Knight KG-630 (Rebuilt and modified).
Introduction: I bought this on impulse (on ebay), it caught my attention because it had been heavily modified with a transistor triggered sweep that I was interested in. If you read this post, you will be taken through my own journey of modifications aimed at improving the original repetitive sweep generator to be accurate enough to allow useful estimation of the applied signal frequency directly from the screen. Here it is showing a tone burst after I had re-built and modified it, (the brightness burst is due to the slow sweep speed interacting with the lighting and shutter speed):
It was very badly packed with minimal protection, here is the wreckage that I received:
The construction is extremely flimsy. You might observe the right angle clips with rubber grommets of which 4 were used to “support” the front of the CRT. It was dropped as evidenced by the chassis, bent downward by the mass of the power transformer:
I set it aside in disgust, then a nice 5EDP1 flat faced CRT came up for a fair price. It will work in place of the original 5UP1 with better sensitivity so I bought it, and that provided sufficient inspiration to take this mess on! I refused to pay for a manual for this, I am glad I did not for I was pointed to the KG-635 manual on Bama and neither the assembly or the operators manual includes schematics! In view of the damage, I decided to remove all the modifications and start by returning it to the original state as far as was possible given some missing parts, this required tracing the circuits and that took a bit of sleuthing. Having done that, I embarked on my own course of modifications. It became the fun modification project I had intended when I purchased it (rather than my usual repair, restoration and calibration activities). After all, it is a toy oscilloscope so why not have fun with it!
Both the PCBs were badly cracked and I had to repair many traces. I had a nice camera bezel from a Tek scope (that I replaced with an original bezel) that fits the front of the 5EDP1 perfectly with felt pads. Conveniently, it also fit the Knight bezel studs so I mounted it on the inside of the front panel to support the front of the new CRT. (The Knight bezel was clearly intended to ape Tektronix.) I also added support ribs either side of the flimsy front panel, a cross rib at the power transformer and braces to stiffen the support for the socket end of the CRT. I placed them the way shown because I did not want angle braces blocking access. The double bolted friction at the joints with the CRT support has proven quite effective:
From a mechanical point of view, this is a rather cheaply designed kit, however in context, it probably provided quite good value. It is AC coupled and I measured the HF -3dB point at 1.1 MHz, again quite good.
Both the X and Y deflection drivers are push-pull, using a 12BH7 for the Y axis and a 12AU7 for the X axis. Both channels have pre-amplifiers, in the case of the Y axis, two cascaded 6BQ7 stages that are biased by contact potential using 10 M grid resistors and no cathode resistors, the X axis uses a single 12AU7 stage. Both axis are provided with cathode followers at the front-end. The Y axis also has a 3 decade compensated attenuator.
The sweep generator is an astable multivibrator originally using a 6BC5 pentode for the switch and a 12AU7 for the timing capacitor charge (flyback). It is designed to run-down about 10% of the capacitor charge voltage which is one way to obtain adequate linearity, the first 10% of the exponential curve is pretty much a straight line. The small run-down amplitude is the reason for a pre-amplifier after the cathode follower in the X system. The flyback pulse is taken from the charge tube plate and amplified using a 12AU7 section, the frequency response of which results in poor flyback suppression at high sweep frequencies. The flyback suppression, which is of low amplitude and negative going at source, is amplified, clipped and inverted, then applied to the CRT cathode. The amplifier has zero bias so it will only produce a positive going output pulse.
How The Sweep Generator Works:
There are two tubes, let’s say left and right. The plate of the left tube is connected to the grid of the right tube while the plate of the right tube is ac coupled to the grid of the left tube. The timing capacitor is in the cathode circuit of the right tube with the run-down resistor (usually variable) in parallel with it.
Consider that the left tube is “on”, its plate is low which turns the right tube “off” effectively disconnecting it from the timing capacitor. In this state, the timing capacitor discharges, via the run-down resistor, this is the sweep state. The run-down proceeds until the potential of the cathode of the right tube falls to the point where it turns on, recharging the capacitor. As the right tube turns on, its plate potential falls sharply and the grid of the left tube receives a negative going pulse via the coupling capacitor, turning thereby turning the left tube off, this is the flyback state. The flyback continues until the cathode of the right tube goes positive with respect to its grid, disconnecting it once again, the turn-off causing a positive pulse at its plate that is communicated to the left hand tube, turning it on again and the circuit returns to the run-down state. The synchronisation signal is applied to the screen grid of the left tube. The pulse at the plate of the right tube is somewhat square and is generally used for the CRT flyback suppression.
While we are in this discussion, it is worth noting explicitly that though this is an oscillator, it is of the multivibrator (MV) type rather than a resonant circuit in a feedback loop. The frequency or repetition rate is given by the reciprocal of the period which is the sum of the run-down time and the flyback time. The run-down time is governed by three parameters: Value of timing capacitor, (frequency range setting) the discharge current (frequency vernier or multiplier) and the amplitude of the run-down. The flyback time also depends on the value of the timing capacitor and the amplitude however, since it is flyback, the capacitor is charging which is driven by the available recharge current.
In brief, the timing capacitor value combined with the discharge & charge current values governs the time the MV takes to go from one state to the next for a given amplitude of the waveform.
Here is the schematic of the (modified) sweep generator which may help:
1. B+ Supply:
On 120VAC, the voltages were too high, in fact the reservoir voltage was over 500 into a 450 V rated capacitor so I revised the power supply using a smaller reservoir capacitor to reduce the voltage and a choke filter,
2. HV Supply: Added a voltage doubler to the HV supply to increase the acceleration potential to around 2.1 kV. (Maximum for the 5EDP1 is 3kV.) Given the substantially higher sensitivity of the 5EDP1 over the 5UP1 this made sense, the maximum sensitivity is a useful 20mV/Div, this change also yielded an improvement in trace definition. To improve safety a little, I also replaced the metal shaft intensity pot with one having a plastic shaft,
3. B+ Regulation: There was a B+ regulator device (for the sweep and X and Y cathode followers) that sampled the HV (CRT supply) that I surmise was intended to modulate the gain of the front end to minimise sensitivity variations with HV changes. Frankly, I did not evaluate it because this scope is not an instrument, it cannot be calibrated. As is normal with cheap scopes of the tube era, the trace tended to move around the screen a bit so I replaced the 6C4 with a OA2 gas tube stabiliser*. This improved the trace stability a lot. I also had the stability and repeatability of the sweep generator repetition rate in mind since, as noted above, the repetition rate is strongly dependent on the ramp amplitude. Note, it is necessary to disconnect the heater supply at pin 4 to use an OA2 in the 6C4 location.
* The voltage at my bench bumps around quite badly, this is actually helpful because if the work I do cannot tolerate this, then speaking frankly, it is unfinished.
4. X and Y Amps: I spent a lot of time fiddling with the X and Y deflection driver supply voltages, aimed at good linearity also, similar deflection plate voltages to minimise trace distortion. In this process, I also found that increasing the cathode (tail) resistors helped, not just with linearity and balance but also bringing the plate dissipation of the tubes within limits,
5. Y Gain Control: The coupling capacitor from the Y cathode follower to the gain control was an extremely leaky 100 µF electrolytic into a 1.5 K “gain” control resulting in the trace zooming off the screen everytime I touched the control, gradually reappearing as the cap charge returned to equilibrium. I decided to eliminate this by making the gain pot the cathode follower load. The gain pot was 1.5K and I replaced it with one that measures 2.2K which is nearer to the original 2.7K load resistor. The new operating conditions resulting for the 6AB4 are fine. This is a view of the Y cathode follower side of the main PCB, also showing the new sweep frequency range switch and the new CRT front support arrangement:
6. X Pre-Amp and Shift Control: The X pre-amp is a 12AU7 amplifier. The 12AU7 is not a particularly linear tube, especially at low voltages and the existing set up with a 10k load at 80V was compressing the sweep to the right. In fact it rendered the linearity of the re-designed sweep generator pointless! I reduced the load to 50k and increased the supply voltage to 340V, this moves the loadline into a much more linear region, the linearity while not perfect is acceptable to me. Similarly to the Y channel, the X CF was coupled to the X gain control via an electrolytic. I changed it to couple to the X preamp via a 100nF cap and re-located the 50k gain control as the load for the X preamp. Here is a picture showing the good linearity and sharp focus:
I also noticed that the shift control was much too sensitive so I reduced the range of the control considerably. It will now move a full width trace fully from one side to the other rather than two or more screen diameters.
7. Sweep Generator Tube Type: The tube for the sweep generator was missing (6BC5), I replaced it with a 6BH6 which has the suppressor grid brought out to the base. I found the generator seemed to work best with the suppressor grounded rather than connected to the cathode. This means that a 6BC5 cannot be plugged in directly in place of the 6BH6,
8. Test Signal: I added a diode clipper from the 10K resistor that supplies a line signal from the heater line to the sync selector. This gives a 1.1 V P-P squarish wave that followed with a 1K / 10K attenuator provides a reasonably accurate 1 V P-P test signal at the front panel.
9. Sweep Generator: The timing caps and fine control were missing due to the previous transistor sweep generator replacement. I worked up a new design with 10, 30, 100, 300, 1k, 3k, 10k, 30k and 100k frequency steps. The fine control is replaced with a variable two-terminal cascode mosfet constant current sink to control the discharge rate of the timing cap and to run-down with excellent linearity. Here is the new sweep frequency range switch:
Here is the frequency multiplier mosfet variable current source:
The idea is to obtain accurate frequency range settings, that when combined with an accurate frequency multiplier control, may be used to obtain synchronisation with minimal signal amplitude (so as to avoid pulling or pushing the sweep frequency), such that the frequency of the subject signal may be estimated directly with a fair degree of accuracy. (This is quite unlike the crude frequency controls on the general run of cheap repetitive sweep oscilloscopes.) This actually worked out after quite a bit of playing around. I arranged the control to multiply up to 4x to cover each 3 to 10 step completely. Having done this and marked a temporary scale, I set a signal generator blind to several different frequencies and used the system to estimate the signal frequency. Most of the time, I was around 13% low improving to 7% low at lower frequencies. The error was consistent which tells me that I possibly could train myself to do a lot better but still, it proves very hard to accurately esimate signal frequency directly using a repetitive sweep oscilloscope. The problem is to accurately count the number of cycles displayed and clearly, I was under-estimating. The traditional method is to use a calibrated marker generator to put blank or bright spots on the trace, however, this is not a direct method. The message is once again, if it does not have an accurate triggered sweep, it is at best, little more than a toy.
If you decide to try this out, there may be a superior current regulator approach that could be a chip, though I would make sure that the frequency response is adequate. I used the cascode mosfet because I have used it in probably dozens of other places (check out my “Audio Notes” on http://www.richardsears1.wordpress.com if you are interested), it exhibits extremely stable current control over a very wide range of impressed voltage. Also note that the mosfets will fail if you somehow break either connection to the current regulator while it is under power. I have done this several times over the years and it is irritating!
The multiplier control requires an inverse power characteristic, to a first approximation, the law of the control looks like R = 35.37X^-1.1 where R is the current regulator sensing resistor in KΩ and X is the multiplier value (in this case from 1 to 4) that ideally corresponds to uniform increments of rotation. I found a reverse log pot to be workable. The actual discharge current ranges from 72 µA to 320 µA over the multiplication range.
Lastly, I added a frequency calibration preset control on the front panel since the calibration does drift somewhat during warm up. I did not want to put it into the cabinet, only to find my efforts rendered useless due to increased temperature. This preset is also necessary to accommodate replacement of either of the sweep generator tubes. It can be set at line frequency by ensuring that 6 whole cycles are displayed on the 10 Hz setting and 2 whole cycles on the 30 Hz setting. I also added a sweep out terminal to permit the use of a frequency meter to set the sweep frequency calibration. The calibration control was implemented by replacing the 1k plate resistor of V5a with an 850R resistor and a 250R RV4 type preset pot, wired as a variable resistor in series with the 850R resistor.
Here are the sweep controls:
The schematics for the KG-630 do not seem to be on the internet so these may help if you have one of these contraptions.