Eico 315 RF Signal Generator Redux

I'm troubleshooting a problem with a Eico 315 signal generator. I described my restoration effort in a previous post

This has a particularly sentimental meaning to me, as Alan Douglas gave me this generator before he passed away. I presume it's the one that was featured in his book "Tube Testers and Classic Electronic Test Gear".

Sometimes the signal generator works fine, but fairly often it drops out of oscillation on band E when tuning below about 17 MHz. Above 17 MHz it oscillates just fine. I put a scope on the oscillator output and noticed that on bands C and above, the oscillator amplitude increases as you tune higher in the band. At the low end of bands C and D, the output looks pretty much like a clipped sine wave (i.e. a square wave), but as you tune higher, the waveform looks to have less third harmonic and more second harmonic. On band E, the waveform looks 2nd-harmonic-ey over the whole band. 

At the bottom end of band E, the amplitude was down to about 250 mV peak to peak. Bands C and D range from about 500 mV to 2 volts peak to peak, while the lower bands have a pretty constant 500 mV square-wave-ish output.

I tried a new 6C4 and that brought the low end of band E up to about 500 mV, so I'm guessing the problem was just a weak 6C4. My tube tester is on the blink, otherwise I'd check the transconductance. The generator seems to be working OK, but it'll take a while to be sure the intermittent drop out is really gone.

The Eico 315 uses a Hartley oscillator design. I've converted the BAMA deja-vu files to PDF's and put them on my web-site, and extracted the schematic as a stand-alone PDF for the readers convenience.

I don't have much RF experience, so I don't understand RF very well. My question is why does the amplitude of the Eico Hartley oscillator decrease as you tune lower in frequency? I don't have a spectrum analyzer, but guessing from the symmetry (or lack thereof) of the waveform, it does seem to transition from primarily third harmonic to primarily second harmonic. Why would the oscillator do that? Why would it not make that transition on band E? 

Is the oscillator power output actually decreasing at the lower end of bands C through E, or is the peak to peak voltage amplitude change due simply to the changing levels of the second and third harmonics?

Why would the oscillator drop out at the low end of the band due to a weak tube, rather than at the high end?

I should get my homebrew AD8313 RF power meter working and try to make some power measurements to eliminate some of the confusion from the changing shape of the waveform. Or else spring for an RF spectrum analyzer, but that's really not in the budget :-(

I noticed that the new tube didn't cure the problem. The drop out condition is thermally-sensistive and doesn't occur until the signal generator is well warmed up, so I was fooled into thinking I had it fixed. So this is no longer simply idle curiosity - I want to finally fix the signal generator.

Over on ARF, Chuck K7MCG had this suggestion:

Just a long shot: can you make the leads of C4 as short as possible. Also try adding a ceramic capacitor with value ~C4/10 (that is, around 0.0047 mf ) in parallel with C4 . And verify that the ground lug is truly well grounded. Small capacitances and inductances introduced by components and lead dress can cause grief in oscillators expected to work over a wide band.

good luck,

Chuck K7MCG

Chuck, I haven't tried bypassing C4 with something low-inductance like a ceramic disk capacitor, but I did scope the B+ line and saw a fair amount of RF. (Drat, I should have taken notes, I don't remember how much. Less than a volt, anyway.) But I didn't see a big increase in RF as I tuned to the bottom of band E, so I don't think that's the source of the intermittent drop out.

I sniffed around with my Measurements Megacycle Meter grid-dip meter and didn't see any parasitic resonances near the bottom of band E, but then again it's hard to get the dipper into the chassis and get it coupled to all of the possible trouble spots.

I read up on grid-leak biased oscillators in William Edson's book "Vacuum-Tube Oscillators". He has a very interesting chapter entitled "Intermittent Behavior". One item discussed therein is the relationship between the Q of the tuned circuit and the time constant of the grid-leak bias for the oscillator. Edson shows that to avoid intermittent loss of oscillation, Q >= pi*f*R*C. 

I checked the resistor and the capacitor in the grid-leak circuit. They are OK and the values are within spec, so something is causing the Q of the tuned circuit to drop at the lower end of band E.

I thought it would be interesting to try to measure the Q of the Eico 315 tank circuit at various frequencies, so I set up a parallel tuned circuit Q measurement along the lines recommended by Wes Hayward in his "Experimental Methods in RF Design" book. Wes recommends using very loose coupling of both the signal generator and the RF power meter to the tuned circuit, which results in a very high insertion loss when making the measurement. The idea is to load the tank circuit so lightly that the measured Q doesn't need to be corrected for the impedance of the RF source or the power meter.

I don't have a sensitive RF power meter on my bench that works up at 15 to 20 MHz. I've got an AD8307 RF power detector board from SV1AFM, but I need to package it up in a good RF-tight enclosure and add a power supply and metering circuitry. In the meantime, I tried using both channels of my Tek 465 in series to get enough gain to make a measurement. 

I discovered two things: first, I've got such bad RFI in my shop that unshielded measurements are worthless. Second, I noticed that despite driving the Eico 315 tank circuit from the relatively clean output of my URM-25F, the resulting signal on the scope had huge amounts of distortion, mostly second harmonic at a guess.

I wasn't able to get Q measurements that I can trust due to the distortion and RFI, but the distortion I observed points to a bad ground connection or bad solder joint that's acting as a rectifier. Why that would only affect the bottom end of band E is still a mystery.

So the next steps are:

1) make some better Q measurements by increasing the coupling in the Q measurement circuit to get less insertion loss and finishing up the AD8307 RF power meter to use as the detector, and

2) inspect the grounds and all the solder joints in the tank circuit, and the contacts on the variable capacitor. (I've already cleaned the band switch, and I did Deoxit the variable cap when I first restored it.)

© Steve Byan 2011-2016