signal is next demodulated by a second light chopper that is synchronized to the modulator by virtue of using the diametrically opposite side of the same chopper wheel. A 50 Hz reject T filter removes any residual modulation resulting in a clean DC signal proportional to the input, that is applied via a cathode follower to the meter.
Kenneth also pointed out that that specialized op. amps having zero dc offset, known as commutating auto-zero amps, are available. What used to take a motor and chopper wheel is now packed in an integrated circuit. I suppose that is progress but not as much fun!
The modulator lamps and the light pipes to the photocells are visible above.
These instruments featured individually photo-calibrated meters and the original tag is still with this one:
I also have the proper probe for it, shown with the unit in the top picture.
The unit arrived in excellent condition. The meter glass was loose due to the retaining clips having fallen out. Actually getting the bezel off was a little tricky since it had been secured using brass break-off head screws, presumably to protect the special calibration. I managed to grasp each one using hemostats and got them all out without any unwanted adventure. I cleaned the glass (but NOT the meter face) and re-tensioned each clip and put a little glue on each also. The chopper bulbs were not burning consistently due to corrosion on the pins, another job for Deoxit. It then worked and applying a DC source set to plus and then minus 0.5000 on my HP 3468A, the meter showed precisely + and – 0.5. My next job may be to construct a precision DC voltage source using something like a GR decade box as a precision divider. It will be powered by batteries for quietness, if, if I do the job!
There are a couple of unusual considerations when using a DC meter of such high sensitivity:
Galvanic and thermoelectric effect between the probe and the terminals of interest. (I have in the past been puzzled by a DC shift on a HP 122A scope that turned out to be a galvanic issue. That one took some lateral thinking on my part to identify.) Also noted in the manual is the possibility of seeing triboelectric effects due to flexing of coaxial cables.
Here is the other side showing the rather daunting range switch:
Here is the underside also showing the range switch:
This is an interesting unit being capable of measuring the frequency response of an audio system over a range of 20Hz to 20kHz. Back in those days, 20kHz was considered enough? My ears give out at 6kHz and I can hear music very well so yes, 20kHz was enough. Maybe we have evolved so fast that our hearing has caught up with that of dogs in this peculiar age of man where technology, fast taking on a life of its own, is getting away from us and the average or even above average human cannot keep up, much less appreciate the ramifications of technological run-away. OK where was I? Oh yes, this ancient audio frequency test unit.
It contains a Wien bridge oscillator, stabilized with a lamp exactly like the original oscillator that Dave and Bill produced to start the amazing company we now know as Agilent. It has three decade frequency ranges, 20 Hz to 200Hz, 200 Hz to 2kHz and 2 kHz to 20 kHz. A push-pull power amplifier (using triode connected 6L6s) is included that allows for an output power up to 5W into loads of 50, 200, 600 and 5000 ohms. Load matching is accomplished by a switched impedance matching transformer. I find the output power useful for in my messing around, sometimes a power signal is handy without having to resort to hooking up an amp. The distortion is given as less than 1% at frequencies above 30Hz. Again, somewhere down the line when transistor amps having large amounts of negative feedback were introduced, our hearing suddenly evolved and we could hear distortion levels of 0.01% or less! Amazing really, how marketing types drive human evolution….. Or not.
There is also an accurate output attenuator, referenced to 0 dBm (1 mW into 600 ohms = 0.7745 Vrms). It has a range from 0 dB to -110 dB in 1 dB steps. A 600 ohm load that can be switched in is provided, allowing accurate setting of the output in dBm. The output is available as balanced or single ended.
To measure the frequency response of the device under test there a VTVM calibrated in dBm from -5 to 8 dBm and in volts from 0 to 2 Vrms is provided. This meter includes an accurate stepped attenuator that extends the meter range to 48dBm and 200 Vrms in 5 dB steps.
Here is the top of the chassis: The push-pull output transformer is directly below the power transformer and has very fine laminations. The tuning condenser is self-evident. To the right is the top of the impedance matching transformer and below that, the switched output attenuator. If you look closely, you can also see the oscillator stabilization lamp that is located between the left-most electrolytic and the 6J7 tube (that has a top grid connection).
It was working as received, I did the usual Deoxit job on the switches and tube pins and lubricated the geared tuning drive. The only other thing required was to re-calibrate the output meter and the input VTVM, pre-set pots are provided for both meters.
Another interesting item from HP, capable of measuring true RMS AC voltages over a range of less than 1 mV to 300 V over a frequency range of 10Hz to 4MHz.The accuracy is specified as:
+/- 1% FSD, 50Hz to 500KHz
+/- 2% FSD, 20Hz to 1MHz
+/- 3% FSD, 20Hz to 2MHz
+/- 5% FSD, 10Hz to 4MHz
My unit after repair does meet this specification, it has a large mirror scale meter that is easy to read. I especially like the old-fashioned arched appearance of the meter, almost architectural!
The input impedances are given as 10 MΩ shunted by 15pF on ranges 1 V to 300 V and 25pF on ranges 1 mV to 300 mV. It uses a 4 stage closed loop amplifier and about 51dB of negative feedback to secure good stability and repeatability, the input stage (outside the loop) is a cathode follower. All the amplifier tubes including the CF are 6CB6 pentodes. The power supply is, of course, regulated. The regulator is a typical series triode with a pentode error amplifier and 85 V gas tube reference. The design incorporates feed-forward to the screen of the error amplifier to secure good line-regulation.
All the manuals I have found are for units having a serial number prefixed with 313 while mine is unit 5641. The overall topology is the same, however there are many significant detail differences. I do not know which came first, I suspect that my unit predates the 313 prefix. So to get things started, here is a picture showing it working:
The main difference is that the attenuator resistor chain forms the load for the cathode follower while the schematic in the manual shows a separate cathode resistor with the CF being coupled to the attenuator via a 5μF capacitor. The attenuator in my unit has a separate 51nF cap for each range, I cannot see any reason why it could not have been implemented using a single capacitor though. Clearly the front end was the subject of some extended development! Here is the schematic of the attenuator/front end of my unit:
And here is a picture of the attenuator and CF:
The wire boards are the precision attenuation resistors that being in series, are used as the CF load resistor.
As you may expect, I had to replace all the coupling capacitors in the attenuator. The other two capacitors that are critical and required replacement were the full-wave voltage doubler capacitors for the meter circuit. Any leakage will result in an unbalanced circuit and a residual deflection of the meter. Unbalance with the penultimate amplifier stage removed left the meter still showing an indication, replacement of the signal rectifier capacitors cured the issue. Other than that, the reservoir capacitor which is one section of a four section twistlock failed. I found a replacement on Ebay.
The main job was sorting the 6CB6 tubes according to microphony, they all have good transconductance. I did this by simply exchanging tubes with the first stage and tapping the unit to note the meter disturbance. Having found the least microphonic tube, I repeated the procedure with the rest of the tubes, still using the first stage location, until I had them ranked. I then installed them, least sensitive in the CF location, next in the first stage location and so on. At the end, I was rewarded with a unit that would show a perhaps 1/3rd scale flick instead of several wild full scale disturbances. As always, Deoxit is absolutely required on all tube pins and switch contacts. I also wrapped the CF and the first tube tightly with PTFE plumbers tape, just one width wide, I did not cover the whole tube, the heat has to radiate! Further, I similarly wrapped the first stage tube shield with PTFE tape (the first stage has the only shield) and slipped thin tubing over the conical spring (with considerable difficulty I note).
And here is the left side:
This arrived filthy inside, with two broken tubes and a perished cord.
The range is from 3Hz to 100KHz at “better than” +/- 2% accuracy. I found that it is indeed better than +/- 2%. Here it is after cleaning up, re-stuffing the two power supply reservoir electrolytics and replacing all the paper in oil caps:
Here are the original caps. One cap (the flat one) looks as though it has already been replaced but looking at a picture in the manual, it seems that all these caps are original. One of the two series connected power supply electrolytics had dried out completely, I re-stuffed them both, the cardboard sleeves make it easy to do this because they cover the results of cutting the aluminum cans open. Many of the caps had cracked cases and you can see that a chunk of the case had fallen away from one of them. I have never seen caps in this state before (and I have quite a bit of experience), this combined with the perished power cord suggests to me that this unit had spent much of its life in a harsh environment:
Here it is after repair showing the 60Hz self-test:
And here at 15KHz:
There is a resistor board that is used to correct each range thereby avoiding the need to use precision capacitors. I spent quite a bit of time working with this using a Tek 180A temperature stabilised crystal time marker as the frequency source:
I think it is worth describing the circuit action of this instrument, as I am not an E.E. I found it both novel and fascinating:
Refer to the block diagram above and the oscilloscope traces below:
The signal is amplified and applied to a conventional Schmitt trigger, the output of which is rectified and differentiated to apply a negative pulse (t1) to the A side of a multivibrator. This turns the A side of the MV off causing a positive pulse at the A plate that is applied to Phantastron run-down circuit, starting a linear run-down. At the start of the run down, the screen grid of the Phantastron rises and stays positive until the end of the run-down. This positive pulse is applied to the B grid of the MV (the constant current generator in the diagram) and is limited against a reference voltage by a clamping diode resulting in a controlled current pulse. The Phantastron run-down is stable and predictable in rate and duration and at the end, the screen voltage drops*, cutting off the current pulse (t2). This action results in a constant current pulse which is stable in duration. At the same time, degenerative action of the MV cathode resistor combined with the grid diode clamp causes the pulse to be stable in amplitude. One pulse is passed through the meter circuit for each input cycle. Since the meter has a large capacitor across it and the meter itself acts to discharge the capacitor, the more frequently the capacitor is charged, the higher the average charge voltage and the higher the indication on the meter.
The traces below were recorded on a Tek 547 with a 1A4 plug in:
Top, MV A plate.
Second down, Phantastron screen pulse.
Third down, Phantastron run-down.
Bottom, MV B current pulse.
I Believe the Phantastron is one of a number of extremely clever devices invented by Alan Blumlien just before or early in WW2 and so named because what it could do was, at that time, fantastic. Such triggereable timing and logic elements included multivibrators and were essential elements of radar and of course code breaking computers.
*It is fairly easy to see how the run-down commences, when a positive pulse is received at the screen grid, the tube starts to conduct, discharging the timing capacitor. The run-down termination process is not so obvious: The level at which the discharge ceases is determined when the potential between the anode and the suppressor grid becomes so low that the anode cannot draw through the suppressor grid the electrons that have passed the screen grid and so the remaining electrons return to the screen grid causing the potential of the screen grid to suddenly drop and so the tube cuts off ending the discharge.
Here is an interesting device that can be used to check oscilloscope Y rise time and trigger jitter. It came with a bunch of Tektronix type 547 oscilloscope parts.
It will change state extremely rapidly, entering a negative resistance region where as the forward current is increased the current suddenly drops causing a sudden voltage increase at the anode. It can be used to build extremely fast negative transconductance oscillators and in the trigger circuits of fast oscilloscopes. The static forward resistance of the diode is just a few ohms and in this case, this small resistance forms the shunt leg of a potential divider, since the input resistance is 3.48k, there is almost no output until the diode changes state. In this case, the change in state occurs at an input threshold of 32.5V. Here it is driven by a sine wave shown on a type 545A oscilloscope:
Here it is on the type 545A at 200nS/cm driven by the square wave calibrator set at 50V P-P. The comparative slowness of the leading edge of the square wave to that of the tunnel diode is very clear!
The slight forward lean just visible at 200nS/cm on the edge of the tunnel diode state change, maybe 15nS, is mostly due to the rise time of the type 545A Y amplifier which is itself, extremely quick. This is why for many years, tunnel diodes were used to generate the trigger pulse for fast oscilloscopes being at least an order of magnitude faster than a triode or transistor Schmitt trigger. Type 545A pre-dated the use of tunnel diodes for triggering and at high frequencies, the timebase is synchronised, not triggered. The later type 547 used tunnel diode triggering allowing the timebase to be triggered at any speed the Y amp could handle (in the case of type 547, greater than 50MHz). Here is the same set up viewed on a type 547 oscilloscope:
Interestingly, the trace on the type 547 is less clear than that of the older type 545A, I think the CRT in this type 547 is getting near the end of its life. Also, disappointingly, I could not get a jitter free trace at a speed higher than 200nS/cm. The type 545A is proving its mettle!
In an attempt to do better, I tried on a type 475 and the result was worse, I could not obtain a jitter free trace above 1µS/cm!
This type 475 is a real disappointment being a 200MHz scope. The triggering is weak and the trace definition poor. I cleaned all the switch contacts and that yielded an improvement, the trace went from very fuzzy to fuzzy and the Y stability improved. However, when I want to see something tricky, the 547, 545A or even 535A and 535 types perform better. If you are expert in restoring the performance of a type 475, or know somebody who is, please leave a comment so that I can contact you!
The 10MHz 535 was introduced in 1954 with the brown box cabinet, updated in 1956 to the blue clam shell cabinet. In 1959, type 535A was introduced having better controls and a 15MHz vertical amplifier versus the 11MHz 535. The A timebase provided 24 settings, the speed range being 5S/cm to 100nS/cm while the B timebase had 18 settings, 1S/cm to 2µS/cm. The timebase ranges of type 545A are the same as those for type 535A. On both models, a 5X multiplier was provided to increase the maximum speed of the A timebase to 20nS/cm, quite fast.
The picture above shows the 545A on the left fitted with a 1A1 50MHz dual trace plug in and the 535A on the right, fitted with a 1A2 50MHz dual trace plug in.
The maximum sensitivity of the 1A1 is 5mV/cm, the bandwidth at that sensitivity being limited to 28MHz while that of the 1A2 is 50mV/cm, Both units could provide the full 50MHz at 50mV/cm. The scopes are showing a 1MHz squarewave from the fast rise output of a type 106 squarewave generator displayed at 100nS/cm, the overshoot and ripple that I thought was due to delay line issues is actually mostly due to termination issues resulting from trying to drive both scopes at once. Thanks to Bruce Baur for pointing this out. Below shows each scope set up the same but driven separately, Type 535A:
Some ripple is still evident however the rise overshoot has gone.
Type 545A:
A very clean trace.
Fairly soon after the 535 was introduced, the demand for greater bandwidth brought the 30MHz 545 and quoting directly from 545-TekWiki:
“Type 545 was introduced February 7th, 1955 along with the Type 541, and superseded in 1959 by the Type 545A, which was in turn superseded in 1964 by the Type 545B. The difference between the 545A and 545 is the control ergonomics, not any major circuit design changes.
Types 545 and 545A have a six stage differential distributed vertical amplifier made of twelve 6DK6 tubes. The vertical amplifier used in the 545A is also used in the 551 and 555. The 545 uses the 154-098 CRT. With the limited CRT technology available at the time, the higher bandwidth of the 541 and 545 came with a tradeoff. The vertical scale is only 4 divisions, 2 above and 2 below the graticule center line. The lower bandwidth 531 and 535 retained the 6 vertical divisions, as used in most other Tektronix scopes.”
This is the 535A Y amplifier:
Here is the 545A distributed Y amplifier:
All these tubes result in a tube count of 74 plus the CRT, with a CA dual channel plug in this would result in a 90 tube room heater!
Here is the distributed Y amplifier schematic:
The distributed amplifier tubes, V1104 thru V1214 are type 6DK6 while the driver tubes, V1033 and V1043 are type 6DJ8.
(Rant mode on: Audiophiles are known to destroy Tektronix scopes for the 6DJ8s, usually to use in frankly lousy circuits. I hope what I am doing helps to generate awareness of and respect for these ground breaking instruments. However, in these crass times, people seem to think that ownership equates to the right to do what they like rather than something more mature like stewardship. Rant mode off.)
The circuit description in the manual says this of the distributed amplifier:
“The output stage is a 6-section distributed amplifier. The tapped inductors in the transmission line between each grid and each plate, isolate each section from the capacitance of the adjacent sections.
The input signal for each tube is obtained from the grid line which is driven by the cathode followers V1033 and V1043. The amplified signal at each plate, fed to the plate line, becomes an integral part of the wave traveling down the line toward the deflection plates.”
And here it is displaying a 30MHz sine wave at 20nS/cm:
Here is a picture showing the entire Y amp side of the 535A over the 545A:
Since the timebases are nearly identical, I have not provided a picture here, if you want to see more go here.
This 1947 vintage radio resides above my workbench and I use it to listen to BBC radio re-transmitted from the internet using a two-tube AM micro-transmitter designed by Fred Nachbaur (deceased) that he named “The Bean Counter”. When the picture above was taken, it was playing BBC Radio 4 Extra from the Bean Counter.
I had this radio as a kid and around 1998 on a trip back to England (from Greensboro, NC, USA) discovered it in my Dad’s shed. It was almost complete except for the two grid caps but the case was scratched and dull and the works were filthy. Thanks to Ray at The Radio Workshop ( http://www.radio-workshop.co.uk ) for providing me with replacement grid caps that contain the correct 100k and 220R grid resistors. The speaker cone was torn and the pole piece was rusty so most of the work went into repairing the loudspeaker. I did this many years ago and no longer know where the pictures are. The repair entailed carefully separating the cone from the surround and frame. I then repaired the tear using toilet paper and paper paste on the back. This worked quite well! I then cleaned up the pole-piece and put it back together. Much time was then spent fiddling with both the diaphragm centering and the pole-piece centering to get it to speak clearly. It is necessary to ensure clarity at both low and high volumes. At low volume, if there is rubbing, it will suddenly break into action as the volume is raised while at high volume, the off-centre motion I noticed (probably due to non axi-symmetric stiffness resulting from the tear repair) can cause rubbing that is not present at low volumes. I diagnosed this movement by carefully touching around the cone at higher volumes when distortion was audible and I noticed that touching on just one region stopped the distortion. It was a tricky and time consuming process however, I was rewarded with success and the radio now plays clearly at all volumes.
Here is the schematic:
The difference between the AC and DAC models is that the AC model uses a auto-transformer rather than a dropping resistor for the heater chain. The advantage of the DAC model in the era of the radio was that it could run from AC or DC. The advantage of the auto-transformer model is much lower heat dissipation. The back is quite often scorched or missing from the DAC models!
Here is the back:
Here are the beautiful Mullard metalised valves (such splendid devices prefer to be called valves, I think):
While performing IF alignment, I noticed that the gain would change a lot when I tapped on the chassis. After some fiddling, I isolated this issue to one IF slug and sighing a bit, removed and disassembled the faulty can. Sure enough once the offending screw-adjuster was removed from the transformer, bits of broken ferrite fell out. The slug had been molded around the end of the screw and it had crumbled apart leaving the actual slug to slide around inside the coil former. Here it is:
I had some heat-shrink tube that just barely would fit over the slug and knowing that the tube will shrink up to 3:1, I repaired it like this:
The increased diameter just fit into the former and so I had my repair and was free to complete the IF alignment.
The design of “The Bean Counter” AM Micro-transmitter” is published at http://www.dogstar.dantimax.dk/tubestuf/amtx-2.htm
Read on before you click the link though because the information that Fred gave on the tank coil isn’t complete.
While Fred is deceased, I believe that somebody is maintaining his work on the internet, whoever you are, THANK YOU!
Here is my version:
The unit is built on a RadioShack box/chassis and I used a Hammond P-T442 power transformer.
I found Fred’s description of the 350µH tank coil a bit puzzling so I searched for an on-line air-cored inductor design tool and found Martin Meserve’s page at: http://www.k7mem.com/Electronic_Notebook/inductors/coildsgn.html and using his program designed a 350µH tank. I used a standard prescription drug container having an outside diameter of approximately 1.27″ (it is tapered) as a former and was just able to get 156 turns of 28SWG enamelled wire onto the 2.3″ length. Using my GR 1650A impedance bridge, it measured at 340µH, close enough. Here is my coil winder (that I built years ago to wind audio output transformers) with the tank coil in place:
Ex RAE mechanical apprentices may recognise the live centre!
Fred’s data for the tank coil states 88 #29 turns on 1″ diameter and 15 #20 secondary turns. The issue that puzzled me was that 88 #29 turns on 1″ theoretically results in an inductance of around 100µH. Doing the math, neglecting the 6BE6 plate capacitance and stray capacitances, Fred’s value of 350µH and the tuning capacitance of 50pF gives a resonant frequency of 1.2MHz. This is 20% high which is most likely “corrected” by stray capacitances, also the capacitance between the tank and the secondary coils. This was enough evidence for me to work from Fred’s stated inductance value rather than his winding data.
Here is the resulting tank coil and secondary:
I used Fred’s winding ratio of 88:15 turns to come up with 26 turns for the secondary which I wound using 20AWG onto a former cut from a Radioshack copper wire reel. The resulting diameter was a nice tight fit into the tank coil. I placed it into the “earthy” end to minimise shunt capacitance at the plate end.
Here is the assembled transformer:
I put a hole in the base of the container to pull the wires through and mounted the transformer by fitting the lid to the chassis using a screw and simply inserting the coil in the lid, twist-lock fashion. The chassis end of the secondary is connected to the earthy end of the coils (tank, B+ and secondary, ground).
I actually found that with the cover on the unit, the oscillator peaked at 50pF (plus strays including the cover). With regard to the design of the transmitter, I did not deviate much from Fred’s design, the only changes being to
1/ Insert a RF choke I had on hand (450uH) and a 22nF capacitor after the full-wave detector that provides audio feedback to the input tube so that I could see the audio feedback signal clearly on the scope.
2/ Reduce the oscillator / mixer cathode resistor to 100R.
3/ Fit a 10nF cap from the 6AU6 screen grid to ground right at the pin, as I had noted a tendency to “sqeg” (blocking oscillation that breaks in and out as the audio signal modulates the Gm of the tube).
I doubt that the RF detector filter makes any difference to the operation of the unit. Initially, I had 1N277 germanium diodes on hand, Fred recommended 1N34 devices. I obtained some 1N34s and sure enough, the linearity of the feedback detector improved. Displayed on a Tektronix 547 (with 1A4 plug-in) is the signal at the antenna (lower trace) and the resulting clean audio signal at the secondary of the AC91 output transformer:
You may notice that the two traces are displayed at different sweep rates, this is a perhaps unique feature of the type 547, the ability to alternate the two timebases.
I did find it necessary to limit the modulation to 50%.
With regard to the setup, I found that the longer the transmit antenna is the better, this is not surprising given the relatively low carrier frequency of 1MHz. I am using around 12′. At the radio end, a short antenna or even no antenna at all works best, I am using around 18″ into the “selectivity” input on the AC91. All this is fed by a Squeezebox internet radio which has variable output allowing the modulation depth to be varied for the best sound and noise compromise. The result is extremely clear and free from distortion, most satisfying after much frustration especially after discovering that I needed to play with the speaker voicecoil alignment again, it has been around 14 years since I initially restored this lovely post WW2 radio. It is just a shame that the BBC Home Service, Light Programme and Third Program are no longer available……
