The Ultimate Spring = Reverb Recovery Amp

The Ultimate Spring Reverb Recovery Amp

Roy = Mallory

In my tech note tech note I describe a = simple and=20 very good reverb recovery amp using a monolithic op amp and only a few=20 components. However, it = does not=20 offer the ultimate in low-noise performance. Low-cost, JFET-input op = amps=20 generally have input-referenced voltage noise of from 8 to 16 nV/rt Hz = in the=20 middle of the audio spectrum, whereas discrete, inexpensive JFET = transistors=20 offer noise voltages of about 2 to 3 nV/rt Hz. In addition, some JFETS often = have noise=20 lower than what their data sheets specify. =20 Therefore, this paper describes a composite op amp that uses a = single=20 JFET for its input stage. = The basic=20 circuit topology looks like this:

Figure 1

We see that the input (via phone jack J4) is = applied to the=20 gate of a JFET, and the output of the JFET goes to the inverting input = of an op=20 amp. The output of the op = amp is=20 fed back to the source connection of the JFET. The feedback connection = combined with=20 the high gain of the op amp acts to keep the current through the JFET=20 constant. Another way to = think=20 about this is that the feedback forces the gate and source terminals of = the JFET=20 to track. This circuit is = really=20 just a non-inverting amplifier whose gain, in its pass band, is A =3D 1+ = R2/R1,=20 just like any other non-inverting op amp stage. One might wonder why I didn't = use a=20 differential pair instead of a single-ended JFET stage at the input of = the op=20 amp as is typically done. = The=20 answer is simple: in a differential pair, the noise of each JFET adds so = as to=20 increase the noise contribution of the JFETS by 3 dB. Of course, unlike a typical,=20 differential-input, op amp stage, this circuit cannot respond down to = DC. In fact, its low-frequency 3 = dB point is=20 defined by R2 and C3, and is

In the example above, F_low =3D 147 = Hz.

Let=92s look at some of the other components. R5 and R1 sets the current = through the=20 JFET, which, in this case, is about 1.25 mA. R4 is chosen to keep the drain = of the=20 JFET near the middle of the available voltage swing. R3 serves no other purpose = other than to=20 provide bias current for the op amp=92s inverting input, and to keep the = DC output=20 of the op amp close to 0 volts. = Its=20 value should be large enough not to lower the gain of the stage, and 10 = megs=20 works nicely. C4 is made = large=20 enough not to influence the gain or low-frequency rolloff of the = stage.

The JFET is chosen mostly for its low votage noise, = but, of=20 course, its break-down voltages limits, I_dss, = transconductance, and=20 other parameters must be taken into consideration. I tested several readily = available=20 JFETS, and many proved suitable. = The venerable PN4393 (or 2N4393) proved to have the lowest noise = of any=20 cheap JFET and is one of the easiest JFETS to find. The J310 is another good = choice. There are some exotic JFETS = with even=20 lower noise that can be used. = They=20 will be discussed later.

There are a number of noise sources in this = amplifier=20 stage. First, is the = Johnson noise=20 of the real part of the reverb tank=92s output impedance, which is just = its DC=20 resistance. For most of=20 Accutronics=92 tanks, this resistance is 200 ohms. The JFET=92s voltage noise is = multiplied=20 by the signal gain of the stage, which, in the example, is 101. The FET=92s current noise is = multiplied by=20 the input impedance of the reverb tank and the signal gain of the = amp. Luckily, because the bias = current of the=20 FET is so low, this noise source can be ignored. The noise of the op amp also=20 contributes, but it is divided by the gain of the JFET stage. The gain of the FET stage = depends on the=20 type of FET that is used, but for the =20 PN4393, we can conservatively call it 40.

As usual, resistors contribute their Johnson noise. = As is=20 typical, resistor noise is modeled as a noise source in series with the = resistor=20 in question. R2=92s = noise is=20 multiplied by the inverting gain, or 100 in the example. Both R4 and R5 contribute = their=20 noise. The easiest way to = calculate=20 their contribution is to note that the feedback loop acts to keep the = current in=20 the FET constant in the face of noise from R5. The output, acting through R1=20 accomplishes this. = Therefore, the=20 Johnson noise of R5 is multiplied by R1/R5, which, in the example, is = about=20 =BC. Similarly with R4, = its noise is=20 amplified by the ratio of R1/R5.

R1=92s noise is amplified by one. In the case of R3, the output = will swing=20 to provide a current equal to R3=92s voltage noise divided by R3. Therefore, the gain of R3=92s = Johnson=20 noise is R1/R3. Let=92s = add up all=20 these noise sources to see which are important, and what the total noise = of the=20 stage is. Remember that = like all=20 uncorrelated voltages, they will add in a root-sum-squared (r.s.s.) = way. That is, each noise voltage = must be=20 squared, then summed, and the sum then square rooted.

Source of noise

Output Noise (nV/ rt. = Hz)

Tank

182

JFET (2.5 nV/ rt Hz voltage noise)

253

8.7

.19

6.9

4.2

Op amp (8 nV /rt Hz voltage noise)

Total

323

We see that the only two noise sources that are = significant=20 are the resistance of the reverb tank and the voltage noise of the = JFET. R2=92s noise could become = significant, but=20 by keeping its resistance substantially lower than the tank=92s it is = not an=20 important contributor.

With the values and parts chosen, the total output = noise is=20 less than twice that contributed by the tank alone. If we wanted to use a more = exotic JFET,=20 Inter-Fet, http://www.interfet.com/ has=20 several types whose voltage noise in the audio band is no more than 1 = nV/ rt.=20 Hz. If we used one of = these FETS,=20 the total output noise would be only 1.25 times larger than the noise of = the=20 tank. That=92s awfully = close to=20 ideal.

So, are we done? =20 Not quite. Like = most=20 single-ended gain stages, the JFET stage in this circuit is susceptible = to=20 power-supply noise. In = addition, we=20 need to discuss stability and how to limit the bandwidth of the = circuit. It=92s also worth discussing = how to reduce=20 ground-loop and common-mode noise. =20 Below is a fully-developed circuit including components that deal = with=20 these issues.

Figure 2

First, let=92s discuss power-supply decoupling for = the=20 JFET. In the circuit = above, the=20 decoupling is accomplished by R8-C5 and R7-C6. These components introduce a = pole at 3.4=20 Hz. Power-supply noise = affects the=20 circuit exactly like noise in R5 and R4. =20 As we have seen, this noise is not amplified by the signal gain, = but=20 appears at the output as the ratio of R1/R5 and R1/R4 respectively. Therefore, it is not hard to = filter=20 out. One could replace R5 = with a=20 current source in order to suppress the effect of voltage noise on the = negative=20 rail, but current sources have their own noise, and the R-C filters are = very=20 simple and cheap.

Next, to frequency response and bandwidth = limiting. The circuit is perfectly = stable if C1=20 and C2 are removed. With = them out,=20 however, the bandwidth of the circuit is very broad, and because we=92re = not=20 really interested in receiving A.M. radio stations, some response = shaping is a=20 good idea. One might = think that=20 adding C1 as would be done in an ordinary non-inverting op amp stage = would be=20 sufficient. However, = adding this=20 component reduces the signal gain at high frequencies, which also raises = the=20 loop transmission at those frequencies. =20 (Loop transmission is defined as the forward open-loop gain of = the=20 circuit divided by its noise gain.) =20 C2 adds some phase lead into the circuit, and that stabilizes = it. The selection of C1 and C2 = depend highly=20 on the selected signal gain of the circuit, of course. With the components as shown, = the stage=20 has a high-frequency 3dB rolloff at 15 kHz, which seems appropriate.

Now, let=92s talk about ground loops and = common-mode=20 noise. Reverb tanks are = notoriously=20 susceptible to magnetically induced hum, but because their output signal = requires a lot of amplification, ground loops and common-mode noise can = also be=20 problems. Looking at the=20 2^nd schematic, we see that J4, the connection to the tank is = not=20 connected directly to ground. =20 Instead, a separate ground wire goes to the tanks=92 metal = shield. R6 provides a ground reference = to the=20 circuit when the tank is unplugged.

By floating J4, common-mode signals are only = amplified by=20 one, not by the signal gain of 101. =20 Therefore, they are substantially suppressed. In addition, floating J4 = breaks any=20 ground loops. Note that = if the=20 reverb tank is conductively mounted to a grounded chassis, the extra = ground=20 connection to the tank should not be needed. Otherwise, it will be = required.

Finally, I have built and installed the circuit = shown in=20 Figure 2 using the values and components as shown. It works just fine and is = noticeably=20 quieter than a conventional circuit using just an OPA134 op amp = configured as=20 shown in my tech note. = The=20 modifications used to reduce common-mode noise and ground loops were = also=20 effective, and the hum was also reduced. =20 In fact, with the reverb gain set to a typical level, the reverb = circuit=20 is quiet and adds no audible noise to the overall amplifier, which is a = claim=20 that not many spring reverb circuits can make..

Home

").append(i.responseText.replace(tb,"")).find(e):i.responseText);d&&j.e= ach(d,[i.responseText,o,i])}});return this},=0A= serialize:function(){return = c.param(this.serializeArray())},serializeArray:function(){return = this.map(function(){return = this.elements?c.makeArray(this.elements):this}).filter(function(){return = this.name&&!this.disabled&&(this.checked||ub.test(this.nodeName)||vb.test= (this.type))}).map(function(a,b){a=3Dc(this).val();return = a=3D=3Dnull?null:c.isArray(a)?c.map(a,function(d){return{name:b.name,valu= e:d}}):{name:b.name,value:a}}).get()}});c.each("ajaxStart ajaxStop = ajaxComplete ajaxError ajaxSuccess ajaxSend".split(" "),=0A= function(a,b){c.fn[b]=3Dfunction(d){return = this.bind(b,d)}});c.extend({get:function(a,b,d,f){if(c.isFunction(b)){f=3D= f||d;d=3Db;b=3Dnull}return = c.ajax({type:"GET",url:a,data:b,success:d,dataType:f})},getScript:functio= n(a,b){return c.get(a,null,b,"script")},getJSON:function(a,b,d){return = c.get(a,b,d,"json")},post:function(a,b,d,f){if(c.isFunction(b)){f=3Df||d;= d=3Db;b=3D{}}return = c.ajax({type:"POST",url:a,data:b,success:d,dataType:f})},ajaxSetup:functi= on(a){c.extend(c.ajaxSettings,a)},ajaxSettings:{url:location.href,=0A= global:true,type:"GET",contentType:"application/x-www-form-urlencoded",pr= ocessData:true,async:true,xhr:A.XMLHttpRequest&&(A.location.protocol!=3D=3D= "file:"||!A.ActiveXObject)?function(){return new = A.XMLHttpRequest}:function(){try{return new = A.ActiveXObject("Microsoft.XMLHTTP")}catch(a){}},accepts:{xml:"applicatio= n/xml, text/xml",html:"text/html",script:"text/javascript, = application/javascript",json:"application/json, = text/javascript",text:"text/plain",_default:"*/*"}},lastModified:{},etag:= {},ajax:function(a){function b(){e.success&&=0A= e.success.call(k,o,i,x);e.global&&f("ajaxSuccess",[x,e])}function = d(){e.complete&&e.complete.call(k,x,i);e.global&&f("ajaxComplete",[x,e]);= e.global&&!--c.active&&c.event.trigger("ajaxStop")}function = f(q,p){(e.context?c(e.context):c.event).trigger(q,p)}var = e=3Dc.extend(true,{},c.ajaxSettings,a),j,i,o,k=3Da&&a.context||e,n=3De.ty= pe.toUpperCase();if(e.data&&e.processData&&typeof = e.data!=3D=3D"string")e.data=3Dc.param(e.data,e.traditional);if(e.dataTyp= e=3D=3D=3D"jsonp"){if(n=3D=3D=3D"GET")N.test(e.url)||(e.url+=3D(ka.test(e= .url)?=0A= "&":"?")+(e.jsonp||"callback")+"=3D?");else = if(!e.data||!N.test(e.data))e.data=3D(e.data?e.data+"&":"")+(e.jsonp||"ca= llback")+"=3D?";e.dataType=3D"json"}if(e.dataType=3D=3D=3D"json"&&(e.data= &&N.test(e.data)||N.test(e.url))){j=3De.jsonpCallback||"jsonp"+sb++;if(e.= data)e.data=3D(e.data+"").replace(N,"=3D"+j+"$1");e.url=3De.url.replace(N= ,"=3D"+j+"$1");e.dataType=3D"script";A[j]=3DA[j]||function(q){o=3Dq;b();d= ();A[j]=3Dw;try{delete = A[j]}catch(p){}z&&z.removeChild(C)}}if(e.dataType=3D=3D=3D"script"&&e.cac= he=3D=3D=3Dnull)e.cache=3Dfalse;if(e.cache=3D=3D=3D=0A= false&&n=3D=3D=3D"GET"){var = r=3DJ(),u=3De.url.replace(wb,"$1_=3D"+r+"$2");e.url=3Du+(u=3D=3D=3De.url?= (ka.test(e.url)?"&":"?")+"_=3D"+r:"")}if(e.data&&n=3D=3D=3D"GET")e.url+=3D= (ka.test(e.url)?"&":"?")+e.data;e.global&&!c.active++&&c.event.trigger("a= jaxStart");r=3D(r=3Dxb.exec(e.url))&&(r[1]&&r[1]!=3D=3Dlocation.protocol|= |r[2]!=3D=3Dlocation.host);if(e.dataType=3D=3D=3D"script"&&n=3D=3D=3D"GET= "&&r){var = z=3Ds.getElementsByTagName("head")[0]||s.documentElement,C=3Ds.createElem= ent("script");C.src=3De.url;if(e.scriptCharset)C.charset=3De.scriptCharse= t;if(!j){var B=3D=0A= false;C.onload=3DC.onreadystatechange=3Dfunction(){if(!B&&(!this.readySta= te||this.readyState=3D=3D=3D"loaded"||this.readyState=3D=3D=3D"complete")= ){B=3Dtrue;b();d();C.onload=3DC.onreadystatechange=3Dnull;z&&C.parentNode= &&z.removeChild(C)}}}z.insertBefore(C,z.firstChild);return w}var = E=3Dfalse,x=3De.xhr();if(x){e.username?x.open(n,e.url,e.async,e.username,= e.password):x.open(n,e.url,e.async);try{if(e.data||a&&a.contentType)x.set= RequestHeader("Content-Type",e.contentType);if(e.ifModified){c.lastModifi= ed[e.url]&&x.setRequestHeader("If-Modified-Since",=0A= c.lastModified[e.url]);c.etag[e.url]&&x.setRequestHeader("If-None-Match",= c.etag[e.url])}r||x.setRequestHeader("X-Requested-With","XMLHttpRequest")= ;x.setRequestHeader("Accept",e.dataType&&e.accepts[e.dataType]?e.accepts[= e.dataType]+", = */*":e.accepts._default)}catch(ga){}if(e.beforeSend&&e.beforeSend.call(k,= x,e)=3D=3D=3Dfalse){e.global&&!--c.active&&c.event.trigger("ajaxStop");x.= abort();return false}e.global&&f("ajaxSend",[x,e]);var = g=3Dx.onreadystatechange=3Dfunction(q){if(!x||x.readyState=3D=3D=3D0||q=3D= =3D=3D"abort"){E||=0A= d();E=3Dtrue;if(x)x.onreadystatechange=3Dc.noop}else = if(!E&&x&&(x.readyState=3D=3D=3D4||q=3D=3D=3D"timeout")){E=3Dtrue;x.onrea= dystatechange=3Dc.noop;i=3Dq=3D=3D=3D"timeout"?"timeout":!c.httpSuccess(x= )?"error":e.ifModified&&c.httpNotModified(x,e.url)?"notmodified":"success= ";var = p;if(i=3D=3D=3D"success")try{o=3Dc.httpData(x,e.dataType,e)}catch(v){i=3D= "parsererror";p=3Dv}if(i=3D=3D=3D"success"||i=3D=3D=3D"notmodified")j||b(= );else = c.handleError(e,x,i,p);d();q=3D=3D=3D"timeout"&&x.abort();if(e.async)x=3D= null}};try{var h=3Dx.abort;x.abort=3Dfunction(){x&&h.call(x);=0A= g("abort")}}catch(l){}e.async&&e.timeout>0&&setTimeout(function(){x&&!E&&= g("timeout")},e.timeout);try{x.send(n=3D=3D=3D"POST"||n=3D=3D=3D"PUT"||n=3D= =3D=3D"DELETE"?e.data:null)}catch(m){c.handleError(e,x,null,m);d()}e.asyn= c||g();return = x}},handleError:function(a,b,d,f){if(a.error)a.error.call(a.context||a,b,= d,f);if(a.global)(a.context?c(a.context):c.event).trigger("ajaxError",[b,= a,f])},active:0,httpSuccess:function(a){try{return!a.status&&location.pro= tocol=3D=3D=3D"file:"||a.status>=3D200&&a.status<300||a.status=3D=3D=3D30= 4||a.status=3D=3D=3D=0A= 1223||a.status=3D=3D=3D0}catch(b){}return = false},httpNotModified:function(a,b){var = d=3Da.getResponseHeader("Last-Modified"),f=3Da.getResponseHeader("Etag");= if(d)c.lastModified[b]=3Dd;if(f)c.etag[b]=3Df;return = a.status=3D=3D=3D304||a.status=3D=3D=3D0},httpData:function(a,b,d){var = f=3Da.getResponseHeader("content-type")||"",e=3Db=3D=3D=3D"xml"||!b&&f.in= dexOf("xml")>=3D0;a=3De?a.responseXML:a.responseText;e&&a.documentElement= .nodeName=3D=3D=3D"parsererror"&&c.error("parsererror");if(d&&d.dataFilte= r)a=3Dd.dataFilter(a,b);if(typeof a=3D=3D=3D"string")if(b=3D=3D=3D=0A= "json"||!b&&f.indexOf("json")>=3D0)a=3Dc.parseJSON(a);else = if(b=3D=3D=3D"script"||!b&&f.indexOf("javascript")>=3D0)c.globalEval(a);r= eturn a},param:function(a,b){function = d(i,o){if(c.isArray(o))c.each(o,function(k,n){b||/\[\]$/.test(i)?f(i,n):d= (i+"["+(typeof = n=3D=3D=3D"object"||c.isArray(n)?k:"")+"]",n)});else!b&&o!=3Dnull&&typeof= = o=3D=3D=3D"object"?c.each(o,function(k,n){d(i+"["+k+"]",n)}):f(i,o)}funct= ion = f(i,o){o=3Dc.isFunction(o)?o():o;e[e.length]=3DencodeURIComponent(i)+"=3D= "+encodeURIComponent(o)}var = e=3D[];if(b=3D=3D=3Dw)b=3Dc.ajaxSettings.traditional;=0A= if(c.isArray(a)||a.jquery)c.each(a,function(){f(this.name,this.value)});e= lse for(var j in a)d(j,a[j]);return e.join("&").replace(yb,"+")}});var = la=3D{},Ab=3D/toggle|show|hide/,Bb=3D/^([+-]=3D)?([\d+-.]+)(.*)$/,W,va=3D= [["height","marginTop","marginBottom","paddingTop","paddingBottom"],["wid= th","marginLeft","marginRight","paddingLeft","paddingRight"],["opacity"]]= ;c.fn.extend({show:function(a,b){if(a||a=3D=3D=3D0)return = this.animate(K("show",3),a,b);else{a=3D0;for(b=3Dthis.length;a").appendTo("body");f=3De.css("display");if(f=3D=3D=3D"none")f=3D"block= ";e.remove();la[d]=3Df}c.data(this[a],"olddisplay",f)}}a=3D0;for(b=3Dthis= .length;a=3D0;f--)if(d[f].elem=3D=3D=3Dthis){b&&d[f](true);d.spli= ce(f,1)}});b||this.dequeue();return = this}});c.each({slideDown:K("show",1),slideUp:K("hide",1),slideToggle:K("= toggle",1),fadeIn:{opacity:"show"},fadeOut:{opacity:"hide"}},function(a,b= ){c.fn[a]=3Dfunction(d,f){return = this.animate(b,d,f)}});c.extend({speed:function(a,b,d){var f=3Da&&typeof = a=3D=3D=3D"object"?a:{complete:d||!d&&b||c.isFunction(a)&&a,duration:a,ea= sing:d&&b||b&&!c.isFunction(b)&&b};f.duration=3Dc.fx.off?0:typeof = f.duration=3D=3D=3D=0A= "number"?f.duration:c.fx.speeds[f.duration]||c.fx.speeds._default;f.old=3D= f.complete;f.complete=3Dfunction(){f.queue!=3D=3Dfalse&&c(this).dequeue()= ;c.isFunction(f.old)&&f.old.call(this)};return = f},easing:{linear:function(a,b,d,f){return = d+f*a},swing:function(a,b,d,f){return(-Math.cos(a*Math.PI)/2+0.5)*f+d}},t= imers:[],fx:function(a,b,d){this.options=3Db;this.elem=3Da;this.prop=3Dd;= if(!b.orig)b.orig=3D{}}});c.fx.prototype=3D{update:function(){this.option= s.step&&this.options.step.call(this.elem,this.now,this);(c.fx.step[this.p= rop]||=0A= c.fx.step._default)(this);if((this.prop=3D=3D=3D"height"||this.prop=3D=3D= =3D"width")&&this.elem.style)this.elem.style.display=3D"block"},cur:funct= ion(a){if(this.elem[this.prop]!=3Dnull&&(!this.elem.style||this.elem.styl= e[this.prop]=3D=3Dnull))return = this.elem[this.prop];return(a=3DparseFloat(c.css(this.elem,this.prop,a)))= &&a>-10000?a:parseFloat(c.curCSS(this.elem,this.prop))||0},custom:functio= n(a,b,d){function f(j){return = e.step(j)}this.startTime=3DJ();this.start=3Da;this.end=3Db;this.unit=3Dd|= |this.unit||"px";this.now=3Dthis.start;=0A= this.pos=3Dthis.state=3D0;var = e=3Dthis;f.elem=3Dthis.elem;if(f()&&c.timers.push(f)&&!W)W=3DsetInterval(= c.fx.tick,13)},show:function(){this.options.orig[this.prop]=3Dc.style(thi= s.elem,this.prop);this.options.show=3Dtrue;this.custom(this.prop=3D=3D=3D= "width"||this.prop=3D=3D=3D"height"?1:0,this.cur());c(this.elem).show()},= hide:function(){this.options.orig[this.prop]=3Dc.style(this.elem,this.pro= p);this.options.hide=3Dtrue;this.custom(this.cur(),0)},step:function(a){v= ar = b=3DJ(),d=3Dtrue;if(a||b>=3Dthis.options.duration+this.startTime){this.no= w=3D=0A= this.end;this.pos=3Dthis.state=3D1;this.update();this.options.curAnim[thi= s.prop]=3Dtrue;for(var f in = this.options.curAnim)if(this.options.curAnim[f]!=3D=3Dtrue)d=3Dfalse;if(d= ){if(this.options.display!=3Dnull){this.elem.style.overflow=3Dthis.option= s.overflow;a=3Dc.data(this.elem,"olddisplay");this.elem.style.display=3Da= ?a:this.options.display;if(c.css(this.elem,"display")=3D=3D=3D"none")this= .elem.style.display=3D"block"}this.options.hide&&c(this.elem).hide();if(t= his.options.hide||this.options.show)for(var e in = this.options.curAnim)c.style(this.elem,=0A= e,this.options.orig[e]);this.options.complete.call(this.elem)}return = false}else{e=3Db-this.startTime;this.state=3De/this.options.duration;a=3D= this.options.easing||(c.easing.swing?"swing":"linear");this.pos=3Dc.easin= g[this.options.specialEasing&&this.options.specialEasing[this.prop]||a](t= his.state,e,0,1,this.options.duration);this.now=3Dthis.start+(this.end-th= is.start)*this.pos;this.update()}return = true}};c.extend(c.fx,{tick:function(){for(var = a=3Dc.timers,b=3D0;b