It looks like you build a compressor. It woiuld be of interest to see how the distirtion changes when the conmpressor becomes active.
BTW Organising the gain structure so that the preamp clips before the poweramp can also be of interest.
Jan
BTW Organising the gain structure so that the preamp clips before the poweramp can also be of interest.
Jan
This amp has two parts (it is integrated type), preamp with no GNFB and OPS with GNFB and gain of two. The volume control is after preamp part.
This anti saturation circuit is to prevent OPS to saturate. The distortion is quite low and start to increase wen zeners start to conduct of course.
Why is important how distortion changes when anti saturation circuit (I would not call it compressor) become active?
This anti saturation circuit is to prevent OPS to saturate. The distortion is quite low and start to increase wen zeners start to conduct of course.
Why is important how distortion changes when anti saturation circuit (I would not call it compressor) become active?
I made this softclipper with simple LED's:
https://www.diyaudio.com/community/...-8-ohm-how-to-limit-power.421941/post-7890159
https://www.diyaudio.com/community/...-8-ohm-how-to-limit-power.421941/post-7890258
Its TDH < 100dB upto -9dB with softly increasing distortion above that, very happy with the solution and you can see when it triggers 🙂
Last edited:
You would want to know how rapid the distortion rises at the onset of overload. How much better is it than just clipping?
Jan
Indeed, looking at datasheets for BC327 and BC337 (several manufacturers) show VBEo MAX as 5V. I've seen reverse-biased B-E junctions of transistors used as noise generators in the 5 to 10V range where they have a zener response, but somewhere I've read advice against doing this, that it damages the transistor/reduces HFE. Then again that may not much matter in this circuit.
You can see from rounded edges when clips that distortion is not abrupt, and the main advantage is that OPS does not clip, as there is no saturation.
If I insert a resistor in series with those zener diodes the clipping is much softer.
First one is just on the brink of the clipping, second one is dip in the clipping.
First one is just on the brink of the clipping, second one is dip in the clipping.
Attachments
From the graphs it apperas that with the series resistor there is less gain, and that results in less clipping.
It is very hard to judge things from a scope screen with many other factors not known.
The best is a graph of distortion versus output level in the various conditions.
Jan
It is very hard to judge things from a scope screen with many other factors not known.
The best is a graph of distortion versus output level in the various conditions.
Jan
Hi Jan,
I built it and this is how it looks like:
The module has two channels.
I also measured it with the threshold set very low ahead of a balanced input, which has a connector mating with this module.
I swapped the module with jumpers back and forth.
Just prior to clipping, distortion was not measurable independent of frequency.
In my mind, it does not make any sense that distortion of this circuit rises with increasing frequency.
I believe the simulation is wrong.
As long as the operational amplifier can drive the other side of the diodes to the same potential as the signal, there is zero voltage across the diodes.
Means the op-amp would need to distort enough that there is a delta voltage that allows a bit to creep through the diodes.
Of course, distortion gets worse with rising frequency, especially with such budget amplifiers, but it would need to be really heavy to generate substantial delta.
I found some fast recovery diodes for the built, but could not record clipping yet, because I gave up improvising an adapter for the measurement.
I will design a proper test jig instead. I have more similar modules sharing this connector and need a proper adapter anyway.
Distortion free Distortion setup?, use the last 10% of amp voltage to compress the signal. Properly biased tube 12AX7 can work well.
Tubes will also "fatten" the waveform before clipping. I estimate the 6922 preamp gives me at least 3db extra headroom
Tubes will also "fatten" the waveform before clipping. I estimate the 6922 preamp gives me at least 3db extra headroom
I agree in principle, except my guess would be the last 50-75% of amp voltage. Seems outrageous at first glance but it's only 6-12dB of cushioning.
Maybe some kind of listening test is in order? Perhaps a Python script to apply test-run some different compression curves?
Maybe some kind of listening test is in order? Perhaps a Python script to apply test-run some different compression curves?
One factor to consider is that opamp input difference voltage is usually assumed zero but it isn't - it is Aol/Vout.
Since the open loop gain falls with frequency, input difference voltage increases with frequency and voltage across the diodes also increases with frequency for the same signal voltage.
That may explain increased distortion with frequency.
What I am amazed at is that nobody takes the trouble to do the one important graph, distortion versus input signal, that can easily show the relative performance of different solutions.
All 'designers' * appear to just waffle about it. Why?
Jan
* someone designing a circuit without showing the factual performance is not a real designer.
Attachments
This is a good & interesting thread.
I haven't read it all.
Has anyone posted the method of using a low distortion chip-amp to drive an incandescent bulb/globe of the appropriate value into
a suitable load to achieve a soft & subtle audio compressor with 'very musical attack & decay times' ?
I haven't read it all.
Has anyone posted the method of using a low distortion chip-amp to drive an incandescent bulb/globe of the appropriate value into
a suitable load to achieve a soft & subtle audio compressor with 'very musical attack & decay times' ?
Last edited:
I've thought about this, regarding guitar amps. One would think "REW" or something could do it as a frequency sweep, incrementing level bit by bit. This would result in a 3D plot, where you can select which harmonic to display. I could probably figure it out in LabView and my HP function generator with GPIB, but that's not something others could easily make use of like REW's FR generating and graphing functions.
They somehow claim to emulate the dynamic behavior of different guitar amps using a fancy box with DSP. I wonder if this "distortion versus input signal" level analysis is what they start off with? If so, someone's done it to get what they want and productize the result.
It's a simple funntion in REW to graph distortion vs F or vs V.
One-click once REW is set up.
But even if it was hard, you'd still do it, right?
What is the point of discussing it without having a clue how it works?
A single freq sine wave with partly flattened tops tells you zilch.
Jan
One-click once REW is set up.
But even if it was hard, you'd still do it, right?
What is the point of discussing it without having a clue how it works?
A single freq sine wave with partly flattened tops tells you zilch.
Jan
