Monday, April 24, 2017

Not fully satisfied with the usability of my curve tracer, especially its 'quick test' function


I discovered that one of the final tubes, a 6146A, had a cracked envelope and the anode was disconnected. The first final tube measured decently, but the second was bad. I then grabbed a tube that appeared to be a 6au8, which is how I set it up, but it didn't measure any plate current at all. I then decided it was a 6au6, swapped the pin settings but still zero. 

Old tubes tend to have faint or virtually nonexistent markings, which would make this risk unacceptable. I will need to try each tube in my TV-3 tester, go through all the shorts testing, and look for leakage before I attempt a tube in the tracer. 

Yikes, when I stuck in yet another tube, no plate current. Doing more testing, I did find that the reservoirs produce good voltage, the tests run and curves plot, although I found several tubes where the current through the tube was much too low at the 'correct' grid bias, but a higher (less negative) bias gave me closer readings. 

Measurements were far enough off to douibt that I happen to have a pair of tubes, one a dual-triode, that are that worn. I knew I had to do some recalibration when I swapped to the power brick from my lab power supply, but these odd results force the issue.

After recalibration, the results were about the same. Everything was running lower at the same settings in these tubes. I fortunately have a set of unused NOS tubes, which I know have never been powered up because the chain of custody is clear. They were shrink wrapped in an old kit (actually, portion of a kit) I bought years ago and I have just removed the parts within the last week. 

The new tubes are producing similarly confusing results - the 'quick test' that should recreate the characteristics defined in the tube data sheet are not working out. I did notice that many of the tube datasheets include a cathode bias resistor in the circuit, which changes the effective grid bias compared to my tracer which hooks the cathode up to 'ground'. 

For example, a characteristic given for -3.5V bias isn't reached until the setting is nearly 0 on the curve tracer. I need to investigate this further. The curves being drawn are more reasonable, although for the new tubes I can find no datasheets online that include curves. 

The cathode resistors are for biasing the circuit and produce the bias voltage by the drop across the resistor from the current flow. Thus, when spec sheet lists the characteristics at some plate and screen voltage with a given cathode resistor value, I should be able to figure out the effective bias. 

That is, if they give the expected plate and screen current, those sum to the cathode current. Using ohms law, the voltage drop is the sum current times the resistance. These tend to give me values close to what I am observing when I experimentally set various grid values to try to achieve the reported currents. 

I found some recommendations in the curve tracer manual having to do with manually setting some ranges and averaging values in order to get a more accurate quick test result. I will experiment with this to see if I can get quick test results that are close enough to spec to give me confidence I can tell a good tube from a mediocre or bad one. 
I got closer, but invariably the specs are at a low negative grid bias and the current and other results are low unless I bump it up a bit. For example, a new 6CL8A was low on both plate and screen current at the spec sheet bias of -1V, but when I set to -0.7 I got almost exactly the right current. The transconductance came out close, but the plate resistance was off by quite a bit and the amplification factor was almost nonsensically high (869x). 

One hypothesis is nonlinearity and inaccuracy at low negative voltages, perhaps combined with too much resistance in the wiring, switches and tube socket, if the grid voltage is not delivered at the value I intended. If it is errors calculating at very low values of grid bias, I am kind of stuck since most characteristic reports I have found are at low negative grid bias voltages. 

I also don't know how precise the tracer can vary the voltages for the quick test - if the grid bias is -0.8V then the variance is 0.08V which may not be achievable by the circuitry. If the delta V is off, all the derivatives will be off. 

Vacuum tube curve tracer completed and working perfectly


I installed the barrier strip, insulated joints, tied down the fuse holders and choke inside, trimmed up connections, and put the entire unit together in its final configuration. It sits in the wooden case of a Century Tube Tester that I repurposed to hold my project, leaving some items like the meter and unused sockets because the cosmetics would be worse with all the gaping holes.

Completed curve tracer in its wooden case
The side panel has the connector for an RS232 serial link to the PC, a switch to choose either anode/screen reservoir voltages or the barrier terminal block to connect to special positions 10 and 11 on the rotary switches.

Side connections - serial port, special voltage switch and external heater block
The power brick and the USB to RS232 adapter cords sit in the slot on the left of the faceplate. The nine rotary knobs choose the appropriate connection for each of the 9 possible pins on a tube socket. Only five of the sockets are connected, the rest are dummies.  The black and red top hat connection cables allow grid and anode sources, respectively, to be connected to any tube with a top hat connector. 
Completed and assembled tracer
The Quality light is actually the power on LED showing the circuit board is powered up. The Short light is actually the "high voltage present" LED which indicates that the screen and anode reservoirs have hundreds of volts present. The small 'power' light is not active, nor is the meter on the top.

I fired up the tracer, hooked it to the PC, fired up the GUI control software, and inserted a 12AU7 tube in one of the active 9 pin sockets. I rotated the appropriate switches to connect the target pins to the voltage they represent. 

For example, the second triode in the 12AU7 (it is a dual-triode tube) has the plate connected to pin 1, the grid to pin 2 and the cathode to pin 3. The filament is 12.6V across pins 4 and 5. Thus, switch 1 was rotated to the position for anode voltage (position 8), switch 2 to the position for grid voltage (position 6), switch 4 rotated to filament A (position 2), etc. 

I ran the 'quick test' on the tube, which is set to the voltages from the data sheet where characteristics such as plate resistance, transconductance and amplification factor are reported. In this case, the manufacturer's data sheet showed that with 250 volts on the anode and -8.5 volts on the grid, it should draw 10.5ma on the plate, have a plate resistance of 7.7K ohms, a transconductance of 2200 uMho and an amplification factor of 17. 

I entered the two settings and the nominal values into the quick test form. It showed the plate current a bit high, the plate resistance exactly on spec, and the transconductance and amplification factor about 9% low compared to spec. Reasonable results for a used tube.

I then ran a curve trace to match the conditions for one of the curves illustrated in the data sheet. The curves for 100, 150, 200 and 250 volts on the anode where shown while varying the grid from -18 to 0 volts, displaying the plate current against grid voltage for the four curves. 
The expected performance of a new 12AU7 tube
My curves are similar but not identical, consistent with the degradation of the used tube but similar enough to be a reasonable test of this curve tracer. I could have moved the plate current scale to the right side and adjusted the start/stop values of the axes to line up exactly with the factory drawn curve, if that was important. I can also hide the measurement points and just leave the interpolated curves. 
Measured curve of one of the triodes in my used 12AU7
This project is complete! Next up I started through all the tubes from my Heathkit HW-100 transceiver, to check them against spec. Might as well play with different curve plots and options too.

Saturday, April 22, 2017

Wrapped up tube checker construction - 99% done


This morning while thinking about testing the tubes in our 026 keypunches at the Computer History Museum, I remembered one limitation of the circuit board I built. It can supply filament voltage only up to the supply voltage of the power brick - just over 19V - which is insufficient to test the 25L6 tubes or any others that have high heater voltage requirements. 

The solution for measuring is to hook up external DC for the filament. At that moment, I was faced with new design choices.

I could add in two more positions on each rotary switch to connect external heater power to the tube socket pins, in addition to the two positions I was about to wire to connect the screen and anode voltage reservoirs for continuous measurements. 

Alternatively, I could use two positions on the switches as a generalized set of special sources, with switches to route the reservoir volts and connectors to route external filament power to the same positions.

I liked the new method and set about adding that. I means that I have to find a suitable toggle switch and mount it, to connect or disconnect the reservoir voltages to these special loops. I also need to find and mount some kind of connector for external heater power. 

While I sought these parts and mulled over locations on the panel, I did some remaining tasks - hooked in the top cap connectors to grid and anode voltage loops plus moved the 'power on' and 'high voltage present' LEDs off the board and onto the faceplate.  

I will have a number of connections and a switch to mount, which I decided to place on the side of the cabinet instead of the faceplate due to restricted space up on top. The items to mount on the side are:

  • RS232 DB9 connector for PC communication
  • Switch to connect reservoir capacitors to special rotary switch positions
  • Terminal block to connect external heaters to special rotary switch positions
The switch to install should be a DPDT, wired so that the reservoir power wires are on one side and the terminal strip are on the other, thus having no risk that high voltage will be present on the terminal strip.

The circuit board is now firmly mounted inside the case. Once I have the connectors added to the side, and the power brick mounted inside somehow, there are some ferrite beads and fuses to add to the filament lines before I hook it all together and close it up, plus a fuse and large choke to install on the line from the power brick to the board. 

When I first mounted the DB9 connector, it stopped communication with the PC. Likely this was caused by a short from some of the handshaking bridge wires that are often soldered onto the back pins of DB9 connectors. After inspection and rearrangement, the connector is back in the case and fully functional.

I now have the ferrite beads on the filament lines, the fuses for filament and main power, the main 300 uH choke and the wires coming from the screen and anode reservoir capacitors. I needed to find and mount the switch, connected to the reservoir lines and the new loops I soldered for switch positions 10 and 11 to deliver either reservoir or external heater power to tube socket pins. 

The reservoir lines can carry 300 or so volts each, which is above the voltage rating of most switches, complicating the sourcing. I needed the switch to be on-off-on, stay in any of those three positions, and have twin poles and double contacts for each pole. 

Having found an acceptable switch, I have it wired in place, although it is only an on-on, no center off position. This is okay because the terminal block that will connect to external heater power can be connected to the rotary switch positions 11 and 10. 

The only risk with this on-on switch would be if I switched a tube pin to 10 or 11 and inside the tube that pin was connected to another pin with screen or anode potential. Even there, the high voltage will only be present for a millisecond at a time during testing.

The power brick is installed, although I have to recalibrate slightly to accommodate the slightly different voltage supplied to the board. Mounting and insulating of a few components and connections, e.g. choke and fuses, will wrap up the work. 

Friday, April 21, 2017

Finished main wiring, tested by measuring a 6AU6


I finished wiring the voltage loops for all the regular connections from the circuit board. After very carefully checking every connection with a continuity tester, I discovered that when I swapped out the original selector switch for a new one yesterday, I miswired the filament connections. With that fixed, it was time to run a trace on sample tube, a 6AU6.

Leveraged tester that has become the curve tracer
The controls are simple - there are nine rotary selectors, one for each of the possible pins on a tube socket. They are switched from positions 1 to 11, but only some of the positions are active. Switching the dial for a given pin to the following numbers hooks it up to the associated voltage from the circuit board inside:

  1. n/c
  2. filament side A
  3. filament side B
  4. n/c
  5. cathode
  6. control grid
  7. screen grid
  8. anode
  9. n/c
  10. n/c
  11. n/c
Only five sockets are wired up - one each loctal, octal, and 7 pin mini, plus two 9 pin mini types. The remainder of the sockets on the plate, as well as the meter, is superfluous. The following picture shows which switch controls each pin, highlights the working sockets, and obscures the unused parts. You can see the connections to the circuit board hanging out of the bottom, as I haven't finished the mount yet.

Marked controls and active sockets only
I fired up the 6AU6 and measured curves - I really like the circuit and GUI of the uTracer 3+. I have some more wiring to accomplish, however. First, there are two top cap connectors, for tubes that have the anode or the grid connection on the top of the tube. I have to hook these to grid and anode voltage lines and remember they are also hot during any measurement. Second, I have to create loops for positions 10 and 11 of the rotary switches to carry the screen and anode voltage reservoirs out to tube pins.

The normal mode of operation is for the circuit board to build up the target voltages in the screen and anode reservoir capacitors, but only connect those to the screen and anode pins for milliseconds at a time while making measurements. Certain unusual measurement situations may require a steady voltage, not a short burst, which is the purpose of positions 10 and 11. 

The boost converters can't supply more than 3ma steadily, so this connection to continuous screen and/or anode reservoirs will only work for low current applications. The archetypical case for using the reservoir voltage is to test magic eye tubes. This will not be a frequent mode for me, in fact I may never use it, but the wiring should be in place. 

Thursday, April 20, 2017

Construction continues on vacuum tube curve tracer


I grabbed some lockwashers and other hardware and attempted to install the three remaining rotary switches. Alas, the chassis opening for a 9 pin mini tube socket is slightly wider than the lockwasher. While I can use flat washers to cover the gap, they don't have any grip to keep the switch from rotating when the knob is twisted. Back to the drawing board (hardware store). 

Armed with new wider star lockwashers and a combination wrench, I tightened on the three new switches and began wiring them up. By midafternoon I had seven switches wired. The next two are easier, because both skip the 7 pin mini socket and one of them also skips the two 9 pin mini sockets. 

In the meantime, I started the lines for each voltage, bridging across the various rotary switches on the same position of each, ultimately leading down to the curve tracer circuit board itself. To do this correctly, I had to decide where the board would be mounted. 

The choice is on small standoffs on the bottom inside the case. This complicates things a bit because wire runs have to be somewhat longer to allow the top to be opened for servicing. I want the wire runs to the rotary switches to be as short as I can, in a loop themselves, to minimize inductance. 

I will also run wires out from the PCB to remotely mount the two LEDs under the faceplate, so show the 'power on' and 'HV present' states to the operator. For this I had to desolder the existing LEDs, put on the wire and place my own LEDs on the faceplate. I don't need to worry about wire length for these thus it won't cause problems opening the unit for service.

This will be powered by a 19+ V DC power brick, also installed inside the case. I had one from an old laptop I used to own, which worked out great for this purpose. It will be fixed inside and the AC cable routed out in the case slot that contained the original AC cord. 

By dinnertime, I had completed wiring all nine rotary switches to the associated pins of the five tube sockets on my tester. I had also wired up the two filament loops - wiring looped through switch positions 2 and 3 respectively of every rotary switch. This worked well and lit heaters on real tubes.

I will wire the cathode, grid, screen and anode voltage loops to their relevant positions 5, 6, 7 and 8 on the rotary switches. By the evening, I was done with the screen and anode loops. A terminal snapped off the old rotary switch that I was leveraging from the original Century circuitry, which required me to remove it, drill out the hole to fit my new type of switch, and wire the new one into place. 

After these are seen to work properly, I can bring the two reservoir capacitor lines in a non-loop to positions 10 and 11 of the switches. I will finish up the cathode and grid lines tomorrow, test out everything carefully with a continuity checker, before hooking it all up to the circuit board and testing with a real tube. 

1401 restoration work and continued curve tracer construction

I spent part of the day with the plumber as he finished up the tankless water heater installation and some other work, then spent the rest of the day with the 1401 restoration team. This evening, I attended a lecture by Paul Wesley at Stanford on the history of Silicon Valley. 


I slowly did the tedious wiring of the sockets - loops between each pin number, around all the sockets being used, through ferrite beads and connected to the commutator of the six rotary switches I had on hand. I also prepared the sites for the next three rotary switches to be installed, now that the stock arrived, for controlling pins 7, 8 and 9. 

Before I can mount the next three rotary switches into the holes where I removed tube sockets, I need to get to a hardware store for some regular and lock washers, since a tube socket is larger diameter than a rotary switch shaft. 

After they are mounted, I can wire up those three loops. Actually, one loop goes to all five sockets, a second loop goes to just four sockets, and the last loop goes to only two sockets, the 9 pin minis. 

Another issue is the mounting for the circuit board, power supply brick and the serial port DB9 connector. Besides working that out and doing the install, I have to wire the board produced voltages to the corresponding positions on all nine rotary switches. 

The voltages on the rotary switches are: 
  1. anode
  2. screen
  3. cathode
  4. grid
  5. filament A
  6. filament B
  7. anode reservoir
  8. screen reservoir. 
The regular anode and screen connections are only connected for very brief intervals during measurements, small fractions of a second, otherwise sit at zero. The reservoirs provide the same voltages continuously. 


SMS cards in the 1401 power supply circuitry drive a set of relays which sequence power on and off at startup or shutdown of the system. All the cards we have are scorched black underneath a pair of resistors that are extremely hot to the touch at all times. We investigated to see if this is a result of failed components or 'normal'.

The card schematics for this card matches exactly what is on the card. It is clear that the pair of 2W resistors are asked to handle more than 3.5 W of power, yet are mounted flat to the card with no space for airflow around them. Thus, the scorching and overheated resistors are by design.

More interesting to me is that the resistors in question are simply providing a voltage divider to bias a transistor, switching it on only if the two power supplies that fed the divider were working. They designed this to draw about 65 ma across the 42 V relative voltage, all to create a minute bias current in the transistor.

We then looked at the ALD schematic page, instead of the SMS card schematic, and found that the design had different and much more appropriate resistor values installed. 3,470 instead of 650 ohms, which dropped the power consumption down to a half watt. The schematic pages had the exact same drawing and part number, but the cards as they exist are the scorchers not the design in the ALDs.

We decided to take one of our spare cards and modify it to match the ALD schematic. I found almost every part needed on hand in the workroom, except that I had to parallel up two resistors to create a 43 ohm part. By the end of the day, I was partway through rebuilding the card. The tools on hand for solder removal were inadequate, so I will bring in my own gear next time.

Tuesday, April 18, 2017

Building curve tracer into case


With the circuit board itself working flawlessly, it is time to build out an enclosure with tube sockets and rotary selector switches, mount the board inside, and wire everything up. I hadn't selected an enclosure yet. I do want to move forward on this.

I initially thought to gut a Century tube checker I own, but the panel has dozens each of octal, 7 pin mini and 9 pin mini sockets, whereas I only need a few slots. Further, it has a large meter and other parts installed which are unnecessary, and lacks the rotary switch openings. Still, I decided to hack it up, as the extra socket openings could eventually hold compactron, novar, nuvistor, acorn and other socket types.

I have five rotary switches with 11 settings, not enough to route each connection type (anode, screen, grid, cathode, and one side of the filament) to any of 12 possible pins (most sockets have 7, 8 or 9 pins but there is a compactron type that has all 12). Thus, I will have to connect in the opposite way, using 12 rotary switches to select among six voltages - anode, screen, grid, cathode, filament A and filament B.

The existing sockets are pop riveted in the panel, thus I need to drill out any that I remove. I can then use screw and nut, or new pop rivets if I can find my gun and supplies. First step was to gut the transformer, unused controls and wiring, Next was to trim back the rats nest of intertube wiring.

I have ordered another batch of rotary switches, enough for 10 pin sockets which will cover everything except the compactron (12 pin type). These are almost exclusively used in tube based color televisions, which is not an area where I expect to receive tubes for testing. Thus, I will ignore the compactron type entirely.

There are a wide range of tube socket types, of which I will support as many as I can easily and inexpensively obtain sockets for. The three workhorse sockets are the octal, 7 pin mini and 9 pin mini. I have plenty of these already on the Century unit.

Loctal sockets are similar to octal but have a center pin that snaps into place in the housing. These were intended for use in automobiles and other high vibration environments. I happen to have two loctal sockets on the unit, otherwise I wouldn't have bothered. Again, mobile radio tubes are not what I expect to be testing. Hifi, computers and test equipment will be the primary source of tubes for testing.

There is a 10 pin socket that is like the 9 pin mini but has an extra connection in the center hole. I happen to have one of these sockets, so it will go in. There is a novar type, which is nine pins but thinner and similar to the compactron style. I think I have one of these as well.

Older radios used 4, 5, 6, 7 and 8 pin sockets, UX4, UX5 etc, that I might stick in if I had them cheap and ready to go, but these are mostly from 30s and older era radios and clearly not going to be needed often.

I have the first six rotary switches (one from the original tester and the five I bought for the project) installed in the unit. Next up is to choose locations for the next three switches, which will arrive in two days. That will give me switches for the nine pins of the sockets already in place (7 pin mini, 9 pin mini, octal and loctal).

I began wiring up the first six switches and the five sockets I am using. It is tedious work, building a loop for each pin across all five sockets, including a ferrite bead to soak up any RF oscillation that might try to build, and hooking each loop to a different switch. I can only build the loops for pins 1 to 6 until the new switches arrive and are mounted.

Monday, April 17, 2017

Tube curve tracer board working and calibrating


I resolved the screen voltage problem - turned out to be two hard to spot solder joint issues. The solder welled up on the component lead in a beautiful shiny cone, but had not flowed onto the pad. Looked great until I went to extreme magnification with the stereo microscope, where I could see only 1 or 2 joints at a time. 

Repaired and the hangup of the main logic is gone, as I expected. It goes through a measurement interval, applying anode and screen voltage to the tube under test (but I had no tube connected) while maintaining the voltage reservoirs at their target level.

In addition to the screen circuit problems, now fixed, the grid bias circuit is not working. It should develop a bias from 0 to -50V during the measurement phase of the curve tracer. The microcontroller produces pulse width modulation on a pin which drives a low noise operational amplifier acting as an low pass filter (or integrator). This yields a smoothed voltage between 0 and 5V, which is fed to a second op amp, LM741, wired as a 10X inverter, to produce the intended 0 to -50V.

I zoomed in to this section of the board and looked quite carefully. There was one solder joint that looked as if it would benefit from reflowing the solder to improve pad bonding, but it is for the unconnected pin on the LM741 op amp package thus it doesn't matter. Time to debug the circuit, measuring and observing, until I figure out what is wrong.

The observation steps to make are:

  1. scope on the microcontroller signal that drives this, which is a 19.5KHz signal whose on time is varied (pulse width modulation) to swing the first op amp output between 0 and 5 volts.
  2. if the pulses are good, look at the first op amp to verify it is filtering/integrating and generating the 0 to 5 volt range expected
  3. if the first op amp is working properly, observe the behavior of the second op amp which should be generating 0 to -50 based on the 0 to 5 of the first device.
  4. if the second op amp is working properly, observe the current mirror transistors that produce the actual grid voltage when driven by the second op amp.
  5. If the mirror transistors work, debug the continuity problem from here to the grid connection.
The microcontroller pulses were behaving exactly as they should, varying the pulse width as the target value changed from 0 to 5. We passed test one above. The second test involved watching the output of the NE5535 low noise op amp to see if it did produce the 0 to 5 level. 

The second test failed! The op amp circuit is not integrating nor acting as a low pass filter to produce the target value. I saw a constant output on pin 6, the output pin. There are only a few components in this portion of the circuit, two capacitors and a couple of resistors. If they are good and the connections and solder joints are correct, it points the finger at the chip.

The sequence of tests to perform for this are:

  1. verify that V+ and V- are making it to the chip
  2. verify the proper resistance for the two resistors and two capacitors on the input side
  3. verify that the input pulses are making it to the non-inverting input
  4. monitor the waveforms on the inverting inputs
  5. check for shorts of the output pin
Test 1 failed - no V+ to the IC, nor am I getting +15V to the other three ICs it feeds. Digging further, I found that my 7815 voltage regulator was bad. I had a replacement in stock, put it into the board, and now everything is working properly! Filament, anode, screen, grid and all the measurement logic doing fine.

Now I have a bit of calibration and then have to build it up with tube sockets, switches and all to make a full tester. Calibration is done easily, with a GUI screen to adjust values to match measured results with the DMM. I worked through this tonight as I collect tube sockets and order any remaining ones I will need to install in my project box/chassis. Every is calibrated and ready to install in a box and begin using.

Op amp lab, vacuum tube curve tracer and disk tool work, plus vacation and other things

There has been an extended gap in this blog, encompassing a vacation in Maui, work on my home, and quite a few other tasks that temporarily took higher priority. While I still have quite a bit of work we plan to do on the house, I was able to get back to some of the electronic projects.


I constructed the uTracer kit before I left for vacation but two areas were not working properly at that time. The board produces anode voltage, filament power and communicates, but neither the screen voltage nor the grid bias voltage are responding. 

The screen voltage section is a mirror of the anode voltage circuitry. It has a boost converter that will build up the voltage in a large 'reservoir' capacitor, driven by pulses from the microcontroller. A analog to digital converter chip samples the voltage present on the capacitor and delivers it to the processor. 

When the sampled voltage is below the target level, pulses are applied to the boost side to increase the voltage. When the sampled voltage is above the target value, other microcontroller pulses discharge some of the voltage. The voltage on the reservoir is thus maintained at its chosen level, to be delivered  to the tube under test in short pulses through a high voltage switch portion of this circuit. 

The voltage on the reservoir capacitor never climbs above 19 volts, the input supply voltage to the board. The logic to issue the boost and the discharge pulses is part of an interrupt processing routine in the microcontroller, fired off regularly by timer pops. The main logic path in the controller has to wait to see the sampled voltage reach the target level before it does any testing, but if the boost converter doesn't reach the target level, it just freezes in the main loop.

I examined the board and its solder joints very carefully, but everything looks good. I then put the scope on the microcontroller pins that deliver the boost and discharge pulses, and saw the logic trying in vain to grow the voltage. 

Since the controller is doing everything right, I turn my attention to the components that might be malfunctioning. The FET that drives the boost function might not be conducting properly. The transistor that implements the discharge function might be stuck on or shorted. The high voltage switch could be malfunctioning and draining power as fast as it builds up, although this is less likely. Other hidden component failures or short circuits are the other, less likely cause.


I continued to pore over the traces and other evidence to figure out where the read operation goes awry when the Alto reads cartridges that I have written with the tool. I see a time interval occuring between the header and label records of the sector that is different by about 8 word times, depending on whether a cartridge is written by me or by an Alto. The outcome is misreading of the label record and a checksum at the end.

The same flaw causes the data record to be misread, thus injecting the wrong data into memory, dooming the bootstrap operation. I haven't yet figure out the cause but will continue to study everything I can until I figure this out.


I attended a one day class at Texas Instruments on designing with op amps - intended for engineers who knew the basics and were designing with them already, but who needed to understand deeper issues such as stability, noise and distortion that weren't clear from the idealized model of an op amp that is used in most textbooks. My friend George from the 1401 Restoration Team also attended.

Part of the course was a board with several types of circuits installed and plug-in modules to try out different op amp part numbers. One could inject various signals to the op amps and watch the results on a scope, to reinforce the lessons of the class. They provided a discount to buy the board for home use.

I have a Digilent Analog Discovery which provides function generators, oscilloscopes and other bench tools to drive and display the op amp results, similar to the NI Virtual Bench used in the TI class but much less expensive. I set it up to experiment further with the op amps.

Another discount from TI gave us six other op amps for use in evaluation. I naively assumed these were premounted on the daughtercards just like the ones supplied with the test board, but instead I received six small SOT-23 chips and a set of empty PCBs to solder them to.

Thus, I had to haul out the stereo microscope to solder these in place, before I could make use of these new op amps on the board. As well, it takes special clamps and tweezers and the steady hand I no longer reliably have. 

Sunday, April 2, 2017

Building tube curve tracer, new meter for TV-3 tester and progress on Alto ethernet gateway


Today Ken is at Marc's house collecting more testing data for his ethernet gateway device. I had to meet a drywall service that is doing some work on my home, so I could not be there during the tests. They did discover the small issue that was blocking the Alto from accepting packets sent by Ken's unit. Now he will focus on building up the protocol stack - file serving, ftp and other functions.


I picked up a physically identical meter that will fit into the panel of the tube tester. It features a removable faceplate allowing me to install the original tube tester plate. This will make it appear identical to the original meter.

All that remains is to validate the shunt and series resistance values so that it yields full scale deflection at 200 uA and offers a total resistance of 2365 ohms. That will make it perform the same as the original meter.

I will need to work inside the meter to deal with the series and shunt resistances, which exposes it to the same risk that I will damage the meter during the manipulation. This was how I ruined the original meter during a 'repair'.


I picked up a kit that allows me to fully test vacuum tubes, producing the various characteristic curves and evaluation the gain at a wide sweep of voltages. It connects via RS232 to a PC and features a nice GUI control panel.

The kit is designed with many testing steps as it is constructed, helping to verify each stage of construction before proceeding to add more components. For example, digital power is validated, then the RS232 communications, before installing the microprocessor circuitry. All that must work before the various high voltage producing circuits are installed.

I am gradually accumulating tube sockets, ferrite RFI suppressors, fuses, rotary switches and other items needed to install the tracer into a cabinet and make it fully usable. I haven't settled on a cabinet to put everything into yet.

I have installed circuits and tested up through the negative power supply. After this, I will be working on grid bias, current amplifier and boost converter/high voltage sections. 

Saturday, April 1, 2017

Work on Alto restoration, progress and a head crash


We met today to work on a number of tasks for the Alto. Among them were debugging my disk write process, debugging the ethernet gateway, and archiving many of the cartridges we borrowed from PARC.

Extensive logic analyzer traces were captured of a good boot with an existing cartridge and of a failed boot on the cartridge I wrote. Unlike past traces, this time the analyzer was also recording time between events and that made a difference visible.

The time between one record and the next within a sector including a quiet period of about 145 uS on cartridges that were written by an Alto, but only 75 uS on the ones I wrote. This appeared to place the sync word and early data words too early, while the Alto was still waiting for a fixed time to elapse, and thus the second and third records in a sector did not start on the sync word. They were misaligned and suffered checksum errors.

More study is needed to verify this, then the flaw must be found and corrected in my disk tool logic. Hopefully it will be an obvious issue that definitively matches the error symptoms and whose correction is unambiguous.

Ken's ethernet gateway is receiving packets but was not able to respond to PUP echo packets from the Alto. This was in part due to a changed protocol in the docs he has versus what is being sent by the Alto. In addition, however, we extended the ethernet board and hooked up scope probes to see exactly what is happening when packets arrive from the gateway.

We found that the first few words of a packet are received but the firmware or software stops listening by resetting the receiver. This is what happens, for instance, if the address of the incoming packet does not match the station address of this Alto, but can be caused by other packet header issues.

We cleaned the nine cartridges on loan from PARC. Two of them had visible damage on the platter when opened up - one shows signs of a minor head crash in the past and the other has a radial scratch deep enough to expose aluminum substrate. They were marked and will not be mounted.

Six cartridges were read and archived, containing source code, Mesa and Smalltalk disks, for example. When the last cartridge was being read, almost at the last cylinder, the power supply we were using hiccuped. We could hear the door latch solenoid cycle several times.

I dived for the Load/Run switch to get the cartridge out of the drive. We opened it up and found burn marks where the heads had rubbed during the time the power cycled on and off repeatedly. Pulling the covers showed burned oxide and binder on the heads too. A head crash!

The contents of the cartridge were archived other than a few sectors at the very end of the pack and it boots just fine in the Alto simulator (Contralto). We added that cartridge to the other two which were scraped or scratched already, so that we don't stick them into a drive again.

Marc and I cleaned the disk heads with lint free cloth and isopropyl alcohol until they appeared to be free of any oxide. Before we turn it on, however, we first must substitute a reliable power supply for the +15 and -15V that feeds the drive.

Then, we will load the cleaned heads on cylinder zero of the pack that had the crash, since the early cylinders are damage free, just to verify that the heads don't damage packs and work properly. As the cartridge is already ruined, but contents archived, any possible damage to cylinder zero is acceptable in order to test our heads. 

Wednesday, March 29, 2017

Continued work on the 1401 (DE) machine at CHM

Have been very busy with many tasks outside of the restoration hobby, which left a bit of a gap in the blog. Starting back up but will still have many distractions.


Today we began testing the 1401 whose Start Reset switch I recently replaced. I had some suspicions that the short circuit weeks ago had damaged more than the circuits I replaced. Alas, I was correct. 

Two of three pins that deliver -6V to the 01A2 A backplane and then to the control panel were dead. Last week I ran new lines to deliver +6V to two of four pins on the same slot (A24), as those were dead. Now, I had to add two wires to provide the missing -6V. All these extra wires run over to the power distribution strip on gate 01A3 which is adjacent to the control panel.

We finished adding the lines, pushed the power on button and were surprised that the system failed to power up. Unhooking the two lines I added didn't fix anything, we still had problems powering up. Some quick checking showed a failure of the +6V supply during the power sequencing.

After lunch, when we were trying to shoot the problem, we saw the +6 supply to the sequence logic yoyo a bit then suddenly the machine was powering up fine. About 30 minutes later, it failed to come up again. This is a problem we will have to address, one that seems to have cropped up independently of the earlier issues.

With the machine up, we found that some of the controls on the control panel are not working properly. Regardless of how the rotary address switches are set, we get all lines high and an error condition. These switches, like most controls, run past a resistor divider hooked to +6 and -6. This tells me that our voltages are not getting from slot A24 to the other A slots that hold the resistor cards.

We need to get access to the resistor cards and paddle cards (connecting wire harnesses to an SMS card stub). These are blocked underneath the backplane in a very very challenging spot on the machine. We discovered a way to swing a cable trough down and out of the way to let us see the cards from the underside. 

Unfortunately, the A24 paddle card is blocked in place by all the cable harnesses running through the cable through and down into the card gates below the control panel. The crowd of attendees built up and the demonstration crew arrived, cutting our work off and requiring us to clean up prior to the scheduled demo at 3PM. 


We will be back together working on the Alto this Friday, after a hiatus due to work obligations for other members. There are nine cartridges from PARC that we will clean, insert, read and archive during the session. Too, we will collect some information to figure out why cartridges written by my tool cannot boot up on the Alto. 

Tuesday, March 21, 2017

1401 Computer and TV-3 Tube Tester progress


Today I dug up the proper manual and schematic for my unit, as I tired of waiting for the printed copy to arrive. The seller of the printed copy promised "fast and free" shipping with delivery tomorrow, but when they actually mailed the copy it went out media mail, the dirt cheap way to send paper items that will take it until late next week to arrive. Grrr. At least I have the online version I downloaded.

I printed the schematic and my guide to the eight push buttons, allowing me to begin tracing the power path from the 83 rectifier tube out to the plate pins of the tube test sockets. I quickly discovered that it was simple oxidation on the rotary dials and push buttons - exercising them a bit cleared up the problem and viola! - 150 plate, 130 screen, 40 bias and other voltages appeared as expected. 

I will apply some deoxit to the contacts of the switches while I wait for my new DMM to arrive tonight, which will substitute for the meter movement and let me further check out the functionality of the tube checker. All is looking great with the exception of my saved then destroyed meter movement. 

Once the DMM was received, I installed a series resistor of 2400 ohms, to substitute for the original meter resistance of 2365. The results were reassuring. even with some signs that contacts are not yet fully cleaned. Setting the line voltage to 117VAC gave a reading of about 73 uA, which is roughly half scale. 

Based on that, I extrapolated the readings I got for several tubes, linearly, and the answers for tubes I own and tested were reasonable. If I can get a replacement meter installed this tube tester will be ready to use. 


I invested some quiet time with the 1401 when nobody was around, just me and the powered down system, in order to trace out the wiring that must be reconnected to the Start Reset button. First I had to decode the locations on a small backplane into which all the front panel components are wired, then trace the wires and match them to the logic diagrams, finally resolve any faults.

IBM's scheme for these computers is called SMS, which defines the cards, backplanes and other aspects of the machine's construction. SMS cards have 16 traces on the edge that are inserted into a card socket on the backplanes. The pins connected to the traces are labeled A thru R (skipping the letters I and O is an IBM 'thing' apparently because of the potential for confusion with the digits 1 and 0.)

Cards populated with transistors, resistors, diodes and so forth are plugged into the card sockets or slots on a backplane. The typical backplane has 26 slots, numbered 1 to 26 and each slot has pins A to R. The backplane connecting the front panel, however, had only 24 slots. It has the slots facing downward below the backplane, with the cards plugged upward from below.

First step was to determine the orientation. Pin A was the furthest in and pin R is closed to the front panel. Slot 1 is furthest to the left, when looking in from the front panel, while 24 is to the right. There is a rats nest of snarled wires wrapped around the 384 pins sticking up on the backplane, making it hard to count rows and pins, or to get to a specific pin with a probe. 

The logic diagrams (ALDs) will list the spots where signals go on or off a particular backplane, gate or module. In the case of signals connected to the front panel, they went through this backplane 01A2 Axxy where xx is the slot number and y is the pin number. 

I found the various locations, hooked on a circuit tester, and then found the wire that corresponds to it hanging loose next to the Start Reset button. In this way, I identified all 8 wires that run to the switch and knew which signal they contained. That also means I know which terminal on the switch they are to be soldered to. 

One problem we had identified earlier was that none of the eight wires had +6V present, yet one of them must for the switch to work properly. On the logic diagram, this is the +6V CK3 line, which means circuit 3 from the +6V power supply. The backplane also had +6V CK1, CK2 and CK4 wired to it. These are all electrically identical, other than using different wires to bring it up into the card slot. 

I found that the pin which brought +6V CK3 was completely open, not connected at all to the other +6V circuits. They were all mutually connected and would measure properly when the machine is powered up. The pin for this power is 01A2 A24M and that means a paddle card plugged into slot 24 (a stub of a circuit card with discrete wire cables connected to the 16 traces) had one wire which was CK3. It either melted or a fuse element somewhere in the line melted open.

Fortunately, there were several other pins that had +6V from other circuits. I moved the wire that runs down from the Start Reset button over from A24M to A22R, which is +6V CK4, and we now have good +6V running to the switch.

Tomorrow (Wednesday) morning we will solder up all the wires, reinstall the switch and verify everything is good. It is likely that the melted line for CK3 occurred when we had a major short between power supply lines at the switch. The short blew 20A circuit breakers on the power supplies, certainly enough to vaporize a bit of wire.

Thursday, March 16, 2017

More tube tester restoration work


With my solid state replacement for the 83 rectifier tube in place, I found that the tester delivered no plate voltage to the circuits - one of the two potentials that are handled by this tube. Time to carefully probe all the voltages present at the tube socket, to determine if I have something burned out in the main power transformer.

The rectifier tube (type 83) is actually a dual diode. Each of the two plates is connected to an opposite polarity AC winding on the power transformer delivering +170V. There is a filament winding where the rectified DC output is the center tap and each side delivers 2.5V AC at opposite phase to light the tube filament.

83 rectifier circuit
I discovered that the filament center tapped winding works fine. I see the 155+VDC at the center tap but it is not getting to the tube socket where I am testing for it. Incidentally, the solid state replacement for the 83 tube uses two rectifier diodes, two resistors and two zener diodes to sit in place of the real vacuum tube.
Solid state replacement for 83 tube - large pins are the filament
I went through the checkout procedures in the manual I currently have, which is for the immediately prior version, the TV-3U and so a few items don't match. In spite of that, I verified all the filament voltages, the screen voltage, grid bias voltage and the small AC signal voltage that drives the gm (transconductance) measurement to validate the tubes gain. The short light tests worked properly as well.

The only thing that didn't check out was the plate voltage on the tube socket when I pushed the gm test button P4 - it should jump to 150V but it did not do that. Having verified the voltage is produced by the rectifier tube, it is just a matter of tracing and checking each step in the circuit from there to the plate pin of the tube socket.

I could have bad contacts on the push buttons or rotary switches that connect the plate voltage to the appropriate socket pin, or it could be a problem with some components in the path. I am overjoyed that this is working so well already, particularly that I don't have a bad power transformer.

Unfortunately, the path from the point where the rectifier delivers 150+ DC is where one of the deviations exists between the TV-3/U and my TV-3A/U unit. As well, there is no "large/small" signal switch on my unit, but the predecessor allows selection of 1V or 5V AC signal on the grid when testing transconductance.

I also discovered that P4 is not the gm test button on my unit - there is a row of pushbuttons with only a couple marked. It think it is better that I wait until I know what is what before I decide whether the plate voltage is delivered or not, plus I need an accurate schematic. I will set this aside until I get the manual and new DMM. 

Working on 1401 computer, restoring tube tester


One of the 1401 systems is down due to a cascade of problems with the "Start Reset" switch on the console. The switch appeared to be bad, someone soldered on a replacement after which the system appeared to work. 

Later, when the switch was tightened into place, it created a short across power supply lines that popped circuit breakers on 20A power supplies. Another set of soldering including some guesswork as to which wires went where, and we now have a machine that won't reset.

Back to basics, tracing the wires and comparing to the logic diagrams until we can get it working right. It is slightly complicated because we have eight wires but only six total switch terminals. We have to figure out where the two 'extra' wires go, but the ALDs are not close enough to physical wiring diagrams to help us with this. 

The switch is actually a pair of single pole, double throw momentary switches. Since the 1401 has two complementary sets of voltage levels, U and T, the switch must generate reset levels at both U and T levels for the various cards that need that polarity.

U level is -12 and 0 volts, with the lower potential equating to logical 0. The T level is -6 and +6 volts, again the lower potential is logically 0. Thus, we have two switch halves, one for U and one for T potential. 

A minor complication is that the actual signals driven are +U Start Reset and -T Not Start Reset - so that the U side of the switch will be at 0 to reset but the T side of the switch will be at -6 to cause a reset.

The U side has -12 on the normally closed contact and ground on the normally open contact. The common pole of the switch is the +U Start Reset signal. Thus, the line sits at -12V (logical 0) until the switch is pushed, when it pops up to 0V. 

The T side has -6V on the normally closed side and +6V on the normally open side. The common pole is the -T Not Start Reset signal. This is an example of the maddeningly bizarre nomenclature IBM used, as the -T means inverted logic levels so a "Not Reset" with inverted levels is 0 when "Not Reset" and 1 to Reset. That means it is really a +T Start Reset line. Thus, the line sits at -6V (logical 0) until it is pushed, when it pops up to +6V.

A second complication is that the machine must reset itself on powerup, so the normally closed contacts (-12V on U side, -6V on T side) are not directly wired to the power supply. Instead, they go through a relay that is part of the power-up sequencing logic.

The initial state of the relay before power comes all the way up will set the normally closed U side to 0V and the normally closed T side to +6V. It reverses after a short time delay to give -12V to the U side and -6V to the T side. The effect of this is to set +U Start Reset to on and -T Not Start Reset (+T Start Reset) to on initially, resetting the machine before the short delay is over and those signals revert to their inactive off state.
1401 Start Reset button wiring
We can trace the center poles of the switches, by finding SMS cards that receive these signals somewhere in the machine and tracing continuity back to the the wires, all of which are unsoldered to trace things properly. We did find and validate the +U Start Reset and the -T Not Start Reset lines.

Finding the +6V, -6V, 0V and -12V seemed simple as well, but we could not find any wire that had +6V on it. That was the condition when we ran out of time Wednesday when the museum visitors arrived for the scheduled demo at 3PM. If there is no +6V available, the reset switch can't activate the +T side of all the reset circuits.


I have to wait patiently for the next session with access to the machine in order to archive nine more cartridges we borrowed and to collect enough debugging information to clearly identify and fix the defect that keeps me from booting cartridges I have written.


I soldered together the solid state replacement for the 83 tube and stuck it into circuit. I powered up to check that no magic smoke or signs of distress arose. All seemed fine, but without a meter I will need to improvise to check things out.

I think I can hook up my modern DVM in current mode, using a 2365 ohm resistor in series with the VOM, and read off the current. Full scale should be 200 uA through the resistor, which is the equivalent of the original meter resistance.

My meter reads 400ma on the scale, which is huge compared to the actual current expected, so I might need a different method for handling this. I decided to upgrade my DMM to a model that has a 200 uA range - will come tomorrow.

Wednesday, March 15, 2017

PARC meeting picture, progress and not on the TV 3 tube tester


Here is a picture of the meeting we had at PARC with some of the original researchers who built parts of the Alto.


I removed the meter from the tester chassis, opened it up and was delighted to find that the series wirewound resistor was open! As long as the movement itself works, I only have to install a substitute resistor to repair this.

Cover off, next to remove faceplate
Shunt wirewound resistor in series with movement

Resistor is an open circuit!
The meter movement moves fine on its own, restricting the repair to a replacement for this. It was helpfully marked on the face with its precision resistance - 1,640 ohms - which is the value I must shoot for in a substitute.

Anchor electronics parts list
If I combine the 1.6K resistor, accurate to .05% in series with the 40.2 ohm resistor accurate to .1%, the target resistance is 1,640.2 with a max deviation of about 0.84 ohms. The higher value resistor can have a value between 1599.2 and 1600.8 ohms, while the smaller resistor will be somewhere from 40.1598 to 40.2402. I will buy several of each and cherry pick the best combination, both on the low side, to get closest to 1640 exactly.

While I am going there, I cataloged the capacitors that must be changed to finish the restoration. One challenge is that I can't locate one of the capacitors shown in the manual I downloaded. It has clear pictures of where it should be, but the location is somewhat different and there is absolutely no capacitor there.

The manual I found was from 1949, for the TV-3/U while my unit is a TV-3A/U from 1950 or 1951. I bit the bullet and ordered a manual reprint from ebay which should arrive on Saturday. It may be that one of the three capacitors I ordered do not even exist in this unit, but I have visually verified two of them.

Parts in hand, I began soldering the replacement components on the tester. Somehow, while preparing to solder the series precision resistors into the meter, I managed to pop the movement off its bearings. I suspect this is now totally broken, by a stupid mistake, when it was almost salvaged.

The capacitors went on, anyway, so that when I either repair or more likely spend a fortune to replace the meter, the tester will then work. Building the solid state rectifier tube replacement now.

Monday, March 13, 2017

TV 3A/U Navy tube tester restoration, still investigating disk tool issuing writing cartridges


Still working through the microcode and hardware to understand each and every cycle I should see on the Alto when it is trying to boot a disk - both successfully with natively written cartridges and failing with the ones written by my tool.


I stopped by an electronics swap meet this weekend and picked up a US Navy tube tester - the TV 3A/U which was built by Hickok. It measures the gain of the tube, not just the static emission strength like most tube testers, which is important to truly verify that a tube is working properly.

The pilot lamp goes on and one of the two tubes inside has it filaments glowing (the lower voltage 5Y3 rectifier tube) but the other 083 tube did not light at all. I may have a failed tube, but need to check voltages and condition to be sure that something more sinister doesn't exist, either to have caused the failure or perhaps inside the power transformer.

Another flaw is that the meter movement does not budge - it should deflect for the line voltage measurement and also to full scale when the tester is set to the ohmmeter function. The meter has a capacitor across it, which may have failed in a short, or I may have a more pernicious problem given that the line voltage check is a very direct connection.

I checked all the AC voltages coming off the transformer and it is good! Further, there is filament voltage on the 83 rectifier tube but it is stone cold dead. I looked for replacement tubes on ebay, found these ran around $30 each but a kit to create a solid state rectifier that plugs in as a replacement was only $12. It should arrive toward the end of the week.

Next up is diagnosis of the lack of meter movement. Separated meter from circuit - wide open circuit. Something burned out inside. Ouch! These are unobtainium parts. My plan of attack is to first open it up and see if I can find the open circuit somewhere external to the meter winding where it might be fixed. Secondly, I would need to find a close enough alternative and swap the faceplates.

I have removed the meter but have to figure out how to open it nondestructively and then inspect it carefully to find the cause of the open circuit.

Detailed oscilloscope images comparing packs that boot and those that don't


On Saturday we had guests over at the Alto restoration meeting - Bob Sproull who wrote the Alto OS among other things, plus John Shoch again. So many great memories, anecdotes and tips. After they left,we captured traces of the data and clock patterns coming from two cartridges - one that I wrote, which fails to boot, and one with original data patterns which boots fine.

Nothing obvious jumped out of the patterns so far, but I will continue studying them for any significant variance. We began at an overall level without sufficient detail, then zoomed in first to the beginning of the header record preamble, subsequently to the end of the header record with its postamble. 
Alto written cartridge reading multiple sectors successfullly
Cartridge which I wrote, during a failed boot try

One thing that is immediately obvious is some data bits that occur right at the front of the preamble on the Alto written packs. By 'spec' the preamble should be nothing but 34 zero words before the sync word. We zoomed in to examine this in more detail.
Good cartridge, zoom into header record of the sector
Failing cartridge, zoom into header record
The views above show the entirety of the header record, which should proceed as follows:

  • Initial delay before beginning to write (lack of clock pulses)
  • 34 words of all zeroes (16 x 34 clocks with no data pulses)
  • sync word of 0000000000000001 (first 1 bit)
  • two words (32 bits total) of the header record
  • checksum word of the data words XORed together and with see of x0151
  • postamble of zero data until clock pulses end
We zoomed further into the beginning of the preamble to compare the two:
Start of preamble on good cartridge boot
The good cartridges have some random data bits - repeatable exactly on the sector but different patterns on each sector - that start the preamble. Reading a pack involves delaying through 21 words of clocked in data, thus these are not looked at as far as we know. The bad cartridges have no one values at all. 

The cause of this is believed to be residual magnetic flux on the media during the first 25 us of the write operation. On the Diablo heads, the erase head poles trail the read/write head poles by a distance that equates to 25 us at the rotational speed of the disk. When both write and erase gates are turned on simultaneously and zero data words are written, the first 25 us of old information is not erased.

We timed the delay from the sector marker pulse until the first clock pulses is recorded and they are essentially identical. The Alto written packs have some very slight variance between sectors and packs I wrote have sectors that are all exactly the same.

Alto written sector, with delay from SM to first clock pulse
Delay on cartridge written by the disk tool
The tail end of the header record was examined next. This should begin with a single one bit value that represents the end of the sync word, followed by 32 bits of the encoded cylinder/head/sector number, followed by a 16 bit checksum, and some postamble or trailing words of zero.
Good boot of sector zero from sync word to checksum and beyond
Cart that fails to boot, sector zero sync, header, checksum and beyond
The two patterns above are identical except that the postamble clock pulses (those after the last bit of the checksum) are four long on a good booting pack and three long on cartridges that fail to boot. 

First pass through all of this flagged only two difference, neither of which 'should' matter. Residual non-zero stuff at the start of the preamble and the difference of one clock pulse on the postamble. Time to study and count pulses much more closely.