RCruz,
I don't know about good looking, but effective.
I have an adcom cross coil cartride. High output moving coil. To me it sounds better then the grado gold and green I have to compare to.
I put the transformer in the box and grounded it. It's an Antec 50 va 32 volt unit. With the needle off the record ther is no hum.
It is the 43 db circuit
The aura caps are c1 47nf. C2 15nf C3 100nf. And 2.2nd output
Yes Takman carbon resistors.
One shunt
All packed in an old VCR chassis to facilitate stacking.
Sharing is fun
With MKT cap 4.7th in the output the shunt acted weird...with the higher value electrolytic in it's place seems to behave well....
This drives an aikido preamp. Tube lab single ended 45 amp and DIY Ariel speakers.
I don't know about good looking, but effective.
I have an adcom cross coil cartride. High output moving coil. To me it sounds better then the grado gold and green I have to compare to.
I put the transformer in the box and grounded it. It's an Antec 50 va 32 volt unit. With the needle off the record ther is no hum.
It is the 43 db circuit
The aura caps are c1 47nf. C2 15nf C3 100nf. And 2.2nd output
Yes Takman carbon resistors.
One shunt
All packed in an old VCR chassis to facilitate stacking.
Sharing is fun
With MKT cap 4.7th in the output the shunt acted weird...with the higher value electrolytic in it's place seems to behave well....
This drives an aikido preamp. Tube lab single ended 45 amp and DIY Ariel speakers.
Try the REFLEKTOR
Imagine about 70% less than IDSS for K170BL under Vbe only in circuit, times the resistor for Vout. If using LED also, you must add its Vf for Vout.
If you fancy a new test instead, there is an idea I am revisiting and studying lately. Maybe you remember a different approach I had shown once, many moons ago in one of the sadly late Babo's threads. That one only Juma took notice and implemented in his BF862 line pre. Also Onvinyl tried with success by 1.2 CCS and mirror reg marriage, soon he solved his 1.2 problems too.
But those were just the core idea, not optimized for performance. Have a look in the attached schematics to see where I am getting at. Q1,Q2,Q3,Q4 comprise the so called improved Wilson current mirror. Q1,Q2 are the heart, Q3 takes care of their base currents impact, restoring equilibrium between the system's both legs, Q4 lends same voltage across Q1, Q2 (1Vbe). Q1,Q2,Q3 need have close hfe. If Q1,Q2 also have same Vbe and temp, we are near a perfect theoretical current mirror function. Its mission is to exactly copy the reference current from its input side to its output side. Its normally used as an active load for LTP input pairs in all kind of amplifiers lending CCS high impedance action to improve gain and PSRR, and as basic building block in op amps. But what a current mirror can do for a shunt reg? Is naturally the question.
Here, R6 is on the Wilson's input side, only thing it can disturb the reference current fed through it is Vout play. Iref=(Vout-(2Vbe+VfD1))/R6 or Vout=(Iref*R6)+(2Vbe+VfD1). R6 can be a 2K trimmer in these 10V examples for your P2P DCB1. What a dynamic load demand does to a PSU? Modulates its output voltage of course. That changes the bias on R6 slightly, hence the input side reference current. Any change will be mirrored (reflected, thus I code named this reg REFLEKTOR) to the output side. R5 does I/V there and drives M2. Also sits the Wilson on Vgs potential. That Mosfet will add or subtract current on the power node, thus restoring the voltage output. A current feedback system watching the output in other words. Elegant.
As for the other parts, R1 and Q5 set the CCS current by ICCS=Q5Vbe/R1. ~150mA in this example. J1 increases the CCS performance by setting its control tail current steady. Can be cascoded even with a 2SK246 or BF245B for example. Depends on max voltage by application specifics. Will not have start up problems with non optimal layout parasitics, since it does not reference to the output node. The whole system looks like giving 60dB open loop gain, also fine Zo frequency extension and stability margin on the simulator, even with strong gfs M2 types and no buffering, keeping simple. Lends to hot rod easily also. IRF640, IRF840, IRFP240 look OK for phase margin and current gain. Keep M1 low Ciss, as 9610. I will make a prototype soon to see it works OK. The other mirror regs did fine back then, so these hopefully improved and optimized ones are looking yummy to check out. Watch Q1,Q2 matching especially, face to face tied is even better. A remote sensing configuration is drawn in the examples, and it can be done 2 wire if very near to load. But you know about that from so many regs here by now. I have many cheap 2N4403 PNP low Rbb' (and low noise) BJTs that I have seen they keep a very tight hfe and Vbe as they came same batch loaded on ammo pack tape. C2 filters R6 and D1 (Vref), and must be 470uF and up, for low 1/f output noise, low VLF Zo, and bypassed as shown. ~4mA charge C2 up to Vref and lend slow start characteristic to Vout. Nice. Many low noise higher Vceo PNP transisors can be used, like BC560, 2SA1015, 2SA970 etc. or even BD140, MJE350, 2SA1381 as Vout and Q3's dissipation demands go up. The Wilson current mirror will have very good balance even with low beta BJTs and that's a plus. If Vin-5V need exceed J1's max in some higher Vo application, a BJT cascode CCS @ 2.5mA can be used instead.
Continuing on practical examples beyond your home brew point to point DCB1, by using a 2x22VAC 50Va Tx, 1A diode bridge, two 4700uF/35V filter caps, but with two green Leds (2xD1) and 5k rheostat connected trimmer R6, a +/-24V system can be made for powering Pass Burning Amp input stages with the transistor types and rest of values exactly as shown in the 10V schematics for instance.
Imagine about 70% less than IDSS for K170BL under Vbe only in circuit, times the resistor for Vout. If using LED also, you must add its Vf for Vout.
If you fancy a new test instead, there is an idea I am revisiting and studying lately. Maybe you remember a different approach I had shown once, many moons ago in one of the sadly late Babo's threads. That one only Juma took notice and implemented in his BF862 line pre. Also Onvinyl tried with success by 1.2 CCS and mirror reg marriage, soon he solved his 1.2 problems too.
But those were just the core idea, not optimized for performance. Have a look in the attached schematics to see where I am getting at. Q1,Q2,Q3,Q4 comprise the so called improved Wilson current mirror. Q1,Q2 are the heart, Q3 takes care of their base currents impact, restoring equilibrium between the system's both legs, Q4 lends same voltage across Q1, Q2 (1Vbe). Q1,Q2,Q3 need have close hfe. If Q1,Q2 also have same Vbe and temp, we are near a perfect theoretical current mirror function. Its mission is to exactly copy the reference current from its input side to its output side. Its normally used as an active load for LTP input pairs in all kind of amplifiers lending CCS high impedance action to improve gain and PSRR, and as basic building block in op amps. But what a current mirror can do for a shunt reg? Is naturally the question.
Here, R6 is on the Wilson's input side, only thing it can disturb the reference current fed through it is Vout play. Iref=(Vout-(2Vbe+VfD1))/R6 or Vout=(Iref*R6)+(2Vbe+VfD1). R6 can be a 2K trimmer in these 10V examples for your P2P DCB1. What a dynamic load demand does to a PSU? Modulates its output voltage of course. That changes the bias on R6 slightly, hence the input side reference current. Any change will be mirrored (reflected, thus I code named this reg REFLEKTOR) to the output side. R5 does I/V there and drives M2. Also sits the Wilson on Vgs potential. That Mosfet will add or subtract current on the power node, thus restoring the voltage output. A current feedback system watching the output in other words. Elegant.
As for the other parts, R1 and Q5 set the CCS current by ICCS=Q5Vbe/R1. ~150mA in this example. J1 increases the CCS performance by setting its control tail current steady. Can be cascoded even with a 2SK246 or BF245B for example. Depends on max voltage by application specifics. Will not have start up problems with non optimal layout parasitics, since it does not reference to the output node. The whole system looks like giving 60dB open loop gain, also fine Zo frequency extension and stability margin on the simulator, even with strong gfs M2 types and no buffering, keeping simple. Lends to hot rod easily also. IRF640, IRF840, IRFP240 look OK for phase margin and current gain. Keep M1 low Ciss, as 9610. I will make a prototype soon to see it works OK. The other mirror regs did fine back then, so these hopefully improved and optimized ones are looking yummy to check out. Watch Q1,Q2 matching especially, face to face tied is even better. A remote sensing configuration is drawn in the examples, and it can be done 2 wire if very near to load. But you know about that from so many regs here by now. I have many cheap 2N4403 PNP low Rbb' (and low noise) BJTs that I have seen they keep a very tight hfe and Vbe as they came same batch loaded on ammo pack tape. C2 filters R6 and D1 (Vref), and must be 470uF and up, for low 1/f output noise, low VLF Zo, and bypassed as shown. ~4mA charge C2 up to Vref and lend slow start characteristic to Vout. Nice. Many low noise higher Vceo PNP transisors can be used, like BC560, 2SA1015, 2SA970 etc. or even BD140, MJE350, 2SA1381 as Vout and Q3's dissipation demands go up. The Wilson current mirror will have very good balance even with low beta BJTs and that's a plus. If Vin-5V need exceed J1's max in some higher Vo application, a BJT cascode CCS @ 2.5mA can be used instead. Continuing on practical examples beyond your home brew point to point DCB1, by using a 2x22VAC 50Va Tx, 1A diode bridge, two 4700uF/35V filter caps, but with two green Leds (2xD1) and 5k rheostat connected trimmer R6, a +/-24V system can be made for powering Pass Burning Amp input stages with the transistor types and rest of values exactly as shown in the 10V schematics for instance.
Attachments
Could be the culprit why it could act up by not having one common reference ground actual same point between 2 boards and one reg. But since it got remedied and its so quiet as you report, I would not bother researching the configuration further if in your shoes.
Implementation could certainly be the problem. Before I changed the cap I tried playing with some different grounding schemes, but I am new to this and may not have come to a proper set up. One thing I have learned is that most things are not as simple as they seem on the surface.....thanks, evan
There is 1uf+2.2R output Zobel non the less. 560 yes. R7+C3=''quality'' bypass, C2 is must. Maybe a 120-220R will be needed on the PNP or NPN Q5's base in practice, but if no local oscillation in the CCS if tight layout, better not.
DCB1 output offset is due to a Vgs of non zero value developing across the upper jFET.
Normally the upper and lower jFETs operate at the same Id and the same Vds and the same Ta and thus have the same Vgs which is zero volts when run at Idss.
If the upper device is run at slightly different from Idss for that combination of operating voltage and operating current and operating Ta, then you can fine tune the output offset by inserting a low value resistor between the upper jFET source and the tapping point for the Audio output.
Look up D. Feucht to see the detailed procedure for sizing resistors around a CCS loaded jFET follower.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.