| Discussing the Urei 1176 (1): Compression Ratio |
 |
Simulation results (PSpice) for Urei 1176, different settings:
Take a look at the bold blue and bold red courve. The selected compression ratio is very different, but with an input level to provide heavy gain reduction, the output level is quite similar in both cases. That's because the lower compression ratio (4:1) starts at much lower input levels, and because even with the highest compression ratio (20:1) the maximum gain reduction of approx. 30dB is not exceeded. It's these 30dB's of attenuation the compressor can "ride", and it will do this along different courves for different button settings. When the 30dB are "used up", there is no further gain reduction, so even with a ratio of 20:1 the output level will eventually increase again after that point. The good thing is that even then, the circuit has enough headroom that there will normally be no clipping. |
| Discussing the Urei 1176 (2): Time Constants |
| Time constants for attack and release times come from charging and
discharging a capacitor over (variable) resistors. This follows an exponential
law. An alternative method is using current sources instead of resistors,
resulting in a linear courve. Signal level (in dB) is a logarithmic value,
and therefore it's easy to build compressors that have attack times and
release times that are independent of the absolute input level: Use a linear
or exponential VCA, linear or logarithmic envelope detector, feedforward
or feedback configuration - no problem with modern components. (If you
choose a feedback configuration, you must use an exponential VCA to get
"constant time constants". In the feedforward configuration, you can also
use a linear VCA.)
Now the 1176 has neither a linear nor an exponential VCA (but a FET with its own kind control nonlinearity), and it uses a feedback configuration. This results in time constants that change a lot depending on the input level, as varying amount of gain reduction means a different point on the FET's nonlinear control input, which changes the closed loop gain of the gain reduction feedback, and therefore changes the time constant. Another important factor is the biasing of the level detectors. The full wave rectifiers will only charge the capacitor over the "attack" variable resistor for a short span of time, when the compressor output signal is larger than the negative bias in the level detector. Now this negative bias is changed with the compression ratio buttons (alongside with a change of feedback gain) to get the behaviour discussed in the section above. As a side effect, the time constants will be affected as well. The time constants you choose on the front panel are just coarse settings (and of rather limited range). The character of the 1176 comes for a limited choice of time constants from the user, and - starting from that - changes of time constants along the "gain ride" that are inherent to the circuit. |
| Discussing the Urei 1176 (3): Abuse - the interesting part |
| So everybody claims to get the best results from the 1176 with multiple
ratio buttons pressed at the same time? Right they are (who would question
the experience of all these veteran audio engineers anyway?), and I gladly
took the advice of using radio buttons, and not a simple rotary switch,
for my clone.
What happens when multiple buttons are pressed? First, let's see what happens not:
This would give us some more nuances for these parameters, some in-between settings - hardly anything more. But there is definitely more, as anybody who has worked with a 1176 will attest. The funny thing is that it's all the result of a little design flaw. The bias voltage for the rectifier and the bias voltage for the FET are derived from the same resistor divider circuit. As long as one compression ratio button is pressed, there is just a different voltage from different taps of a divider string selected. Just like a potentiometer with different settings, the total value from one end to the other remains constant. Not so, when multiple buttons are pressed: Now part of the divider string is shorted. At the tap there is a sort of middle value between both settings, but the resistance of the whole divider string (the "potentiometer value") is changed! As we're speaking of all passive divider circuits here, there is some side effect on the output voltage of the trimpot that is used to adjust the FET bias voltage. The FET that performs the Gain Reduction in the compressor has a limited range: At 0V (or slightly above) you get maximum attenuation (approx. 30dB). At the FET threshold voltage (approx. -2.5V) there is no gain reduction (0dB). That's the useful range. A lower control voltage, like -4V, will have the same effect as -2.5V: no Gain reduction. With the trimpot the FET bias is set such that there is "just no gain reduction" without an input signal. That is, the FET is waiting at the border of its useful range for any audio signal to reach the compression threshold (set in the full wave rectifier), and will then immediately start its gain reduction duty, as the circuit designers had intended it. Given that, it's easy to see the effect of a "wrong" adjustment, as it comes from pressing more than one button. The FET starts way outside of its usefull area, with an strong audio signal approaching the CV has to race up (in open loop configuration!), but nothing will happen before the -2.5V threshold is crossed. Then, suddenly, as the threshold is crossed, gain reduction begins, the feedback loop is closed, and everything is changed. Do not ask me to calculate the effect this has on the time constants. It's enough to state that these are very different in open loop and closed loop configuration, so you can imagine what bouncing between these conditions means for the transisents. And by the way, a positive (red) meter reading just means that the FET is below its threshold voltage. It is adjusted to read "0dB" when the FET is at the edge of its useful range, after all. Is this the whole truth? I don't know. It's what I think I figured out from the circuit. Keep in mind that I have not worked on an original, but un my own clone. Comments, corrections, all kind of useful feedback is welcome. |
| Discussing the Urei 1176 (4): Distortion |
| Don't expect me to go into the transformers here. It's known that transformers
cause some subtle and often pleasant distortion. I have no doubt that this
plays some role in the 1176 as well, but I surely have no expertise to
tell how big or how small of a role that is. I cannot tell a 1176 with
input transformer from a 1176 without input transformer, simply because
I have not played with either. Yes, I never played with an original, and
all of the above is from the schematics, from simulations, and from my
attempt to build a clone. So take it all with a grain of salt.
But back to the distortion: There is one source of distortion that, while obvious, is very interesting for its sonic quality. The FET - the main gain control element - is a source of distortion. Distortion increases with higher audio voltage. A circuit trick (feeding -6dB of the audio voltage across the FET back to its gate) is used to get the highest possible audio voltage across the FET for some tolerable low distortion figure which I will call "nominal distortion" or "THDnom" for now. It's not zero, but it's so low that it won't be unpleasant. With small input levels, you will even get less distortion than THDnom. With increasing input level, at some point THDnom is reached. But that's when the compressor starts to work anyway: The FET starts to conduct, reducing the voltage across the FET, and thus limiting the FET distortion as well as the output level. So THDnom would never be exceeded. Pretty clever, isn't it ? But this is only true for the steady state. Compressors are mostly used with longer attack times in order to preserve some of the "kick" of a transient. That means, there is some delay before the compressor has its full gain reduction. That is, before the FET conducts enough to limit the voltage across the FET. So during the attack transient or "kick" there is way more distortion than THDnom! We could now conclude that the 1176 adds distortion during the attack phase, compared with a modern all-linear compressor. An explanation that sounds good to describe the character of a vintage piece of equipment, no? But wait. There's a different point of view. What if I say the 1176 might produce less distortion than modern devices that are built around ultra low THD VCAs? It's just a different point of view. It starts to make sense when you look at the headroom of your whole signal path. Take the modern "linear" compressor and set it to have this "kick" from a slow attack time. Where will this exceeding signal level go? In most cases, it will hit the ceiling somewhere. If not inside the compressor, then in the mixing console, in other signal processing devices, you name it. Unless you're setting all the following input levels so low that even these peaks will pass (which sort of contradicts the use of a compressor in most cases). There may be exceptions (like playing into tube amps), but normally the rather soft distortion from the FET will be way more pleasant than other types of distortion. In a modern "linear" compressor, the remaining "kick" might be running into the supply voltage rails of an opamp or a BBD chorus, to name the worst case. I really don't have the experience of playing with many different compressors. In practice, I'm basically comparing two extremes: A cheap Behringer compressor that uses excellent VCAs but sounds like crap, and the clone of an old FET compressor that uses a handfull of transistors and sounds so smooth. I'm not entirely sure about the role this kind of FET distortion plays, but to me it makes some sense. I find myself being much more strict with attack times on the Behringer, in order to avoid transients to distort elsewhere. Which means more permanent distortion from modulation with short time constants. |
| The Clone, in Stereo |
| Schematics, page 1
Schematics, page 2 Schematics, Page 3 Schematics, page 4 |
| Links - Originals and Clones |
| Univeral Audio |
| Purple Audio |
| Urei Schemos at Waltzing Bear |