Go Back   Home > Forums > Amplifiers > Solid State

Solid State Talk all about solid state amplification.

Please consider donating to help us continue to serve you.

Ads on/off / Custom Title / More PMs / More album space / Advanced printing & mass image saving
Reply
 
Thread Tools Search this Thread
Old 23rd March 2014, 02:45 PM   #1
peufeu is offline peufeu  France
diyAudio Member
 
Join Date: Mar 2001
Location: Lyon, France
Default Experimentations on Regulators

I need to fine-tune the power supplies for my ES9018 project.
So, let's go measure some regulators.

I'll start with the VDD supplies. This means LDOs with an output voltage of 1.2V, and a rather high output current, let's say 150mA to be safe.

Those will be powered from nearby +3.3V digital supply and decoupling capacitors, which leaves 2V of dropout, not a problem.

Power dissipation should not exceed 0.5W worst case, so SMT packages like SOT89 or SOT223 are adapted. I would rather avoid QFN or smaller if possible.

Local decoupling of this 1V2 supply will need to be strong enough to maintain a nice low impedance across the frequency range, something in the tens of mOhms should work well. A high impedance would make a noisy power supply, which can couple into nearby victims.

I will start with ceramic capacitors on the backside of the board, under the ES9018. Perhaps something bigger will be needed, maybe a POSCAP. Due to layout tightness I'd prefer to avoid big capacitors.

So, here's our LDO search criteria.

- Vin about 3.3V, Vout 1.2V, Iout 150mA
- SOT89, SOT223, or similar
- Stable with low-ESR caps at the output
- Minimize capacitor footprint
- Aim for flat and low output Z across the range
- Noise isn't really important

Some digikey search digging, I selected a few candidates.

AP7217D-12YG-13
NCP566ST12T3G
NCP694D12HT1G
NCP4687DH12T1G
RT9183-12GG
RT9166A-12GXL
TLV117112DCYT

After soldering them on little test PCBs, I can mesure them.

Click the image to open in full size.

I am using an E-MU 0202 USB soundcard. It goes to 192ksps, and the bandwidth goes up to where it should, so measurements up to 80-90 kHz are fine. The SNR on this soundcard is good below 48kHz, but when using 192ksps, the high frequencies are not so clean. Not a problem though.

The test setup is simple :

- E-Mu headphone output, through a 130R resistor and a 220F cap, drives AC current into regulator input and output. It can drive at least +/- 15mA.
- E-Mu analog input measure regulator's input and output voltage.

Then, I had to find some software to do the impedance measurements. The requirements are simple, it has to work, obviously it has to be scriptable, and not cost an arm. I found nothing, so I just wrote the stuff in Python. The code is a bit ugly but at least it does what I want.

It does a log sine sweep and records the audio. Calibration is via the usual calibration standards (short, thru, 1 ohm resistor, etc), and then it spits out impedance, frequency response, distortion harmonics, blah blah. No problem.

Test conditions for all regulators :

- Cin : 1F 0603 X7R // 10F 1206 X5R ceramic capacitors at the input
- Cout : same at the output
- Vin : power supply, 2 or 3 AA batteries in series, it has a rather flat impedance of about 1 ohm up to the upper measurement range
- Rin : possibility to add a 10 ohm resistor at the input, this makes it easier to inject a signal in the input to measure PSRR
- Rout : load resistor, 15 to 68 ohms, so DC test current is 17-80 mA
Attached Images
File Type: jpg ldo-test-fixture.jpg (29.6 KB, 711 views)
  Reply With Quote
Old 23rd March 2014, 02:52 PM   #2
peufeu is offline peufeu  France
diyAudio Member
 
Join Date: Mar 2001
Location: Lyon, France
Default First victim : LM317

I will show the same curves for all tested regulators.

Here, impedance is plotted :
- with/without 10R resistor in the supply (different color)
- for all values of load resistors (all curves of the same color).

Its dropout is a bit high, and when Rin is set to 10 ohms, lowering input voltage at the higher test current, it stops working, as shown by the uppermost blue curve.

Also, this old fart was born before the age of ceramic decoupling capacitors. It does not like modern caps with low ESR. Here, it oscillates nicely, as shown by the huge impedance peak.

It will be unstable with virtually all modern capacitors, btw. I just tested it because I had one around. Using LM317 means the caps have to be selected appropriately, as per datasheet, with high enough ESR to make it stable, and bypass caps have to be applied with caution.

So, LM317 is out.
Attached Images
File Type: png Output Impedance vs Frequency for all Rin-Rout - LM317.png (238.2 KB, 682 views)
  Reply With Quote
Old 23rd March 2014, 03:08 PM   #3
peufeu is offline peufeu  France
diyAudio Member
 
Join Date: Mar 2001
Location: Lyon, France
Default Next victim : P-type LDOs

The typical LDO has an internal architecture like this :

Click the image to open in full size.

To achieve low dropout, the pass element must be either a PMOS or a PNP transistor, which acts more or less as a current source, driven by the error amplifier.

Extremely low dropouts can be achieved (in the 10s of mV for some LDOs). PMOS also allow extremely low ground currents.

However, the output, being a drain or a collector, is high impedance before feedback is applied. The output impedance therefore depends entirely on feedback.

This means we can expect it to rise at high frequencies... but how much ?

Here are 4 different LDOs measured, same as above, at various input voltages and load currents.

Some of them have impedance varying with input voltage and load, others not, but all have the same characteristic : impedance is very low at LF but starts to rises at not so high frequencies.

At 100 kHz all of them don't look so good.

The AP7217 impedance rises enough to show the 10F output cap taking over at about 60 kHz.
  Reply With Quote
Old 23rd March 2014, 03:32 PM   #4
peufeu is offline peufeu  France
diyAudio Member
 
Join Date: Mar 2001
Location: Lyon, France
However, not all PMOS LDOs are slow.

Those two by Richtek have a bit higher dropout that the others, therefore I do not show the curves with 10R resistor in the power supply, which reduces the input voltage too much. They're perfectly happy at 3V input, so no problem.

Their impedance is just flat. At low frequencies, it is higher than the previous bunch, but at high frequencies, those two start to win.

Especially interesting is the second one which is ruler flat at 30-40 mOhm from DC to 80 kHz. Nice.

EDIT: fireworks: I added the forgotten pics !...

Last edited by peufeu; 23rd March 2014 at 03:55 PM.
  Reply With Quote
Old 23rd March 2014, 03:43 PM   #5
diyAudio Member
 
Join Date: Feb 2010
Quote:
Originally Posted by peufeu View Post
However, not all PMOS LDOs are slow.

Those two by Richtek have a bit higher dropout that the others, therefore I do not show the curves with 10R resistor in the power supply, which reduces the input voltage too much. They're perfectly happy at 3V input, so no problem.

Their impedance is just flat. At low frequencies, it is higher than the previous bunch, but at high frequencies, those two start to win.

Especially interesting is the second one which is ruler flat at 30-40 mOhm from DC to 80 kHz. Nice.
Can you post the graphs for those two also ?
  Reply With Quote
Old 23rd March 2014, 03:54 PM   #6
peufeu is offline peufeu  France
diyAudio Member
 
Join Date: Mar 2001
Location: Lyon, France
And finally, we have an outsider, NCP566.

What is inside isn't known, the datasheet mentions that it has an "output stage". Well, yeah. The usual graph of Dropout voltage versus current is missing. There must be something super secret in there...

This one seems to require a minimum load current. The soundcard volume is set to +/- 12mA peak, and the 68R load resistor draws 17.6mA... in this case the regulator switches in and out of some sort of powersaving mode, or just switches on and off (its output is distorted).

The 33 ohms resistor makes it happy. It could probably work with less.

Anyway, once this minimum current is satisfied, the performance is just incredible. 3 mOhm at 10kHz, 6 mOhms at 70 kHz, I had to check if the output was short circuited, but it isn't...

The datasheet just understates "The fast loop response and low dropout voltage make this regulator ideal for applications where low voltage and good
load transient response are important." and "Ultra Fast Transient Response (1.0 s)"

But all other datasheets have words like "ultra fast"... this one has impressive datasheet transient response plots though.

It has low noise, too.

To be continued...
  Reply With Quote
Old 23rd March 2014, 04:10 PM   #7
diyAudio Member
 
Join Date: May 2011
Location: Silicon Valley
NCP566 max dropout voltage = 1.3V supports the hypothesis of a Nchannel MOSFET source follower driving the output pin.
  Reply With Quote
Old 23rd March 2014, 06:09 PM   #8
peufeu is offline peufeu  France
diyAudio Member
 
Join Date: Mar 2001
Location: Lyon, France
That could explain the excellent performance !

I have added :
- 470uF Panasonic LF ultra-low-ESR polymer capacitor at the output
- large 2700uF Nichicon PW low-impedance electrolytic at the niput

The LF's are really nice caps. I have used some at the output of a switching regulator, a tiny 220uF cap eats 2 amps of ripple, doesn't heat at all, it doesn't care, and the output is pretty clean.

Anyway. More curves.

The dashed lines are previous measurements, with only ceramic caps. The full lines are measurements with capacitors. Same color, same regulator.

Left plot : the slow ones. Here we got some problems. Three are definitely unstable. The TLV1171 looks ok, except, well, it would need an enormous capacitor to have a flat impedance curve.

So, I will no longer consider these regulators.

Right plot : the fast ones. Very different picture...

- The two Richtek regs deliver a smooth transition from regulator to capacitor between 10-20kHz.
- The NCP566 desn't seem to care.

Those three are quite spectacular...
  Reply With Quote
Old 23rd March 2014, 06:30 PM   #9
peufeu is offline peufeu  France
diyAudio Member
 
Join Date: Mar 2001
Location: Lyon, France
Next, I'll look a bit further to the right on the frequency scale... tomorrow.
Attached Images
File Type: png vna.png (17.3 KB, 78 views)
  Reply With Quote
Old 24th March 2014, 11:33 AM   #10
peufeu is offline peufeu  France
diyAudio Member
 
Join Date: Mar 2001
Location: Lyon, France
Stitching together the soundcard and network analyzer data provides a whooping frequency range of 5 Hz to 60 MHz.

Of course, above a few MHz, a test with a real layout and appropriate ceramic capacitors would be needed, so please don't look too much to the right.

Here, for each of the 3 winninig regulators, 4 tests are performed.

Black trace :
Cin = 1 X75 // 10 X5R // 2700 low-ESR
Cout = 1 X75 // 10 X5R

Red trace :
Cin = same Cin as Black, plus 470F LF polymer
Cout = 1 X75 // 10 X5R

The two Ricktek regs show no difference between those two. Thus, the ceramics at the input are sufficient.

For the NCP566, the adding a local polymer cap at the input seems to cause some problems. I would have liked to look at the waveforms, but my scope died... However, with more output decoupling, the problems disappear.

With only ceramic caps, its output impedance is still very low, 10mOhms at 200kHz, impressive !

Blue trace :
Cin = same Cin as Red
Cout = 1 X75 // 10 X5R // 100 Nichicon R7 polymer

Cyan trace :
Cin = same Cin as Red
Cout = 1 X75 // 10 X5R // 470 Panasonic LF polymer

Blue/Cyan comparison :

Here the difference is the output capacitor, 100 versus 470. Both are high performance polymer caps, with ridiculously low ESR and excellent HF performance.

NCP566 seems to like the 470u cap better. It provides a smooth transition from regulator to capacitor, whereas the 100 cap creates a small peak.

This is due to the output impedance of the regulator becoming inductive at HF. It behaves like an inductor having an extremely low ESR (1-2 mOhms). So, in parallel with a cap, it can resonate depending on the value and the ESR/ESL of the cap. The same thing happens when parallelling capacitors. 470F seems to be the sweet spot.

For the two richteks, no such problem, since their output impedance is higher, they behave like an inductor with more ESR, which dampens the resonance. The caps lower the impedance above 10-100kHz.

Capacitor choices :

For the NCP566 I would stay with the 470u LF cap which provides a nice transition. However its ESR is extremely low, which means it creates a not so good looking antiresonance peak with the 10 ceramic. This means more low-ESR caps would be needed to fill the holes, maybe around 50-100F, or more large ceramics. More tests needed.

The Richtek regs would be better associated with a capacitor having higher ESR, perhaps 15-20 mOhms, instead of the 4-7 mOhms of the LF/R7. This would flatten the impedance curve and soften the antiresonance peak with the ceramic caps. Thus, less capacitors would be needed in total.


So, these regs offer two design options.

The NCP design target could be :
- 2mOhms up to 10k
- < 5 mOhms up to 100k
- < 10 mOhms up to HF
This option would use many capacitors.

The Richtek design target could be :
- flat 20 mOhms up to HF
This option would use less capacitors than the NCP.

Decisions, decisions...
  Reply With Quote

Reply


Hide this!Advertise here!
Thread Tools Search this Thread
Search this Thread:

Advanced Search

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off


Similar Threads
Thread Thread Starter Forum Replies Last Post
Series Pass Regulators driven by 3-terminal regulators SGK Power Supplies 8 13th March 2014 04:10 PM
What about these regulators? albamaiuscola Power Supplies 4 23rd March 2012 07:42 PM
+/- with pos and neg regulators vs 2 pos regulators mr.duck Power Supplies 8 12th October 2006 04:27 AM
Regulators carlmart Solid State 28 11th March 2003 01:55 AM


New To Site? Need Help?

All times are GMT. The time now is 07:40 PM.


vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright 1999-2014 diyAudio

Content Relevant URLs by vBSEO 3.3.2