Solid state vs Tube

This debate is very, very old. In the beginning there were vacuum tubes. These electronic devices were used to build audio amplifiers. When the electric guitar was invented, naturally , tubes were used to build amplifiers for the guitar signal. In the 1950's, the transistor was invented. It took many years for transistors to percolate up to guitar amplification, but by the 70's ,  transistor based amplifiers were common. However, there was a complaint coming from the musicians. They claimed that the transistor amps "had no tone", "lacked  warmth", or just plain "sounded shitty". Audio engineers would look at the specifications and scratch their heads. "What are these long hairs talking about.". The THD is low, the amplification head room excellent, the frequency response stellar  etc.  However, the musicians were adamant that tubes sounded better.  So much so that they continued to buy tube amps even as they climbed in price as tubes were becoming obsolete for every other electronic application. Was this just a case of being nostalgic for old technology? That might account for some of it, but when musicians are willing to pay 3 times as much for the same audio power if it is delivered by tubes, then there has to be more to it. And it turns out, there is. Engineers tend to focus on the linear models used for amplifier design. Gain and Bandwidth are the primary linear modeling parameters for an  audio amplifier. Distortion is characterized by how much is produced via the 'Total Harmonic Distortion' spec, but the spectral composition of the distortion  is not considered with THD, only the  total spectral power generated by non-linear amplification. The reason is simple, distortion is bad, and when designing an audio amplifier you want to minimize it. However, guitar amplifiers are not used as linear amplifiers. They are often driven into deep non-linearity by the guitar player to generate the characteristic "fuzz" of rock guitar. So the gap was that standard amplifier design methodology did not consider in detail the spectral composition of harmonic distortion. When they did, they found out one key difference between the transistor and tube amplifiers. Tubes tend to produce even order harmonic distortion and transistors produce more odd order harmonics. This means that if the string frequency is f, a tube amp will produce overtones at 2f,4f... Where as a transistor amp will produce 3f,5f. Even order harmonics are musical  and pleasing to the ear. Odd orders sound dissonant. This was the first breakthrough at understanding something that was only qualitative before.  The even order harmonic generation by tubes occurred because tubes are what are called a "square law" devices. The tube transconductance is expressed approximately as a I=K*V^2.  Transistors on the other hand were exponential.  It turns out that a certain class of transistors had square law behavior. These were MOSFETS. Guitar amp manufacturers started designing transistor based amps with MOSFETS and achieved some pretty good results. The day of the "buzzy" transistor amp was gone. Still musicians weren't satisfied and pros continued to buy tube amps. Amateurs now had a lower cost alternative that produced fairly good overdriven tone. In the end, it would not be possible to completely replicate the dynamic characteristics of vacuum tubes. In addition, some of the effects produced by tube amps  resulted from the high output impedance tube amps connected to a low impedance speaker via a transformer. The dynamic characteristics of the transformer also helped shape the sound due to a phenomenon called core saturation. Sharp transients into the magnetic transformer would be softened by the magnetic saturation of the transformer core. There are many more differences such as open loop design vs negative feedback. But the point is that trying to make a transistor based circuit sound exactly like a tube amp just isn't possible. So in the end, replication of tube amp sound by solid state electronics would not happen until DSP based amp modeling became available. Here a digital model of the tube amp and speaker produce very accurate sound.   So to conclude, if you are a gigging musician, what type of amp should you use? If you are playing solo or at relatively low volume, use a solid state amp with an amp modeler. Using a solid state amp has many advantages : they weigh less, are more reliable and also more durable. The amp modelers are so good now that you can get any guitar sound you want. If you are a rock band operating at high volume you still need tube amps. Amp modeling won't work  because if you are already simulating an overdriven stack you can't run that through an overdriven stack!  In the studio, there is no competition. Amp modeling all the way.

Digital vs Analog

Having spent the last couple of weeks delving into some guitar effects research, I have come across a lot of mis-understandings on the analog vs digital guitar effects debate. Of course people can decide what they like to hear for themselves, though I'd still like to see some double blind sound tests. Those that claim they can tell a real tube amp from a modeling amp should be put to the test.

The number one myth I keep seeing as I read articles on internet is that digital signal processors cannot truly represent an analog signal because they are taking a finite number of samples per second. Now I don't want to go into sampling theory for band limited systems here since there are many well written discourses on this subject. I'll simply state, and leave it as an exercise for the reader to verify, that this belief is completely false. Analog systems also have finite resolution in amplitude because of electronic noise and finite resolution in time because of inherent system bandwidth. The belief that analog signal representation provides for infinite resolution is completely false. If one samples an analog signal at or greater than the highest frequency component in that signal, the reconstruction of that signal is complete and there is no loss of information. This is fundamental to digital signal processing and is why it works.

So what about stomp boxes? which is better analog or digital? Well, it depends on the application. If you are driving a real tube amplifier then an analog pedal is better. Why? because of simplicity. A pedal like the Tube Screamer is a simple OP-AMP circuit that provides for some clipping and equalization. Its purpose is to drive a tube amplifier to get saturated tone without having to crank up the amp all the way. However it is only functioning as a preamp. You cannot take the output from the Tube Screamer and go direct to a recording device. Well, you can, but you won't get a good sound. There is no point in doing a digital implementation of this trivial analog circuit. So, its not because a digital implementation would sound inferior, its because it doesn't make since to implement a preamp in DSP.

Now, how about amp emulation for direct recording. This is an area that was pioneered by Tom Sholze of Boston fame. Tom Sholze was obsessed with direct recording of saturated guitar amp tone. That is, recording guitar distortion sounds without using microphones set up around a real overdriven amplifier. An early approach he employed was to build a dummy load for a guitar amp. A speaker presents usually around an 8 ohm impedance to the output stage of a guitar amplifier. This impedance represents the conversion of the electrical energy from the output stage of the amplifier to sound energy transmitted to the air by the speaker cone. However, one can build a 'dummy' load that will convert most of the amplifier output to heat instead of sound. And then recover the signal as it appears across the dummy load to create a low volume version of the amplifiers output wave form. This way the output stage of the amplifier is saturated as it would be driving a speaker and you can direct record this waveform without even using a speaker. Now these devices are called 'power soaks'. The second generation of direct saturated guitar tone consisted of an analog circuit that would later appear in the 'Rockman' product developed by Tom Sholze. As it turns out, simple solid state clipping circuits, like OP-AMPS with diodes, can be made to sound remarkably like an overdriven tube amp if the proper frequency equalization is employed. This is what Tom Sholze next accomplished. He found the right combination of clipping and frequency equalization to get a very realistic sounding signal that could be recorded directly without using speakers and microphones. For those of you that have played with the Rockman, I think you will agree that it sounds pretty good direct to headphones. As you can see the purpose of an amplifier emulator is quite different than that of a distortion stomp box. The stomp box is a preamp that will augment a tube amplifier into producing saturated guitar tone. The amplifier emulator is a complete representation of a tube amp and speaker cabinet.
So where does digital come in here? The Rockman is pure analog and can emulate an amplifier. Well, actually this area is where the digital approach really shines. The rockman is cool, but the problem is that the range of equalization that can be created is very limited. The Rockman uses real analog active filters based on OP-AMPS and as such cannot implement complex low pass responses that are easily done on a digital signal processor. In addition, the type of nonlinearity used to produce clipping is completely programmable in a DSP system. The companies that build amp modeling units use a combination of white box analysis ( studying the circuits in the amplifiers) and measurement characterization to encode very detailed models of the amplifier they are emulating. These units , when used for direct recording or listening, sound amazingly realistic. There is simply no way for simple analog circuits to accomplish that degree of system modeling, without actually building the real system ( amplifier + speaker cabs).

Musicians tend to be retro oriented and still favor the stomp box to multi-effect units. However that says more about the desired user interface than the guts of the pedals. Digitech, for example, offers it's effects in stomp box form as well as integrated even though the stomp box's are still using DSP. The gigging musician prefers the separate pedals perhaps for ease of actuation, but they are , in many cases, still using DSP.

Just as audiophiles have rigid beliefs about vintage gear sounding better, so do guitar players have strong opinions on how to get killer tone. In the end, each musician uses what gets them that tone. But sometimes we have to have an open mind about the new technology and ask ourselves, Are we using some retro piece of equipment because it really sounds better, or because it's cool to be retro? Recording without microphones is a great thing and digital guitar processors make this possible without giving up the tone.