• These commercial threads are for private transactions. diyAudio.com provides these forums for the convenience of our members, but makes no warranty nor assumes any responsibility. We do not vet any members, use of this facility is at your own risk. Customers can post any issues in those threads as long as it is done in a civil manner. All diyAudio rules about conduct apply and will be enforced.

21st Century Maida Regulator

Member
Joined 2009
Paid Member
If you want the ultimate in ripple rejection and ease of use, this regulator is for you.

Key features:
  • True floating regulator design
  • Phenomenal ripple rejection (20 uV output ripple in my setup!)
  • Soft start
  • Stable with capacitive load
  • No need for expensive and bulky high-power resistors
  • 2x2 inch (50x50 mm) board footprint

It's now 32 years since Mike Maida authored National Semiconductor Linear Brief 47, describing a high voltage regulator based on the LM317. Lots have happened since then. National Semiconductor was acquired by Texas Instruments for one... And semiconductors have improved tremendously since the 1970'ies. So given that it's been 32 years to the month since LB-47 was published, I figured I'd do an update of Mike Maida's original regulator.

The main drawback of the original Maida regulator is that it requires at least 5 mA (10 mA worst case) to flow in the regulator for it to regulate properly. Typically, this current flows in the feedback network. For lower output voltages, this is no big deal. But for higher output voltages - such as the ones typically used in tube circuits - the power dissipated in the feedback network becomes quite significant, necessitating the use of 5~10 W rated resistors.
In addition, implementing soft start on the original Maida regulator is actually a bit of a challenge as it requires the use of high-voltage PNP or PMOS devices. These are becoming increasingly hard to source.
Modern voltage regulators also have much lower drop-out voltages than the LM317, hence, less power is dissipated in the regulator. As a result, the only heatsink needed is for the cascode device.

My "21st Century Maida Regulator" is based on the same topology as the original Maida Regulator; a low voltage regulator with a cascode in front to drop the voltage. I chose the LT3080 as it has a low drop-out voltage and needs only 300 uA (typ; 500 uA worst case) to operate. As described above, this minimizes the amount of power dissipated in the feedback network. Hence, only 2~3 W rated resistor types are needed.
The LT3080 is a low dropout regulator and only needs 1.4 V (worst case) across it to regulate. This minimizes the power dissipated in the LT3080. It doesn't even need a heat sink.
For the cascode I use a beefy NMOS - STW12NK95 (10 A, 950 V). I've used these in my other regulators and they work well. They're also capable of surviving the conditions present at regulator start-up without running into SOA limits.

My prototype regulator was adjusted to 420 V out @ 200 mA. There is no measurable ripple on its output. With 16 V RMS (50 Vpp) ripple in, I measure 20 uV (yes, micro volts) RMS of ripple and noise on the output of the regulator. Attached pictures show the transient response as function of load current and load capacitance. It looks rock solid to me...
The start-up time comes in at about 10 seconds. This does, however, require a resistive load. Without load, the start-up time is about one second as the output capacitor is charged through zener diode D2. The start-up is smooth without tendency to overshoot.

Using the values in the schematic, I only get about 1 mA running in the feedback network. In order for the LT3080 to regulate properly, at least 300 uA must flow in the LT3080. Hence, with Iout = 0 A, the current flowing in zener diode D2 must not exceed 700 uA. With R1 = 68 kOhm, I get 700 uA when Vin-Vout > 48 V. This isn't enough to guarantee reliable start-up across worst case mains variation, hence, the 330 kOhm "minimum load" seen across the output terminals in the image of the regulator prototype. I'll probably end up burning 1.5~2 mA in the feedback path to ensure that the regulator will start up and regulate properly. I'll also increase R1 from the value in the schematic.

After running for a few hours feeding 200~210 mA into my 300B amplifier, the LT3080 has reached 35 deg C. Clearly, no heat sink is needed for the LT3080 - the cascode still needs a heat sink, obviously. I have no audible hum in the speakers and the amp is dead quiet. I like it....

I plan to offer boards for sale on my website. I haven't done the cost calculations yet, but I figure I'll land around $10.

~Tom
 

Attachments

  • IMG_8576_LT3080_0uF_200mA_14mAstep.JPG
    IMG_8576_LT3080_0uF_200mA_14mAstep.JPG
    94.7 KB · Views: 12,182
  • IMG_8577_LT3080_2uF_200mA_14mAstep.JPG
    IMG_8577_LT3080_2uF_200mA_14mAstep.JPG
    93.6 KB · Views: 11,674
  • IMG_8578_LT3080_10uF_200mA_14mAstep.JPG
    IMG_8578_LT3080_10uF_200mA_14mAstep.JPG
    86.3 KB · Views: 11,391
  • IMG_8579_LT3080_47uF_200mA_14mAstep.JPG
    IMG_8579_LT3080_47uF_200mA_14mAstep.JPG
    82.9 KB · Views: 11,237
  • IMG_8583_LT3080_0uF_14mA_14mAstep.JPG
    IMG_8583_LT3080_0uF_14mA_14mAstep.JPG
    82.6 KB · Views: 11,016
  • IMG_8584_LT3080_2uF_14mA_14mAstep.JPG
    IMG_8584_LT3080_2uF_14mA_14mAstep.JPG
    80.6 KB · Views: 1,263
  • IMG_8586_LT3080_47uF_14mA_14mAstep.JPG
    IMG_8586_LT3080_47uF_14mA_14mAstep.JPG
    82.4 KB · Views: 2,432
  • MaidaReg_1p0_Schematic.png
    MaidaReg_1p0_Schematic.png
    48.8 KB · Views: 9,729
  • MaidaReg_1p0_Prototype.jpg
    MaidaReg_1p0_Prototype.jpg
    142.9 KB · Views: 9,074
  • IMG_8585_LT3080_10uF_14mA_14mAstep.JPG
    IMG_8585_LT3080_10uF_14mA_14mAstep.JPG
    85.3 KB · Views: 4,534
Last edited:
Hmm, decent regulators are hard to come by ;)

Surely you don't know what the ripple will be until you play music through the amp it is connected to?

Compared to music signal ripple surely mains ripple is almost irrelevant?

What is the dynamic impedance (in the audio band 20-20kHz) of the output?

ETA: What was the Vin for your testing at 420V 0.2A output?
The board looks very neat BTW.
 
Last edited:
Member
Joined 2009
Paid Member
Surely you don't know what the ripple will be until you play music through the amp it is connected to?

Compared to music signal ripple surely mains ripple is almost irrelevant?

What is the dynamic impedance (in the audio band 20-20kHz) of the output?

I think you're intermingling questions here.... :)

The regulator needs to provide good attenuation of mains ripple AND low output impedance.

Mains ripple: The ripple attenuation is important even (or perhaps especially) without music playing as any ripple on the regulator output will result in hum on the amplifier output. With music (or tones) playing, it may also result in IMD. Those who have followed my threads over the past couple of years will know that I've tried many regulator topologies and as a result have had the opportunity to listen to my amp with varying levels of ripple on B+. I find that even with my relatively inefficient speakers (87 dB @ 1W, 1m) I can hear the hum from my listening position if the ripple on B+ exceeds 2~3 mV. Ripple below 1 mV is acceptable and below 2~300 uV not audible in my setup.

Output impedance: The output impedance can be viewed as a measure of how much voltage sag should be expected for a given change in load current. This is relevant when the amp is reproducing music through speakers. For accurate reproduction, the supply impedance should be as low as possible.
I have not measured the output impedance yet. It's on my list... Also, getting a setup going that can measure the output impedance as function of frequency is on my list as well. I've looked at the setups used by several people, but question if they will actually provide accurate results.

ETA: What was the Vin for your testing at 420V 0.2A output?

About 475 V (with 5 Vpp ripple). For the test with 50 Vpp ripple in, I reduced the supply capacitance (external to the board) by a factor of 10 and raised the input voltage to a bit over 500 V to ensure that the regulator had enough drop-out voltage to regulate properly.

I measured the drop-out voltage to be about 15 V. I'd probably use 25 V as a minimum in my designs.

The board looks very neat BTW.

Thanks. I took my time placing the components. It came together pretty nicely.

~Tom
 
Last edited:
Member
Joined 2009
Paid Member
I like it! So it reduces 50v pp ripple to 20uV? At 200mA/420V output?
Looks like bye bye for chokes....

That's how I see it...

I don't think I've actually found the actual ripple rejection yet. Even with 50 Vpp ripple in, the measurement accuracy is limited by that of my 6.5 digit AC voltmeter... It may just be the noise floor of the regulator I'm measuring. The voltmeter has about 1 MHz bandwidth, so 20 uV is quite good, actually.
At 20 mV/div (highest sensitivity with a 10x probe) on my oscilloscope, I get a straight line when measuring the AC voltage.
I think I'll try to measure it with my HP3581A wave analyzer (frequency selective voltmeter).

I don't see the 10u and 100uF electrolitics on the board?

There aren't any. Anything labeled "load" in my charts is applied externally. I.e. the Cload = 47 uF, for example means that I attached a 47 uF capacitor to the output of the regulator. I'm just verifying that the regulator can tolerate it if someone decides to add more capacitance to its output, that's all.

~Tom
 
Last edited:
I believe Paul was refering to C1b and C3 in your schematic.


Originally Posted by pauldune
I don't see the 10u and 100uF electrolitics on the board?


There aren't any. Anything labeled "load" in my charts is applied externally. I.e. the Cload = 47 uF, for example means that I attached a 47 uF capacitor to the output of the regulator. I'm just verifying that the regulator can tolerate it if someone decides to add more capacitance to its output, that's all.

~Tom
 
I believe Paul was refering to C1b and C3 in your schematic.

Oh, my bad... I get tunnel vision some times... :)

C1 allows for the use of one 10 uF, 450 V electrolytic can for input voltages of 400 V or below OR one 1 uF, 875 V polypropylene cap for input voltages up to 800-ish V. I'm running over 400 V in, hence, have C1a populated. The footprint for C1b is actually covered by C1a to avoid confusion - or perhaps confuse everybody... ;)

The 100 uF in the schematic (C3) does not need to be a high voltage cap, as it only "sees" 1.0 V in the circuit. In my prototype, I actually used a 10 uF, 25 V tantalum cap (near the edge of the board, by the regulator) as I didn't want excessive start-up time. With a 10 uF cap, the start-up time is supposed to be 10 seconds; a 100 uF cap extends this to 100 seconds. Sorry for not being clear on that.

~Tom
 
before you sell it, you'd best run a phase-gain plot to make sure it's stable -- this is, after, an LDO

I do agree that a gain/phase plot with phase margin, gain margin indicated would give me a warm and fuzzy feeling inside. However, the same information can be inferred from the transient response and output impedance plots.

Do you have any suggestions for how to take the gain/phase measurement on this regulator? I have the gear...

My idea is to use a small common mode transformer to inject a signal into the feedback loop and measure Vout. Do you have better suggestions?

~Tom
 
Tom,
I'm a little bit worried about the buffer cap behind the regulator. Its an important part of the ac current loop, its the ac/signal return path to ground. If its not large enough, I expect a part of the signal current wil find its way through the regulator itself instead of the cap.
I dont have it really clear in my mind how the signal current loop goes with a regulator.

I read about maida regulators, that they dont do well with capacitive loads, but would this particular design work with say 10uF? I think that would be enough.

Or am I completely mistaken?

Greetings, Paul
 
I'm a little bit worried about the buffer cap behind the regulator.

I don't know which cap you mean. Which part is it on the schematic?

I read about maida regulators, that they dont do well with capacitive loads, but would this particular design work with say 10uF? I think that would be enough.

I have had great trouble with my original Maida regulator (based on National Semiconductor LB-47 using an LM317) driving a capacitive load. That started my quest for a better regulator to begin with. With this 21st Century Maida Regulator, I have loaded it with up to 47 uF on its output without start-up issues as long as a resistive load was present. I need to ensure that the same is the case when starting up into a purely capacitive load.

My goal with this regulator is to get the best ripple rejection I can get. I would prefer that the regulator is also capable of starting up into a capacitive load (as you say, at least 10 uF would be nice). Ripple rejection is a "MUST HAVE". Start-up into a highly capacitive load is a "nice to have" in my mind.

~Tom
 
Last edited:
I meant c3, 2,2uF in your schematic. I' d like it to be a bit bigger, thats the 10uF I was talking about.
Do i understand you correctly, as long as a resistive load is present, that would be ok?
How does this translate in a reallife situation, where for example a typical 300b se amplifier is powered through your regulator? During warmup the regulator sees only the buffer cap, because the tubes are not conducting. Or is this why you have a slowstart build-in?
 
I meant c3, 2,2uF in your schematic. I' d like it to be a bit bigger, thats the 10uF I was talking about.

That makes sense. I think you mean C2 = 2.2 uF, though.

C2 is a critical component in the design. In order to ensure stability of the regulator, it must have low ESR - preferably well below 100 mOhm. This drove me to choose a polypropylene film cap. 2.2 uF was an acceptable trade-off between size, price, and performance.

I would like to be able to add a 10 uF electrolytic cap in parallel with C2. But I need to make sure the regulator can start up with such a cap in place. Stay tuned.

I don't think I would include a footprint for the electrolytic cap as the regulator actually works just fine without it. I'd rather suggest that people place a 10 uF electrolytic cap by the B+ connection in the circuit that's being powered by the regulator. This should result in the best performance.

Do i understand you correctly, as long as a resistive load is present, that would be ok?

That's the case presently. However, I'm working to characterize this fully. Stay tuned. The goal here is to have a regulator that can start up with at least 10 uF added on the output (in parallel with C2) without requiring any additional resistive load. 47 uF would be even better.

How does this translate in a reallife situation, where for example a typical 300b se amplifier is powered through your regulator? During warmup the regulator sees only the buffer cap, because the tubes are not conducting. Or is this why you have a slowstart build-in?

The slow start serves many purposes. Graceful start-up into a capacitive load being one of them. It is also required in order for the cascode device (Q1) to survive start-up.

I would model a typical 300B amplifier during start-up as an open circuit. This is why I would like the regulator to be able to start up into a purely capacitive load. My application happens to be a 300B amplifier, so that'll be my test bench (in addition to lab testing, of course).

Right now, I need to do more characterization before I can commit to start-up into a purely capacitive load.
However, I had no issues with startup into 47 uF in parallel with a 30 kOhm resistor (Iload = 14 mA, Cload = 47 uF).

~Tom
 
Last edited:
Tom,

I have been experimenting with solid state regulators for use in tube amps for some time. I want to make one that's suitable for experimentors which imposes a new set of constraints. You need an adjustable output over a fairly wide range, and it needs to be stable for all loads including a short.

The one that I could never make right with the Maida design is the ability to handle an accidental (or in the Tubelab sense deliberate) short circuit on the output when the regulator is powered up and operational. Have you tried this. If not, and you do try it, shield yourself from mosfet schrapnel. The Fuji's that I have been using explode violently. The mosfet Maida with a few well placed zeners will start up into a short OK, but it self destructs when shorted
 
My goal with this regulator is to get the best ripple rejection I can get.
This is a fascinating topic -- I think that what you have to do is craft a board on which you can substitute a number of devices.

My guess is that below -80dB ripple rejection is going to become less relevant than Zout. (It's a somewhat educated guess based upon another project I've been working on). The corollary is that the amplifier in the regulator has to be pretty quick, high bandwidth. I would suggest looking at some of the fast transient response regulators from LLTC and not overly damping the device for low noise. Some of the LLTC regs allow for remote load sensing and this is a MAJOR IMPROVEMENT!