motorboating

[jayme]

I've solved all of the issues with my 6SN7 Aikido preamp except for the very low frequency motorboating. It is below audible range, but causes the woofer to swing in and out (wobble). It is visible in both channels, so I assume that it has to do with the power supply.

The PSU is CLCLC. Upping the middle C from 20uF to 120uF seemed to help, but upping the last C from 20uF to 120uF didn't do anything. I've also read that increasing PSU capacitance only lowers the frequency of the oscillation (moving it below the cutoff frequency of the output coupling caps) and doesn't actually solve the underlying problem.

I don't have a scope, unfortunately. Just a DVM, so I need to resort to a try-it-and-see approach to determine the cause of the oscillation.

I searched a number of places. One post suggested placing 100R 10W resistors in parallel with the chokes to see if that has any effect.

Any other thoughts or possible avenues of exploration?

[SY]

That's an interesting problem; your Aikido is built to Broskie's exact schematic? Values of the cancellation string cap and resistors? Supply common to all stages? What's going on with the heater supply? Stable? Biased to DC properly?

A scope would help see if the underlying cause is an ultrasonic oscillation, though a 6SN7 is a less likely candidate for that sort of trouble.

[kevinkr]

Are you using a tube or solid state power amplifier? (Solid state amp with response down to or near DC?)

Could be classic motorboating, but there are other possibilities.

A scope would be most helpful. You might be surprised to find that the source of your low frequency problem could be very low frequency amplitude variations in your ac mains. I can see this quite clearly on heavily filtered high voltage dc power supplies loaded right into resistors.

Should this prove to be the cause regulated supplies may help - I'm typically able to get at least a 30dB improvement below 20Hz by going to tube based regulation, and much greater improvements are possible with a monolithic like the LM317.

Filament supply ok? Have you subbed a different set of tubes, and if so is the problem the same? (Some tubes generate more LF noise than others.)

Smaller coupling capacitors may help here as well. Other potential causes SY has already mentioned.

[jayme]

My Aikido is based on Bas's boards from this forum.

The HT supply is a 250V mains transformer (which also has heater windings), into:
5AR4
3.3uF cap
15H choke
120uF cap
then splits into separate Left and Right filter stages of
15H choke
20uF cap
which gives about 320V B+

The LT supplies are all AC for the 6.3V 6SN7s and the 5V 5AR4. I use the heater windings on the mains. One 6.3V winding is used for all of the 6SN7s, and it is biased up to about 90V to meet heater-cathode requirements.

There are additional heater windings which are not used (another 6.3V and 5V). They are not connected to anything.

As far as the Aikido circuit, it is all pretty standard, using the recommended values by Broskie. I'm using Kiwame resistors for the signal path, and I used some carbon comps for the 1M safety resistors. Coupling cap is .47uF, and the cap for the voltage divider is 0.1uF.

My amp is an EL-34 based tube amp (VTL ST-85) with an input resistance of 100k ohms.

When I hooked the preamp outputs up to my computer soundcard, I can actually record the signal waveform using SigView (trial version). I can record a steady sine wave that oscillates at about 2Hz between 0V and -.6V



If the problem IS actually a mains problem, what is the best way to measure it safely?

(Also, I'd rather not reduce the size of the coupling caps. That would just treat a symptom and not the actual problem. Besides, this preamp has amazing bass slam, and I'd hate to move the corner frequency up and reduce that...)

[pinkmouse]

Every time I've had problems with LF oscillation, it's been a grounding issue. I'm not saying there aren't other causes, but it's worth checking thoroughly.

[SY]

PSUD sims (I guessed some values) don't look promising for B+ supply stability. Have you looked at this, plugging in the actual values of DCR and the like for your components? I think you can end up simplifying that PS considerably and improve performance.

[Brian Beck]

I think SY has found the point in PSUD. LC...LC supplies can be "dangerous" because every time an L sees a C, a resonance is created. You have to make sure that the Q's are low. Any line noise or amplifier signal in the resonance frequency region can keep a high-Q resonance ringing like a bell. A regulator like Kevin suggests would probably stop it dead.

Here's a link to a previous post where I showed the impedance peaks of a typical LC...LC supply, showing the frequency domain view instead of PSUD's time response:

LCLC supply impedance

[jayme]

I have actually simulated with PSUD quite a bit. And, honestly, my real-world experiments do not necessarily sync with it.

First, my final stage in SY's model should actually be H=7.5H and C=40uF, since the L/R are in parallel. The Hs should be halved, and the Cs should be doubled.

As far as modelling, my performance with my INITAL PSU design looks quite decent in PSU:
5AR4 -> 3.3uF -> 15H (60R) ->20uF -> 7.5R (30R) -> 20uF

[jayme]

Originally posted by jayme
I have actually simulated with PSUD quite a bit. And, honestly, my real-world experiments do not necessarily sync with it.

First, my final stage in SY's model should actually be H=7.5H and C=40uF, since the L/R are in parallel. The Hs should be halved, and the Cs should be doubled.

As far as modelling, my performance with my INITAL PSU design looks quite decent in PSU:
5AR4 -> 3.3uF -> 15H (60R) ->20uF -> 7.5R (30R) -> 20uF

Accidentally posted early.

Anyways, the CLCLC with a 20uF middle cap shows dramatically lower ringing during a step change, a better voltage response on PSUD2, but a heck of a lot worse performance in real-life: dramatically more woofer movement. I actually REDUCED the woofer movement by increasing the middle C, which is non-intuitive if you model in PSUD2.

And, attempts to smooth out the ringing in PSUD2 by increasing the 3rd C have little to no effect on the actual woofer movement.

Hence my confusion on how to solve this problem. PSUD2 doesn't seem to relate to my situation.

That said, I am still concerned about Q factor of my power supply. What is the Q calculation for a CLCLC filter?

[SY]

My goodness, are those really your choke DCRs? They must be quite large. In any case, you can't really talk about the Q of the whole shmeer, just of bits and pieces of it. Likewise, the supply, having two branches will not be able to be modeled accurately, but half-@$$ed attempts like mine do point to the root cause, which is too many filters running around loose for a preamp with good PSR.

What you really want to do (assuming your goal is to have a quiet, reliable, good sounding preamp) is knock this supply down to about half its current complexity and size. I've got a 6SN7 Aikido using a very simple supply (ss rectifiers, CLC) that is dead quiet.

If you're just too much in love with the topology and the sight of all that iron to clip about half of it out, then add some series resistance to the chokes and you might be able to smooth things out at the expense of some B+.