B+ Ripple

[loudthud]

I ran across this web page linked on diyaudio: AC Ripple in a Class AB Power Amp

I would have prefered more scope photos and maybe some sound clips. Anyway, I have a different way of looking at things. Below is an X-Y scope shot of a prototype Class A power amp I am working on. The rails are about +/- 21 V at the peak of ripple, the load is 8 ohms. The left photo is with 2200uF per rail. The right photo is with 4400uF per rail. Still could use more capacitance to compare to the 5F6A scope photo also posted below and the 5E3 pics I posted earlier in different thread. They don't show very much ripple at all except for a bright comet shape along the left and right edges. The class A amps have almost constant supply ripple so they won't give the comet shape.

I don't really hear much difference between the two experiments with the class A amps. I actually like it more when running from a regulated supply. Anybody got an AC30 I can borrow?

[R.G.]

I have learned to be deeply suspicious when someone in the audio field says things like "the character of the amplifier changed completely".

I'd really like to see some spectral analysis of the amplifier's output. If there's 100Hz affecting things, it'll be in there, perhaps as a 100Hz difference in two spectral lines, or some such.

The other thing I'd like to see is whether an amplifier running from heavily regulated power rails can have the missing "character" added by impressing some 100Hz (or 120Hz) ripple on the input signal artificially. After all, the electrons on the output don't know how they've been treated or mistreated. If the output voltage waveform has "character" with 100/120Hz ripple impressed on it along with the sag envelope, that can be done before a non-sag, not-ripple amplifier too. And for much less cost and effort than running a tube amp. In fact, if that envelope effect is what give it "character", then producing the envelope in any way should give the sound the same "bite" and "character".

Electrons have no memory and no soul.

[diagrammatiks]

I have learned to be deeply suspicious when someone in the audio field says things like "the character of the amplifier changed completely".

I'd really like to see some spectral analysis of the amplifier's output. If there's 100Hz affecting things, it'll be in there, perhaps as a 100Hz difference in two spectral lines, or some such.

The other thing I'd like to see is whether an amplifier running from heavily regulated power rails can have the missing "character" added by impressing some 100Hz (or 120Hz) ripple on the input signal artificially. After all, the electrons on the output don't know how they've been treated or mistreated. If the output voltage waveform has "character" with 100/120Hz ripple impressed on it along with the sag envelope, that can be done before a non-sag, not-ripple amplifier too. And for much less cost and effort than running a tube amp. In fact, if that envelope effect is what give it "character", then producing the envelope in any way should give the sound the same "bite" and "character".

Electrons have no memory and no soul.

A class A amp isn't going to exhibit the same effects. That article that pentode press talks about concerns AC modulation of power supply DC that's dependent on the AC signal voltage.

Overlaying a fixed ripple frequency would have the same issue that the study points out.

[trobbins]

Electrons do have memory!! I change the spin of the electrons with a special switch on the front of the amp - it gives a very different character ;-)

[R.G.]

Sorry - someone else invented that first. The Electron Spin Processor (ESP) is a feature of the top class of amps from a Very Small Amp producer, and is available for another $10k as an add-on mod upon request. There's a two year waiting list. They just didn't post the schematics to the internet because someone might steal them.

[R.G.]

Overlaying a fixed ripple frequency would have the same issue that the study points out.

Yep. That's what I was getting at. If it's really the overlaid ripple, that's easy to add without the bother of the whole amp to do it. And if that's (really...) the secret of an amp's "character", then it can be done solid state, simply. I can envision a version of the NE571 compander IC set up to do this, perhaps including the sag feature if you worked at it.

Sigh. So many effects, so little time.

[Steve Conner]

As I always point out when this subject comes up... I built a guitar amp with fully regulated power supplies, and it didn't sound anything like as bad as you'd expect.

[R.G.]

Steve, this only means that your counter-example will be dismissed as anecdotal evidence, since it claims no epiphanial glimpse into the intentions of the Tube Gods, and their anecdotal evidence will be seen as evidence that you're wrong and misguided.

[loudthud]

Playing the amp with a regulated supply does not sound bad, but it laks a little something. An earlier version with a cap coupled speaker sounded a little better. The recovery of the cap after hitting the speaker with a non-symetrical clipped waveform added a little thump when a note was cutoff. The next version I'm going to build should be 50W at 4 ohms. It'll use a fan cooled CPU heatsink.

I have a power supply big enough to run a 5F6A, 525V at 500mA. For sag, a simple RC on the output should do the trick. But simulating the regulation of the screen supply is not so easy. You would have to monitor the current on the main B+ to dial in the voltage drop and then figure out some way to do the ringing of recovery when current drops.

I want to build a circuit to clamp the speaker voltage at the resistive load peak voltage to prove that the extra voltage that a tube amp makes available to the speaker in fact makes the amp louder. Note that the +/- 20V peaks of the 5F6A would take a 100W at 2 ohm amp to reproduce without clipping.

[diagrammatiks]

As I always point out when this subject comes up... I built a guitar amp with fully regulated power supplies, and it didn't sound anything like as bad as you'd expect.

See that's my problem with that article...

I don't know if they really set out to prove that it sounds bad...but it seems to be the implied conclusion.

What they do say is that it doesn't sound anything like the original...

but that seems to be begging the question...why would it? the entire power supply is different.

So

also.

Is your fully regulated power supply for your tube amp using tube hv regulators?

if not you are out of the clubhouse.

[Gingertube]

Sorry Loudthud, I could'nt resist, you said "anyone got an AC30 I can borrow?"

--- I thought you were testing/comparing Class A Amps!!!! ----

I had thought we had quite thoroughly put a stake through the heart of the AC30 Class A myth.

Cheers,
Ian

[loudthud]

Sorry Loudthud, I could'nt resist, you said "anyone got an AC30 I can borrow?"

--- I thought you were testing/comparing Class A Amps!!!! ----

I had thought we had quite thoroughly put a stake through the heart of the AC30 Class A myth.

Cheers,
Ian

I'm curious about the supply ripple on the output when an amp clips, hence the title of this thread. I'd like my amp to sound like the best tube amps, if it sounds like an AC30, that's not a bad thing. I like the simplicity of class A amps. The current prototype is 50W at 4 ohms and only uses 6 transistors. It uses a big transformer, a bigger heatsink and fan but has no dissipation protection circuit.

The philosophy of a class A amplifier (note 1) breaks down when the amplifier clips if it has negative feedback (which the AC30 doesn't). Even without negative feedback, there is usually enough drive voltage available to drive the output device(s) to cutoff the current. The AC30 is not a perfect class A amp, but it's pretty close. There are flaws in the arguments presented against the AC30 IMHO. Nobody disputes that a single ended amp is class A (ie 5F1), but current in the output tube is cutoff when the output clips and the supply current could go down depending on bias point and how much current the speaker is demanding when the tube is saturated.

In my design there is enough open loop gain that when the output clips, the drive to the output MOSFET slams the rails (note 2). So cutoff of current in the active device is unavoidable. The supply current in my amp actually goes down about 10% when the output is clipping.

My design follows Nelson Pass's design of the Aleph series amps with a couple of exceptions. Refering to figure 5 in the patent (see link below), I use a single PNP transistor (like the JLH amps, note 3) instead of the differential pair. I do this because of Douglas Self's comments (note 4) that differential pairs produce odd order harmonics and in my experiments, it seems to produce a non-symetrical square wave when overdriven. This makes DC coupling to the speaker a problem. A servo could be used but when the integrator is hit with a non-symetrical squarewave, an offset would be produced. Capacitor coupling the speaker solves the problem of DC offset when the supply voltage changes, when the output devices heat up and when the bias current changes.

Looking upward from the output device Drain, (patent fig 5) the current source looks like a negative resistance in parallel with the load. As the voltage across it increases, the current through it decreases. This raises the efficiency over what a straight current source would provide. I do it in a different way than the patent, found in TUT 2.

Note 1: The output device(s) conduct at all times in a class A amplifier. Paraphrased from RCA tube manual RC-19, page 13.

Note 2: Comment on this thread: Ultra linear and phase inverter drive . This is why you should always have gate protection zeners. If enough drive voltage is available, the gate will breakdown if the source can't follow the gate when the drive slams the rail.

Note 3: JLH refers to John Linsley Hood who designed a class A transistor amp in 1969. His first amp only used 4 transistors. See this site: The Class-A Amplifier Site

Note 4: See this article: Distortion In Power Amplifiers

Link to Pass patent: Amplifier having an active current source

[diagrammatiks]

I'm curious about the supply ripple on the output when an amp clips, hence the title of this thread. I'd like my amp to sound like the best tube amps, if it sounds like an AC30, that's not a bad thing. I like the simplicity of class A amps. The current prototype is 50W at 4 ohms and only uses 6 transistors. It uses a big transformer, a bigger heatsink and fan but has no dissipation protection circuit.

The philosophy of a class A amplifier (note 1) breaks down when the amplifier clips if it has negative feedback (which the AC30 doesn't). Even without negative feedback, there is usually enough drive voltage available to drive the output device(s) to cutoff the current. The AC30 is not a perfect class A amp, but it's pretty close. There are flaws in the arguments presented against the AC30 IMHO. Nobody disputes that a single ended amp is class A (ie 5F1), but current in the output tube is cutoff when the output clips and the supply current could go down depending on bias point and how much current the speaker is demanding when the tube is saturated.

In my design there is enough open loop gain that when the output clips, the drive to the output MOSFET slams the rails (note 2). So cutoff of current in the active device is unavoidable. The supply current in my amp actually goes down about 10% when the output is clipping.

My design follows Nelson Pass's design of the Aleph series amps with a couple of exceptions. Refering to figure 5 in the patent (see link below), I use a single PNP transistor (like the JLH amps, note 3) instead of the differential pair. I do this because of Douglas Self's comments (note 4) that differential pairs produce odd order harmonics and in my experiments, it seems to produce a non-symetrical square wave when overdriven. This makes DC coupling to the speaker a problem. A servo could be used but when the integrator is hit with a non-symetrical squarewave, an offset would be produced. Capacitor coupling the speaker solves the problem of DC offset when the supply voltage changes, when the output devices heat up and when the bias current changes.

Looking upward from the output device Drain, (patent fig 5) the current source looks like a negative resistance in parallel with the load. As the voltage across it increases, the current through it decreases. This raises the efficiency over what a straight current source would provide. I do it in a different way than the patent, found in TUT 2.

Note 1: The output device(s) conduct at all times in a class A amplifier. Paraphrased from RCA tube manual RC-19, page 13.

Note 2: Comment on this thread: Ultra linear and phase inverter drive . This is why you should always have gate protection zeners. If enough drive voltage is available, the gate will breakdown if the source can't follow the gate when the drive slams the rail.

Note 3: JLH refers to John Linsley Hood who designed a class A transistor amp in 1969. His first amp only used 4 transistors. See this site: The Class-A Amplifier Site

Note 4: See this article: Distortion In Power Amplifiers

Link to Pass patent: Amplifier having an active current source

Yes but the issue is that the AC30 isn't a class A amp in any way shape or form.

The argument against the ac30 only breaks down if you try to classify output stages in clipping.

By definition any classification can only happen under clean signal conditions.

Otherwise you'd just drive the amp into squarewave and call it class D.

[loudthud]

Yes but the issue is that the AC30 isn't a class A amp in any way shape or form.

The argument against the ac30 only breaks down if you try to classify output stages in clipping.

By definition any classification can only happen under clean signal conditions.

Otherwise you'd just drive the amp into squarewave and call it class D.

Has anybody posted an oscilloscope photo of the cathode current of the output tubes and output waveform? It would be a little tricky, but I can do it. Bring me an AC30. The only argument I've seen was based on voltage across the cathode resistor. Next you'll say that a 5E3 isn't class A. It's about the same as an AC30.

[diagrammatiks]

Has anybody posted an oscilloscope photo of the cathode current of the output tubes and output waveform? It would be a little tricky, but I can do it. Bring me an AC30. The only argument I've seen was based on voltage across the cathode resistor. Next you'll say that a 5E3 isn't class A. It's about the same as an AC30.

when the output waveform is clean the 5e3 is class A.

when the output waveform isn't clean then it is by definition operating within no amplifier class.

You might as well call it a function generator. You'll end up with a square wave eventually no matter what amp it is.

Is the Vox AC-30 really class A?