Crossover distortion scope shots and questions

[waspclothes]

As far as looking at the crossover notch, there is no point looking after the signal is clipped. You can take the notch right out and it may come back at (or beyond) clipping. It will be different for various circuits.
Once you are hard clipping like that, you probably won't discern which part of the distortion is the crossover distortion.
Do you have a shot of what the notch looked like before clipping?

I don't have a shot of it before clipping, but yes I did notice that the crossover points only started to distort once the waveform started clipping at the top and bottoms. As the volume went up, the clipping at the top and bottom got worse along with the crossover points getting more pronounced and spikier. At volume levels below where clipping started happening, the waveform looked really good and only slightly started to "step out" of alignment as it approached the clipping point. I did notice that Changing the bias to increase the plate current (hotter bias, right?) did clean that stepping out up a bit.

But you're saying that once the waveform starts clipping at the top and bottom then all-bets-are-off anyways for getting that crossover point to not look terrible? My concern was initially that the amp sounded "buzzy" when cranked to distortion and I thought that the crossover spikes might be causing that harshness and maybe that was because of a problem with the amp itself.

[Alex R]

Your shot is of an overdriven signal as has been stated. You will usually see some crossover-ish stuff on a clipped signal. Try an input more like 100mV, and just tweak the volume control till the amp cleans up.

The crossover notch method kind of works but is imprecise and confusing - this is a good example of the confusion, an artefact of clipping being interpreted as an artefact of cold bias. I would be surprised to hear of a working repairman who uses the method to bias amps. Most of us have got used to bypassing the OT with an ammeter, it's quick and pretty accurate though also quite easy to make sparks and blow fuses with

[Audiotexan]

As far as looking at the crossover notch, there is no point looking after the signal is clipped. You can take the notch right out and it may come back at (or beyond) clipping. It will be different for various circuits.
Once you are hard clipping like that, you probably won't discern which part of the distortion is the crossover distortion.

Oh wow. Yes, excellent point.

I just completely assumed (incorrectly) that the OP'd know that it was done with a 'maxed signal' (read as: clean sine) just prior to clipping. And worse still, didn't think about mentioning that aspect, even after looking at the wave provided. My apologies. =(

...but yes I did notice that the crossover points only started to distort once the waveform started clipping at the top and bottoms.

This is what's expected. And there's the point you want to check your bias.
My huge error was thinking you were having close to this level of notch while being just before the point of clipping the peaks. :X

The crossover notch method kind of works but is imprecise and confusing - this is a good example of the confusion, an artefact of clipping being interpreted as an artefact of cold bias.

So true. It works, but you have to know when to "apply it" or rather when to look for it being an indicator of a possible issue. :X

I would be surprised to hear of a working repairman who uses the method to bias amps.

To me, it's a tool like the other methods...But I tend to use all 3 methods relative to the overall picture as general indicators.
Negative voltage can be telling...as can crossover distortion, with the most emphasis on current draw (idle and active).

But you're saying that once the waveform starts clipping at the top and bottom then all-bets-are-off anyways for getting that crossover point to not look terrible?

The goal is to have your max output being clean (no clipping on the peaks), and crossover distortion minimized, at rated power, while being in spec (no excess current draw). So if you have that, you should be golden.

Breakup starting at 6.5-7 isn't unheard of in my experience. Especially on smaller amps with a rectifer tube.

I apologise sincerely for not mentioning (and not even realizing I made the leap/assumption that I did) that crossover distortion is only important below clipping. It's an error I'll try not to repeat.
That would have been quite the time saver, and not have you chasing ghosts. I feel horrible for that!

[MarkusBass]

I'm using a 1Vpp 2khz sine wave going into the 1 input of the vibrato channel....... with the volume on the amp cranked to 10.

The signal that you use is to high. Try rather 0.2Vpp, or even 0.1Vpp (especially that the amp is cranked to 10).

Mark

[pdf64]

Some ideas -
Get the scope on to each stage along the signal chain, advance the levels and learn how each stage overdrives and what the headroom limits are.
Find out what tone control settings give the flattest audio response; try a squarewave input for a very quick indication of this, interpreting this can save running a frequency sweep.
The input stage cathode voltage on the DR should be over 1V, so the stage accommodate an input of nearly 2Vp-p before reaching saturation / cut-off.
Don't mistake grid conduction of the following stage for a lack of headroom in the stage you are investigating.
Pete

[teemuk]

This is scoped with a reactive, not resistive, load, right?

The enhanced spikyness and the "asymmetry" of crossover points strongly suggests that: voltage and current are not in same phase, not to mention frequency response is likely far from flat so clipping edges have exaggerated spikyness. I'm quite sure it would look more "normal" when scoped with a purely resistive load.

What is your "dummy load" to be exact?

I don't have a shot of it before clipping, but yes I did notice that the crossover points only started to distort once the waveform started clipping at the top and bottoms. As the volume went up, the clipping at the top and bottom got worse along with the crossover points getting more pronounced and spikier.

This is usual for many, many classic tube power amps. When you overdrive the output tubes their grids begin to conduct and draw current. This actually alters the bias voltage at the grids because the grids are usually capacitively coupled and driven from a high-ish impedance source. It's typically not an instantenious effect but on scope screen you will pretty much see a condition of "settled" DC voltages.

You are probably more interested in whether there is significant crossover distortion when the amp still runs clean, producing its rated output power. When it overdrives it seems to behave exactly like how most tube power amps behave at such point: Hard clipping and plenty of crossover distortion due to bias shift at grids.

[pdf64]

Regarding bias shift when overdriven, see Grid Bias Excursion Calculator

Note the disclaimer in the text -
'The calculations focus on the clamping action of the coupling capacitor CG due to grid current, so they assume that the cathode bias voltage or fixed bias supply voltage remain constant. In reality these are not perfect voltage sources. The grid bias supplies also react to being overdriven and have their own time constants'
Pete

[waspclothes]

Oh wow. Yes, excellent point.

I just completely assumed (incorrectly) that the OP'd know that it was done with a 'maxed signal' (read as: clean sine) just prior to clipping. And worse still, didn't think about mentioning that aspect, even after looking at the wave provided. My apologies. =(

I apologise sincerely for not mentioning (and not even realizing I made the leap/assumption that I did) that crossover distortion is only important below clipping. It's an error I'll try not to repeat.
That would have been quite the time saver, and not have you chasing ghosts. I feel horrible for that!

Hey Audiotexan, first of all thanks for all your help. I'm sorry I didn't make myself clear, I wasn't trying to use the scope to set the bias.. I was trying to investigate a "buzzy" distortion. So that's why I was trying to blast a 1Vpp signal through it at max volume

Teemuk: I am using a Weber Mini Mass set to 8ohm and max attenuation. The Mini Mass has a speaker motor inside as well as some passive resistors.

pdf64: Thanks for the idea, I will do that next.. it will be nice to visually see the signal at each stage and at one point the clipping starts.

Thanks for the help on this one everyone. Am I right thinking that the spikey parts at the crossover points at maximum volume are normal, but are elements of the distortion that would not be present on a single-ended amp?

[Mick Bailey]

Find out what tone control settings give the flattest audio response

Good point; quite often a glitchy-looking waveform (other than crossover or hard clipping) can be corrected with the tone controls. Particularly transistor amps with graphic EQs. I always begin by getting the best looking flat-response output with least distortion at a fairly low-to-moderate level. This establishes that the amp is capable of producing a relatively clean output at all and gives a basis for further investigation.

[teemuk]

I am using a Weber Mini Mass set to 8ohm and max attenuation. The Mini Mass has a speaker motor inside as well as some passive resistors.

Yes, I would have been surprised if the load would have not been reactive.

Basically, because the current and voltage with such load are in different phases then current won't be 0A when voltage is 0V. Hence the crossover notch is "delayed" from the conventional "zero crossing point". Therefore the asymmetry.

Because the frequency response of a high output impedance amp (like a tube amp) driving a reactive load is not flat all waveforms composed of several different frequencies (e.g. square waves or square-ish waves will be "deformed" as amp's gain throughout its effective bandwidth varies. Hence shapes of waves composed from multiple frequencies will also "skew".

If you scope that amp using a purely resistive dummy load (which weber mass + resistors isn't) you would see somewhat different waveform on the scope screen. Amp's gain across the effective bandwidth would remain constant so there would be less "spikyness"; both the crossover region and the waveform tops would be much flatter. As current and voltage would also be in phase the crossover notch would shift to the 0V area. Like in that scope capture posted previously.

Now you are also seeing the effects of reactive loading. They might look surprising if you are accustomed to seeing scope captures of amps taken with resistive loads.

Am I right thinking that the spikey parts at the crossover points at maximum volume are normal, but are elements of the distortion that would not be present on a single-ended amp?

That is correct.

Crossover distortion like that is typical for many overdriven push-pull amps, where output tube(s) of one half of the push-pull pair seizes conduction while the other half takes over. Hence the cross-overing effect. If you have a single-ended amp the output tube practically conducts throughout the entire waveform and there won't be such crossovering effect. Due to other differences in circuit performance the waveform output, when overdriven, will usually be radically different from those of push-pull amps. Long story, I won't go there.

But heavily overdriven SE tube amp output will usually resemble more something akin to this:

There is usually slight "crossover" looking effect which comes from the overdriven tube going into cutoff where current flow comes to abrupt stop and the collapsing of the magnetic field in inductive elelements of output transformer and load causes high voltage surges. The same phenomenon will take place in push-pull amps too but SE amps, due to their design, are even more prone to it. This is the phenomenon that can kill cranked tube amps: The voltage arcs through tube sockets, circuit board tracks, or in worst case, through OT's winding insulations.

To get back on previous topic of overdriven push-pull amps, the ovedrive-induced bias shift is not inherent to all such designs. It generally requires that ovedriving can shift the bias by some mechanism. If you have high current, low impedance source (like a cathode follower) driving the output tubes, or if the output tube grids are transformer coupled then such bias shifting effect will not occur, or at least will occur in much lesser extent. However, if you look at traditional guitar amp designs you generally won't find many of such designs.