Hi Andrew,
I'm glad we have it somewhat sorted.
I see what you are trying to achieve with the pre-regulator. Does setting the pre-regulator to ~3V (ground referenced) above your final regulated output give a similar performance advantage?
As for higher voltage operation, I had no problems starting with 100 mA loads when fed ~50 VAC, looking for 65VAC out with 2K for R10 (.19W in T4) I'm going to try my higher voltage version fed with DC and see what happens on startup for KSA/Leach front ends (my reasoning for wanting something scalable in the first place.)
EDIT: 18mm 1,000 uF 100V caps will fit the board with a little interference, as the layout calls for 16mm diameter. Using these, my 65V ouput regulator's noise and ripple were buried in the residual noise of my scope. You may not need to prefilter your front end supplies. 😉
You're right, the current in the differential must be reduced at higher voltages. I still need to review it once more, but I have a guide nearing publication.
In the meantime, here is a spreadsheet that helps choose component values for other than 15V operation. Hopefully it will be useful without the supporting verbage.
Remember, if you use MPSA42/92 you must insert them opposite the way shown on the silk screen.
I'm glad we have it somewhat sorted.
I see what you are trying to achieve with the pre-regulator. Does setting the pre-regulator to ~3V (ground referenced) above your final regulated output give a similar performance advantage?
As for higher voltage operation, I had no problems starting with 100 mA loads when fed ~50 VAC, looking for 65VAC out with 2K for R10 (.19W in T4) I'm going to try my higher voltage version fed with DC and see what happens on startup for KSA/Leach front ends (my reasoning for wanting something scalable in the first place.)
EDIT: 18mm 1,000 uF 100V caps will fit the board with a little interference, as the layout calls for 16mm diameter. Using these, my 65V ouput regulator's noise and ripple were buried in the residual noise of my scope. You may not need to prefilter your front end supplies. 😉
You're right, the current in the differential must be reduced at higher voltages. I still need to review it once more, but I have a guide nearing publication.
In the meantime, here is a spreadsheet that helps choose component values for other than 15V operation. Hopefully it will be useful without the supporting verbage.
Remember, if you use MPSA42/92 you must insert them opposite the way shown on the silk screen.
I think you lose the protection offered by the chip regulator. Floating means the chip reg ALWAYS maintains 2.3V across the pass device so in overload the scalable Vout voltage could fall either slightly or substantially and the dissipation remains proportional to output current. The chip reg now absorbs the output current AND the extra volts drop plus the original volts drop. It's internal protection limits current to about 1.5A and has temp shutdwon if dissipation is exceeded longterm. Ground referenced chip reg loses most of this dissipation protection and would have to be configured for excess voltage due to 37V limitations.
R10 sets the LTP pair current. R1 is effectively a constant current source.
t3 dissipation is fixed by Vout and R1
t4 dissipation is fixed by Vout and [Ir10-Ir1]. Increasing R10 reduces Pd of t4 but makes the starting worse. Hence the need for improvement.
Thanks for confirming that 100mA and r10=2k0. I shall series connect my lab supply and try to replicate the numbers.
What make are those 100V caps? or even 80V. My big stock of 820uF63V don't quite stretch to a Leach front end when all the tolerances go to max Vin.
The 1,000 uF caps are nichicon UVR2A102MHD, digikey part number 493-1152-ND
the largest true fit is 470 uf 100V, Panasonic FC, Digikey part P10874 (and others in similar packages)
There is an 18mm 100V Panasonic FC, but only 680 uF. Digikey P/N P10875
For the output caps I used 100 uF/100 V, P10775-ND - no room to squeeze bigger in there.
Agreed, except that T3 dissipation is set by Vunreg, R1 (sets current) and R5
With a floating preregulator, you have a fairly quiet input voltage. How about just omit R4 in this case? Or to be really overboard, replace it or R3 with a constant current source and resistor as needed.
This arrangement feels something like cascoding a device in the signal path, dramatically improving its bandwidth/linearity by keeping a constant voltage across the device.
There seems to be two schools of thought on cascoding - one that you reference the source/emitter and the other referencing ground.
So, a ground referenced "cascode" could be a fairly simple preregulator with a TO-220 and a string of zeners, like Pass used in the balanced zen preamp (BOSOZ). It doesn't have to be a mosfet. What do you think?
the largest true fit is 470 uf 100V, Panasonic FC, Digikey part P10874 (and others in similar packages)
There is an 18mm 100V Panasonic FC, but only 680 uF. Digikey P/N P10875
For the output caps I used 100 uF/100 V, P10775-ND - no room to squeeze bigger in there.
Agreed, except that T3 dissipation is set by Vunreg, R1 (sets current) and R5
With a floating preregulator, you have a fairly quiet input voltage. How about just omit R4 in this case? Or to be really overboard, replace it or R3 with a constant current source and resistor as needed.
This arrangement feels something like cascoding a device in the signal path, dramatically improving its bandwidth/linearity by keeping a constant voltage across the device.
There seems to be two schools of thought on cascoding - one that you reference the source/emitter and the other referencing ground.
So, a ground referenced "cascode" could be a fairly simple preregulator with a TO-220 and a string of zeners, like Pass used in the balanced zen preamp (BOSOZ). It doesn't have to be a mosfet. What do you think?
Hi,
the advantage of the chip reg (LM317/337) is overload protection built in and it only adds three components. I think with care I can fit all three onto the PCB in that spare space where Jens' credits are printed (below C4 & C27). The best place for it is above C5 & C28 but there's less space before Vout
Besides, it's the recommended way of WJ and he's always right.
Using floating pre reg and omitting R3 turns scalable into super regulator (well, almost).
the advantage of the chip reg (LM317/337) is overload protection built in and it only adds three components. I think with care I can fit all three onto the PCB in that spare space where Jens' credits are printed (below C4 & C27). The best place for it is above C5 & C28 but there's less space before Vout
Besides, it's the recommended way of WJ and he's always right.
Using floating pre reg and omitting R3 turns scalable into super regulator (well, almost).
It will be interesting to see what the measured and audible differences in performance are with a preregulator. I don't have access to equipment to measure any improvements in noise or frequency response. Please post if you do measure in your modified state.
Perhaps omitting R7 and D6 will improve startup in your situation.
While you're at it, lift the appropriate end of R2 and take your feedback from the load point to be even closer to super-regulator configuration. Hopefully you won't find the instability issues that some have had with the AD797 based super-regulators.
Perhaps omitting R7 and D6 will improve startup in your situation.
While you're at it, lift the appropriate end of R2 and take your feedback from the load point to be even closer to super-regulator configuration. Hopefully you won't find the instability issues that some have had with the AD797 based super-regulators.
Makes me wonder just how many intend to use the regulator boards for the KSA and Leach fronts ?
Hi,
we should not need to make this modification
we should not need to make this modification
but it is forced upon us. Fortunately it is very easy to carry out.
I'm not sure what you mean by that Andrew. The standard configuration is well suited for most applications.
If you are saying that we should have had the option of taking feedback from the load point rather than the regulator output, we might have been able to accommodate that had a request been made in the design process. It would have added a 2 pin jumper, that most user would have simply connected with a link.
However, if I understand what I have read about the various remote sensing super regulators, the remote sensing that gives such accuracy also gives rise to increased oscillation potential.
The originally stated design goal was not a super regulator, just clean quiet power that fits within the confines of Eagle's free version board limitations. I believe most users bought this supply looking for a "plug and play" power supply that works without being overly sensitive to the particulars of their build layout. I think Jens did a fine job meeting that objective with this design and layout.
Is there room for improvement in the design? Of course there is, nothing is perfect. But given the board size constraint, you'd be hard pressed to fit much more in there. It might be interesting to see what a CCS in place of R10 and R4 does to the performance. As it is, I could not measure any improvement without buying a lot of new test gear.
I still look forward to seeing your results in your discrete super regulator quest.
Perhaps you can quantify the minimum current in the differential for reliable starting. I have found that 4.7 mA works reliably, while you had problems with 2.6 mA, although since you had R7 down to ~1K this may be OK, too.
If you are saying that we should have had the option of taking feedback from the load point rather than the regulator output, we might have been able to accommodate that had a request been made in the design process. It would have added a 2 pin jumper, that most user would have simply connected with a link.
However, if I understand what I have read about the various remote sensing super regulators, the remote sensing that gives such accuracy also gives rise to increased oscillation potential.
The originally stated design goal was not a super regulator, just clean quiet power that fits within the confines of Eagle's free version board limitations. I believe most users bought this supply looking for a "plug and play" power supply that works without being overly sensitive to the particulars of their build layout. I think Jens did a fine job meeting that objective with this design and layout.
Is there room for improvement in the design? Of course there is, nothing is perfect. But given the board size constraint, you'd be hard pressed to fit much more in there. It might be interesting to see what a CCS in place of R10 and R4 does to the performance. As it is, I could not measure any improvement without buying a lot of new test gear.
I still look forward to seeing your results in your discrete super regulator quest.
Perhaps you can quantify the minimum current in the differential for reliable starting. I have found that 4.7 mA works reliably, while you had problems with 2.6 mA, although since you had R7 down to ~1K this may be OK, too.
Hi Bob,
I would rather not pursue that issue any further, having upset another with my views.
Locating the Vout end of R2 under the output terminal does not seem that difficult and I cannot see the stability risk.
I shall see how my experiments go and report back. Just don't hold your breath.
Congrats on the spreadsheet. Easy to understand and use.
I would rather not pursue that issue any further, having upset another with my views.
Locating the Vout end of R2 under the output terminal does not seem that difficult and I cannot see the stability risk.
I shall see how my experiments go and report back. Just don't hold your breath.
Congrats on the spreadsheet. Easy to understand and use.
Andrew,
I just did some simulations and measurements.
I changed R4 from 1k to 1k2 to 1k5.
The results of startup is as follows. All sims are with 15 Ohm load
First R4 original value of 1k
Startup at about 19V DC
Note the difference in startup voltage between the three simulations.
Now for R4 = 1k2
And finally R4 = 1k5
From the simulation I would (Re)choose R4 = 1k2 to have some noise margin for ripple on the DC feed. This is only to solve your problem with startup, I still recommend to keep the original value for everybody that does not have a start up problem.
I hope this helps
\Jens
I just did some simulations and measurements.
I changed R4 from 1k to 1k2 to 1k5.
The results of startup is as follows. All sims are with 15 Ohm load
First R4 original value of 1k
Startup at about 19V DC
Note the difference in startup voltage between the three simulations.
An externally hosted image should be here but it was not working when we last tested it.
Now for R4 = 1k2
An externally hosted image should be here but it was not working when we last tested it.
And finally R4 = 1k5
An externally hosted image should be here but it was not working when we last tested it.
From the simulation I would (Re)choose R4 = 1k2 to have some noise margin for ripple on the DC feed. This is only to solve your problem with startup, I still recommend to keep the original value for everybody that does not have a start up problem.
I hope this helps
\Jens
Jens,
That is interesting, thanks.
Does the startup voltage vary with load?
Do you plan to run simulations to determine the minimum differential current? For the spreadsheet I picked 2.5 mA just basing it on twice the current needed in the T3 leg.
That is interesting, thanks.
Does the startup voltage vary with load?
Do you plan to run simulations to determine the minimum differential current? For the spreadsheet I picked 2.5 mA just basing it on twice the current needed in the T3 leg.
Bob,
Lower loads seem to let the regulator start at lower input voltages.
This is strange as Andrew has problems at low loads.
\Jens
Lower loads seem to let the regulator start at lower input voltages.
This is strange as Andrew has problems at low loads.
\Jens
Hi Jens,
thanks for taking the time to run those simulations.
The last graph (1K5) shows a start up that mimics what I found at low currents (100Rload) when R4 was removed (infinite), except that without R4 it starts up at a much lower voltage and then tracks Vin until regulation is achieved just as your sim shows.
It does appear that the R3:R4 ratio and Rload are interdependent with regard to start up.
Your graph also confirms the LDO values to be expected, less than a volt at 1A of output is pretty good.
thanks for taking the time to run those simulations.
The last graph (1K5) shows a start up that mimics what I found at low currents (100Rload) when R4 was removed (infinite), except that without R4 it starts up at a much lower voltage and then tracks Vin until regulation is achieved just as your sim shows.
It does appear that the R3:R4 ratio and Rload are interdependent with regard to start up.
Your graph also confirms the LDO values to be expected, less than a volt at 1A of output is pretty good.
Hi,
just posting my circuit values for those that want/need to know.
r1=910
r2,4=1k05
r3=10k7 was 100k & does not start
r5=1k0
r6=4k7 for Vout <=25V
r7=10k tested @1k3 and did not start
r8=22k
r9=100
r10=1k06 was 3k6 & does not start
r11=5k1
p1=5k
c1,2,8,9=47n pes
c3,12,13=100n cer
c4,5=820u E63V
c6,14=1.5u tant50V
c7=100p cer
c10,11=330u E63V
My current requirements are well down on design specification. Hence smaller electros @ higher voltage.
Iout=200mA max continuous and 500mA peak. 50V tantalum will only suit KSA50 and not Leach.
Some further testing of start up shows little dependance on Vout setting.
Bob,
setting my current limit deliberately low does not interfere with start up. I tested at 50mA, 100mA, 250mA and 1000mA. All behave exactly the same into open circuit and 100Rload.
just posting my circuit values for those that want/need to know.
r1=910
r2,4=1k05
r3=10k7 was 100k & does not start
r5=1k0
r6=4k7 for Vout <=25V
r7=10k tested @1k3 and did not start
r8=22k
r9=100
r10=1k06 was 3k6 & does not start
r11=5k1
p1=5k
c1,2,8,9=47n pes
c3,12,13=100n cer
c4,5=820u E63V
c6,14=1.5u tant50V
c7=100p cer
c10,11=330u E63V
My current requirements are well down on design specification. Hence smaller electros @ higher voltage.
Iout=200mA max continuous and 500mA peak. 50V tantalum will only suit KSA50 and not Leach.
Some further testing of start up shows little dependance on Vout setting.
Bob,
setting my current limit deliberately low does not interfere with start up. I tested at 50mA, 100mA, 250mA and 1000mA. All behave exactly the same into open circuit and 100Rload.
Hi,
I just tried a new higher voltage version.
Vin=51.6Vdcmax, Vout=49.5Vmax, Iout=250mAmax,
Dissipation in the pass device ~=1W, runs cool with a 21C/W sink. Tc about 30degC to 35degC.
I disconnected the circuit to check component temperatures.
The PCB is a lot hotter than the sink, I'd guess over 40degreesC.
This was after about 15mins at various input voltages 28Vdc to 51.6Vdc and output voltages from 25V to 49.2V and Rload =197r5
Iinput max (at Vout=49.5V) about 320mA.
Total input power about 16.5W, output power 12.4W that's 1w in pass device and 3.1W in all the other components. No wonder the PCB was hot. There was some leakage of heat from the load back into the PCB, the output end was noticeably hotter than the middle and similarly the diode bridge end was of an intermediate temperature.
I have never run a PCB at these kinds of elevated temperatures.
Any comments?
I just tried a new higher voltage version.
Vin=51.6Vdcmax, Vout=49.5Vmax, Iout=250mAmax,
Dissipation in the pass device ~=1W, runs cool with a 21C/W sink. Tc about 30degC to 35degC.
I disconnected the circuit to check component temperatures.
The PCB is a lot hotter than the sink, I'd guess over 40degreesC.
This was after about 15mins at various input voltages 28Vdc to 51.6Vdc and output voltages from 25V to 49.2V and Rload =197r5
Iinput max (at Vout=49.5V) about 320mA.
Total input power about 16.5W, output power 12.4W that's 1w in pass device and 3.1W in all the other components. No wonder the PCB was hot. There was some leakage of heat from the load back into the PCB, the output end was noticeably hotter than the middle and similarly the diode bridge end was of an intermediate temperature.
I have never run a PCB at these kinds of elevated temperatures.
Any comments?
For 65V out I increased R3 to 200K and R4 to 20K to reduce their dissipation. If you've left R4 at 1K I'm surprised it didn't burn out.
Similarly R2 was increased to 18K6, with R11 at 3K3 to give lower dissipation in the feedback string.
As mentioned earlier, R10 was increased to 2K.
The spreadsheet in Post 341 prompts you to increase resistor values when dissipation exceeds .125W
Don't forget the dissipation in R6.
Similarly R2 was increased to 18K6, with R11 at 3K3 to give lower dissipation in the feedback string.
As mentioned earlier, R10 was increased to 2K.
The spreadsheet in Post 341 prompts you to increase resistor values when dissipation exceeds .125W
Don't forget the dissipation in R6.
Hi Bob,
I increased resistor values, although not as far as your spreadsheet, which suggests limiting dissipation to less than half maximum rating.
I was on 40Volt values but kept increasing output to 49.5V to see if regulation kept going OK and to force more output current into my load.
Jens has confirmed that 40degC or so is not a problem. So I am happy.
BTW. LDO = 0.66V at 49.5Vout and 250mA. There must be a fair bit of gain in that LTP/Common Emitter combination.
I increased resistor values, although not as far as your spreadsheet, which suggests limiting dissipation to less than half maximum rating.
I was on 40Volt values but kept increasing output to 49.5V to see if regulation kept going OK and to force more output current into my load.
Jens has confirmed that 40degC or so is not a problem. So I am happy.
BTW. LDO = 0.66V at 49.5Vout and 250mA. There must be a fair bit of gain in that LTP/Common Emitter combination.
Hi,
finally got the oscilloscope on the output.
My vertical and horizntal calibration is well off so all my numbers are going to be approximate.
Vin=44Vdc to 50Vdc, Vout42V, Rload~=200r 20W, as you have read my caps are not as recommended.
R4=10k, R3=3k9, Vout starts when Vin is about 11.5V and tracks with Vin until regulation is achieved.
Now to try and describe what I see.
Bursts of 5MHz oscillation with a mark space ratio that varies from 60:40 at very low LDO ~=800mV across T1 that changes to 20:80 by the time Vin is at 50Vdc (still using my lab supply). The mark on time seems to stay nearly constant, the space time seems to increase markedly as Vin rises.
The space has a tapering (damped) waveform that goes down to noise that is too low for me to measure and then just as the burst is starting again the waveform comes back above noise and goes on to about 2mV when Vt1=800mV and increases to about 6mV when Vin=50Vdc. The 50MHz voltage holds fairly constant while the burst is on and it's just the ends of the burst that rise and fall into noise.
It took a while to find these bursts, using AC input, 1uS to 10uS/div and 5mV/div. The scope locks onto them quite easily.
Are they artifacts due to interaction between lab supply output control and scaleable's control?
ps. the T3 to T4 base to base voltage is 20mV on negative and 10mV on positive. T4 Ic=5mA and T3 Ic=1.3mA. T3 Ib ~=2uA.
Comments?
finally got the oscilloscope on the output.
My vertical and horizntal calibration is well off so all my numbers are going to be approximate.
Vin=44Vdc to 50Vdc, Vout42V, Rload~=200r 20W, as you have read my caps are not as recommended.
R4=10k, R3=3k9, Vout starts when Vin is about 11.5V and tracks with Vin until regulation is achieved.
Now to try and describe what I see.
Bursts of 5MHz oscillation with a mark space ratio that varies from 60:40 at very low LDO ~=800mV across T1 that changes to 20:80 by the time Vin is at 50Vdc (still using my lab supply). The mark on time seems to stay nearly constant, the space time seems to increase markedly as Vin rises.
The space has a tapering (damped) waveform that goes down to noise that is too low for me to measure and then just as the burst is starting again the waveform comes back above noise and goes on to about 2mV when Vt1=800mV and increases to about 6mV when Vin=50Vdc. The 50MHz voltage holds fairly constant while the burst is on and it's just the ends of the burst that rise and fall into noise.
It took a while to find these bursts, using AC input, 1uS to 10uS/div and 5mV/div. The scope locks onto them quite easily.
Are they artifacts due to interaction between lab supply output control and scaleable's control?
ps. the T3 to T4 base to base voltage is 20mV on negative and 10mV on positive. T4 Ic=5mA and T3 Ic=1.3mA. T3 Ib ~=2uA.
Comments?
- Status
- Not open for further replies.