Chip amp musings

[Morocotopo]

Hi guys, I´ve been toying on and off with an LM1875 chipamp design for a while. Tried several pres with it, a tonemender, JFEt ones, etc. The chipamps are great, but one condition to make them sound good is to never, ever, let the power amp clip. That sounds awful. I´ve read several posts about that, mostly R.G´s. So, investigating the ways of avoiding that, I´ve come with the following variables to take into account. Let me know if I´m missing something:

- Supply voltage: the higher the supply voltage, the more headroom.
- Input signal amplitude: If this is too high, it will clip
- Chipamp gain: If set too high via feedback R, higher than the power supply allows the signal to swing, it will clip.
- Bandwidth: If limited to the guitar´s range, less chance of oscillations/bad behavior

Well, do I miss something else? I don´t fully understand the relationship between current consumption and the other variables. I tried the amp with 24V 1A and 30V 2A transformers (single supply), the latter giving more headroom. None got hot (neither did the heatsink). I was testing at moderate levels with a clean signal, to hear any distortion produced, and with gain set at 27 (220K and 8K2 R´s). I DO remember, in previous testings, getting the thing to heat quite a bit. Probably had the gain set higher.
My conclusions so far are that you have to use the chipamps NOT close to their limits to get good sound, for example, the LM1875 datasheet states that it can produce up to 20W, but at those power levels it´s easy to get clipping, overheating, bad sound.  At least in a guitar amp. Also, the pre signal has to have some form of limiting to avoid clipping the power amp. I´ve read R.G.´s idea to use the limiter in some Vox amp, but not sure how to implement that. Maybe a simple diode pair to ground? A simple compressor (Orange Squeezer)?
Well, this are my thoughts on the subject. Anything you can comment is welcome.

Regards

[R.G.]

It's pretty simple. The Vox scheme was to apply a signal with the (adjustable in their case) limiter substantially turned off. The gain was set to drive the amp to clipping into a load. Then the limiter was turned on and set to pull the signal back down from power amp limiting so the limiter limited the signal, not the amp.

For a chip amp, the process involves adjusting gain, not limiter level unless you want to replicate some kind of adjustable limiter.
(1) know your incoming signal level and the output level of the chip amp. An 1875 might put out +/-15V peaks, depending on power supply, as an example. The incoming signal level is probably guitar (about 100mV at a guess) that has been preamplified some.

(2) Pick a limiter. You can use silicon diodes to get +/-0.7V peak limiting, a stack of them for +/-1.4V, +/-2.1V, etc.; or LEDs, MOSFETs, whatever. But pick a limiter and know by measurement and design how big the signal can get when limited by your limiter.

(3) Adjust the gain of your chip amp so that it is lower than the amount that lets the power amp clip. Let's say you use a two-diode stack of silicon diodes. That will clip at roughly +/-1.2V to 1.4V peaks. The output of the chip amp is (in the example) +/-15V peak. So a gain of ten on the chipamp will amplify a 1.4V peak up to 14V peak, and the chipamp will never get a signal that can push it into its own clipping. The limiter limits first.

[Mark Hammer]

IIRC, a lot of the smaller Crate amps do precisely that, with a limiting of the input signal to the power amp chip, directed by the output of the power amp.

[R.G.]

That description sounds like a "sag" or compression setup. And it would be useful in simulating that part of a tube amp.

The soft limiter is another piece of it, I think.

[PRR]

> - Supply voltage: the higher the supply voltage, the more headroom.
> - Input signal amplitude: If this is too high, it will clip
> - Chipamp gain: If set too high via feedback R, higher than the power supply allows the signal to swing, it will clip.
> - Bandwidth: If limited to the guitar´s range, less chance of oscillations/bad behavior

Chipamp is just a specific case of general audio management.

There is NEVER "enough" supply/headroom. We had 10W amps, then 40W then 100W amps, then 300W monsters, and now 1,200W in half a shoebox.... musicians still clip.

You do-and-don't have control of input amplitude. You always have a volume control. This always gets turned-up more and more, axe strummed harder and harder, as the night goes on.

Chipamp gain is just a part of overall gain/level management. If you cut the gain of the chipamp, the user will turn-up more. If cut enough so that a player "can't" overload it, they can't get the power they paid for, and will complain that it is hard-work to get what they can.

Bandwidth is a minor issue. A stage amp ought to be very stable, needn't have large bandwidth (2 decades), but there isn't much out-of-band crap from guitar and any other junk (ultrasonics, buzz) is bad amplifier implementation.

And BTW a guitarist "wants" distortion even into significant clipping. Voice, brass, and woodwinds change timbre from soft to loud. The actual power may not change much, "forte" is often a few db more fundamental and many db more overtones. Long steel strings do not have so much increase of overtones with amplitude. Amplifier softness and breakup is musically useful.

The stand-out feature of "chipamps" and transistor power amps generally is that they have ABRUPT transition from clean to bent. A 50W amp may be 0.1% distortion (insignificant) to 49W, rising suddenly to 10% at 60W (2db increase).

In most of the "natural" overtone rises the increase is square or cube law; something less than abrupt clipping.

Simple "low fidelity" amplifiers, particularly some tube amplifiers, sorta do this naturally. Most transistor amps, even the simple LM386, clip abruptly and rudely.

The obvious technique is to pre-bend the signal so there is a ~~10db zone between the onset of audible distortion (added overtones) and abrupt clipping.

The obvious problem is how to do this. I don't think there is any perfect answer.

[Morocotopo]

Guys, thanks for the answers. The problem, seems to me, is getting a limiter circuit that works gradually and sounds nice, and adjusting all parameters to never let the power clip.
I tried two Ge diodes to ground before the power in, works for limiting, gets progressive distortion at higher levels. Also makes a slight unpleasant "buzz" at the beggining of notes when driven hard (besides the expected distortion of the waveform). Maybe needs a buffer before and after.

Anyone remember the LHX2 fender simulator? That had a string of GE diodes to ground (14, 18, or so). Might try that...

http://home3.netcarrier.com/~lxh2/frankmod.gif

[Gurner]

The problem, seems to me, is getting a limiter circuit that works gradually and sounds nice, and adjusting all parameters to never let the power clip.

Every guitar amp comes with one - the owner ...his ears (clipping is easily heard) & then his finger on the master volume! 

[DDD]

Clipping indicator (red LED for example) can help a lot.
It's, by the by, very simple: comparator, detector, 1 BJT and LED.

[R.G.]

The problem, seems to me, is getting a limiter circuit that works gradually and sounds nice, and adjusting all parameters to never let the power clip.

I liked diode-connected MOSFETs. They never really fully flat-top.

[Morocotopo]

Interesting R.G. I´ll try that. You mean that they never really hard limit to an absolute value or that the transition curve between no conduction and conduction is more gradual?

[R.G.]

Interesting R.G. I´ll try that. You mean that they never really hard limit to an absolute value or that the transition curve between no conduction and conduction is more gradual?

The first, primarily, although I really meant both. It is difficult to get diode-connected MOSFETs with any significant source impedance driving them to really flat-top.

[Mark Hammer]

The problem, seems to me, is getting a limiter circuit that works gradually and sounds nice, and adjusting all parameters to never let the power clip.

Every guitar amp comes with one - the owner ...his ears (clipping is easily heard) & then his finger on the master volume! 

Yes and no.  I seriously doubt whether a great many here, myself included, would be able to judge, by ear, the wattage produced/required for peaks in response to pick attacks.  The average volume one aims for may require a very modest wattage to produce, but that initial transient may require many times in excess of that.  One of the many reasons why bass players generally require much more wattage than guitar players: those thumb-popping peaks have to be pristine.

So, I agree with you in principle, but I think there is much about the moment-to-moment demands on power amps, and power supplies, that we are not at all well-positioned to detect or anticipate by ear, and control by hand.  Sometimes we really do need those automatic devices to compensate for that last 5% we can't do on our own.

Stupid imperfect humans!!  And stupid dynamically variable wide-ranging guitar signal!!

[Gurner]

So taking that to the ultimate conclusion .....what we're trying to fix is something we can't perceive?

I'd like to therefore steal/edit a useful maxim "if it ain't broke, don't fix it"....to...... "if we can't perceive it, then don't worry about it".  

For example, I dabble a lot with PICs....often with low frequency PWM theres audible whine spillage into the signal path - the solution - put the PWM up in frequency where we I hear it...the problem is still there, but I can't hear it so it's no longer a problem! The local dog population gets mighty peeved though!

However if we can perceive the aforementioned distortion, reach for the gain & knob and turn it down!  

[Mark Hammer]

Oh no, we can perceive it alright, when clipping happens.  Our problem is we can't tell in advance when we're gonna produce that clipping by how we play, because the clipping is occurring in response to events that may last less than 100msec, and which we can't make our fingers avoid.

My teeth are a bit missshapen.  I'm not homely, but I've got some sharp bits that stick out where they oughtn't.  I bite my lip and cheeks and tongue from time to time while chewing food.  It is a brief incident - taking only a moment to puncture - but I have no way of knowing when it's gonna happen.  And I hate when it happens.

Same thing with chip amp clippipng.  It doesn't happen constantly, but its ugly when it happens and you have little way of controlling it to NOT happen, apart from circuits, because your fingers and your motor cortex and cerebellum just ain't that smart.

[Gurner]

Plug your guitar into your amp - set your preferred tone, now thrash it (ie pluck/twang those strings as hard as you're ever likely to) tweak the preamp gain & master vol until you can't perceive even one littlle bit of distortion......make a mental note of the settings - don't go past them ....job done! If you seek distortion free sound (in the 'perception' definition of the word), there should be no way it can sneak up on you if it's your own guitar rig....unlike your tooth/jaw reflex, the guitarist actually controls the supporting parameters/dynamics of the situation....though I accept that most guitarists in full 'solo constipation wince', maybe aren't focusing too much on their amp settings and can stray into no-go land  

[teemuk]

Same thing with chip amp clippipng.  It doesn't happen constantly, but its ugly when it happens ...

Then again, a good handful of forumites over here have likely built something LM386 -based just to overdrive that chip amp to clipping. 

[Mark Hammer]

Same thing with chip amp clippipng.  It doesn't happen constantly, but its ugly when it happens ...

Then again, a good handful of forumites over here have likely built something LM386 -based just to overdrive that chip amp to clipping. 

True dat!