I have a question about grounding/return paths.
My next 2 pictures will show the 2 possibilities I'm thinking about, and I don't know which is good.
The 1st uses some kind of "floating ground", the ground of each component (PCB) is taken from the component (PCB) before it.
When I'm thinking about return paths, this one looks better.
But if somewhere in your circuits you have to feed multiple loads, or loads that have other sources connected to it, you'll have ground loops
The 2nd connects every ground to star ground, but return paths create loops :/
Help!
My next 2 pictures will show the 2 possibilities I'm thinking about, and I don't know which is good.
The 1st uses some kind of "floating ground", the ground of each component (PCB) is taken from the component (PCB) before it.
When I'm thinking about return paths, this one looks better.
But if somewhere in your circuits you have to feed multiple loads, or loads that have other sources connected to it, you'll have ground loops
The 2nd connects every ground to star ground, but return paths create loops :/
Help!
Attachments
No, I don't think so, but you can order the full series here: http://www.audioxpress.com/magsdirx/audelex/aabilist.htm, 1995 series.
Alternatively, Walt Jung may have an electronic copy.
Jan Didden
Alternatively, Walt Jung may have an electronic copy.
Jan Didden
Re: A picture may not not beat a thousand words
I had already privately asked ALW about the substitution of the 10 k with the current sink, and to possibly use a better voltage refference, guess where I got that idea from...
ok ok I'll tell you: it was the CCS for the AlephX you posted years ago,
As I recall ALW seemed to think the led-resistor combo was good enough, but quote me on that.
Decoupling the OPAMP PS looks like a good idea, but what do I know?
Fred Dieckmann said:
I had already privately asked ALW about the substitution of the 10 k with the current sink, and to possibly use a better voltage refference, guess where I got that idea from...
ok ok I'll tell you: it was the CCS for the AlephX you posted years ago,
As I recall ALW seemed to think the led-resistor combo was good enough, but quote me on that.
Decoupling the OPAMP PS looks like a good idea, but what do I know?
Just extreme design I guess......
"ALW seemed to think the led-resistor combo was good enough."
With a pre-regulator I absolutely agree.
When you look at an unregulated supply for the regulator it might be a different story. Assuming about 40 ohms for the LED impedance (which is function of bias current), 10K gives about 4 mV per volt of change for every volt of change on the input voltage. With a 100 ohms degeneration this gives a change in current of 40 uA per volt of change on the unregulated supply. This gives an effective output impedance of 25K for the current source, which is very low for a current source. This is MUCH lower than the impedance of the same current source with a well filtered reference voltage for the current source and the bottleneck for the performance of this part of the circuit.
Using a 1.8 mA JFET current source like the J507 will give about 1 Megohm impedance which is a couple orders of magnitude better than a 10K bias resistor for the current source LED voltage reference.
All this points out the obvious question of how good is good enough. Design in this manner doesn't require a lot of math and gives insight to circuit changes to might be impossible to measure in the finished circuit. Methods like these often bridge the gap between subjective and measurement oriented design. This example illustrates something easily measured in the subcircuit, but who's influence might be below the noise floor in power supply rejection measurements at the regulator output.
"ALW seemed to think the led-resistor combo was good enough."
With a pre-regulator I absolutely agree.
When you look at an unregulated supply for the regulator it might be a different story. Assuming about 40 ohms for the LED impedance (which is function of bias current), 10K gives about 4 mV per volt of change for every volt of change on the input voltage. With a 100 ohms degeneration this gives a change in current of 40 uA per volt of change on the unregulated supply. This gives an effective output impedance of 25K for the current source, which is very low for a current source. This is MUCH lower than the impedance of the same current source with a well filtered reference voltage for the current source and the bottleneck for the performance of this part of the circuit.
Using a 1.8 mA JFET current source like the J507 will give about 1 Megohm impedance which is a couple orders of magnitude better than a 10K bias resistor for the current source LED voltage reference.
All this points out the obvious question of how good is good enough. Design in this manner doesn't require a lot of math and gives insight to circuit changes to might be impossible to measure in the finished circuit. Methods like these often bridge the gap between subjective and measurement oriented design. This example illustrates something easily measured in the subcircuit, but who's influence might be below the noise floor in power supply rejection measurements at the regulator output.
Fet CCS
"On the replacement of the LED current source with a JFET current source'
Actually I was taking about replacing the resistor biasing the Green LED reference for the voltage reference for the BJT current source. Replacing the PNP with a ZVP3310 and the zener with two series Green LEDs for the voltage reference would make fet current source with low dropout voltage if one wanted to replace the PNP.
"On the replacement of the LED current source with a JFET current source'
Actually I was taking about replacing the resistor biasing the Green LED reference for the voltage reference for the BJT current source. Replacing the PNP with a ZVP3310 and the zener with two series Green LEDs for the voltage reference would make fet current source with low dropout voltage if one wanted to replace the PNP.
Hi Fred,
On the subject of line regulation, if we assume no preregulator, it seems that the AD825 is not the best choice for line regulation performance at 120 Hz. Because the dominant open-loop pole of this op-amp is at about 10 kHz, the open loop gain at 120 Hz is not nearly as high as it could be, given its gain-bandwidth product. In looking at my 90 Volt version of this thing, which has no pre-regulator, I found that (in simulation at least) one can get about 40 dB better line regulation at 120 Hz by using an op-amp having about the same gain-bandwidth product as the AD825, but with an open-loop pole below 120 Hz (giving about 40 dB higher open-loop gain at this frequency).
On the subject of line regulation, if we assume no preregulator, it seems that the AD825 is not the best choice for line regulation performance at 120 Hz. Because the dominant open-loop pole of this op-amp is at about 10 kHz, the open loop gain at 120 Hz is not nearly as high as it could be, given its gain-bandwidth product. In looking at my 90 Volt version of this thing, which has no pre-regulator, I found that (in simulation at least) one can get about 40 dB better line regulation at 120 Hz by using an op-amp having about the same gain-bandwidth product as the AD825, but with an open-loop pole below 120 Hz (giving about 40 dB higher open-loop gain at this frequency).
Bricolo said:
Are you good for drawings?
No, but I'll try...
An externally hosted image should be here but it was not working when we last tested it.
Here's something off the internet:
An externally hosted image should be here but it was not working when we last tested it.
First, I'm not sure what part you don't understand. You don't know why I choose scheme 2 or you don't know why I indicated that the sense lines allow you to move the star connection to the circuit block. So, I'm not sure which question I'm trying to answer.
As far as scheme 1 and scheme 2, it really depends on where the sense circuitry is connected (where the Kelvin connections are located). At first scheme 2 looked right to me, but it's dependent on the sense connection.
As in a four-wire ohm measurement, there are two "Kelvin" connections: upper Kelvin (Ku in diagram) and lower Kelvin (KL). These are the sense lines, of the greatly simplified Regulator circuit. The regulator will regulate the voltage across those two points. Because of series resistance, the voltage at other points in your whole circuit will be not as well regulated (though perhaps the difference is small).
The reason this helps is because the current in the relatively low impedance conductors of +Reg and +Reg Return have voltage drops (again perhaps small). The sense lines +Reg Sense and +Reg Return Sense are relatively high impedance connections to the voltage reference. The Regulator is looking/sensing/regulating across the Kelvin connections.
To the best of my knowledge, the best place to place the Kelvin sense connections is where your most sensitive circuitry is (ie the input to the amplifier circuitry). Someone at diyAudio wrote that the most sensitive circuitry should be closest to the power supply. That way fluctuations in the power section of your circuitry are downstream of your sensitive circuitry. Rather than the sensitive circuitry being down stream of the power section.
Simply put using the two sense lines properly compensate for the series resistance loss of the power and return lines.
I would expect that the Kelvin connections are the place to have star connections. The +reg side is starred as the +reg return is a star.
Comments welcome.
Also, if anyone is familiar with something called a 16-wire Kelvin connection (where each of the 4 traditional Kelvin connections each have 4-wire connections), I'll appreciate some information. I'm sure that ran across that once, but can not find any reference to it (thought it was from a Texas Instruments page).
JF
That's not helping your sales.
"What does this mean in real life, I mean at the regulator output? Hardly 10-100 times better?"
Build it and see! Who do you think I am?
Your mother?
Hint: it depends on the frequency, and the loop gain of the op amp used. When do you think you will have the measurements to share with us? We are all extremely lazy and I thought maybe you would do the work for us for a change.
"What does this mean in real life, I mean at the regulator output? Hardly 10-100 times better?"
Build it and see! Who do you think I am?
Your mother?
Hint: it depends on the frequency, and the loop gain of the op amp used. When do you think you will have the measurements to share with us? We are all extremely lazy and I thought maybe you would do the work for us for a change.
LM317s and level shifting
There are 2 aspects of the ALW reg which concern me: the LM317 pre reg and the level shifter.
ALW in his application note states that decoupling the ref pin on the LM317 pre-reg is a must do (BTS!!). This indicates that the SR (downstream of the LM317) is somewhat sensitive to what it is fed with. If this is the case then surely some of the "greyness" people attribute to the 317 will afflict the SR?
Fred has obviously thought about this and made some excellent suggestions to improve the PSRR if ommiting the pre-reg. The level shifter is the target of these improvements (although I would have to think about the merits of using a FET for the level shifter and if splitting the 10K sink resistor into 2 X 5K I would decouple the mid point to the +ve supply rail not to 0V).
My approach would be to get rid of the LM317 and the level shifter, if you don't have the level shifter the pre-reg is (possibly?) not required.
The clever bits of the Jung topology are that the voltage reference and the op-amp supply are powered from the regulated o/p. I agree that the reference will benefit from this approach but the op-amp should be quite happy working off a zener/emitter follower with local decoupling for the op-amp (eg. 10R and 100uF) off the raw dc supply. The op-amp can then drive an emitter follower (with suitably high beta) directly without the need for the level shifter. The reference can still be powered by the output as per the Jung topology.
Dave
There are 2 aspects of the ALW reg which concern me: the LM317 pre reg and the level shifter.
ALW in his application note states that decoupling the ref pin on the LM317 pre-reg is a must do (BTS!!). This indicates that the SR (downstream of the LM317) is somewhat sensitive to what it is fed with. If this is the case then surely some of the "greyness" people attribute to the 317 will afflict the SR?
Fred has obviously thought about this and made some excellent suggestions to improve the PSRR if ommiting the pre-reg. The level shifter is the target of these improvements (although I would have to think about the merits of using a FET for the level shifter and if splitting the 10K sink resistor into 2 X 5K I would decouple the mid point to the +ve supply rail not to 0V).
My approach would be to get rid of the LM317 and the level shifter, if you don't have the level shifter the pre-reg is (possibly?) not required.
The clever bits of the Jung topology are that the voltage reference and the op-amp supply are powered from the regulated o/p. I agree that the reference will benefit from this approach but the op-amp should be quite happy working off a zener/emitter follower with local decoupling for the op-amp (eg. 10R and 100uF) off the raw dc supply. The op-amp can then drive an emitter follower (with suitably high beta) directly without the need for the level shifter. The reference can still be powered by the output as per the Jung topology.
Dave
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.