Ya know... I think I just confirmed that, yes, I am a geek. Woke up this morning thinking about R7, C3 and realized their behavior is best characterized by a first order differential equation. Boy... It's been a while since I dealt with those and it's way too early in the morning to solve differential equations. But my thinking continued... Neglecting the trickle current through D2, the startup is set by a 10 uA current source, R7, and C3 - all three in parallel. R7 and the current source is a Norton source. This can be converted into a Thevenin source, in which case, it becomes a voltage source with Rout = R7 into C3. That will behave like the Vfinal * (1-e^(t/tau)). So neglecting the current through D2, that's correct.
The intuitive way to look at this is that the SET current charges C3. As the voltage on C3 rises, current will flow in R7 so now the charging current into C3 becomes less. There's only 10 uA total current available. So as the voltage across C3 rises, it will charge slower. Just as the exponential describes.
Insight! Yes!
Thanks for pushing on this!
The good news here is that this eliminates my final concern - that the start-up behavior would change by changing R7, C3. But it looks like as long as the product of R7*C3 is constant, the start-up behavior should not change. That's very good news as guaranteeing graceful start-up of these regulators with low voltage parts in a high-voltage circuit is not trivial.
~Tom
The intuitive way to look at this is that the SET current charges C3. As the voltage on C3 rises, current will flow in R7 so now the charging current into C3 becomes less. There's only 10 uA total current available. So as the voltage across C3 rises, it will charge slower. Just as the exponential describes.
Insight! Yes!
Thanks for pushing on this!The good news here is that this eliminates my final concern - that the start-up behavior would change by changing R7, C3. But it looks like as long as the product of R7*C3 is constant, the start-up behavior should not change. That's very good news as guaranteeing graceful start-up of these regulators with low voltage parts in a high-voltage circuit is not trivial.
~Tom
Agreed, Tom. Good luck on your measurements. I’m looking forward to hearing of them.
Now, about that lousy D2 current. It must be a real pain on start-up with light loads. We could replace D2 with a very low current shunt regulator (or a few in series), like the LM385-2.5 which works on only 20uA. Or the REF1112 that works on 1.2uA! Their dynamic impedance beats that of the zener and the R1 bias resistor could be raised to a much higher resistance. The combination would further improve line rejection and reduce startup problems with light loads without adding additional output capacitance.
Now, about that lousy D2 current. It must be a real pain on start-up with light loads. We could replace D2 with a very low current shunt regulator (or a few in series), like the LM385-2.5 which works on only 20uA. Or the REF1112 that works on 1.2uA! Their dynamic impedance beats that of the zener and the R1 bias resistor could be raised to a much higher resistance. The combination would further improve line rejection and reduce startup problems with light loads without adding additional output capacitance.
Attached sims rather well. I'm not able to see any difference in the start-up behavior between this and the original circuit. The output impedance is improved (lowered) by an order of magnitude (10x). These are the changes compared to the original BOM:
R4 (pot) = 20k
R5 = 6.65k 1%
R6 = 16.2k 1%
R7 = 1M 1%
C3 = 1uF, X7R ceramic
I still need to test this in practice, but I see no reason why it wouldn't work.
I increased the adjustment range on the pot to roughly +/-7.5 % from the previous +/-5 %. Just to get a bit more margin in case you end up with resistors that are on the outskirts of the statistical distribution of resistance.
Me to...
My sim says that the regulator has 130 dB of line rejection from DC to 20 kHz and a +20 dB/dec up from there. At 1 MHz it flattens out at 90 dB and rolls off from there. This with 2 mA load current. So in practice, the line rejection is probably limited by things such as parasitic capacitances, board layout, etc.
I deliberately sized the feedback network to absorb both the current required to make the regulator work and the trickle current through D2. If the current in the feedback network doesn't exceed the current through D2, the regulator doesn't regulate properly. The output voltage rises and the line rejection goes to zero.
~Tom
R4 (pot) = 20k
R5 = 6.65k 1%
R6 = 16.2k 1%
R7 = 1M 1%
C3 = 1uF, X7R ceramic
I still need to test this in practice, but I see no reason why it wouldn't work.
I increased the adjustment range on the pot to roughly +/-7.5 % from the previous +/-5 %. Just to get a bit more margin in case you end up with resistors that are on the outskirts of the statistical distribution of resistance.
Me to...
My sim says that the regulator has 130 dB of line rejection from DC to 20 kHz and a +20 dB/dec up from there. At 1 MHz it flattens out at 90 dB and rolls off from there. This with 2 mA load current. So in practice, the line rejection is probably limited by things such as parasitic capacitances, board layout, etc.
I deliberately sized the feedback network to absorb both the current required to make the regulator work and the trickle current through D2. If the current in the feedback network doesn't exceed the current through D2, the regulator doesn't regulate properly. The output voltage rises and the line rejection goes to zero.
~Tom
Attachments
Last edited:
Thanks for posting your results.
I got something very close when I adopted the schematic in your post: 128dB from DC, dipping to 138dB at 100Hz, rising above 128dB at about 15KHz. Sound right? Can you post a schematic showing your source and if there is anything else connected to the right of C8?
Now, I know I'm being ridiculous here. But R12/D4 in the schematic is the weak point in the design for PSRR. Delete R12 and replace D4 with a 10V battery. Simulate, stand back, and catch your breath.
For something more practical, make R12 10Meg and replace D4 with two micropower 5V LT1634-5 in series. 172dB from DC up to 150Hz, rising to 128dB @ 15KHz.
Is it worth it? Well, you get a lot more low frequency rejection (theoretically), and the D2 current steals virtually nothing from the output.
It's fun to be ridiculous! Let me know what you think.
I got something very close when I adopted the schematic in your post: 128dB from DC, dipping to 138dB at 100Hz, rising above 128dB at about 15KHz. Sound right? Can you post a schematic showing your source and if there is anything else connected to the right of C8?
Now, I know I'm being ridiculous here. But R12/D4 in the schematic is the weak point in the design for PSRR. Delete R12 and replace D4 with a 10V battery. Simulate, stand back, and catch your breath.
For something more practical, make R12 10Meg and replace D4 with two micropower 5V LT1634-5 in series. 172dB from DC up to 150Hz, rising to 128dB @ 15KHz.
Is it worth it? Well, you get a lot more low frequency rejection (theoretically), and the D2 current steals virtually nothing from the output.
It's fun to be ridiculous! Let me know what you think.
Sounds about right.
Stare at attached.
You'll have to ensure that the parts would survive during startup, power-down, a momentary short circuit of the output, as well as the high di/dt pulses caused by loose connections in tube sockets, etc. Zeners are pretty rugged. ICs are not...
The other question is if better regulation is actually needed. In my 300B amp, I'm running 525 V DC with 50 Vpp ripple into the regulator and measure 20 uV of noise+ripple on the output. I don't actually know how much of that 20 uV is ripple and how much is noise. I suspect most of it is noise.
Another point is that a zener diode is 50 cents. The LT1634s available at Digikey are $10/each and you'll need two of them. The LT1634s take up more board area as well.
~Tom
Attachments
Fair enough. Although, you gotta wonder if it's real or if the line rejection of the LT3080 just isn't modeled very well. I know with opamp models, the CMRR is often not handled well as it depends in part on mismatches between the various parts of the opamp. These mismatches are, generally, not included in the model.
Also, the board layout and other parasitics come into play. Few people realize how much 130 dB really is... Never mind 170 dB...
One could increase the resistor in series with the zener. My sim says the line rejection would improve by 6 dB by increasing R2 (schematic in post 165) from 100 kohm to 270 kohm. Running that experiment would indicate whether the performance of the implemented regulator is limited by the zener leakage path or by something else. Just be careful... You don't want to starve the zener too much.
~Tom
How about these: Low Current Zeners from ON. Look at the 10 V zener Vz vs Iz (page 5). Nice...! That would allow R2 to be increased by quite a bit, hence, improving the line rejection further.
~Tom
~Tom
Can R3 and R9 be mounted off the board, say on screw terminal strips? I'd like to set up an adjustable regulated supply with the ability to deliver 120 to 250 volts at 20 to 100 ma for experimentation. Can the potentiometer be a 10-turn panel-mount type instead of a trimmer type for easier adjustment? Everything would be mounted in an enclosure, but I would like some adjustability. I could live with a small penalty in performance, but not stability.
I don't see why not... Just keep the wires as short as possible.
I would select between different R9s using a rotary switch on the ground side and have a pot for the fine tuning of the output voltage. The adjustment range on the pot is +/-5 % if you use the values in the Schematic_BOM .pdf on my website. +/-7.5 % if you use the values in my post above (#163). You could widen that by changing R4, R5, R6 or you could use a regular single-turn pot and just have more fixed resistors to choose from. Options, options.
Make sure to ground the metal case and shaft (if applicable) of the pot.
R3 is a bit trickier. You do NOT want to break the connection between the regulator IC and the cascode Q1 when the power is on. So I would implement R3 as one fixed resistor set for 20 mA and switch in resistors in parallel for the higher currents. The current limiter is a rather soft limit, so I don't think it makes sense to have more than 2-3 ranges. Maybe 20, 50, and 100 mA.
~Tom
Last edited:
A simple full-wave rectifier with a reservoir/smoothing cap will work just fine. I see no reason to build anything fancier than that.
The only thing to watch out for is that the regulator requires at least 15 V from input to output to regulate properly. I.e. the input voltage must be at least 15 V higher than the output voltage at all times. So make sure the ripple voltage doesn't cause the instantaneous input voltage to drop below Vout+15 V.
In my 300B amp, I run 380 V RMS into a full-wave rectifier and a 50 uF cap to get about 520 V. The ripple voltage is about 50 V peak-to-peak (Vpp). There's 20 uV (yes, microvolt) of ripple+noise on the output of the regulator.
If in doubt, use the Maida Calculator spreadsheet I have on my website for calculating the needed transformer voltage and reservoir cap for a given output voltage. The spreadsheet calculates the supply ripple and takes mains variation into account.
~Tom
Last edited:
Excellent point about whether it’s needed.
It’s not immediately clear to me that raising R1 with the existing zener or moving to the lower current zeners will increase attenuation (regardless of what the sims say.) Zener dynamic impedances are higher at lower currents. Of course, if line rejection improves with this experiment, you’ll know you’re on the right path. But if it doesn’t, it may be that you need another approach.
At low frequencies where it’s most important, the lower dynamic impedance of an IC may be needed to get the effect. Cheaper IC’s than the LT’s I mentioned, with lower dynamic impedances at low frequencies than zeners, are:
from TI, TO92 packaging, 1k piece pricing:
[To allay concerns about the ICs surviving, a test could be done by replacing the 10V zener with a 12V one, tacking a pair of wires across the zener, and paralleling off board a series string of regulators set to a lower voltage. If line rejection gets better and it seems worth it, the regulators could be added to the PCB. Or, an abusive test could be done first without the zener to see if it was necessary to retain it.[LM385-2.5 ($.22), with min. regulating current of 20uA (at 2.5V you’d need 4);
M404D41 4.096V, ($.32), with min. regulating current of 73uA (at 4.096V you’d need 2 or 3);[
M404D41 4.096V, ($.32), with min. regulating current of 73uA (at 4.096V you’d need 2 or 3);[
But one deficit of ICs is their rising impedance with frequency, a characteristic that zeners do not share. So what if we did a double-zener attenuator? Cut R1 in two and connect a second zener between the cut point and the LT regulator output. With a lower current zener, we’d probably get a lot more attenuation and draw less current than now.
I’m not saying that there would be a benefit to further improvement in line rejection. It may not even be possible for the reasons you suggest (parasitics, etc.) I’m just brainstorming on how it might be done. So please forgive the presumption.
I’m not saying that there would be a benefit to further improvement in line rejection. It may not even be possible for the reasons you suggest (parasitics, etc.) I’m just brainstorming on how it might be done. So please forgive the presumption.
No worries. I have quite a bit on my plate right now, so any experiments will be a ways out.
~Tom
Hello
My maida regulator is ready and working--
I use it to my Lampizator 6n2p ,output from my dac
the caps to input is 140uf-100ohm resitor-150uf..all polyprop caps
256vdc regulatet to 240vdc at the maida
There is a great lift in sound---more open ,dynamic and better bass
A wery fine regulator---Thanks Tom
best
Bjarne
My maida regulator is ready and working--
I use it to my Lampizator 6n2p ,output from my dac
the caps to input is 140uf-100ohm resitor-150uf..all polyprop caps
256vdc regulatet to 240vdc at the maida
There is a great lift in sound---more open ,dynamic and better bass
A wery fine regulator---Thanks Tom
best
Bjarne
Hi tomchr,
Finally I found something really small and effective. My system will use forced cooling (reduced rpm fan) though, so a smaller heatsink will still do the job, I guess...
Just because the ease of purchase, replacing the STW12NK95Z with STW12NK90Z version wouldn't do much difference, right?
Do you have only SMD version?
Best regards
Matej
Finally I found something really small and effective. My system will use forced cooling (reduced rpm fan) though, so a smaller heatsink will still do the job, I guess...
Just because the ease of purchase, replacing the STW12NK95Z with STW12NK90Z version wouldn't do much difference, right?
Do you have only SMD version?
Best regards
Matej
Using the STW12NK90Z should be fine.
I can't make a regulator this good using leaded-only parts. Hence, I offer a version that uses SMD components. They're pretty big by SMD standards. Many people (even first timers to SMD) have build this regulator successfully. For those who prefer to not solder SMD parts, I do offer to do it for them. I have boards available for sale that have the SMDs populated.
~Tom
I can't make a regulator this good using leaded-only parts. Hence, I offer a version that uses SMD components. They're pretty big by SMD standards. Many people (even first timers to SMD) have build this regulator successfully. For those who prefer to not solder SMD parts, I do offer to do it for them. I have boards available for sale that have the SMDs populated.
~Tom