Higher voltage swing into lower?

[MrStab]

hi,

i've not been posting that often lately, but i'm still obsessed with making stuff. maybe i'm spending more time actually trying to figure things out than annoying you guys!

lately i've been tinkering with a boost pedal that uses a charge pump to get up to around 31V (under load) through an OPA2134. when i was working with a TL072 at around 27V, my SS test amp could handle it fine. now, it clips a tiny bit. the amp is filled with TL072/NE5532s and has a 15V regulator for either rail, so 30V essentially, reduced further by the swing of those ICs. i have a tube amp i could test on, but it's not easily-accessed at the moment.

so my question is this, and i guess it may be confirmation of the obvious: are higher voltage swings inhibited by the voltage of whatever comes after it? i've never read any mention of such problems with voltage-multiplied circuits, presumably because the diode drop & cap values of the multipliers (and the op-amp swings) keep it below a safe threshold. but surely people have run into problems putting circuits before purely-9V pedals?

i'd imagine 31V is just a drop in the ocean when it comes to the hundreds of volts passing through tubes. i realise peak-to-peak is sometimes preferred in this regard, but don't have the means to measure that (and it confuses me anyway!).

cheers for any insights.

p.s: as of Monday, i'm now a qualified Portable Appliance Tester. just wanna mention it because it was entirely pedal-building (and all your help) that led me to do it. it's for the charity music shop i volunteer at, which is cool for coming across all kinds of vintage gear. it's not that big a deal considering some of you are ultra-qualified in electronics, but it's my first electricity-related qualification, so this humble string-basher is pretty chuffed. so aye - thanks, guys!

[R.G.]

so my question is this, and i guess it may be confirmation of the obvious: are higher voltage swings inhibited by the voltage of whatever comes after it? i've never read any mention of such problems with voltage-multiplied circuits, presumably because the diode drop & cap values of the multipliers (and the op-amp swings) keep it below a safe threshold. but surely people have run into problems putting circuits before purely-9V pedals?

i'd imagine 31V is just a drop in the ocean when it comes to the hundreds of volts passing through tubes. i realise peak-to-peak is sometimes preferred in this regard, but don't have the means to measure that (and it confuses me anyway!).

Here's the deal: Everything has some limit beyond which it won't respond linearly, and possibly some other limit where it will be damaged. On top of that electrons have no clue where they came from and where they're going - the circuits they run through do that for them.

A little thought will give you the answer. Imagine you have a magic voltage source, which has one knob. The knob turns the voltage from 0.000000Vac up to 1,000,000Vac. It can supply any amount of current whatsoever. Turn up the knob, the voltage appears. Imagine you have the leads of this thing connected to a piece of iron. With the knob at 0.000V, no current flows, and so nothing happens. as you turn up the knob, the magic voltage source increases its voltage. The steel may have a resistance of, say, 50 milliohms. So when you hit 50mV on the knob, 1A flows. The power the steel dissipates is 1A times 50mV, or 50mW. Turn the knob some more, up to 500 millivolts. The current is now 10A and the power is 0.5V * 10A = 5W. The steel can hardly tell it's being heated.
Up to 5V. Now the current it 100A and the power is 500W. The steel makes a nice "burner" and you can cook eggs on it. Up to 50V - the power is now 5kW, and the steel is starting to glow a bright red. At 500V, there's 50kW going into the steel and it's beginning to turn white hot, liquify and drip into a puddle.
The steel does not limit current much, so power goes up hugely with incoming voltage.

But let's say we put a 1M ohm resistor across the leads. Turn up the voltage. Nothing much happens until the power (voltage times current) gets up to the power rating of the resistor, which in a 1M resistor is many many volts.  At 500V the 1M resistor is letting through 500uA, and the power it dissipates is 250mW. The steel melted, the 1M resistor barely gets hot. So ***load impedance matters***.

What if we put the steel plate in series with the 1M? The resistance is 1M plus 50mOhm, which is about the same as the resistor by itself. If we consider the 1M resistor to be part of the magic voltage source, then it has an *internal impedance* of 1M ohm, which severely limits its output current under all normal loads. So ***source impedance matters***.

Let's say we hook the magic voltage source across a kitchen toaster, put in some bread, pup the lever down and turn up the voltage. Nothing much happens until we turn it up to 120Vac (or 240Vac if we're using a non-USA toaster) at which point the bread starts browning. But twist the knob up a lot and the toaster heating coils vaporize. The *** ability of the load to absorb voltage and current matter***.

If we connect the MVS to a bipolar transistor with no emitter resistor, it does not much until the voltage gets to about 0.6V, when current flows. Above that, the current flow is unlimited and massive current flows. So *** the nonlinear characteristics of the load matter***.  If we sub in a triode, nothing much happens till we hit about 1.5V (for a 12AX7 and no cathode resistor) where grid current starts to flow. Still, this is pretty tame till we hit a few hundred to a few thousand volts, when the tube electrodes flash over and vaporize. Again, *** the nonlinear characteristics of the load matter***.

The moral of the story is that you cannot know what will happen unless you know the source impedance of the signal source, and the linear and nonlinear nature of what you're driving. Some things stay linear. Some distort. Some get hot. Some evaporate explosively.

Can one horse run faster than another? Absolutely - but which one? You have to know the horses.

[MrStab]

thanks, RG - i think i understand what you mean from the analogies. so i need to be more meticulous than just banging things with a big voltage hammer.

the goal is to make this boost adaptable to whatever would come after it. before reading your explanation, i'd considered a simple toggle switch to change voltage, but now i'm thinking there might be a better way to go about it - one without such a big "pop". so it could be as simple as changing the value of a series resistor (and possibly pulldown resistor) on the output? that sounds too simple, no doubt i've gone wrong somewhere. i should study more amp schematics to try and find functional averages - i'm surprised at how few solid-state amp varieties i can name.

you also got me thinking of another caveat to consider here:



Edit: i misinterpreted this graph (read it backwards without applying common-sense), but i'm sure there's something in it i need to consider...

thanks!

[MrStab]

...not that simple because increasing that resistor would have an effect on treble and Nyquist noise. and other big words i don't really understand. but points for at least trying to counter my naivete!

[R.G.]

Good - with the analogies under your belt, you're up for understanding what comes after.

If you drive the input to a tube amp, there is nearly always a 68K resistor or two in series with the input - although not always! - so things get tamed a bit. Tubes do this odd trick where you can take the input to minus a few hundred volts without anything changing much except the tube turns fully and totally off after the first few volts minus. However, you can only pull the grid a volt or two positive until it starts conducting current and rapidly saturates the tube's plate. There's an input nonlinearity, but the series 68K (or whatever) keeps it from being damaging.

Solid state amps can be lots of things. Some of them are the base of a transistor inside a DC blocking cap, and this can often follow an input signal several, or sometimes many volts positive if there is an emitter resistor. If you get more than one diode drop above the power supply on the collector, the base turns on to the collector and any further input signal is clamped to the power supply, through any series resistances that exist in that path. For raw emitter followers, this can be destructive. If you can pull the base more than about 7-9V negative (for an NPN in this discussion; same stuff applies to PNPs, but backwards in voltage) then the base-emitter junction breaks like a zener and the transistor's noise performance is permanently damaged.

You can use series resistors and bigger, burly power diodes to clamp the input signal to the power supplies and prevent damage. There is some discussion of this at geofex in "what are all those parts for?" and "when good opamps go bad".

Opamps used to be killed by funny stuff on the inputs, then the manufacturers started putting clamping and protection diodes on them. They're better these days, but the big thing is to prevent the currents from getting large. This is why you'll sometimes find 1K-10K resistors in series with a high-impedance input even on a JFET opamp. The resistor limits current through the protection diodes.

Notice that in all these cases, the circuits will start to sound UGLY long before damage hits. There is a concept called "compliance", the idea that a circuit node (like an input pin) can only be pulled up or down X volts before things get hurt. The input compliance range of a noninverting opamp input is typically the power supply voltages minus a couple of volts. The input compliance range of the INVERTING input is typically millivolts, as this pin is nearly always driven to the same voltage as the noninverting input pin. The differential mode compliance for opamps may be a volt or two before damage, or for some few opamps may be the entire power supply. In general, pulling any pin of an IC outside the power supply range forward biases a diode to one or the other power supply, and may cause damage unless the current is limited.

But back at your problem. You want to make BIG signals and send them into circuits.

First, peak to peak. An AC signal swings some amount positive, then reverses and goes some amount negative. Even if it's riding on top of a DC level, the positive-going voltage goes up to some voltage, then reverses and goes down. That most-positive voltage is... wait for it... the positive peak voltage. It goes down to some most-negative voltage, and that's the negative peak voltage - as far negative as it goes. If you take the positive peak voltage and subtract the negative peak voltage (which really means "add the total of the two peak swings) you get the peak to peak voltage.  A sine wave has a positive and a negative peak that are the same size, so a 1V peak sine wave has a peak to peak of 1+1 = 2 volts peak to peak.

There are several ways to talk about signal size. A 1V peak sine wave is 2V peak to peak. It has a zero average level (it spends the same amount of volt-seconds positive as negative) and the equivalent heating power of a 1V peak/2V peak to peak sine wave is the Root Mean Square value, which happens to be waveform dependent. For a sine it's 0.7071 times the peak, or for a 1V peak wave, 0.7071Vrms.  A 120Vac RMS house current wave has a peak of 120Vrms/0.7071 = 169.7V peak or 339.4V peak to peak.

[PRR]

I don't know any "amplifier" which needs more than 1V-2V input to clip.

Extreme power rails should not be needed?

[R.G.]

There are situations where driving the input of a tube amp with a low impedance source can produce pleasingly soft overdrive on the positive-going side because of the peculiarities of the tube grid voltage enhancing the plate current flow - if you can provide the current the grid conduction needs.

It's also possible that certain side effects of the clipping can be pleasing in special situations. I've seen guitarists drive amp inputs with up to 8-10Vrms with not-horrible sounding results.

[MrStab]

okay, i think i'm following. you've outlined the failure modes of various input stages in the following amps/pedals, RG, and that makes me think that it might be unwise to rely on the user to set the conditions, if there are different thresholds of failure or damage for different inputs.

maybe there's some complex way of using a comparator to detect impedance differences vs. a test split. but probably an excessive and impossible solution.

i could maybe figure out how to play safe for BJT/FET inputs (though any pointers for parameters could help!), but do you have any resources on the behaviour you mention with tube inputs, that's relevant to this situation? maybe there's something at Geofex, i'll have a look later on.

i have a feeling Ohm's Law is calling... at least i hope, but i'll probably have to wade through heftier equations.

[MrStab]

okay, so i've spent at least 15 hours straight reading (i have no life) and i'm still no further ahead. what confuses me the most is that there are pedals sold on the market that do these high voltages, at least 27V, and i have no clue how they get away with it. the MI Audio Boost & Buff, for example (as i understand). no schematics available. well, except for the John Hollis Titan Boost, but that may deviate in design too much. probably the same core principles, though.

if it makes any difference, i'm only going for a gain of 6 here.

it's not a very good test as my multimeter can't show the waveform, but the AC measurement i'm getting from the output at full blast (with 10k series resistor) is 2.1V, so about 2.86V RMS...? inconclusive until I get a guaranteed 1V P-P signal to compare it to on a scope or a computer, but it's a step.

i'm thinking of building a bunch of common input stages in solid-state amps, and stress-testing them. as transistors can go bad without it being immediately obvious to the ear, any ideas what before/after conditions i should take note of?

[R.G.]

Bear in mind that overrunning the linear input range of an amplifier stage doesn't necessarily damage it - but it's wise to remember that it can.

Damage to parts is nearly always a thermal issue. "Thermal" means "internal power generated can't get out fast enough to keep temperatures down". For electrical stuff, power is always current times voltage.

Restating the earlier stuff another way, an input with a signal coming in that's bigger than its ability to respond linearly will at some level have a nonlinearity - turning on a substrate diode, clamping to power supply, or simply breaking over a junction or arcing. For voltage inputs, the current in is usually quite low, so when a higher voltage drives it nonlinear, the current goes up. It can go up a lot or a little depending on how much resistance there is to limit the current and how far overdriven the input is. If the current times voltage stays under the level which overheats something on the input, then there's no damage.  Or in the case of bipolar base-emitters, there is only the generation of microscopic regions of crystal defects, which I've read is behind the increased noise. In situations like this, the damage is tiny, but cumulative.

There's another item hiding here. The reason people like big-headroom signals driving (especially tube!) inputs is because they like the way the nonlinearity sounds. They're using the input as a set of "clipping diodes".  The hidden issue is the not all inputs are identical, so what sounds good driving a tube amp input with a 68K in series and a 12AX7 tube as the first element will sound different into a bipolar-transistor input, and different still to a JFET-input opamp. Presumably there's a reason you're generating a big signal voltage, and it's likely to be sound. Well, it will sound different - not necessarily worse, just different - on different inputs that you didn't get to pick.

[PRR]

> the AC measurement ... is 2.1V, so about 2.86V RMS...?

Most general-purpose AC meters read in RMS. So "2.1V" is already 2.1V *RMS*.

Actually, low-price meters determine the Average of an (rectified) AC wave, and multiply by 1.1 to give the RMS.

This assumes the wave is Sine. Most power waves (wall-outlet, power transformers) are so Sine-like that this is valid. Musical waves may be very un-sine-- however it can be shown that nearly any likely musical wave-shape will read the same Average or RMS within 15%. And in most audio work, 15% is close enough for almost anything except government contract testing.

If your output is very near 2.1V RMS, then your *Peak-to-Peak* is in the area of 6V p-p. Here we can't be too sure. A narrow spike "buzz" can peak higher. However there are practical limits to what ugly waves we can generate simply. I would suspect your "2.1V RMS" will fit inside a 9V power rail (i.e. 7V or 8V p-p).

> input stages in solid-state amps, and stress-testing them

Ugh. When you test 99 of them and none dies, what do you know about the 100th? There are many designers and some of them do NOT think about user abuse; OTOH some have been forced to over-obsess stresses because past warranty costs were hurting their company. (I love finding 19-cent diodes protecting 5-cent resistors... some guys must get paid by the part.)

Listen to R.G. Often it is not the voltage but V*I or Power. Not ZAP but cook.

"Nearly all" transistor and tube pins can stand 10mA, but maybe not a lot more. Assume this goes with a 10V break-down (Si B-E is 7V). 10mA at 10V is 100mW. While early Germanium parts might cook at 50mW, nearly all later stuff is packaged for 300mW.

*NO* input needs a whole mA to clip. Normal (undistorted) input current is more like 0.01mA (1V in 100K). A tube grid might need that mA for a fully saturated positive-grid slam (however the common 34K input R won't allow that at sane voltages).

True, *ONE* breakdown can cause DC shift and increased hiss. However I suspect customers looking for a BOOST either accept the chance, or just don't know/care.

27V supply for audio waves is 14V each side. 14V at 1mA is 14K. So 15K output impedance "should(?)" be "safe" in almost any case. For long lines (100 feet) this is kinda high; for "guitar cord" work it is fine.

[MrStab]

hmm. i think this is sinking in. no heat dissipation-related pun intended. i have a better idea of what to avoid. your description of how FET/voltage inputs can fail makes me uneasy, RG, but i'm starting to see how the V=IR relationship (and particularly power) applies here.

thanks for clearing up the RMS/P-P thing, Paul - got that the wrong way round. you're right - no amount of testing can substitute solid mathematics and understanding, and even that can only go so far when it comes to the real world. i assume this sort of procedure would help identify the spikes more, and render my DMM method redundant: http://www.muzique.com/lab/pick.htm. a tone generator probably wouldn't go amiss either. i'll get round to that.

the whole reason for this pursuit, i suppose, is that i really like the combination of that chip and supply voltage through several amps. though as you guys point out, i don't have access to every amp in the world.

dumb question (i'm full of those), but i should be basing my calculations on the chip's voltage swing and not the actual supply voltage, right? because that is about 14V on either rail, iirc (though will need adjusting as i'm not using the absolute max. supply).

off-topic, but w/ regards to user screw-ups: the first "self"-designed (ie. YATS) distortion i sold to someone a few weeks ago was instantly killed by an AC adapter, despite my excessive warnings. just earlier, my girlfriend turned off her computer mid-update, regardless of the "Do not turn off" message, because she "didn't want Windows 8.1". *sigh*. true foolproofness is a pipe dream.

sometimes i don't understand what i've just been told until days or months later, in case i've totally misunderstood something. just for the record. lol
thanks!