evilcat
Well-known member
Yep, I'm ok with Script. I can't wait to start & of your projects!!!
etheory - a bunch of projects in progress
- Thread starter etheory
- Start date
Help Support GroupDIY Audio Forum:
etheory
Well-known member
Initial results of eJFET comp compressor - revision 01 (as it is, sitting on the breadboard):
RMAA using RME Fireface 800:
Noise Level dBA = -103.9dB
Dynamic Range dBA = 104.1dB
Frequency Response 20Hz to 20kHz = +0.11dB, - 0.06dB
Total Harmonic Distortion (THD) = 2.869% (I'm rather annoyed by this figure - it's not very good at all, so I'll have to investigate what could be causing this. This simulator suggested much lower, but I already have a new design revision to test anyway....)
(EDIT)
Reference, which was connecting up the amp without any jfet-related attenuation, but maintaining the up-front passive input attenuation and make-up gain:
Noise Level dBA = -103.8dB
Dynamic Range dBA = 103.7dB
Frequency Response 20Hz to 20kHz = +0.01dB, - 0.06dB
Total Harmonic Distortion (THD) = 0.003% (it's pretty clear from this impressive value directly from the e2510 DOA, that I still have some work to do on the jfet circuit, but also that the e2510 is pretty nice....
)
Revision 1
With compression:
With a few modifications I'm down to 0.195% distortion at full gain.
Will work throughout the week to bring that down as low as possible, in amongst finishing off the e274.
(2nd EDIT)
Revision 2:
Noise Level dBA = -97.4dB
Dynamic Range dBA = 97.1dB
Total Harmonic Distortion (THD) = 0.15%
Revision3:
Frequency Response 20Hz to 20kHz = +0.05dB, - 0.01dB
Noise Level dBA = -90.4dB
Dynamic Range dBA = 90.5dB
Total Harmonic Distortion (THD) = 0.0027% !!!!
OK, now we are getting somewhere.... That's an awesome distortion measurement - now to make everything else match that quality level, including noise and dynamic range.
This was using a 2N5458 jfet (because that's the only one I can get in Australia from element14 that fits my purpose, though I will also try J111, VCR2N and a few others I found with a high Vgs(off) from futurelec whilst they still have them....). Though it doesn't seem too sensitive a circuit since it automatically linearizes the JFET response for any given CV range, and can be tuned for different JFETs with different Vgs(off).
WHAT a learning experience this is....
I found that applying 15dB gain reduction using the jfet makes it SUPER linear. From 15dB to 60dB of gain reduction there are almost no harmonics at all.
But when the input signal rises, or the gate voltage moves back towards zero, the distortion soars. It seems to need a little gate voltage love for the lowest distortion values.
The graphs look almost the same as those of the sound card. It's VERY close to being completely transparent (except for the elevated noise level due to the up front passive attenuation)....
Frequency response is ruler flat across the entire range.
Bass response is ridiculously good - even a square wave at 100Hz looks almost perfect out the other end.
This is all just for the vca with no gain reduction. In the latest version gain reduction produced no increase in distortion all the way to maximum attenuation, which is pretty cool.
RMAA using RME Fireface 800:
Noise Level dBA = -103.9dB
Dynamic Range dBA = 104.1dB
Frequency Response 20Hz to 20kHz = +0.11dB, - 0.06dB
Total Harmonic Distortion (THD) = 2.869% (I'm rather annoyed by this figure - it's not very good at all, so I'll have to investigate what could be causing this. This simulator suggested much lower, but I already have a new design revision to test anyway....)
(EDIT)
Reference, which was connecting up the amp without any jfet-related attenuation, but maintaining the up-front passive input attenuation and make-up gain:
Noise Level dBA = -103.8dB
Dynamic Range dBA = 103.7dB
Frequency Response 20Hz to 20kHz = +0.01dB, - 0.06dB
Total Harmonic Distortion (THD) = 0.003% (it's pretty clear from this impressive value directly from the e2510 DOA, that I still have some work to do on the jfet circuit, but also that the e2510 is pretty nice....
)Revision 1
With compression:
With a few modifications I'm down to 0.195% distortion at full gain.
Will work throughout the week to bring that down as low as possible, in amongst finishing off the e274.
(2nd EDIT)
Revision 2:
Noise Level dBA = -97.4dB
Dynamic Range dBA = 97.1dB
Total Harmonic Distortion (THD) = 0.15%
Revision3:
Frequency Response 20Hz to 20kHz = +0.05dB, - 0.01dB
Noise Level dBA = -90.4dB
Dynamic Range dBA = 90.5dB
Total Harmonic Distortion (THD) = 0.0027% !!!!
OK, now we are getting somewhere.... That's an awesome distortion measurement - now to make everything else match that quality level, including noise and dynamic range.
This was using a 2N5458 jfet (because that's the only one I can get in Australia from element14 that fits my purpose, though I will also try J111, VCR2N and a few others I found with a high Vgs(off) from futurelec whilst they still have them....). Though it doesn't seem too sensitive a circuit since it automatically linearizes the JFET response for any given CV range, and can be tuned for different JFETs with different Vgs(off).
WHAT a learning experience this is....
I found that applying 15dB gain reduction using the jfet makes it SUPER linear. From 15dB to 60dB of gain reduction there are almost no harmonics at all.
But when the input signal rises, or the gate voltage moves back towards zero, the distortion soars. It seems to need a little gate voltage love for the lowest distortion values.
The graphs look almost the same as those of the sound card. It's VERY close to being completely transparent (except for the elevated noise level due to the up front passive attenuation)....
Frequency response is ruler flat across the entire range.
Bass response is ridiculously good - even a square wave at 100Hz looks almost perfect out the other end.
This is all just for the vca with no gain reduction. In the latest version gain reduction produced no increase in distortion all the way to maximum attenuation, which is pretty cool.
First of all, thanks for this thread. For me this is one of the most inspiring ones at the moment.
In medias res:
In the e274, is there a particular reason why C5 is 0.22uf? Could it be 1uf or even 2uf? Or does that mess up something here? I've just been playing around with some resistor values in my mind to get a virtual "feel" of the HPF (IN, cap, followed by resistor to ground, OUT) and because I thought, if I were to build only one channel of this, perhaps I'd prefer a 10K pot (with 1uf for corner frequency at 15.9 Hz) instead of a multi-position switch. But then again, for equal left/right channel settings, pots are probably not precise enough.
In medias res:
In the e274, is there a particular reason why C5 is 0.22uf? Could it be 1uf or even 2uf? Or does that mess up something here? I've just been playing around with some resistor values in my mind to get a virtual "feel" of the HPF (IN, cap, followed by resistor to ground, OUT) and because I thought, if I were to build only one channel of this, perhaps I'd prefer a 10K pot (with 1uf for corner frequency at 15.9 Hz) instead of a multi-position switch. But then again, for equal left/right channel settings, pots are probably not precise enough.
etheory
Well-known member
Thank you, that's very nice of you to say. Glad you are enjoying my education ;DScript said:
Yes, it had the least CV feedthrough when I measured it. I tried 1uF and 2uF and they weren't as "good" for me. For you, maybe you are OK with the added thump, depends on what you want to use it on.Script said:
You should try it yes, the circuit is rife for experimentation. One huge reason I want the boards to be 2mm instead of 1.6mm thickness and to use a higher weight of copper is for experimenter abuse. I want people to mess with the circuit, solder and desolder, and have the PCB able to cope with some learning. Because there are a lot of different things you can do.Script said:
I'll post the LTSpice simulation too so people can mess with it. There are a lot of things you can gather from the simulation, since it's so accurate and close to the physical unit, that you can save yourself a lot of de-soldering by playing with it and the values in it.
No, it won't mess anything up, it just reduces the speed that the circuit can cancel out the cv feedthrough with, so you get longer duration "bumps" of cv coming through.Script said:
I've specifically designed the circuit to be able to take 10K Audio Log pots for all controls, or a multi-position switch. The choice is up to you. Precision isn't the aim of this circuit, it's really just a modern clone, so it has a lot of the quirks of the Siemens Sitral U274, good and bad, but it's slightly more modern, and slightly better behaving, and a very different choice of tone than what most people have access to currently.
One of the things I'm most curious about is revisiting a lot of designs that a lot of people haven't seen, or are rare and unheard, and re-incarnating them in hopefully interesting ways. It's good to see people are interested in this.
Sorry for the plodding speed of this thread also. Time is a at a premium for me. I am currently working full-time on some deliveries at my day job. We recently finished the first press-release for the Lego Movie that I'm working on at Animal Logic. Here is the huge render we did: http://www.hdwallpaperstock.in/walls/the_lego_movie_2014-wide.jpg
evilcat
Well-known member
It seems that you're in good progress with your eFETproject, weel done. I hope we could here some samples soon !
Ben
Ben
etheory
Well-known member
A couple of other measurements of the eJFET comp.
With negligible distortion at all times, here are some signals, at 1kHz
-31.31dB in is the threshold where distortion across the JFET drops to nearly 0%.
The output of the amp at this point is -6.68dB.
The gain range from 0V CV into the VCA to 1.943V is -6.68dB to -43.17dB or a range of over 36.49dB.
Not too bad. As per usual, almost none of this matches the simulation THAT well, but it's close-ish. I'm still quite impressed. In theory this means it would be considerably cleaner sounding than a 1176. Still not sure if that's a good or bad thing yet. More soon....
(EDIT) Hmmm..... going to have to redo some of these measurements.
Unfortunately my soundcard seems to be outputting -6dBu when set to 0dBu. Godda figure out why then redo these measurements.
(EDIT) I've been doing some experimenting, and I think it's going to be possible to provide the options for +-15, +-24, and +24V rails on the same board (switchable, not simultaneous). This should allow for BigFetBloke, 2520, 2510, Melcor, NTP M100, OA10, Neve BA338/340/640, and other opamp/DOA varieties, which would be very useful for experimentation and "finding your own sound".
I'm toying with a few possible ways to provide this level of reconfigurability, but it seems like the right thing to be doing....
So far I think the NTP M100 or BigFetBloke is the one I will go with (when I finally get it working!), but I can't wait to test with the BA338/340 2520-footprint variants I am going to have going soon for some class-A FET action 8).... Looks like I might also have to provide a way to fit a sub-board for surface mount FET's now that a lot of through-hole ones are becoming scarce....
So much to do, so much to do.
(EDIT) - I think I've settled on the JFET comp input arrangement now.
-20dB cut-down on the input consisting of a 10K and 100R voltage divider.
0.013% distortion at the output with a 1kHz +4dBu input signal.
The spectrum from 20Hz to 20kHz also gives that same distortion figure at that input level.
Lower levels distort much less (down to about 0.003% for half that input level).
All distortion is 2nd harmonic.
Frequency response is -1dB at 12Hz and -0.6dB at 40kHz, so rather good.
Noise Level is -96.4dB and dynamic range is 97dB, sitting on the breadboard, so I would expect much better than that once the pronounced 50Hz produced by the mains cable sitting near it drops down once it's inside of a shielded box.
OK, now to go back to, and finish off, the e274 diode compressor....
With negligible distortion at all times, here are some signals, at 1kHz
-31.31dB in is the threshold where distortion across the JFET drops to nearly 0%.
The output of the amp at this point is -6.68dB.
The gain range from 0V CV into the VCA to 1.943V is -6.68dB to -43.17dB or a range of over 36.49dB.
Not too bad. As per usual, almost none of this matches the simulation THAT well, but it's close-ish. I'm still quite impressed. In theory this means it would be considerably cleaner sounding than a 1176. Still not sure if that's a good or bad thing yet
. More soon....(EDIT) Hmmm..... going to have to redo some of these measurements.
Unfortunately my soundcard seems to be outputting -6dBu when set to 0dBu. Godda figure out why then redo these measurements.
(EDIT) I've been doing some experimenting, and I think it's going to be possible to provide the options for +-15, +-24, and +24V rails on the same board (switchable, not simultaneous). This should allow for BigFetBloke, 2520, 2510, Melcor, NTP M100, OA10, Neve BA338/340/640, and other opamp/DOA varieties, which would be very useful for experimentation and "finding your own sound".
I'm toying with a few possible ways to provide this level of reconfigurability, but it seems like the right thing to be doing....
So far I think the NTP M100 or BigFetBloke is the one I will go with (when I finally get it working!), but I can't wait to test with the BA338/340 2520-footprint variants I am going to have going soon for some class-A FET action 8).... Looks like I might also have to provide a way to fit a sub-board for surface mount FET's now that a lot of through-hole ones are becoming scarce....
So much to do, so much to do.
(EDIT) - I think I've settled on the JFET comp input arrangement now.
-20dB cut-down on the input consisting of a 10K and 100R voltage divider.
0.013% distortion at the output with a 1kHz +4dBu input signal.
The spectrum from 20Hz to 20kHz also gives that same distortion figure at that input level.
Lower levels distort much less (down to about 0.003% for half that input level).
All distortion is 2nd harmonic.
Frequency response is -1dB at 12Hz and -0.6dB at 40kHz, so rather good.
Noise Level is -96.4dB and dynamic range is 97dB, sitting on the breadboard, so I would expect much better than that once the pronounced 50Hz produced by the mains cable sitting near it drops down once it's inside of a shielded box.
OK, now to go back to, and finish off, the e274 diode compressor....
etheory
Well-known member
More eJFET comp distractions.... (I can't stop thinking about it....)
I just made another circuit change that takes that 0.013% distortion down to 0.0094%.
But another one on top of that has me completely stumped.
Every piece of literature I've read so far indicates that feeding a JFET a floating CV signal summed with 1/2 of the audio signal gives the least distortion.
Well, I can't explain this, but my measurements indicate that feeding about 1/4rd the audio level (instead of 1/2) reduces that distortion even further. I'm still kind of confused by this, but it's something I can measure repeatedly with the same results....
But whatever, a measurement is a measurement, and it's better, so I'm keeping it.
I just made another circuit change that takes that 0.013% distortion down to 0.0094%.
But another one on top of that has me completely stumped.
Every piece of literature I've read so far indicates that feeding a JFET a floating CV signal summed with 1/2 of the audio signal gives the least distortion.
Well, I can't explain this, but my measurements indicate that feeding about 1/4rd the audio level (instead of 1/2) reduces that distortion even further. I'm still kind of confused by this, but it's something I can measure repeatedly with the same results....
But whatever, a measurement is a measurement, and it's better, so I'm keeping it.
etheory
Well-known member
The eJFET, jfet-based compressor prototype, is now operational.
I'll post some audio after more tests soon.
Best description I can think of so far is extremely transparent (with fast attack and long decay).
It's also very fast with attack times down to 100uS! Very excited by this one!
It's also my first totally original compressor design.
I'm going to start tuning the attack and release times to various ones that sound good to me and start to integrate that switching into the design.
Man I have to get back to that e274
I'll post some audio after more tests soon.
Best description I can think of so far is extremely transparent (with fast attack and long decay).
It's also very fast with attack times down to 100uS! Very excited by this one!
It's also my first totally original compressor design.
I'm going to start tuning the attack and release times to various ones that sound good to me and start to integrate that switching into the design.
Man I have to get back to that e274
Attachments
etheory
Well-known member
Here is a recording of the eJFET Compressor/Limiter using a side-chain heavily inspired by the Fairchild 670.
Timing constants are 100R driving a 2.2uF timing cap with a 150K release resistor (to approximate the 2uF and 220K + 68K of the 670).
This provides approximately 0.9ms attack and 0.08s release by my calculations (calculating to 90% charge or discharge, so to the usual 63.2% it's a lot quicker....).
This approximates the Timing setting 1 on the 670 (ish).
11 drum loops are processed, first dry, then through the eJFET.
After this is a long solo, followed by the same solo processed, in order to hear how nicely it can level a longer nuanced performance.
The same settings were used throughout. I haven't measured the threshold or ratio yet, but gain make-up was set to 15.5dB, so a lot of compression was going on.
To me this sounds kind of vintage is some cases, and just harsh on others, but it's quite a precise accurate sound - very high frequency response and no CV feed-through at all.
A +-15V BigFetBloke (thanks Tamas!) was used for the DOA (with a 24K resistor replacing the 1.1mA diode current source).
What do you think? 8)
I'll try some more dynamic material with longer attack settings like vocals and see how it sounds (I'd also love to try it with some of the Haufe iron I have here on the output.... These recordings were simply unbalanced in/out straight to the DOA with AC coupling, that's it):
http://www.evolutionarytheory.com/downloads/wav/eJFET_Comp_Limiter/eJFET_Limiter_Comp_Test_Recording_v01_01.mp3
Timing constants are 100R driving a 2.2uF timing cap with a 150K release resistor (to approximate the 2uF and 220K + 68K of the 670).
This provides approximately 0.9ms attack and 0.08s release by my calculations (calculating to 90% charge or discharge, so to the usual 63.2% it's a lot quicker....).
This approximates the Timing setting 1 on the 670 (ish).
11 drum loops are processed, first dry, then through the eJFET.
After this is a long solo, followed by the same solo processed, in order to hear how nicely it can level a longer nuanced performance.
The same settings were used throughout. I haven't measured the threshold or ratio yet, but gain make-up was set to 15.5dB, so a lot of compression was going on.
To me this sounds kind of vintage is some cases, and just harsh on others, but it's quite a precise accurate sound - very high frequency response and no CV feed-through at all.
A +-15V BigFetBloke (thanks Tamas!) was used for the DOA (with a 24K resistor replacing the 1.1mA diode current source).
What do you think? 8)
I'll try some more dynamic material with longer attack settings like vocals and see how it sounds (I'd also love to try it with some of the Haufe iron I have here on the output.... These recordings were simply unbalanced in/out straight to the DOA with AC coupling, that's it):
http://www.evolutionarytheory.com/downloads/wav/eJFET_Comp_Limiter/eJFET_Limiter_Comp_Test_Recording_v01_01.mp3
shot
Well-known member
Those samples sound awesome!
Did I understand right - there's one fixed setting of attack/release?
Huh! Can't wait!
Eagerly waiting for you to make pcb's avaliable!!!
Did I understand right - there's one fixed setting of attack/release?
Huh! Can't wait!
Eagerly waiting for you to make pcb's avaliable!!!
etheory
Well-known member
Thanks! Yeah I got pretty excited. I like that in some of the samples the compressor really adds more punch and excitement.shot said:
I think in other cases it just sounds mushy
, but still, overall, it does a good job at levelling.For that set of samples only one fixed setting was used. I am currently working on what settings I like, which will likely be quite a vast array, so don't worry, it'll be far more than a one trick pony ;-)shot said:
Me neither! But sorry, you guys will have to have some patience. I hope it won't take too long, but I really want to make sure I get this right. I'll have to do a few revisions before I'm totally happy. I am a real stickler for high quality, so I won't be shipping anything without being 100% sure it will operate precisely to spec for all component tolerances I am working with.shot said:Huh! Can't wait!
I am also working on the input and output options, which will include various iron combinations, and various electronically balanced options.
I want to take care of the boring but essential stuff like precise matching between stereo channels (not easy with the variance of JFETs, but the linearization process I have used should massively simplify that and prevent the need for JFET cherry-picking) and a reasonable assembly .pdf etc. etc.
These things all take time. Whilst it is exceptionally exciting, and I will be offering kits for sale, I want to make absolutely sure the quality of the result is as high as you can possibly get.
They will definitely be available, and I'll keep you all posted on the progress of the e274 diode comp and the eJFET comps. Both should not be too long. I think within a couple of months for the eJFET and a couple of weeks for the e274.shot said:
cheers!
evilcat
Well-known member
All of this sounds good man, good job !
Best,
Ben
Best,
Ben
etheory
Well-known member
Thanks evilcat!
Does anyone happen to have a Fairchild 670 that they would be OK running over the original audio?
I would dearly love to have the real thing for comparison....
Even if just for the hell of it?
If anyone is up for it please do let me know and I'll send you an audio file to use.
To be honest I'd love to hear any vari-mu on these loops.
It would be great to have a processed file for each of the 6 timing settings and a few different DC and AC theshold settings.
Any takers.... :
Does anyone happen to have a Fairchild 670 that they would be OK running over the original audio?
I would dearly love to have the real thing for comparison....
Even if just for the hell of it?
If anyone is up for it please do let me know and I'll send you an audio file to use.
To be honest I'd love to hear any vari-mu on these loops.
It would be great to have a processed file for each of the 6 timing settings and a few different DC and AC theshold settings.
Any takers.... :
etheory
Well-known member
What load is the side-chain output running in to on a Fairchild 670?
I'm curious what the load is on the output of the timing network so I can calculate better estimates of the attack and release times.
From what I can tell the SC load impedance is the transformer plus the input impedance of the tube yes?
I might do some simulations initially, but if anyone has done these calculations before I'd love to hear what you found out.
Thanks!
I'm curious what the load is on the output of the timing network so I can calculate better estimates of the attack and release times.
From what I can tell the SC load impedance is the transformer plus the input impedance of the tube yes?
I might do some simulations initially, but if anyone has done these calculations before I'd love to hear what you found out.
Thanks!
evilcat
Well-known member
etheory said:Thanks evilcat!
Does anyone happen to have a Fairchild 670 that they would be OK running over the original audio?
I would dearly love to have the real thing for comparison....
Even if just for the hell of it?
If anyone is up for it please do let me know and I'll send you an audio file to use.
To be honest I'd love to hear any vari-mu on these loops.
It would be great to have a processed file for each of the 6 timing settings and a few different DC and AC theshold settings.
Any takers.... :
I have a Tube Tech LCA 2B, which is vari-mu and have the 6 timing settings of the 670. But it have a continusly variable ratio. Interested ?
Just thought I'd drop one of those supportive-can't-wait-posts! 8)
These look ace, I'm already on board for a pair jfet and u274 comps... What sort of size are the pcbs of each? I'm in a shoehorn-it-into-a-51x slot phase!
Thanks,
Ian
These look ace, I'm already on board for a pair jfet and u274 comps... What sort of size are the pcbs of each? I'm in a shoehorn-it-into-a-51x slot phase!
Thanks,
Ian
etheory
Well-known member
evilcat said:
That would be awesome, I'll contact you with details on how you can get hold of the same, cheers!
Much appreciated.
etheory
Well-known member
And I greatly appreciate it, thank you! Sorry again for taking so long, but work is getting busier and busier as production on the film I am working on (day job) ramps up.irfrench said:
That's fantastic! The PCB for the e274 should be approx 5cm x 15cm in size, since this gives me an excellent price/quality ratio for the PCB and more options.irfrench said:
The eJFET is still theoretically in the design stage, so no info on that yet, it really depends how much more complex I grow the side-chain. Currently it's got transformer in/out and 2 x DOA's for the signal chain, but the side-chain might also have 1 or 2 DOA's in it. I make all my stuff for 1U, so I might have to carefully consider how to fit the eJFET into 500 series.
I feel the eJFET comp will be the one that gets the most attention overall though, since it should be pretty damn cool when I'm finished with it
I recently figured out a threshold/ratio technique I'm finally happy with, so I still have more prototyping/measuring to go on that one.
etheory
Well-known member
I just got 30 x VSR2N's from Electronic Goldmine for the eJFET comp prototype to see what the fuss is about with these.
My intention is to draw up some graphs of CV voltage vs gain curves for the VSR2N, J111, 2N5458, 2N5457, 2N5459, 2n3819, and whatever else I have handy and contrast that against distortion performance to see which is the right one for me.
I'll post all the results.
I'm curious mostly about the 2N5457/2N3819/VCR2N since they are the ones I see the most in schematics. But I've had excellent results already from the 2N5458, so it's probably also true that with the right design you might be able to equally use any of them to good effect.
What I think would be REALLY cool would be to somehow provide enough adjustment parameters to be able to use any of them, since all of them are getting harder and harder to find.
My intention is to draw up some graphs of CV voltage vs gain curves for the VSR2N, J111, 2N5458, 2N5457, 2N5459, 2n3819, and whatever else I have handy and contrast that against distortion performance to see which is the right one for me.
I'll post all the results.
I'm curious mostly about the 2N5457/2N3819/VCR2N since they are the ones I see the most in schematics. But I've had excellent results already from the 2N5458, so it's probably also true that with the right design you might be able to equally use any of them to good effect.
What I think would be REALLY cool would be to somehow provide enough adjustment parameters to be able to use any of them, since all of them are getting harder and harder to find.
etheory
Well-known member
OK, moving along towards finishing the e274 I have come across some interesting findings.
After reading up on (mostly inspired by the JLH Class-A amp) and doing simulations of some interesting class-A topologies I came across something very very interesting....
Firstly, I had the idea of the biasing for the original U274 completely wrong....
The purpose of the 20K trimpot is actually to remove the DC offset at the output of the two output transistors running in class-A, NOT to produce symmetrical output clipping as I had originally guessed.
There are a couple of interesting outcomes of calibrating this way:
1.) The output transistors perfectly share the load, so each one heats up the same.
2.) The output of the class-A compressor amplifier before my balancing output stage could be DC coupled, no need for a cap when the signal can be calibrated to zero DC offset.
3.) The distortion starts to fall drastically.
I also went over the simulation with a fine-toothed comb and think I can further drop the distortion even more by changing the 30K resistor labelled as R15 in the original schematic to about 62K. This increases the headroom of the class-A amp from +12V to -8V to well over +-12V, and should drop distortion a LOT.
More soon, but I'm getting closer and closer to finishing this now ;-)
After reading up on (mostly inspired by the JLH Class-A amp) and doing simulations of some interesting class-A topologies I came across something very very interesting....
Firstly, I had the idea of the biasing for the original U274 completely wrong....
The purpose of the 20K trimpot is actually to remove the DC offset at the output of the two output transistors running in class-A, NOT to produce symmetrical output clipping as I had originally guessed.
There are a couple of interesting outcomes of calibrating this way:
1.) The output transistors perfectly share the load, so each one heats up the same.
2.) The output of the class-A compressor amplifier before my balancing output stage could be DC coupled, no need for a cap when the signal can be calibrated to zero DC offset.
3.) The distortion starts to fall drastically.
I also went over the simulation with a fine-toothed comb and think I can further drop the distortion even more by changing the 30K resistor labelled as R15 in the original schematic to about 62K. This increases the headroom of the class-A amp from +12V to -8V to well over +-12V, and should drop distortion a LOT.
More soon, but I'm getting closer and closer to finishing this now ;-)
Similar threads
