[aos]

ppl, you should look at the latest voltage regulators. The classic 3-pin 1970's designs are way inferior to everything being out now. I spend tons of time over the last months looking at them for my DAC. There are references out there which have all the key parameters - load, line, temperature, dropout and noise - **SEVERAL** orders of magnitude better than those old regulators. For example, just couple of hours ago I placed an order for Analog's ADP3303-xxx voltage regulators. Don't knock out them just yet. Besides, what are you going to use in a discrete regulator to give you reference voltage? A LT1027/LT1036 or another one of many of voltage references made by LT, AD, NS, TI and others. 1ppm temp coefficients, less than 1uV p-p noise, virtually independent of line/load, ultra high initial precision etc...

However, I would like to look at your discrete regulator. Not for the portable DAC, but for the stationary it's always good to have the option. The one I used before (shunt reg) has only a resistor in the rail path and MOSFET as parallel regulating element, and opamp is used to compare the voltages. I suppose opamp can be replaced with a Darlington transistor or something, but you say you have your own design which has low temperature drift, so I'd love to see that too.

 

[ppl]

My discreet Regulator is your basic Series pass transistor witn a Gain transistor and Current sourcing Everything. The Great tempature performance comes from Current sourcing but more importently using an LED as the Voltage reference. LED temp Coe. is just the Oppisit of silicon so positive changes in one are canciled by negitive changes in the Other. LED's as a voltage reference is not New it is used in BJT current sources so as to maintain A constantant Current with Tempature as well as Lower noise and lower dynamic impedance over a rather wide bandwidth. I haven't seen an LED used as a Voltage reference in Voltage regulators yet. So with me always trying new unproven Ideas i used a radio Shack Blue LED as the Primarry voltage reference and Red led's to Bias the Current Sources whitch Resulted in Greatly improved Tempature stability. LED's also have lower noise than a Zener as well as Substantualy wider bandwidth that the Buried Zener devices like the BB (Ref-01) Current sourcing any voltage reference is required to obtain The best possible Performance out of any Ref. Most people still use a resistor rather than a current source and the Performance of the Reference Suffers as a result. Constant Current Sourcing eather in the form of BJT or FET will improve the rejection ratio about 30dB over just a resistor. The shunt Regulators like the TL-435? are alot better than the Typical Three termanal devices, However most people use a resistor to drop the voltage. Again about 30dB of additional PSRR can be obtained and 60dB is possible if the Current source is a Cascode type. this is because of the Higher output impedance the cascoding supplies. Ill chek out the AD device you mentioned.

 

[aos]

Some TI and AD references have PSRR of 60dB still at 1MHz and about 40 at 10MHz. AD regulator I mentioned provides up to 100mA and is stable with any output capacitor (so they call them AnyCAP (TM)). This is quite non-typical for classic references which loose any semblance of regulation when they reach about 100kHz.

But hey, current sourcing the reference, that's a cool idea. Although the ones I'm looking at have such extremely low dependency on load and line - just a couple of parts per million! - that it's not really necessary. Well, partially true, line regulation is really almost perfect but load regulation would still probably get improved by using the cascoded current source... Mmmm... Gotta try it.

 

[ppl]

These devices may have that kind of performance under static conditions, However i would like to see how thay change with large transients causing them to heat up and thus Change. Now i understand that the Feedback used will correct for this somewhat but unless the open-loop bandwidth is verry large then we can get a similar effect to TIM in amplifires where the Amp looses feedback for a moment and is thus momentarly overloaded. This showes up as ringing on Fast transientts and High frequency Sq waves. Now regarding the temp stability issue evean if these References did not drift at all and are used as a reference for another silicon based device the Change of the other device will still Change regardless of wether or not the Reference is stable and this is where the LED comes in.as it also changes but in the Other direction and can cange quickly well into the MHz. range. Oh ya Try the current sourcing unsteed of a resistor Regardless of what ref you are using dramiticlc improvement so if your chosen reference has 60 dB of PSRR to start with then Current sourcing will add another 30dB or so to that. I have Seen References combined with cascode J-fet/BjT current sources aproach 120db of PSRR from DC-200MHZ providing the ref is that fast whitch LED's are.

 

[aos]

All the references come with transient response graphs. Some are better than the others. I was looking specifically for high perfomance at high frequencies and found AD and TI. TI even advertises its thing as "fast transient, RF", for use in digital assistants and cell phones. I'm using these for digital supplies.

I like the idea, anyway. Now even the solid-state equipment, like my Kevin Gilmore amp, are glowing. Who needs tubes

.

 

[ppl]

What? are you refering to in the Coment about Kevins Amp. that you used Current sources or the References you mentioned?

 

[ppl]

Walt jung discribes the advantage of Current sourcing in the EDN artical about his supper regulator. http://archives.e-insite.net/archive...97/01di_03.htm

 

[aos]

Kevin's amp uses LEDs as voltage references for current sources. So there is a glow coming from inside the amp when it's working

.

 

[ppl]

Ok i understand. It is true that LED's have been used for voltage references in Current sources for a long time. Thay were used for the Reasons i cited in my above Post. Walt jung also dose the same and for the same Reason. I just wanted to point out about the Bennifit of using LED's as the Voltage reference in Voltage regulator circuits as well as point out the advantages of Current sourcing these Voltages References. Take for example the Classic BJT Current source using a LED as the Voltage reference. Note that the LED is still receiving it's current Via a resistor. This Resistor can be easly changed to a CRD Diode or jfet Current source. But I gess that's just me I like to cover All possible Bases in Circuit Design and leave nothing to Chance.

 

[ppl]

Attached is a copy of the schematic to Walts supper regulator. This is also available from the Link i provided above

 

[transducer]

I have a bunch of 2SK389's and 2SJ109 dual JFET's sitting around. Can I use these for your cascaded current source? Also, which leg does the little dot on the FET represent?

How do I calculate the current supplied to the opamp?

Can this be used on an output opamp, that is one that is not buffered by another opamp as you show in your schematic? I was thinking of doing this to my OPA627 line stage which also doubles as a headphone amp, it's just one opamp feeding either the amp or the headphones.

Thanks,
RonS

 

[transducer]

I have a bunch of 2SK389's and 2SJ109 dual JFET's sitting around. Can I use these for your cascaded current source? Also, which leg does the little dot on the FET represent?

How do I calculate the current supplied to the opamp?

Can this be used on an output opamp, that is one that is not buffered by another opamp as you show in your schematic? I was thinking of doing this to my OPA627 line stage which also doubles as a headphone amp, it's just one opamp feeding either the amp or the headphones.

Thanks,
RonS

 

[ppl]

transducer: If you are refering to the Dot on the Schematic i posted as an attachment, Then that dot represents the source of The FET. The gate and source on an N Ch Fet go twards the Negitive rail. The amount of Current supplied by a FET connected as a Current source is determand by the Idss specification of the FET in normal situations However since the Gate to source Voltage on the Cascoded fet is nor 0 Volts the actual amount of Current supplied by the cascode Current source will be less than the Devices nominal Idss Spec. What you Need to do is Connect the Dual N Ch fet up as a cascoded Curent as depicted in the Schematic with the Dot on the Schematic being Represented by the Source Lead. Then conect the Compleated Cascoded Current source up to a Battery with the Source twards the Negitive Battery termanal. then connect a Meter to the remaining drain lead and the battery Positive termanal. This set the Meter to measure on the 20 Ma range. The Meter will then read the actual amount of Current being supplied by the Current source. I then recheck this with a 1.2-1.5 volt Battery to see if the Curent remains the same. This is importent for battery operated amps, But an AC powered amp operating on + & - 15 to 18 Volts then just using the 9 Volt battery is Enough. The 2SK389's are an N channel device and thus are referenced to the Neg. Supply. the 2SJ109 are P Ch Devices and suould be referenced to the Positive rail. I don't have a data sheet on the 2SK389 but i Reviewed the 2SJ109 and as far as DC parameters go Idss of 2.6-20 Ma should be about right for a 1-2 Ma Cascoded Current source. The Capacitence of this device is quite High 95 Pf so if it were me i would use another Fet. A Pair of MPF-102's ar Better yet a BF-256 or the American equiv 2n5484 with a capacitence of 1.0 Pf would be alot better. As far as what the results in sound quality would be doing this to an opamp Drictly driving Headphones I think that since any reasonable Volume level is going to Require more current than the Current source is suppling and thus no longer Keep the Opamp into Class A it may be a Moot point. As far as an opamp driving another Active device like a Power Amp then the Bennifits will be realised because it would still be driving a High Impedance Load. Most power Amp's have a input impedance of 50,000 Ohms or more thus allowing the OPamp from the Line stage driving it to remain in Class A. On my other Posts i stressed the fact that i dont like to require an Opamp to drive low impedance loads like Headphones and Thus A Buffer will Dramiticaly improve the sound of the Opamp under these Conditions Regardless of wether or not the Opamp is into Class A. Hope this Helps.

 

[aos]

Ha! I knew those 2N5484 would come in handy one day!

 

[transducer]

Hi PPL,

Thanks for the reply. The 2SK389 has a capacitance of 25pF, so still to high I guess. I'll get some of the FET's you recommend.

I'm wondering what you think of this schematic which was posted at the Elliott Sound Products page, it shows an OPA2134 feeding a low impedance input on a THS6012 CFB opamp. I had thought of trying this circuit with a Class A biased input opamp. The article is also intereting if you haven't read it before.

Cheers,
RonS (transducer)