• WARNING: Tube/Valve amplifiers use potentially LETHAL HIGH VOLTAGES.
    Building, troubleshooting and testing of these amplifiers should only be
    performed by someone who is thoroughly familiar with
    the safety precautions around high voltages.

Oscillation in tube amps

Just a couple of observations:

cathode:
- RF reactances in the cathode are common-mode to input and output, and are likely to make oscillation more likely. So I think that it's best to remove the ferrite bead, and add ca. 10Ω in series with the 1nF cathode bypass.

Grid, G1
3.9kΩ is too much for the G1 stopper, and can encourage feedback through the internal A→G capacitance route (by increasing the grid's shunt load impedance).
I would try with a simple 100Ω resistor for G1, and check that the grid wire has only a short route before meeting a lower impedance drive circuit or impedance.
The grid wiring must not run in parallel with any G2 or Anode wiring - use some grounded shielding foil to keep them apart.

You can also try adding 47pF to 100pF + 10Ω in parallel with the grid-leak R, mounted on the circuit-side of the grid stopper R.

Grid, G2
Since the G2 draws substantial RF current during VHF oscillation, ferrite beads are very effective.
Choose the bead to suppress the most likely spectrum (50MHz - 350MHz). A leaded muti-turn bead is worth trying if the type you have now does not control oscillation, since these reach down to lower frquencies.

Anode:
The leakage inductance of the IT can make trouble. A 4.7 to 10Ω wirewound resistor or multi turn bead are worth trying here.
Decouple the anode power supply near the IT with a 100µF electrolytic - try it without any small bypass caps, since the Electrolytic's ESR can be helpful fr damping RF currents.
The ground from cathode R ( and 1nF bypass) to the decoupler's negative side should be short!
 
  • Like
Reactions: 1 user
Member
Joined 2019
Paid Member
Thanks @Rod Coleman , I'll give your ideas a try.

Simple changes I made this morning - removed cathode ferrite, reduced G1 stopper to 100ohm, and placed a slip-on ferrite between the two 100ohm stoppers on G2 yielded similar results, still oscillating. I have some 10ohm 1/8W CC resistors on the way.

In regards to placing 10ohms in series with the 1nF bypass on the cathode, I assume you mean like below as a HF RC filter, as opposed to using 10ohms unbypassed as a cathode stopper.

10ohm series.png

I did at one point try a 100uF electrolytic bypass as opposed to the PP cap feeding the IST thinking the more lossy electrolytic might help. Results were similar, however my ground leads to the bypass cap were still quite long admittedly. I'll pick up some multi-turn beads and try the grid RC today as well.
 
Member
Joined 2019
Paid Member
Yes, I did try both configurations, along the way I tried G3 to anode via 100R CC resistor with the same result. I think G3 to anode actually increases transconductance slightly. Dave (Slagle) uses G3 to cathode, so I've decided to move forward with that. Also internal shielding of the D3a is connect to G3. Very little of the things I've tried has changed behavior measurably. I had hoped my changing to ceramic sockets with shields would help, but alas it did not.

I'm keen to try the 10R CC in series with the 1nF cathode bypass. I'd also like to try a 10R cathode stopper left unbypassed. If nothing else, it will knock down the transconductance of the D3a a little bit, maybe to my benefit!

I tried 56pF in series with 10ohm across the grid leak resistor without improvement. Will try some other values as well, believe I have some 100pF caps on hand.

Also noted that behavior is the same with or without the AVC in place. Straight from my DAC to the amplifier with -36dBFS gives the same result as with AVC attenuation.
 
Last edited:
Member
Joined 2019
Paid Member
@Potentiallyincorrect no GNFB loop in this amp, only feedback is local in the cathode follower output stage. But yes, it is relatively low frequency, which I assume is being modulated by a higher frequency oscillation.

1nF tried across the grid resistor no improvement unfortunately. Just for the heck of it, I disconnected the cathode bypass cap, same oscillation behavior persists. I shortened by ground connection from the PP power supply cap with its ground return to the center pin of the 9-pin tube (same as the cathode resistor), no change.

Will try 10R cathode stopper and the 10R-1nF RC on the cathode when I have some 10R CC resistors next week. Will try the multi-turn ferrite beads as well and some shielding for G1 wiring.

If that doesn't get me anywhere, I may try operating points similar to the datasheet - i.e., putting a cap on the input, putting a 9V battery under the grid leak resistor, and increase Rk to 820R, as below. I've read a few comments across the forum that this may help stabilize.

HY69 Schematic.png
 
Last edited:
Member
Joined 2019
Paid Member
I wish that were the case, but I've measured many amps with my current measurement rig, I feel very confident it is oscillating. Low frequency noise performance is quite good in the amp, but there is a spray of high frequency noise on FFT, just as is seen in the FR plot on the previous page. And I can see active oscillation in the audio band on square wave on my scope.

I tried the 9V battery setup, no bones, same behavior. Going to give @Rod Coleman 's ideas a go, placed an order for a range of multi-turn beads. Far wider effective frequency range and substantially higher impedances. They are larger, that's the trade off. Hoping they get the job done, will report back next week.
 
Does your scope allow you to expand the region where oscillation is visible, to allow a better estimate of oscillation frequency?

If it is the a.m. radio range, then a simple transistor radio may provide an independent confirmation, and even some direction sniffing. Or alternatively, use your scope probe tip as a sniffer - perhaps with some heat shrink over the spiky bit in case you accidentally touch something.
 
  • Like
Reactions: 1 user
This is a bit of a tough one!

Another thought....
The IT may show high impedance & gain at oscillation frequencies, or even unintentional resonances.

Maybe a simple series RC across the primary could be added, to keep impedance low above 100kHz or so, and taper the gain to low levels at HF & VHF. You already have the circuit in LTSPICE so it will be easy to optimise.
 
Last edited:
Are we fersure, fersure that the regulator isn't the culprit? C5 is big enough that it may be inductive at a frequency where the regulator could oscillate, and it looks a lot like an oscillator. Ham radio guys called the low frequency results of high frequency oscillation "squeeging", where oscillation builds to the point where the valve (in those days) cut off, then relaxed, then resumed oscillating, with a repeat time in the audible, or so, range.

All good fortune,
Chris
 
I missed the Maida in the anode supply. Yes, that definitely needs checking out - the stacked power devices in the Maida scheme results in precariously low phase margin.

This regulator needs at least 5 ohm resistor in series with the output capacitor, and added capacitors for local decoupling should also have series Rs. Simply trying a 10 Ohm resistor in the supply to the load might be worthwhile.
 
Member
Joined 2019
Paid Member
@trobbins my scope is a somewhat basic Siglent, rated to 200MHz. I've tried to nail down a higher frequency oscillation but unsuccessful so far, I will look again.

@Chris Hornbeck that thought had occurred to me, I've used these regulators successfully in many other builds (including a recent one that used the same cap at C5), but that doesn't mean it isn't oscillating. Two things I've done so far - I tried a 100uF electrolytic at C5 in place of the PP cap without any difference, I've also bypassed the 2K resistor and current sink with a 5K resistor to rule out any possible contribution from the current source, with again no difference. I'll try some other values of electrolytic, and also try @Rod Coleman 's suggestion of adding some series resistance.

I'm working over the weekend so won't be able to get to these things until Monday, I appreciate the input, thank you.
 
Member
Joined 2019
Paid Member
I did a number of things and collected some data, the oscillation persists.

I added multi-turn ferrites to the plate and G2 to plate connections. Also added a 10R (metal film) to 1nF RC to the cathode as discussed. No change after those moves. I then I added a multi turn ferrite to G1 and increased the G1 stopper value from 100R to 330R without any improvement. Then replaced the 1nF cap in parallel with the grid leak to a 10R to 100pF RC. Again, no change.

Here's how the socket is wired currently.

HY69 Schematic.png PXL_20240108_234659982.jpg

Next, I moved toward ruling out the Maida regulator. I added 10R series resistors at the output of the regulator and from the D3a stage cap to the interstage transformer. No change. I took it one step further and removed the regulator from the circuit - I fed the amp directly from the reservoir cap bypassing the regulator with the same oscillation pattern. So I feel very confident now it is not the regulator.

Then I worked out nailing down the higher oscillation frequency. With the scope probe on the HY69 grid, if I pulse the amplifier with a 1kHz square wave, the audio frequency range oscillation is easily teased out.

SDS00027.png

Looking at higher frequencies, I saw that when the amp is connected to the DAC/ADC I use for testing, I see a mix of frequencies in the 600kHz - 1.7MHz range.

SDS00020.png SDS00023.png

Interestingly, these don't show up when connected to my signal generator, only the DAC/ADC. But again, a square wave pulse from either results in audio range oscillations. I did detect some 150-200MHz hash but it is present whether the amp is on or off, so I think it is simply noise, whereas the 600kHz-1.7MHz frequencies are only present when the amplifier is powered and the DAC is connected to the RCA input.

So perhaps I should be targeting a lower frequency range with my efforts. One thing I have not pursued is DC heaters. Right now the heaters are AC, but DC elevated with ferrites at the leads and bypassed to ground with 1nF ceramic caps. The leads are quite long though and the grounding point for the DC elevation voltage divider is in the back of the chassis.

Voltage divider-2.jpg

Thanks for your input.
 
Last edited:
I see a mix of frequencies in the 600kHz - 1.7MHz range.
That's too low a frequency for ferrites to be of help, and unusual for self-oscillation.

If it only shows up with the DAC, is the loading on the DAC capacitive? Long cables? Many opamps lose phase margin under fairly small C loads, unless there is a 51 ohm resistor at the opamp output.

Could also try rolling off the gain of the D3A stage with a RC across the IT primary.
 
  • Like
Reactions: 1 user