Amp Resonance

.

Do you ever consider input and output impedances when you make your circuits?

CMOB-2S appears to a 600 ohm 1:1 transformer. This is a rather low value, so how sure are you that your circuits can actually drive it properly without running into problems?

NE5534 has a push-pull output stage that can at best handle 35-40 mA. At say, 12V peak into 600 ohms you don't need more than 20 mA, so it will work. It will however have a bit more distortion than if you had use a higher impedance load. It can be unstable for gains lower than 3, unless you add a compensation cap.

Regarding your single stage amp, who knows... draw up schematics of both

While listening tests are important, you can get more objective tests of the frequency response, distortion etc. using software like RMAA, REW, ARTA, HolmImpulse and many others.

Guitar amp tone and especially distorted sounds are the result of gentle and incremental filtering and clipping. More specifically you boost certain frequencies like in the human voice range and maybe a little higher like 300Hz to 7kHz such that it sounds bright and has nice detail and you can hear hammering of individual frets and little harmonic tones from picking and such. Then you amplify that just enough for it to clip a little, then gently low pass < 10kHz so that the squared edges of the straight clipped waveform turn into little spikes on the leading edges. Those little spikes give you the bite that you hear in fender or marshall style amp.

Having said that, I have tried very hard to create a transistor / op amp based circuit that sounds like a tube amp (hasn't everyone?) and I was never satisfied with the results. The problem is the "amplify that just enough for it to clip a little" part because the onset of clipping is relatively quick with transistors / op amps compared to a tube. Meaning when you ultimately play through one of these circuits it will tend to be either clean or distorted but not both. In this situation I always felt compelled to just add more gain in which case you end up with a distorted mess.

IMO the threshold at which a tube amp starts to distort is what makes a tube amp special. And that is very difficult to do with solid state circuitry. It might be impossible.

80hinhiding said:

Anyway, here are my circuits. I just drew them up.  I think I should probably add a resistor off the + input on the op amp too.

Click to expand...

Interesting. As usual, seeing the real thing explains a lot.

In the first amp, you have given up trying to convert balanced to unbalanced? The poor negative input is just shorted to ground through a cap. Is that because it comes from another transformer? In that case, just ground it. Otherwise don't connect it at all. With the given values the transistor may start clipping at around 1V peak. Decreasing the 470k bias resistor that goes to ground down to, oh.. maybe as low as 100k, should give you more headroom. For the output, I can't figure out what the 470k's and the additional capacitor to ground is good for. Leave them out in both amps? Maybe you have a good reason for doing it.

The second amp is a good attempt at converting the balanced signal to unbalanced and provide gain. In the first amp, the input capacitor is grounded at the input side with 470k, here it is not. Either way will work, but the reason for doing it is to keep the input side at ground even when nothing is connected, so that you don't get a *click* when something is connected to the input. The coupling caps are only 3.3u, which seems a bit low compared to the 100u you used for the other amp. Moving on... 2.2k and 22k in series? Ok, if that's the exact value you wanted, go for it. Then you have 4.7k + 4.7k for some reason

You have biased both inputs, but that isn't necessary. The opamp will do its best to keep both inputs the same by changing its output. So out with the two 100k's on the - input side.

The + input is shorted to ground through a 3.3u cap. That will lowpass the signal at ~3 Hz. That's probably not what you wanted? Any hope of unbalancing the signal goes out the window. Remove that cap.

Now, to actually unbalance the signal properly, the gain between the +/- sides need to be equal, which means the resistor ratios must be the same. They are not:
- side 470k / (2.2k+22k+4.7k+4.7k) = 470k / 33.6 = ~13.98
+ side (100k || 100k) / (2.2k+22k) = ~2.05

To correct it, I would remove the 4.7k+4.7k and the two 100k, then create a stable 15V "reference" using, say, two 10k and a cap from the midpoint to ground. I would use a bit more than 3.3u. Then you can put a single resistor from the midpoint to the - input to bias it. The resistor should have the same value as the feedback resistor. Gain calculations now start making sense again, being 470k / 24.2k = ~19.42 times (25.8 dB). Maybe a bit high, since the first amp would have little to no gain.

PS: Adding a resistor off the + input on the opamp would do (almost) nothing at all.

80hinhiding said:

Oh, 3.3uF for the inputs on the op amp.  I remembered seeing those values on another schematic a while back and decided to try them for this one.  I'll increase those.

Click to expand...

You could, but you don't have to. 3 Hz as the lower cutoff is good enough for rock'n'roll (and most other types of music I would imagine).

80hinhiding said:

I've found input impedance and the bias definitely impacted that for me.  So I think this might tell me I like some clipping or maybe I like the compliment of a clean tone mixed with a slightly clipped/distorted/compressed one.

Click to expand...

Maybe you should make something (loosely) based on the tube screamer pedal? With a low gain there's not much distortion and the clean sound is always mixed in.

I guess maybe this is why I've seen bias controls on some gear.

Click to expand...

Depends on what gear you have in mind. For power amplifiers, what they call bias is more about how much idle current that flows with no signal being input. This is done to avoid crossover distortion, but here it's a single transistor in class A, so it's more about getting it to the point where it operates without clipping.

I'm kind of confused as to what I should do.

Click to expand...

Figure out what you actually like? Then analyze the signal to bits That's what I do anyway - it might not work for you. Run test signals through it and look at the results - model it in spice, etc.