• 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

Member
Joined 2019
Paid Member
@gorgon53 I'll consider trying that, thanks.

@trobbins today I stacked four AA batteries across the heaters, left the 1nF bypass caps in place, left the batteries floating (did not ground the negative terminal). Same result. All of the testing I've done has been in only one channel, since it's impractical to test so many things in both channels simultaneously. They are also isolated from one another after the reservoir cap, i.e., each channel has it's own Maida regulator. The images from my scope above are at the IST secondary (HY69 grid connection) without the HY69 in its socket to isolate testing the D3a by itself. The only time I have placed the HY69 is when doing an FR sweep.

@Rod Coleman I have to wonder if the low frequencies I'm seeing from the DAC are a red herring. The cable connections are not long, they are Blue Jeans cables as well, which are good quality. I check FR using my DAC/ADC, square wave testing from my signal generator. The 600kHz-1.7MHz noise is not seen using my signal generator, but the same audio band oscillations are present when I pulse the amp with a square wave. I tried an RC across the IST primary (10ohm - 1nF in LTSpice gives a good drop off beyond the audio band), with no change.

@euro21 shorting the input and measuring at the secondary of the IST, I can measure RF 70-100MHz noise. This is the same I can measure simply holding my scope up in the air. Placing a loop of wire to make an antenna out of my scope, same frequencies are measure all over the chassis, higher near the ISTs. I live in a densely populated area and there is a cell tower less than a mile from my home, so perhaps I live in a very high RF area and the D3a cannot function well here.

SDS00035.png

Other things I tried today:
  • Straight battery bias with input cap, a single AA 1.5V battery under the grid leak and a 10R CC resistor on the cathode, cathode grounded with no bypass caps. Thought maybe grounding the cathode would have some stabilizing effect, it did not.
  • To rule out any interaction of the IST and the D3a, I disconnected the primary and loaded the D3a with a 5.1K resistor. Measured at the plate, same oscillatory pattern
  • Worth noting I see the same oscillation behavior using other high gm tubes like E810F and 6E6P, which are both pin compatible with D3a and can function in this circuit.

Possible this is being communicated through the mains? I have a low capacitance Topaz isolation transformer I could try connecting in front of the amp. Again, thanks for the input, sorry the results have been poor. At this point I am feeling very discouraged, maybe I need to change the input tube altogether, although this circuit requires something high gain and low enough Rp to drive a 60H IST primary, which by default means at least moderately high gm. D3a is the perfect fit, or even E810F, but I did not anticipate them being near impossible to stabilize.
 
To rule out any interaction of the IST and the D3a, I disconnected the primary and loaded the D3a with a 5.1K resistor. Measured at the plate, same oscillatory pattern
It is very difficult to imagine how such low frequencies can be generated by self oscillation, with such a plate load .100MHz, sure, but not 1MHz... Time constants for 1MHz will not be found from ordinary stray L and C, with a R load, especially now you have tightened the layout up.

Perhaps the spurious audio spectrum in square wave testing is generated in the power supply?
 
Member
Joined 2019
Paid Member
@Rod Coleman my gut feeling was to look elsewhere given no change whatsoever has occurred despite all of my efforts. I have made a meaningful observation.

I removed the D3a and did an FR sweep of just the HY69 output stage with the signal input at the HY69 grid and got the below.

HY69 FR (1).jpg

The audio band oscillations/ringing persist despite the D3a being out of circuit. I then removed the HY69 and did an FR sweep of just the D3a stage with the output taken at the IST secondary using my Linear Audio Autoranger for signal input and got the below.

D3a FR.jpg

Which is quite an improvement. This would seem to suggest the issue is primarily in output stage.

I first tried a Zobel network across the OPT secondary, which improved the overshoot on square wave, but the noise from 7kHz-20kHz and beyond is the same on an FR sweep. I then tried a 22ohm resistor at the plate top cap of the HY69 like you might see in an 807 amplifier with the same result. Increasing the value of the HY69 G1 stopper had no effect either. I will try adding a cathode stopper next.

Possible it is some interaction between the output transformer and the filament regulator, Rod?

Not shown in the schematic, but G2 is tied to the plate by 100R CC and G3 is tied to cathode at the negative end of the filament regulator, as is the top of the OPT primary.

HY69 Output.png
 
Member
Joined 2019
Paid Member
I'm willing to entertain that, for FR right now I'm using a MOTU M4 DAC/ADC paired with Jan Didden's Autoranger. But I've been meaning to pick up the Diglent Analog Discovery 3, I've invested a lot into this build, so seems as good a time as any to upgrade my audio analyzer setup. Worth noting that I've measured several amplifiers with this setup and have not seen these artifacts.
 
Not dissing your equipment. It’s just that any equipment can mislead. Not even confident that equipment is your problem. Rather than something assumed is not as believed. The symptoms as described are not responding to treatment that should at least significantly change them based on the diagnosis and the treatment attempted. It’s not adding up. In the end it will be something simple that has been overlooked. I wish I could tell you what it is.
 
Post #345 had measured sweep results of each stage, with some anomalous signal levels approaching 20kHz in the HY69 stage FR. Given focus on output stage, can the output stage be tested using the same measurement setup but with perhaps some base level configurations, to see if the same anomalous response occurs. Eg. output stage powered, but input and output probes at the same location (such as input and output). Another configuration could be FR from input to output but with anode connection removed. Another configuration could be FR from input to output but with HY69 grid grounded. Also not sure why the FR plot stopped at 20kHz ?

Given you are using a MOTU M4 and autoranger (powered from a USB feed), can you clarify if either or both are effectively powered from the mains ac, or say solely from a battery in say a laptop ? Stray noise and earth loops via the mains can be a concern, as well as quirky USB based comms and some forms of response aberrations (eg. such as in the FR plot of the D3a stage).
 
Perhaps it will be worth looking again at the impedance of both the grid and cathode-filament circuits from ground.
Both G1 and f are only returned to ground via large inductances - the impedance rises with frequency.
This can make the stage vulnerable to noise pickup - especially the filament side, since the transformer leakage is still present (even with the improved Toroidy). Leakage current is a common mode noise current.

Adding capacitors to ground might upset the performance, but could indicate whether this route is worth pursuing.
 
Member
Joined 2019
Paid Member
@trobbins sorry, yes the top FR plot was stopped at 20kHz, I adjusted the sample rate and FR sweep to max 48kHz for the second plot. Both the MOTU M4 and Autoranger are powered from a laptop.

@Rod Coleman this is my suspicion as well, that the noise I am seeing in the output stage is due to the filament having a high impedance pathway to ground. Even on 1kHz FFT, there is a spray of very ugly harmonics in the 6kHz-20kHz region, same as is seen in the FR plot. I recently built a cathode follower power stage with a similar schema, however it was using indirectly heated pentodes, and I'm afraid using directly heated filament may be a bridge too far.

Gentlemen, I am going to throw in the towel and convert the output stage to a conventional common cathode stage. The output transformer is well suited to load the plate of the HY69 (6.5K primary relative to the HY69's plate resistance of ~1.8K). The additional gain of the HY69 (mu 6.5) will allow me to convert the input stage to a more stable tube, likely 5687 / E182CC with the sections paralleled, which I have used even recently in a very similar circuit successfully.

I will battery bias the 5687 / E182CC input stage with a 3V CR2 battery and ground the cathode, the end result will look like this.

I'm confident this will solve my problem, whether it be self oscillation in the input stage, or high frequency noise in the output stage due to a high impedance ground pathway for the filament. Thank you again for helping me troubleshoot, it's helped me come to this conclusion, which I think is the best path forward.

HY69 Schematic.png
 
Member
Joined 2019
Paid Member
I took one more stab at the issue above and narrowed it down to the toroidal filament transformers. I substituted an EI core transformer with electrostatic screen instead of the toroidal and got a perfect FR without any high frequency noise. Problem is, these EI core transformers have so much leakage flux, the circuit is flooded with 60Hz noise, the reason I switched to toroidal in the first place.

FR using the EI core with signal input at the HY69 grid wired as a cathode follower.

EI Core FR.jpg

I went back to the toroidal, thinking perhaps maybe it is ringing and a better snubber might help. I used Mark Johnson's Quasimodo to dial in a perfect secondary snub, but alas the same noise persists.

FR back on toroidal with the snubber.

Toroidal Core FR.jpg

I've also tested the circuit in a common cathode configuration with the transformers in the anode and cathode bias rather than the cathode follower and this noise does not show up, it is only when wired as a cathode follower with the toroidal filament transformers.

These toroidals have primary-to-secondary electrostatic screening, overall the quality seems very high from Toroidy. Any thoughts on what it could be if it isn't coupled noise from mains? If the toroidals can't be salvaged set up as a cathode follower, I'll either have to go to EI and accept a 60Hz hum, go common cathode with the toroidals, or try a third round of filament transformers.
 
Toroids have much better frequency response than they have a right to as mains trafos. If the noise is coming in off the line you may have to filter it on the primary side. Common mode choke, X-capacitors. You may have a rectifier ringing somewhere else that is getting through this transformer.
 
  • Like
Reactions: 1 user
The problem with the EI solution appears to be pickup of stray field in a high-impedance cathode (plus high-Z grid circuit); and it's possible that the trouble with the Toroidy solution is similar.

The impedance of both grid and cathode rises with frequency (inductive); this may explain why only high frequency couples through from the Toroidy.
It is relatively easy to try countering the impedance rise with RC networks across them. It may be a matter of finding a balance between noise suppression and risk of FR limitation or sound degradation, but I suspect that low values of C will suffice, since the noise is coupling through low values of C.

It's possible that the HF noise couples from the Toroidy primary to grid or cathode wiring, so judicious foil shielding or shielded cable may help. The whole of the filament heater circuit is part of the cathode, so the coupling distance of all components of this must be considered.

The other possible method is to feed the Toroidy with a 1:1 mains transformer - an encapsulated type that lives outside of the enclosure. Now one side of the primary of the Toroidy can be connected directly to ground, and a cap placed across it. This should put an end to external HF noise.
 
  • Like
Reactions: 1 user
Member
Joined 2019
Paid Member
Toroids have much better frequency response than they have a right to as mains trafos. If the noise is coming in off the line you may have to filter it on the primary side. Common mode choke, X-capacitors. You may have a rectifier ringing somewhere else that is getting through this transformer.

I do have a filtered IEC in place from Schurter, for what it's worth.

Mains Filter.png

Perhaps an EI with mains winding screen, and with a secondary winding screen, or a C-core with each winding on its own side of core.

Thanks, will try to make these toroidals work first, then may have to investigate other transformers if not. This project has been quite the odyssey.

@Rod Coleman thanks as always, I have tried RC and C on both the grid and the cathode without success. Yesterday for the sake of seeing if it worth pursuing, I put 0.1uF and 0.05uF from cathode to ground, which effects HF bandwidth but does not seem to alleviate the noise.

Here is the 0.05uF FR. I've had similar results using various RC combinations.

HY69 FR 0.05uF.jpg

Here is a 1kHz spectrum of the output stage at about 1mW output into 300ohms. As you can see, very ugly noise floor and harmonics. With the EI core, it is clean and low distortion. Low frequency stuff is being picked up by my measurements, no hum in practice.

HY69 FFT.jpg

I like the idea of using an isolation transformer so I can ground one end of the secondaries of the toroidals. If that works, it would be the simplest solution for me. What value of cap do you think would be best across the secondaries?

Thanks again for the input everyone.
 
Last edited:
Member
Joined 2019
Paid Member
Sorry that was a typo, yes grounding the primary, obviously the secondary must be floating. I'll give it a try today with my Topaz transformer today, thanks. If it works, can get a smaller Topaz in the 250-500VA range. Will probably check for any leakage current from chassis to earth as well.
 
L0rdGwn,

You said: "Problem is, these EI core transformers have so much leakage flux, the circuit is flooded with 60Hz noise, the reason I switched to toroidal in the first place."

A few things that I have experienced that cause Power Mains frequency hum,
Any EI power transformer plus any EI output transformer (single ended air-gap is more sensitive than push pull).

Magnetic Steel Chassis:

Spacing between power transformer and output transformer

Orientation of EI laminations of the power transformer versus EI lamination of the output transformer.
Magnetic Steel Chassis conducts magnetic fields, and it can "turn" the field orientation, which can change the Best orientation of the EI power xfmr versus the EI output xfmr.

My one steel chassis has 1mV to 2mV hum.
My aluminum chassis have 100uV hum.
(Added advantage: I prefer to drill, nibble, and chassis punch: aluminum, not steel).
 
Member
Joined 2019
Paid Member
@6A3sUMMER thanks, it is an aluminum chassis. The 60Hz noise was due to magnetic coupling between the mains/filament and output / interstage transformers.

@Rod Coleman I tried your isolation transformer suggestion today with my MGE / Topaz transformer. Grounded one end of the toroidal primary, 1uF across the secondary. No difference unfortunately. Maybe this design just isn't viable with the cathode follower and toroidals given the high impedance path to ground. Works fine in common cathode series feed. Pains me to change it but perhaps it is the best option. Seeking out new filament transformer I might just find myself in the same predicament - LF noise with non toroidals, or HF noise with toroidals.

Here is the EI core spectrum for comparison.

HY69 Spectrum EI core.jpg
 
Last edited: