etheory - a bunch of projects in progress

[etheory]

To get myself back on track I put together my ESP Project 66 - inspired mic pre, and, unlike the comp, it worked perfectly from first powerup.
There are a couple of differences with this one - the major being fully balanced output and phantom power on the input.

OK, now back to the comp. Will post schematics in a sec - but I'm going to first principles this sucker from the breadboard one section at a time to see where I've messed it up.

 

[etheory]

OK, here is a schematic of the U274 with the compressor disabled.
Since it's not relevant to this discussion.
What's left is the amplifier sections.

This schematic provides context as to the meaning of my previous post about Ts1 etc. etc. without needing to download the schematic on Kubarth's site.

 

[Kingston]

I see. I thought you had something like the GSSL input going on with the "one side of an NE5532 to replace the input transformer" but now there's a transformer instead. With a strange bias, too.

Seems like there's a lot more going on with those designs than you are posting here and we get to see something akin to torn off pages from your private notebook.

 

[etheory]

Seems like there's a lot more going on with those designs than you are posting here and we get to see something akin to torn off pages from your private notebook.

I know what you are gunning at, and it's not like that. More like my "private notebook" is full of crappy designs that aren't working in real-life, so not worth sharing just yet.

Yep, this is the original, and, due to the absolute mess I got the circuit in through my "improvements" which were more like NE5532-based oscillators, I decided not to post until I actually fix the mess I've made Especially at the risk of someone else building a horrible design. It seems from my current analysis that what I've done is accidentally produce positive feedback around the input stage, completely accidentally, which of course will make it squeal. So time to correct that....

I jumped the gun since the simulation was so perfect. In fact, despite prior warnings, this is, even after quite a few successful builds, the first ever case for me where the simulator doesn't even come close to reality. So I have to finally concede I don't know enough, so I've gone back to a part-by-part breadboard version where I can tweak it.
I even bought the second edition of "The Art of Electronics" to help me, since I know that whilst this should be working, I'm missing some basic details I should know (but don't).

I'm actually heading back more towards the circuit I just posted and attempting to replace the transformer input in a different way.

I will post something when I start heading towards something that actually works. Thanks.

 

[etheory]

Quickly back to the preamp I just built.... (as an aside for a project that actually works).
From the notebook. Here is the built and tested (except for phantom power) transformer-less pre-amplifier design I was messing with, that started with the excellent ESP Project 66 preamp.
Nothing particularly special or even original, but I learnt a lot studying and building this thing:

(edit, just noticed a couple of mistakes in this schematic - will edit some corrections in tonight)....

 

[etheory]

OK back to the diode limiter/compressor.

I've been rebuilding the e274 on a breadboard and slowly and thoroughly analysing the results.

Interestingly the front end section that was oscillating on the PCB version, when built in isolation (photo below), is now perfectly fine.
It's a unity gain buffer like this and seems to work perfectly as the simulator says.

So I think my accidental addition of positive feedback due to the opamp circuit I added was the issue.

The reason I feel that the simulator didn't pick it up is the solver it uses being a numerical solver, and hence tending to smooth out results.
Therefore it's possible that a higher order solver would also oscillate. At least that's my theory....
I'll post a schematic of the incorrect circuit for reference, but it looks like I should be able to fix it now with what I've learned.

 

[etheory]

It looks like perseverance has paid off.
After multiple redesigns, I finally have the e274 compressor/limiter audio circuit operating exactly as it does on the simulator, even though I had to go back to an entire implementation on breadboard to do it. No more oscillating!
I was not going to let this one beat me! 8)
The crazy original-style biasing is working, all voltages are matching what they should be matching from the original schematic, and I'm getting some exceptionally clean amplification happening:

Next step is to snip up the PCB that wasn't working to match this new design, which shouldn't be too hard, and then to test the compression element.
Seems like it should be good however.
The output signal is very clean, and by the looks of the analog scope I've been investigating it with has a nice symmetrical clipping at about +-10V output and an exceptionally high slew rate with a really clean square wave output. Looking forward to this!
Here is the fixed schematic, which is part of the whole design, with a modified front end just for the testing phase. I'll complete the design next:

 

[Kingston]

You could have a line level unity gain balanced input for the same price. See the original GSSL schematic, or Igor sontec (sans the servo). I've used this all over the place, works with a DOA as well. Common mode rejection isn't quite transformer-grade, but it's still in range of 70dB and otherwise very high performance.

 

[etheory]

Totally, thanks for the tip.
That's what I'm basically shooting for.
For the breadboard test I just kept it simple, and that's what that schematic is for.
I'm probably going to go for the balanced input and then an impedance balanced output, and build one up to see how it performs.
cheers,
e

 

[etheory]

Ladies and Gentlemen....
WE HAVE COMPRESSION!

Will hook it up to the computer for some audio soon.

But it's looking very good on the scope.
Exceptionally clean signal, amazingly. Probably a lot cleaner than I was expecting.

There is still the odd occasion where it breaks into oscillation, but I think I understand why now. I think the higher hfe transistors that I am using are a little bit too jumpy, so I might try the BC109B and see if they are less prone to falling into oscillation than the BC109C. (edit) - The compressor is now unconditionally stable with a very wide range of hfe transistor. Should work perfectly with BC109C, BC109B, BC550C and various others. Will test each one for comparison.

More soon....

 

[etheory]

Here is a sound sample of 3 different drum loops running through the e274 Limiter with approx a 1ms attack and about a 5:1 compression ratio.
This was recorded with 4 unmatched 1N4148 diodes in the voltage controlled attenuator element, so it's pretty clear you can get excellent results without even bothering to match the diodes, or even using the fancy UC3611 matched diodes arrays that I've specified:

http://www.evolutionarytheory.com/wp-content/uploads/2013/04/e274_Limiter_Recording_v01_01.mp3

I'm working on a few little bits and pieces to make it sound a little better, but I'm pretty happy with how it's progressing.
The frequency response of the original has been pushed out to around 10Hz to 25kHz, and I've cleaned up some of the signal path a little bit more.
The output transistors now run much cooler than the originals, and I've completely removed the feedback issues from before so there is no chance for sporadic oscillation any more.

I've just added a high pass filter option (very simple RC filter) to the side-chain and a control for compression amount (basically a ratio control).

The recording was done with the following bread-board version of the circuit, and considering it has around -75dB SNR on the breadboard, it should be pretty good once PCB'ed and boxed:

More soon, but next I have to update the PCB board design, build another one, and see whether it all works on the board. Then I'd get boards and kits made up for anyone interested. Shouldn't be more than a month away at the latest.

 

[Dylan W]

Etheory, those clips sound awesome. I was going to suggest adding a sidechain HPF... and then saw that you're already on it. These will kill on drum rooms/loops.

Dylan

 

[etheory]

Etheory, those clips sound awesome. I was going to suggest adding a sidechain HPF... and then saw that you're already on it. These will kill on drum rooms/loops.

Dylan

Thanks Dylan! Yeah I'm pretty excited!

I've tried 1N4148's, 5.1V Zeners and the UC3611N, and so far, bizarrely, the 1N4148 unmatched set sounds the best!

I'm messing around with the crazy idea at the moment of making a switch to choose between these three sets of diodes - should make for a large palette of smashing sounds!

Also I've tried a few different time constants, will probably still go for the fastest possible as a preference, but having 10uF, 22uF, 33uF, 47uF and 100uF time constants looks like it will be pretty easy.

I've neglected worrying about metering. This thing has too much personality and variance for accurate metering anyway!

 

[Kingston]

Great clean sound, good smashing! Reminiscent of the Alesis 3630, but somehow cleaner.

What do you mean 1N4148 unmatched? I haven't heard a convincing reason to match these bulk diodes in rectifying stages, whether in SSL variants or mastering vari-mu compressors.

If you have time, I'm very interested in examples.

 

[etheory]

Great clean sound, good smashing! Reminiscent of the Alesis 3630, but somehow cleaner.

Thanks Kingston! Yeah I'm pleasantly surprised how well this little circuit is performing. Especially considering how simple it is.

What do you mean 1N4148 unmatched?

The 4 diodes that comprise the voltage controlled attenuator were 1N4148's I just picked up out of my stack of diodes without bothering to measure or match them. I thought this would result in a lot of distortion, but it didn't. Either I got lucky and picked up 4 well matched diodes, or, it just doesn't matter and the circuit doesn't care.

I haven't heard a convincing reason to match these bulk diodes in rectifying stages, whether in SSL variants or mastering vari-mu compressors.

For rectifiers I agree 100%, but for the U274 voltage controlled attenuator, which is kind of similar to a diode bridge, I thought things would be different.

If you have time, I'm very interested in examples.

Cool. No problem. I want to record the same loop compressed using different diodes - 4 x 1N4148's, 1N34A, 1N4002, UC3611N and 5.1V Zener so people can hear the differences between their attenuation action. After more messing around I've found I can get each one to sound pretty similar, but there are some interesting differences - like the minimum achievable attack in this circuit (which changes a lot). Now that I have the HPF in the side-chain working, the sound is MUCH clearer and more controlled (less muddy and less quenched by the kick), and the UC3611N is starting to shine due to the schkotty diodes having a very low turn-on voltage (smoother onset of compression). The inherent distortion in the bass region is also doing some wonderful things to bass material, and the extended frequency response I've given the circuit compared to the original is allowing bone-crunchingly low frequency signals to pass through with a little added warmth and quite a few interesting harmonics.

 

[Triode-Joe]

Hello All,

I have a couple Siemens u274s that I use to give "character" to drums, electric piano and Hammond organ.

It looks like your main work here is on the diode bridge and associated caps; on the U274 these parts are on a small daughterboard. I have always considered the U274 to be a one trick pony, but as I read through this thread, I wonder if your work could be applied backwards to the U274? The idea of a simple method of adding variable comp ratio, side chain filter and simple attack/release times makes me wonder if a new daughter board might make the original modules far more useful. I would appreciate any thoughts.

Joe

 

[etheory]

I have a couple Siemens u274s that I use to give "character" to drums, electric piano and Hammond organ.

Awesome. Interestingly I actually have two originals I got off of ebay. But I still haven't fired them up yet. I actually bought them as reference for this project. I definitely want to fire them up at some point soon. But I almost feel this project now has a bit of a life of it's own.

It looks like your main work here is on the diode bridge and associated caps

Yes and no. There was an equal amount of work on removing the input and output transformers for ease of construction/parts availability and size/weight.
I also modified the audio circuit for different and higher power rails for more headroom, much lower power consumption and the option of different modern equivalent transistors such as BC550 + BD139 etc.

I have always considered the U274 to be a one trick pony, but as I read through this thread, I wonder if your work could be applied backwards to the U274?

It might be a tiny bit messy but absolutely. That should definitely be possible.

The idea of a simple method of adding variable comp ratio, side chain filter and simple attack/release times makes me wonder if a new daughter board might make the original modules far more useful. I would appreciate any thoughts.

I'd be happy to provide some mod ideas and how to do them. I'll make sure to post that here + some proper diagrams over the new few weeks. But that should be no issue, aside from snipping wires in your compressor a little bit.

Just so you know, here are a few pointers you could use for the original:

HPF: The filter is a high pass filter based around running a resistor between 2K and 10K from the point between the diode bridge and the 0.22uF side-chain feedback capacitor to ground. I'll draw you a diagram when I get home, but it should be obvious from the schematic of the original (on Kubarth's site) where that point is. That turns the side-chain feedback path into a simple RC filter. It actually works a lot better than you might expect ;-) 10K is around 72Hz and 2K around 361Hz, with other values filling the space between. 10K there makes a huge difference to the clarity of compression from the original.

TIMING: I haven't been clever enough yet to separate attack and release, but simply switching out the two 50uF caps for 10uF, 22uF, 33uF, 47uF and 100uF should provide a range of different useful timing constants. Careful with oscillation if the cap is too low however. This depends a lot on the diodes you have.

RATIO: Not sure if ratio or "compression amount" is a better name for this, but a resistor between the Ts1/Ts2 connection and the 2.2uF capacitor will reduce the amount that the diode bridge can affect gain. Hence driving the diode bridge hard will have less effect the higher this resistor is. Values between 0R and 100K give you approx 1:Max to 1:1 compression. The lower the resistance here, the higher the "ratio".

That's all I have so far. But they do all work. You will definitely have to experiment however.

(edit: this edited image might make all the text clearer - If you want me to take this down Kubarth, just lemme know, sorry if it's cheeky to crop an image of yours directly, but I thought this would make things infinitely clearer)

On another tip.
If you want to improve the high frequency response of the original, swap C1 from 5uF to 22uF and drop the 220pF C6 entirely. This gives a tad more high end when under compression.
That, along with the HPF resistor brightens up things rather nicely and still retains a lot of the original compression action.

Also remember a feedback compressor is really just a well-adjusted oscillator. So if any modifications create a high frequency oscillation (even when no audio is being processed), it's because your feedback loop is too intense from the output transformer, and through C5. One remedy for this is a resistor between 100R and 6.2K sitting in series with the side-chain feedback path and before C5. This limits the signal passing through the feedback loop, and, hence, reduces, and when high enough eliminates, the chances of oscillation.

 

[etheory]

BTW I saw a couple of posts on other sites about people asking about resistor R14 (a 20K variable resistor on the original schematic).
This sets the bias, and, indirectly, the output symmetry of the amp.
Setting this to around 11-12K is a pretty much set and forget process.
If you don't have a scope try 11K - which works pretty well for me.
If you do have a scope, run in a 100Hz sine wave until the amp clips (frequency is arbitrary but I did it successfully with 100Hz), and then adjust for clipping symmetry.

 

[tv]

Ladies and Gentlemen....
WE HAVE COMPRESSION!

P.s.: Instead of 2N2219, you may also want to test-drive 2N1711 (or 2N1613 with lower Hfe) and Euro, BC140

 

[etheory]

P.s.: Instead of 2N2219, you may also want to test-drive 2N1711 (or 2N1613 with lower Hfe) and Euro, BC140

Nice tip, thanks! Though the 2N2219A (I chose the low-A hfe-grade that seems quite similar to the ones you've specified, at least from the data I can find) is working pretty well in that spot.
But they look like excellent alternatives for sure. The 2N1711 looks like it has the potential for better power handling (or cooler operation, depending on which way you look at it).
Also the noise figure maximum for the 2N2219A is 4dB, the 2N1711 is 3.5dB min to 8dB max and I can't find a noise figure for the BC140.

All appear to have rather similar specs for power handling, hfe and potentially noise.

All very interesting for sure, thanks! I'm sure all would be excellent options for the output transistors. I'll have to give them a go. They would also be worthy candidates for my next project - the W295 EQ, which I'll get to in a month or two after this one is done.