diod compressor..will this work?..

[Johan]

just trying to see if I got things right...but like title says..will it work?..
http://www.aronnelson.com/gallery/main.php?g2_view=core.DownloadItem&g2_itemId=40844&g2_serialNumber=1

obviously it will need opamps that don't oscillate with gain below 1, so probably something like a 741 or 4558 and a 5534 for signal path..and the values for resistors and caps will have to be figured out..the meter circuit, if it works as I hope it does, should be sensing the current going through the diods...but will it work?...there will be a dc offset in the opamp in the signalpath, but I know from the stompbox world, that is no problem as long as it is decoupled before the next stage..
at the moment I am not even sure I'll try and build it, just thought I'd get some feedback on it...
j

 

[JohnRoberts]

It depends on how you define "work".  Diode clamps will limit voltage but you get high distortion.  In some very low voltage swing circuits you can play games with varying the diode impedance by biasing it with DC current but not of much use for audio compression IMO.

JR

 

[abbey road d enfer]

I suggest you take a look at the schematics for neve 33609. The diode attenuator is implemented in the nicest way. Symmetrical in order to cancel out 2nd order distortion, transformer-based for DC control voltage elimination.

 

[Johan]

thank's JR
..is that why the Neve 33609 (the inspiration behind me starting to mess with this..) is ballanced in the GR stage too..so the distortion is cancelled?
..in that one the signal is dropped down -31d too..I guess to reduce that distortion?..
j

edit: thank's for that...you answered my q while i was typing.. back to drawing board then..  ;D

 

[Johan]

ok..second try...what if the main signal path looked like this?..and the signal was attenuated to ~-30db just before the diods?

j

 

[abbey road d enfer]

You need a resistor from +in of 1st opamp to ground, for two reasons:
a ) proper balance of diff amp
b ) referencing DC operating point
Apart from that, it works. You need a real low-noise opamp there.
You can't use too low value resistors in this stage, because the diode attenuator needs to see a relatively high value. As a consequence there will be noise due to noise-current of the input transistors, unless you use a VLN FET opamp. You also need to fit a balance trim in order to properly reject DC CV.
Tha ttenuation you need to operate at ca. -30dBu can be done either by a pad before the XFMR, a pad after the xfmr, or just raising the value of the input resistors. It depends on max. level specs of xfmr.

 

[Johan]

thank's Abbey..
but doesnt the opamp get its bias from the 0.5 V point between the diods( constantly set at one diode drop)..I know it will be offset some in regard to ground, but I also know that works from my experience building stompboxes. I thought that would be enought for a DC reference?....and it keeps the tiniest signals away from the crossover-point of the opamp output...
..I cant argue about the balance of the diff amp thou...
the pad would be the tree resistors closest to the transformer, working double together with the diodes too..
j

 

[abbey road d enfer]

I agree that you would have a stable operating point, but probably not exactly the one you want, and temperature drift would not be compensated (not a big problem anyway, but ther's already too many erratic factors in any design).

 

[Johan]

how about now then?..

the opamp is decoupled from the dc offset and a ground reference resistor is added both for dc reference and diff amp happiness
j

 

[abbey road d enfer]

Yes. Caps are not indispensable though. But I reckon it's always good practice to restrain VLF contents.

 

[PRR]

If signal voltage is small, it's not a "diode clamp".

If done right, it works fine. Within limits.

Here's a problem with your first one. Say that a diode needs 600mV to flow 1mA and hit 30 ohm impedance. Then your "off" bias might be 0.01mA for 3,000 ohm resistance. From 1mA to 0.01mA is a 120mV change of diode voltage. Your max signal level is 28mV, and probably much less. A rapid off/on change (limiting a sudden transient) causes 120mV of bias shift on 30mV signal, an awful thump.

Also, your upper diode seems to do nothing, because it is fed from a high impedance. To make it useful, you buffer the CV to give a low audio impedance to the diode.

CBS used a diode bridge between a CT winding and a diff-amp. This cancels most of the thump and gives low-impedance paths to shunt audio. I ferget if this was Volumax or Audimax, but you'll find it. Your post #4 has the general idea.

> signal is dropped down -31d too..I guess to reduce that distortion?

Shockley's Law. The current in a diode doubles every 28mV. Conversely, the dynamic resistance halves every 28mV. In a single-ended diode shunt, the peak signal must be MUCH less than 28mV or distortion is gross. Bridges give a first-order cancellation, and also two diodes in the path. They won't be horribly bent at 40mV peak, though if you have a specific THD spec you may need to go lower (OTOH, in military intercom duty, where CBS patented the idea, you can go past 50mV).

Overall I think the S/N will never be better than 60dB-70dB. This was excellent for AM and FM broadcast applications. It is still excellent for any real music. It does look "poor" next to 16-bit specs.

 

[Johan]

thank's PRR and Abbey..
..as I said in my first post, I am just trying to learn something new and might not even end up building it, but anyway..

am I correct in assuming the noise would be mostly voltage noise from the resistors. and if so, could I get around that with a different diff-amp?..something sensing current instead of voltage?..or is it just the same..perhaps the current throu the diodes is worse?...

I was thinking something like this...or am I just wasting opamps and resistors?...

j

 

[abbey road d enfer]

No, the main noise sources are the diode's junction noise (which depends on the current) and the noise voltage of the opamps, which will be amplified because you're dealing with a low operating level (-30dBu, so you have to apply 30dB gain). That's why I mentioned you should use low-noise opamps. Neve use discrete FET stages in the 33609.

 

[Johan]

No, the main noise sources are the diode's junction noise (which depends on the current) and the noise voltage of the opamps, which will be amplified because you're dealing with a low operating level (-30dBu, so you have to apply 30dB gain). That's why I mentioned you should use low-noise opamps. Neve use discrete FET stages in the 33609.

thank's
..is it fair then to asume the noise gets worse as the gain reduction gets deeper?..no G.R.= lower noise?..if so the diodes should be biased "just off" or very, very cloose to it, right?..
I was going to use a FET input opamp after you mentioned it a few replies above..perhaps an opa2604?
j

 

[abbey road d enfer]

Since you'll be dealing with medium impedance (less than 10 k) good old 5534's will give you better noise performance. With 3.5nV/sqrtHz, the 5534 has a 10dB advantage over the 2604 (10nV/sqrtHz).
The discrete FET used by Neve in the 33609 have 3.5 10nV/sqrtHz noise voltage.
Unfortunately the FET's in monolithic IC's are not as good.

 

[PRR]

Let's take an example: max limiting 30dB at 1mA diode current.

(You have to aim high because when these systems do overload, they become Diode Clamps as JR says.)

And skip the series and push-pull details because they make "only 6dB" differences.

At 1mA the diode impedance is ~~30 ohms. For 30dB limiting we need ~~1,000 ohms series resistor.

The hiss of 1K is like 0.4uV. This will appear when NOT limiting. (When limiting, the impedance falls low, and hiss drops. The actual heavy-GR hiss is complicated by diode action; but in real life, we do not care, we can't hear hiss with all that racket.)

The max level per diode is well under 30mV.

Hmmm... this actually gives 97dB between average idle hiss and peak output. Not bad.

BUT this is from 0dB to 30dB GR. So the OUTput is only a 67dB range from hiss to clip. Or if you schedule maximum 20dB limiting, 77dB. Ample for final squash on FM transmitter or <30ips tape. Looks shameful in a "96dB" digital system.

Also you would not run as hot as 1mA in modern diodes. They are near-perfect down to very small current, get very non-perfect at 100mA, and at 1mA you can't be too sure that pairs will cancel well. And 1K is a distressingly low impedance. As abbey says, a few K. 5532 can give low added hiss to 5K sources- a clean-sheet design could start with 5K, work the diodes from 0.002mA idle up to 0.1mA or 0.2mA for hard GR.

In practice, once Blackmer et al VCAs became practical, and optimized, this became "the" way to tame wide input signals. Other techniques survive for "color", cheapness, or patent fees (and I think all the Allison/Blackmer patents are now dead).

 

[Johan]

...I guess my grey cell is dimmer than I thought, becouse figuring out the values for the resistors seems to be a much more delicate balancing act than I anticipated...

if I am going to get any action from the diodes, the R2 needs to be at least 10 times bigger than the lowest diode-Z, right?..something like at least 300-500Ohm
if so, if I need  to make R1 ~5k to get the signal down to around -30db(assuming-10db input )..right?

if I make R3 too small, wouldn't R2 end up...well..almost bootstraped?..if so, will the input pad still work?..or do I need to make R3 about the same size as R1?..

when calculating the gain at the opamp, is it R4 devided by R3?..or is it really R4 devided by (R3+R1)?...then if so, doesnt R4 end up too big to support the bias current requirement of a 5532?

..it seems the deciding factor would be how low the diodes can go...any hint's?
..perhaps it would solve a few problems if I put a source follower on each leg between the pad and the opamp?

..right now I feel like Homer Simpsson trying out science..

j

EDIT: right now I'm trying 10k for R1 and R3, 1k for R2 and 220kk for R4..looks to be working on my breadboard...havn't listened yet thou..only looking on my oscilloscpoe..

 

[abbey road d enfer]

You don't need R2.
Then the attenuation is R3/R1+R3

 

[PRR]

The one-opamp diff-amp is a terrible cheat.

Use a good high-impedance 2-amp or 3-amp diff-amp.

No, you do NOT want to push diode impedance super low.

Make your R1s a few K each (well matched).

The limiting level will be in the area of 30mV. Under 20dB-30dB limiting, the input to the R1s will be 300mV-1V. If your source is larger, pad before the R1s; you probably want a pot in front.

 

[Johan]

thank's again both of you
designing a new better diff-amp is ouside of my capabilyties, and probably not necessery ...but I could adopt the design from something else...do you think the input from an URIE 1176, modified for more gain would be a good candidate?..
j