5E3 X-Y display of output

[loudthud]

I haven't found much on plots of dissipation limiting circuits, one from an old Crown power amp is posted below. Can anyone else remember seeing one?

[J M Fahey]

Yes, I remember seeing them.
In fact there are different methods of designing SOAR compliant protection circuits.
The best follow quite closely the SOAR boundary without crossing it.
In theory those amps should be "unburnable" ; in practice they are far from it.
I spent long hours designing them, discovered that making them "good" meant limiting possible output too much when using real world speakers.
The time honored technique od using a simpler protection and doubling the amount of output devices works very well (the "Peavey style")

[Steve Conner]

I built a SOAR protection circuit according to Doug Self's design method, and plotted the locus of it.
scopeblog » Blameless short circuit protection
http://scopeboy.com/scopeblog/wp-con...less-locus.png
(the above graph was for one pair of output devices only)

JM and Peavey are right, if you try to use the SOAR protection circuit to keep an undersized output stage alive, you end up with an amp that is good for nothing but driving a dummy load. On a real speaker load, the protection will kick in and cause extremely nasty distortion. I think this is what Rich Koerner of Time Electronics means when he talks about "hang time blowing up speakers".

I aimed to drive a load of 4 ohms resistance in series with 4 ohms of inductance without triggering the protection, and I ended up having to use two pairs of output devices, and rails of only 35V. And, there are some high-end hi-fi speakers that present even more reactive loads than that.

So, I think that to make an amp that works in the real world, you have to go "Peavey style" anyway, irrespective of what kind of protection you use. An undersized output stage will always either fail or sound crappy.

PS: The Crown DC300 looks pretty unimpressive, with a short-circuit current of only 2A.

[J M Fahey]

Cool.
Thanks for doing your homework so often (beyond the callof duty).

As of Crown´s "unimpressive" short protection, well, that´s the point!!!!
Limit short circuit current to 2 A.
A "perfect" protection should limit it to 0 A , since it´s wasted current anyway .

In fact, 99% of protections exactly *do* protect to 2A, that.s where the "magic" 0.33 ohm emitter resistor comes from.
1 diode junction across it= 0.65V/0.33 0ohm= 2A.
Not to forget that most power transistors SOAR (which really is a series of values) is specified at 2A or thereabouts.
A lot of popular power transistors in fact can only do , say, 1.67A (not even straight 2), but designers (me included) can afford the risk for 2 reasons:
1) the classic +/- 40V rail *drops*significantly under load.
2) as you noted, the duty cycle is 50%. It helps.

[Steve Conner]

Well, you would think so, but it's not wasted current, it's necessary to drive a reactive load.

The essence of a reactive load is that it stores energy, therefore current continues to flow through it after the amp's output voltage has reached zero and even become negative. If the SOAR protection doesn't allow this, it will trip, and the output will slam to the opposite rail. (See Figure 4 here: VI Limiters in Amplifiers - Rod Elliott explains it a lot better than I can)

Therefore, the higher an amp's short-circuit current rating, the better it is at driving reactive loads. (Mine does about 10A.) I think this is the reasoning behind the audiophile habit of leaving the protection off entirely.

[J M Fahey]

Well, I usually don´t use current limiting protections at all, for the reasons posted above.
Although all my amps (including the bass ones) have mixed feedback to achieve damping=1, so in the event of a short, current never reaches catastrophic levels.
If anything, I have a current triggered relay which disconnects the load if it doesn´t like what it sees.
Specially for those dumb people who listen to the amp struggling, barely putting out a couple useful watts into the shorted speaker, and thinking "oh well, it´s gone bad, I´ll keep playing as long as it lasts" instead of searching for the defect, 99% of the time a bad cable. Oh well.
Now they ring me telling: "hey Juan !!! the amp is running very hot and turning off after a minute or two, should I use a bigger fuse? "
Mind you, it *can* be cheated if the musician turns the amp on, full signal blasting, into a shorted load, because for a couple seconds the protection is still stupid.
Probably will redesign it to add a , say, 10 or 20 second slow turn on mute.

[trobbins]

For comparison of voltage swing, then probably need to do the testing with the same Wrms - which would require a specialist wattmeter if you can get one.

I don't think the results can infer that plate connected catch diodes would actually be active at all. But what would be great for an overdriven PP output stage is to measure the plate voltage/current X/Y operating loci - perhaps by passing the plate leads through a wideband CT (eg. I have some LEM LA25NP) to give a net AC signal, and just sense one anode to 0V level - then manually plot the VI curve onto a standard datasheet VI characteristic to show how the curve bends on aproach and during saturation to indicate screen and B+ voltage sag influences.

Ciao, Tim

[J M Fahey]

plate connected catch diodes would actually be active at all

Short answer: not visibly active with resistive loads.
There is a very small front end spike, almost negligible.
Now, on complex impedance real world speaker loads you have a quite visible front end spike, which is almost killed by plate backwards connected diodes.

[Chuck H]

Hmmmm... I've never used the "protection diodes" but I am aware of the spike. I used a shunt filter on the OT primary consisting of a series resistor and cap. About 1.25X pri Z for the resistor and a 1500pf cap (pri Z 8k resistor 10k in case you want to figure the knee frequency) And this just about levels the spike. But if I can get the same benefit AND protection from the diodes I may need to try it.

[J M Fahey]

Your snubbering data is interesting, thanks.
It computes to about 10KHz.
It should be somewhat noticeable from around one octave below.
Matches glove-in-hand the idea that it should be the right value to be just enough to kill the spike but not so much as to muddy the sound.
If you arrived to this by ear and trial and error, congratulations, your ears are working very well
In theory it shouldn´t be noticeable, since guitar speakers die a sudden death (call it 18 to 24 dB/octave) above 3500 to 4500Hz , so anything above that should be imperceptible, and yet ..... snubbers are noticed to smoot sound, so .......
Diodes perform a similar spike killing function, with the difference that it´s not frequency dependent.
It works on *any* peak that goes beyond 2X B+ , no matter what.

[trobbins]

Measurement on secondary could well be subtlely different than what is exprienced at plate - especially any overshoot behaviour. Do you have any plate voltage measurements?

My preferred OT protection is a MOV-R on each half-primary. The effective RC time can be pushed a lot higher by using small-size disk MOV's (eg. 7mm), and can even be placed to dampen transformer dominent self-resonance. The protection level is inherently higher, as it acts directly on the winding that is vulnerable to the overshoot.

[loudthud]

I dug out a 1000X probe I got off ebay some years ago. The scope is set at 200mV per division so with the probe, the deflection factor is 200V/div. The center vertical graticule line is the zero volt reference. The probe is connected to the plate of one of the 6V6 tubes with a clip lead. You can see voltage spikes lasting about 30uS can reach almost -800V. The last pic is the probe sitting atop the amp.

Note the VI plot from the Crown amp is from the earliest version, DC300{nothing} of the amp. In later versions of the manual, the plot is edited out.

[trobbins]

That looks like some serious overdrive - was the amp just with mid level pot settings as before, and the sound from the speakers ok with more volume still in reserve?

The top left screen shows a cut-off to saturation transition that is very fast. There doesn't seem to be a situation where both output valves are in cut-off. To me that indicates the input signal slew rate was high, or the speaker wasn't connected.

[J M Fahey]

Very interesting.
I usually got much longer leading edge pulses.
These are so narrow as to have a frequency spectrum absolutely unrelated to the actual note being played.
Of course, the typical limited bandwidth of a guitar speaker will make them inaudible.

[loudthud]

I've often wondered about the affect comutating diodes would have. I ran into a problem once with the diodes used to kill spikes on relays. They short the pulse out and that causes the pulse to last a long time. If you put a resistor in series with the diode, (or just an RC) the voltage will spike up but the pulse will be shorter. Does this affect the motion of the cone in a loudspeaker?

Volume control was up about half for the test. You can't read the numbers on the amp, in the picture you can sorta see the rust on the front panel. The amp was in sad shape when I got it. Tweed was gone and OT was open on one side. I think it's a Mercury Magnetics transformer. I got it from Gerald Weber. The speaker is original as far as I know.