2 stage Stereo/4 Stage Mono Triode Preamp with independent drive control & inverters

[sampleaccurate]

Submitted for your (dis)approval, a stereo dual triode preamp with cascaded common cathode stages that are purposfully biased to compress, add second order distortion without audible clipping, and limit transients as a transparent alternative to traditional compressors. It can be adjusted from less than 1% THD to well over 20% without any significant distortion products higher than 2nd order. When the channels are cascaded it becomes a 4 stage common cathode preamp that can operate in two fundamentally different modes depending on the setting of the inverting switch.

The opamps between the tube stages serve useful purposes and without them some of the features such as distortion free inversion, selectable inverting/non-inverting configuration, and input and output buffering would not be possible or at least not nearly as easy to impliment.

I'll know in a few days how she sounds. The main purpose is signal conditioning in a digital audio workstation to actually process the tracks through a real analog device that isn't perfectly linear.

[DF96]

In order to compress (i.e. reduce the input signal) by non-linearity alone, which I think is your aim, you must have odd-order distortion not even. However, cascading two even-order stages will give odd-order so achieve the desired effect.

[sampleaccurate]

Originally Posted by DF96

In order to compress (i.e. reduce the input signal) by non-linearity alone, which I think is your aim, you must have odd-order distortion not even. However, cascading two even-order stages will give odd-order so achieve the desired effect.

The goal is to be able to do either which is why I put the inverter switch in there. One triode will compress, but it only compresses "one side" of the signal, the negative or the positive (unless you drive it hard enough). I may add another switch to the second opamp inverter to be able to create odd order distortion with only a single channel as you suggest.

I almost have the circuit board drawn out and ready to etch. Hopefully I'll have it working in a day or two. I just need to find one more OPA134 in my junkbox.

[DF96]

Second-order will expand the other side just as much as it compresses the first side, so the original signal amplitude is left unchanged. All you have done is add second-order distortion too. As I said, to compress a signal you need odd-order.

[sampleaccurate]

Originally Posted by DF96

Second-order will expand the other side just as much as it compresses the first side, so the original signal amplitude is left unchanged. All you have done is add second-order distortion too. As I said, to compress a signal you need odd-order.

There is no expansion going on here, only simple compression of "one side" of the signal due to the non-linearities of the triode. I can level/amplitude shift the input and output signals and overlap them to compare and prove that the "negative" side of the waveform is not being altered in any way and maintains almost exactly the same shape. It's the top portion of the waveform that is being compressed when the tube approaches cutoff (but never actually gets cut off) as the tube operates at very low currents during positive excursions of the waveform. The negative excursions are unaffected.

The sonic result of this is yet to be determined. You can speculate all you want to, but until I listen to it working, it's pointless really. It's fun to guess, but in reality I won't know until I plug it in and crank it up. I would concede that the uncompressed negative half cycles will reduce the overall perceived compression, but perhaps with the sonic advantage of preserving at least some of the dynamics of the original signal since only half the waveform is compressed.

I don't claim to know, but within a day or two I'll be able to apply as much or as little drive to multiple cascaded triode stages with the bias adjustable for even or odd distortion, or anywhere in between. If it sounds good I'll post some wave or flac files of examples of the effect (and it IS an effect).

[DF96]

I was not speculating, but merely stating mathematical fact. If you are getting compression on one side, and no change on the other, then you have a mixture of roughly equal amounts of odd and even order - not the pure even order you think you have.

[sampleaccurate]

Originally Posted by DF96

Second-order will expand the other side just as much as it compresses the first side, so the original signal amplitude is left unchanged. All you have done is add second-order distortion too. As I said, to compress a signal you need odd-order.

Second order does not expand the other side, and the overall peak to peak amplitude is reduced so overall the signal is slightly compressed. At best half of the waveform remains unchanged, but it certainly isn't expanded.

Originally Posted by DF96

I was not speculating, but merely stating mathematical fact. If you are getting compression on one side, and no change on the other, then you have a mixture of roughly equal amounts of odd and even order - not the pure even order you think you have.

You have your facts wrong. The "soft" compression of one side of the waveform yields predominantly second order distortion, not equal amounts of even and odd. If the signal is inverted then passed through another triode stage then yes, the circuit is capable of odd harmonic distortion is amounts greater than the even distortion it can produce when configured to do so.

I suggest you look at the harmonic spectrum of the distortion products in the FFT. The 3rd is far below the 2nd. It's common knowledge that a single common cathode triode amplification stage (single ended) generates more even harmonic distortion than odd (in general when not clipping), that cascaded triode stages generate odd distortion products more than even (or mixed depending on the drive levels - both sides of the waveform are being compressed), and that a push pull output stage will generate more odd than even due to cancellation of the even products, so I'm not sure where you get:

[/QUOTE]

Originally Posted by DF96

"If you are getting compression on one side, and no change on the other, then you have a mixture of roughly equal amounts of odd and even order - not the pure even order you think you have".

That's not correct. Here's the spectrum of two cascaded 12ax7 stages with the signal inverted between stages so that the even harmonics add:

[sampleaccurate]

The partial build - one channel plus the op-amp balanced supply.

[DF96]

It is a simple fact of trigonometry and power series that even order cannot compress, as even order generates no products at the fundamental frequency. If you have a triode stage which uses some new mathematics unknown to science then I suggest you write it up for publication.

What may be happening is that the compression is being caused by a bias shift, as even order does create DC which will tend to bias the valve back away from its quiescent point and so reduce its gain. This means that your steady-state and transient responses will be different: it will compress sustained signals but not transients, as the bias will not have time to shift.

[sampleaccurate]

Originally Posted by DF96

It is a simple fact of trigonometry and power series that even order cannot compress, as even order generates no products at the fundamental frequency. If you have a triode stage which uses some new mathematics unknown to science then I suggest you write it up for publication.

What may be happening is that the compression is being caused by a bias shift, as even order does create DC which will tend to bias the valve back away from its quiescent point and so reduce its gain. This means that your steady-state and transient responses will be different: it will compress sustained signals but not transients, as the bias will not have time to shift.

Explain to me why the FFT shows very low odd order distortion, almost all 2nd order, and at the same time the peak to peak amplitue of the output waveform is attenuated by at least 4dB, all on one side, but 4dB from peak to peak nevertheless. I call that compression. I agree that asymetrical signals can cause a bias shift but that's irrelevant to this discussion. I don't think you've looked at the schematic. The bias is stable and the compression on transients should be no different than steady state.

If you post your mathmatical proof that partially truncating "one side" of an audio signal doesn't reduce the peak amplitude I'm all ears. And if you can explain how the other side is magically "expanded" I'll really be paying attention. If you want to dispute the results of the FFT take it up with ORCAD.