I was playing around with FET compression circuits recently, after I read this article - VCAs (section 7). The FET circuit does indeed work but I was only using a fairly general purpose 2N3819.
Does anyone know any resources I could use to further my knowledge specifically in FET compressor design? i am interested particularly in finding how to select FETs for great performance (even though I realise they are now hard to find) and how I can modify the design. I noticed in particular these designs are quite noisy and the switching can be a bit rough, it would be good to know how great FET compressors get around this
any master FET selectors out there? 🙂
Does anyone know any resources I could use to further my knowledge specifically in FET compressor design? i am interested particularly in finding how to select FETs for great performance (even though I realise they are now hard to find) and how I can modify the design. I noticed in particular these designs are quite noisy and the switching can be a bit rough, it would be good to know how great FET compressors get around this
any master FET selectors out there? 🙂
Check out GroupDIY - Index I remember there were some custom designs / analyses of old designs back several years ago. Should be wealth of info there and relatively more people involved in the technology than on this forum. Have fun!🙂
You could take a look at Urei/UA 1176 info. There is plenty available, including schematics and descriptions of circuit operation. The 1176 was a recording studio standard and is highly regarded today with clones and DAW plugins.
FETs are not hard to find and there are low noise ones available. Look for JFET in electronic suppliers catalogues (Mouser/Digikey). JFETs are depletion devices and there are new depletion types which don't always get listed with JFETs.
The key to understanding FET compressors is rooted in the FET devices themselves. FETs are more variable than BJT and are chosen according to the circuit operating conditions. The production spread bias of a given JFET can vary by 500% or more, unlike a BJT which has much tighter bias variance. Therefore you may be faced with device selection or smart circuit to accommodate. Some JFETs are available in coded bias bands, but the spread is still significant. Between the different models, bias can vary from 0.3V to 10V. So circuit design has to be evaluated with a given device in mind. With BJT you can design a circuit and select devices to suit after the fact. Apart from noise, you can't really say one FET is better than another.
FETs are not hard to find and there are low noise ones available. Look for JFET in electronic suppliers catalogues (Mouser/Digikey). JFETs are depletion devices and there are new depletion types which don't always get listed with JFETs.
The key to understanding FET compressors is rooted in the FET devices themselves. FETs are more variable than BJT and are chosen according to the circuit operating conditions. The production spread bias of a given JFET can vary by 500% or more, unlike a BJT which has much tighter bias variance. Therefore you may be faced with device selection or smart circuit to accommodate. Some JFETs are available in coded bias bands, but the spread is still significant. Between the different models, bias can vary from 0.3V to 10V. So circuit design has to be evaluated with a given device in mind. With BJT you can design a circuit and select devices to suit after the fact. Apart from noise, you can't really say one FET is better than another.
thanks all.
if i looke for low noise FETs on Mouser, I get surface mount ones which advertise low noise but don't tend to have any measurements for noise in the datasheet (i.e. MMBF5103) and then ones which are stated for use as audio FETs which cost a lot more, and generally do have noise ratings (i.e. 2N4867). what makes the latter more suited to audio, and what is a good rating for noise as far as FET compressors are concerned?
if i looke for low noise FETs on Mouser, I get surface mount ones which advertise low noise but don't tend to have any measurements for noise in the datasheet (i.e. MMBF5103) and then ones which are stated for use as audio FETs which cost a lot more, and generally do have noise ratings (i.e. 2N4867). what makes the latter more suited to audio, and what is a good rating for noise as far as FET compressors are concerned?
Low noise amplifier is NOT the same as low noise attenuator.
In gain-reduction the key concept is signal/noise ratio. A low-hiss JFET is usually a low Vto device which will have a low overload point.
The ancient 2N3819 is not a bad part here. (Also now hard to get.)
In gain-reduction the key concept is signal/noise ratio. A low-hiss JFET is usually a low Vto device which will have a low overload point.
The ancient 2N3819 is not a bad part here. (Also now hard to get.)
I wouldn't pay attention to the audio specific nature of what is available. The 'audio' types are with reference to mic front ends. In a typical compressor design, after the mic preamp, noise of the FET itself will not be much of an issue.Through-hole parts are disappearing, so magnifying glass and small tipped iron needed for DIY. PCB SMD to DIL boards available on ebay can be used.
2SK3557 15V has been measured at around 1.2nV/rtHz, 1dB in datasheet.
2SK879 50V noise 0.5dB. Anything that is around 2dB or less in the datasheets will be fine typically. Circuit topology is the deciding factor. Note that these FETs have very different transconductance which will impact design.
FET behaviour converts to a resistance. Noise contribution is calculated from that. The noise of the FET may be inconsequential.
2SK3557 15V has been measured at around 1.2nV/rtHz, 1dB in datasheet.
2SK879 50V noise 0.5dB. Anything that is around 2dB or less in the datasheets will be fine typically. Circuit topology is the deciding factor. Note that these FETs have very different transconductance which will impact design.
FET behaviour converts to a resistance. Noise contribution is calculated from that. The noise of the FET may be inconsequential.
Very true. For the moment, the J111, J112, and J113 N-channel JFETs are all still available, in through-hole TO-92 packaging.
They are labelled as switching FETs, but I've been using them as general-purpose audio FETs for guitar electronics, and they work just fine in those applications.
I suspected that, particularly with low on-resistance JFETs.
-Gnobuddy
> FET behaviour converts to a resistance.
In the usual FET limiter, the hiss is ALL about the series resistor. Typically 20k. So 2uV.
The maximum signal level across the JFET for tolerable THD may be 20mV or less.
So already 80dB S/N.
The OUTput hiss will drop as signal is attenuated, but that is no help because signal is already huge.
Either live with 20th century limits or get a new calendar and do it 21st-century ways.
In the usual FET limiter, the hiss is ALL about the series resistor. Typically 20k. So 2uV.
The maximum signal level across the JFET for tolerable THD may be 20mV or less.
So already 80dB S/N.
The OUTput hiss will drop as signal is attenuated, but that is no help because signal is already huge.
Either live with 20th century limits or get a new calendar and do it 21st-century ways.
Out of curiosity, what is the 21st century way to implement compression? Change the multiplication factor in the code running on a DSP chip?
I have never been happy with any of my long-ago attempts at making a JFET-based guitar compressor, though one of them (entirely by accident) made a nice subtle harmonic enhancer, warming up the guitar tone just a wee bit.
-Gnobuddy
😀😀😀😀 It's also the way to get the singer to sing in tune. 😀😀😀😀 Okay, so "I believe" (still) that Cher should sing in tune
Apply enough heat and pressure to coprolites, and they turn into diamonds. Back in late 2012, Ella Henderson took that awful turd of Cher's, and actually managed to polish it into a gem: YouTube
I was serious about the new lower noise approach to compression - is it lower-noise to attenuate digitally in the DSP code? At first glance I would have thought the noise would resurface at the D/A converter - but at least there's no thermal noise from a voltage-controlled resistive element.
-Gnobuddy
It's certainly cheaper!
Your fundamental problem (headroom/clipping artifacts vs noise ) remains the same. Which is really an ADC problem.
In general (if you're not messing with resampling and avoid the gross errors of clipping or aliasing) the catch in discrete number systems is in the rounding errors and almost-zero maths. You can easily throw away a bit every time you add two numbers together as you have to shift-right to do it. (and multiplication is often implemented as a series of additions).
Likewise, if you subtract two similar numbers, you end up with (mostly) noise*.
But for a straight-forward compressor, this should not be a problem.
*For those playing at home, try plotting a tractix horn in Excel. First time I tried it, I managed to get something that looked more like Nessie rather than a trumpet.