Quote from: joecool85 on April 08, 2006, 08:30:05 PM
Looks like earlier when I had posted stuff saying about 60v for input on the lm3886, I was right, but not for the way the chipamp is setup. We only need 1vRMS if I'm correct.

Sort of. The typical supply voltage for the chip is about 60V when the negative rail is used as the reference: That´s +- 30V per rail when referenced to the ground. However, the chip's input sensitivity, which is a whole another thing, is totally dependant on the circuit's gain set. Now if i remember right, the LM3886 required at least a gain of 20 to operate stabile. With a one volt that should equal 20 V on output, which gives 100W of power to 4 ohm load - way more than LM3886 can even handle. If you have a notably lower input signal amplitude just raise the gain. This is easy since the chip basically operates just like a high power opamp. I think there is a limit for the maximum gain too (as there is such for opamps as well), which is the point where the chip starts to operate unstabile. Let's just say that i wouldn't go higher than a gain of 100 for example. These two factors limit the input sensitivity to a sensible range. I'd say something from 400mV to 1V should be good.

Sounds simple but 1:1 buffer is not the best circuit though. What you haven't considered are the losses caused by the EQ circuit which can be quite big even with efficient marshall-type tonestack. I'd say it's best practice to amplify the input signal slightly before the EQ circuit in order to improve SNR. There's really no point of feeding the PA stage with a noisy signal.

Quote from: joecool85 on April 06, 2006, 06:21:30 PM
We could do this all on our own. However, if we design the PCB, then print 5 as prototypes, thats $100.

I don't see any reason for such a high cost of prototypes - assuming we speak about the preamp here with very cheap parts etc. Almost anyone who owns, or has access to, etching equipment can make five prototype boards that cost only a few bucks. If those boards are ok then he/she can share the design here. Maybe i'm just missing the point here...?

What you might be after RDV is not neccessarily a bunch of diodes acting as a limiter but a variable gain or attenuation control. This kind of circuit stops from ever driving a gain stage too hard and therefore it has a much smoother tone - in best circuits it's almost inperceivable. If you want to keep things simple, here's a very simple example of compressing/limiting the signal with a LDR:

http://machines.hyperreal.org/categories/DIY/info/compressor

With an improved detector, circuits like these provide quite good results.

So where should such a circuit exist? Consider this: The easiest way to keep the signal clean is to keep the gain low in every stage. However, this isn't always possible. Whenever there´s a chance that a stage might be overdriven, there should be a limiter circuit - either within the stage or before it. In most simple form the preamp has a very low gain throughout and the limiter is the last link in the chain before the power amplifier - perhaps located outboard in the effects loop.

It would be good to locate the bunch of diodes we have been discussing about right before the PA input. There they would act as "hard" limiters for the maximum amplitude of the input signal, (which should be the input voltage required for full power aka. input sensitivity). The job of this stage is to clip the transients that the softer limiter is too slow to handle. Why so? Limiting (with, for examle, LDRs) is preventive; meaning you can prevent clipping from ever occuring. However, clipping with diodes is not preventive; therefore the "hard limit" offered by diodes should be the final link in the chain. If you aim for a quite clean tone the hard limit should always be the final resort. Of course you can always use a hard limiter after softer limiting to remove the transients that are caused by the slowness of the detector circuit. Located in this place the hard limiter should be nearly "transparent".

A little expansion beyond preamplifiers: Most of the chip amps, like LM3886, have the benefit that they operate basically like opams - thus feedback/gain is very easy to set and control. Adding a limiter even to this path is therefore quite easy. Here's how Vox did it in the 80's with the compromise of loosing some headroom:

http://www.schematicheaven.com/voxamps/busker.pdf

Quite nice considering that you add only four extra components. The output will clip earlier and the amp will never reach the full power as "clean". However, the clipping has a much more rounder knee. Little less simple, but much more ear-pleasing, way would be to add a LDR somewhere either to control the gain or attenuate the input signal.

R.G. also represented some very good points. I especially liked the one about input protection: Although a typical DIYer probably plans to use the device only by him/herself, (and therefore thinks nothing will ever happen to it), sooner or later there will come a point when someone else get's a chance to hassle with it. Unfortunately, people who are not familiar with basic electronics can do utterly stupid things with good intentions. Another chapter are the hazards caused by pets, little children or simple carelessness. For example, feedbacking guitars have blown numerous amps and speakers. Even "normal" use can ruin some devices: Plugging jacks in and out or leaving the coords that are unplugged from the other end into the amp while the device is on. Even some static charges can do destruction if the device is not protected. A good design practice is to at least consider the possible risks.

The standard marshall style stack should fit in as is and there shouldn't be notable frequency losses due to output/input impedance mismatches. If you build this preamp be sure to have enough capacitance in order to minimize the ripple voltage. The circuit is quite unforgiving in amplifying it.

Here is how the preamp looks so far:
http://img465.imageshack.us/img465/8698/preamplifier6od.gif

I really can't take credit for the design since it's mostly based on Flavio Dellapione´s 60W guitar amplifier's preamp found from Red Free Circuit Designs. http://www.redcircuits.com/
Main mods are a higher input impedance and a simpler tonestack that replaces the way too complicated tonestack/harmonic modifier of the original circuit. The power amp circuit that i will use will swing to max voltage with a 200 mV output signal from the preamp - (i don't know if such is the case with your LM3886 PA) - so the headroom of about +-24 V with a +-1.2 V input is quite overkill for the design. Anyway, i wanted a clean preamp. I don't know whether i will implement any more features since i'm afraid that they would ruin the simplicity.

A word of warning: Though the circuit seems very simple it operates on a "high" single supply voltage and relies mostly on a Sziklai pair for gain - which doesn't have a very impressive SNRR figure when compared to, say, an opamp. Unless the power supply has huge capacitance or is well regulated you will definitely get a lot of hum with this circuit.

Quote from: RDV on April 05, 2006, 10:35:03 AM
I've got a few of those ROG circuits and I tried to use them as a preamp for my LM3886 amp and the problem was headroom because of the 9v supply. If I could figure out a way to power them with bipolar 15v power I'd try one again.

Jfets are one option but i think they are highly overrated for what they really are. The problem, that is quite evident when you further examine all Runoffgroove circuits, is that most of them only provide a semi clean gain stage that distorts on higher signal amplitudes. Listen to the samples on the page: How many of these circuits you can honestly call "clean". Actually, a totally clean fet amplifier would have a very poor distortion tone: I built a clean and simple jfet booster few months ago. It is basically a simple common source amplifier circuit with some source bypassing to make up a variable gain. Honestly it doesn't sound any good when it distorts probably because of the lack of any EQing besides the low decoupling that provides a high pass filter. ...But as a clean booster it's really awesome! 95% of the "fet-tone magic" lies in the surrounding circuitry. This is why i would rather go with having two channels: Clean and dirty.

Well, enough about my opinions on jfet tone - i still prefer them over opamps in simple circuits hands down.

Anyway RDV, converting to higher voltage bipolar supply shouldn't be that difficult: Just connect drain resistor to positive supply and gate and source resistors to negative supply. As a downside, you probably have to bias the gain stages again and this could mean almost complete rebuild. While you're at it you might as well extend past the +- 15 V limit for even more headroom. I'm currently planning to build an amplifier that has a preamp which swings to almost +- 25 V (50Vpp) on a 2V input signal. Both power amplifier and the preamp run from a 60 V supply. Now, that starts to be some serious headroom. This one doesn't have fets: I couldn't find ones that could take this high voltage.

I prefer simple circuits too but i'm afraid that the words simple and Solid State really don't go together that well in this context. The transitional curve of a transistor has a quite hard knee (and opamp's has even harder) which results into a situation where the clean signal turns into a harsh distortion very quickly. I really can't take credit from what you refer to as my ideas about diodes since they aren't that uncommon. However, to voice diode clippers correctly you need a lot of diodes and/or other components. Diode clipping alone sounds quite horrible unless you're really into a 60's fuzzy bass tone. Slightly overdriven diode clipper sounds even worse. This is the most important thing and still far too overlooked: Pre- and post distortion EQ makes a big deal!

I think RDV is on the right track about limiting since guitar's dynamic range is quite tricky. I think the best way to achieve this would be to limit and perhaps even compress the signal to get rid of the most severe transients. You could experiment, for example, with LDR circuits - they are quite easy to keep simple. The limited signal could then be used to overdrive a stage that has a very soft knee in the transitional curve. There was an article somewhere in the internet about "knee shaping" with resistors and diodes. I constantly forget where since it's not on my bookmarks but it seemed far more advanced than other ideas about diode limiting i have seen so far. Even the LDR limiting alone might do the trick for some people.

Is this simple? Well, in my opinion not really - plus you also need some EQing which is different for both clean and overdriven signals. I don't say it can't be done, i just say i think it can't be done simple and that's why i prefer external pedals over complicated preamplifer too: They sound far more better than the average attempts to make analog circuits sound nice when overdriven. My ideal preamplifier would just limit the transients, have a tone control, effects loop and a spring reverb. All foot switchable on/off, of course.

The diodes D1 and D2 shunt voltage peaks exceeding +/- 15,65 V to the supply: This is a common way to protect high input opamps from fatal transients.

Edit:
This article concerning opamp failure modes is very educative reading:
http://www.geofex.com/circuits/when_good_opamps_go_bad.htm

I drawed the circuit in quite a hurry and already find some places to tweak: The first gain stage might benefit from a bit more gain, as could the second. As is, the poweramp will require perhaps too much gain in order to compensate pickups that have a low output signal. Adding a trimmer (in a test purpose only) in series with the R3 might help in determing the right amount of gain/headroom required. I wouldn't exceed a gain of 10 in the first stage though. After the voltage drop caused by the tonestack there should be much more headroom for higher gains. The point is this: It's hard to say how much the right amount of gain should be since a) I don't know the maximum voltage swing of the power amplifier and b) I don't know the maximum output swing of the instrument plugged in nor the guitarist's preference for headroom. Using trimmers to set the gain might be the easiest solution that pleases everyone. Reliability aspect would require that they have series resistors so that the opamps do not operate on a gain level that causes them instability issues.

The potentiometer R12 could also be eliminated by replacing R14 with a potentiometer that is in series with the decoupling capacitor C8, (although this solution might cause instability issues too). Unless the parts count is an issue i do not recommend this.

This is what i had in mind:
http://img427.imageshack.us/img427/5465/preampproject2dz.gif

Quote from: RDV on April 01, 2006, 09:45:57 PM
*Here's one I did design with some improvements*.

I do not wish to sound like i'm nagging, but isn't the gain of 22 quite high for the first stage regarding the fact that it has no gain control? Allthough guitar signal is usually around 20mV - 200mV, a humbucker pickup might give an output of even 2 volts. 2 x 22 = 44! Sounds like serious distortion to me. I know the circuit's benefit lays in it's simplicity but consider this option: Lower the resistance of either 4.7k or 100k resistor and connect it in series with a suitable potentiometer or trimmer to have a gain control. With it you should be able to get rid of any unintentional preamp distortion.

I didn't too any calculations but the 100n capacitor in the gain set of the first stage seems rather small. There might be some issues in producing the full bandwidth. Then again, maybe not. As i said, i didn't do any calculations.

As more practical additions, i'd leave PCB a place for a high frequency NFB capacitor in the second stage (in order to avoid any disappointments considering oscillation), add an RF filter in front of the first stage and protect the opamp inputs from static charges with diodes. These little additions only should lift the circuit into a way more "professional" and reliable level.

The stages might also benefit from a mild diode clipping. In my opinion it's always a better option than driving the opamps to clip. (Unless you find an opamp with fast recovery or nice clipping characters of course). Connect diodes, leds or zeners in series to the feedback loop so that their voltage drop will go just below the opamp's voltage swing per rail. (Which unfortunately depends on the opamp model. 13V is a quite safe bet, 14V might be a good value too). The diodes could be made switchable for an "extra clean" output. An effect loop or reverb would be a nice addition also.

The OD channel seems to be pretty solidly implemented into the design, i wonder if it can be ditched without affecting the tone. ...And i really doubt the need of a buffer, what use would it be?

Instead of using a buffer i would replace the first opamp stage with a jfet gain stage that has a variable gain and bigger input impedance, something as simple as the schematic attached. I would leave the protection diodes and RF filter intact and maybe alter the coupling capacitor values if needed. Might need a NFB cap too, (10pf to 100pf). With correct resistor values this simple circuit should give a nice variable gain from 2 to 20 easily: Lot more gain than from the 5x of the first opamp stage (if you happen to need it). Also provides a suitable high impedance input for a guitar. Works either as booster or "transparent" gain stage. You could also make the bypass capacitor switchable ie. for pedal use. The unused half of the TL072 can be used to buffer either the output or the tonestack: Lot more use for a buffer there.

Here's my little LM386-based "practice" amplifier. It's almost too loud for apartment use but could benefit from little more headroom when i'm playing in duo with an acoustic guitar.  Too bad i cannibalized two 5" 4 ohm speakers after building the case... Maybe i'll use them for the next LM386 project.

Nice idea. My guess is that the audience could tell a difference between amps but could not tell whether they are tube or solid state. The whole debate is actually quite funny considering the fact that at least 60% of a typical audience is not educated in electronics at any way and could not tell a difference between a SS and tube amplifier - even if they saw one.   I go to many shows and see both SS and tube amps used by the musicians regularly: Some of them use budget line Crates and some of them top quality tube amplifiers - without hassle. The way to ruin a performance has more to do with horrible balance between instruments and the singer than with the used equipment.

Everytime i find myself jumping too much on the "wonderful tone of tube amplifiers" bandwagon i take a look at this picture and return to ground level... (I believe that's a Crate GX-65)  
http://www.trollech.com/gallery/gigs/firstgig7.jpg

The effect of SPiKe protection on sine wave can be seen at LM3886 datasheet page 9. It quickly decreases the gain after the sine peak leaving sort of "spikes" to the signal. Ever heard how a CD rip sounds when the CD is corrupted? I guess it might be quite close of describing the sound. I think some of these spikes are already filtered out by the guitar speaker itself but really the protection should never even go on - after all, it is a protection. You got that right about massive heatsinks. Top quality SS amplifiers need lot of heatsinking iron, especially if they're put on a 35 degree celcius warm stage with a stage light blasting on them.

I believe we have run into a collision of terms in this thread: I mean, how do define concepts like "transparent amplifier" or "painfully high levels"? The later actually tells nothing about a tone of an amplifier: The level can be painfully high with huge distortion or painfully high with no distortion at all. Maybe we should try to be more accurate in choosing the terms we use in order to leave less room for speculation.

If the power amp is indeed "transparent", which i guess in this case means that it never distorts in a way that can be perceived as "distortion", i think it shouldn't be all that difficult to mimick, for example, a high gain tube amplifier or a speaker system's frequency response. Hardly anyone, even the "all-tube" - fanatics deny this. The common argument coming from almost anyone, myself included, is that the modelling sounds good only at "bedroom" levels. This only implies that the power amplifier is not up to it's job.

To get back to the original topic, i'd like to see a new study about solid state and tube amplifier differences too. The question is how should the study be made: Wouldn't the result be biased if the circuits aren't "identical"? Testing a triode gain stage against, say, common emitter and common source gain stages is pretty straigtforward but how about against an opamp, which is a complicated IC? And how about if we're talking about a complete amplifier system?