Author Topic: Sziklai pair for voltage regulation?  (Read 4047 times)

0 Members and 1 Guest are viewing this topic.

Offline ElectronSurfTopic starter

  • Regular Contributor
  • *
  • Posts: 104
  • Country: aq
Sziklai pair for voltage regulation?
« on: April 21, 2020, 02:18:45 pm »
While studying different lab power supply schematics, I noticed that every one of them used a darlington configuration.

In this article the author mentions that the Sziklai pair has lower voltage drop, better thermal stability and better linearity compared to darlington pair.
https://sound-au.com/articles/cmpd-vs-darl.htm

I know it's an old discussion comparing this two configurations but the discussion was usually for audio related circuits.

Is the Sziklai pair performs better than darlington pair for voltage regulation in lab PSUs and if yes why it's not used widely?
« Last Edit: April 21, 2020, 07:21:58 pm by ElectronSurf »
 

Online langwadt

  • Super Contributor
  • ***
  • Posts: 4857
  • Country: dk
Re: Sziklai pair for voltage regulation?
« Reply #1 on: April 21, 2020, 02:27:08 pm »
a Sziklai pair with a big NPN acts like a PNP, with that comes all the stability issues of an LDO
 
The following users thanked this post: ElectronSurf

Offline ElectronSurfTopic starter

  • Regular Contributor
  • *
  • Posts: 104
  • Country: aq
Re: Sziklai pair for voltage regulation?
« Reply #2 on: April 21, 2020, 02:31:36 pm »
a Sziklai pair with a big NPN acts like a PNP, with that comes all the stability issues of an LDO

How about with an NPN driver and big PNP?
 

Offline fcb

  • Super Contributor
  • ***
  • Posts: 2135
  • Country: gb
  • Test instrument designer/G1YWC
    • Electron Plus
Re: Sziklai pair for voltage regulation?
« Reply #3 on: April 21, 2020, 02:33:18 pm »
The output pass transistor arrangement of a typical 'lab supply' is within a main feedback loop - so the Sziklai benefit of less variable voltage drop, thermals, linearity is largely irrelevant.

The effective conversion from NPN to PNP could be useful if you want to use the same power devices in a split-rail supply.  Also the single Vbe drop would help if you are designing a tracking post-regulator.
https://electron.plus Power Analysers, VI Signature Testers, Voltage References, Picoammeters, Curve Tracers.
 
The following users thanked this post: ElectronSurf

Offline ElectronSurfTopic starter

  • Regular Contributor
  • *
  • Posts: 104
  • Country: aq
Re: Sziklai pair for voltage regulation?
« Reply #4 on: April 21, 2020, 02:37:50 pm »
The output pass transistor arrangement of a typical 'lab supply' is within a main feedback loop - so the Sziklai benefit of less variable voltage drop, thermals, linearity is largely irrelevant.

The effective conversion from NPN to PNP could be useful if you want to use the same power devices in a split-rail supply.  Also the single Vbe drop would help if you are designing a tracking post-regulator.

Less voltage drop and better thermal characteristics and still designers choose to use darlington configuration, why?
 

Offline xavier60

  • Super Contributor
  • ***
  • Posts: 3034
  • Country: au
Re: Sziklai pair for voltage regulation?
« Reply #5 on: April 21, 2020, 03:25:27 pm »
The output pass transistor arrangement of a typical 'lab supply' is within a main feedback loop - so the Sziklai benefit of less variable voltage drop, thermals, linearity is largely irrelevant.

The effective conversion from NPN to PNP could be useful if you want to use the same power devices in a split-rail supply.  Also the single Vbe drop would help if you are designing a tracking post-regulator.

Less voltage drop and better thermal characteristics and still designers choose to use darlington configuration, why?
To complicate things a bit more, designs of the ever so popular Harrison type topology use Darlingtons almost exclusively. Because it's driven as common Emitter, it has only one Vbe drop. Although I cant easily explain why, from my experience, Darlingtons seem to have faster response.  Sziklais used in certain situations are said to be prone to self oscillation.

I use a Sziklai in my bench supply to great advantage, It is easy to drive from the lowside via a level shifter.
https://www.eevblog.com/forum/projects/linear-lab-power-supply/msg2388873/#msg2388873
HP 54645A dso, Fluke 87V dmm,  Agilent U8002A psu,  FY6600 function gen,  Brymen BM857S, HAKKO FM-204, New! HAKKO FX-971.
 
The following users thanked this post: edavid, ElectronSurf

Offline ElectronSurfTopic starter

  • Regular Contributor
  • *
  • Posts: 104
  • Country: aq
Re: Sziklai pair for voltage regulation?
« Reply #6 on: April 21, 2020, 03:34:54 pm »
To complicate things a bit more, designs of the ever so popular Harrison type topology use Darlingtons almost exclusively. Because it's driven as common Emitter, it has only one Vbe drop. Although I cant easily explain why, from my experience, Darlingtons seem to have faster response.  Sziklais used in certain situations are said to be prone to self oscillation.

I use a Sziklai in my bench supply to great advantage, It is easy to drive from the lowside via a level shifter.
https://www.eevblog.com/forum/projects/linear-lab-power-supply/msg2388873/#msg2388873
That's great, do you have an scope to share the performance of your PSU please?
 

Offline xavier60

  • Super Contributor
  • ***
  • Posts: 3034
  • Country: au
Re: Sziklai pair for voltage regulation?
« Reply #7 on: April 21, 2020, 03:51:04 pm »
To complicate things a bit more, designs of the ever so popular Harrison type topology use Darlingtons almost exclusively. Because it's driven as common Emitter, it has only one Vbe drop. Although I cant easily explain why, from my experience, Darlingtons seem to have faster response.  Sziklais used in certain situations are said to be prone to self oscillation.

I use a Sziklai in my bench supply to great advantage, It is easy to drive from the lowside via a level shifter.
https://www.eevblog.com/forum/projects/linear-lab-power-supply/msg2388873/#msg2388873
That's great, do you have an scope to share the performance of your PSU please?
There are various test results scattered through a few threads, I just use load transient tests to gauge stability, this is an early one,
https://www.eevblog.com/forum/projects/linear-lab-power-supply/msg2092534/#msg2092534
It's presently compensated for 10µs.
Although a Harrison design can go much faster, It's not necessary for my uses.
This is the output stage's linearity, totally unexpected.
https://www.eevblog.com/forum/projects/linear-lab-power-supply/msg2274876/#msg2274876
« Last Edit: April 21, 2020, 03:57:03 pm by xavier60 »
HP 54645A dso, Fluke 87V dmm,  Agilent U8002A psu,  FY6600 function gen,  Brymen BM857S, HAKKO FM-204, New! HAKKO FX-971.
 
The following users thanked this post: ElectronSurf

Offline xavier60

  • Super Contributor
  • ***
  • Posts: 3034
  • Country: au
Re: Sziklai pair for voltage regulation?
« Reply #8 on: April 21, 2020, 04:13:02 pm »
This post shows the 10µs response and also a test while loaded with a 470µF poly cap.

https://www.eevblog.com/forum/beginners/lm324-power-supply-with-variable-voltage-and-current/msg2405490/#msg2405490
HP 54645A dso, Fluke 87V dmm,  Agilent U8002A psu,  FY6600 function gen,  Brymen BM857S, HAKKO FM-204, New! HAKKO FX-971.
 
The following users thanked this post: ElectronSurf

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 17428
  • Country: us
  • DavidH
Re: Sziklai pair for voltage regulation?
« Reply #9 on: April 21, 2020, 05:03:21 pm »
While studying different lab power supply schematics, I noticed that every one of them used a darlington configuration.

The Tektronix PS500 series of power supplies use Sziklai pairs as shown below.  I am not sure why they did not use the Darlington configuration.

Quote
Is the Sziklai pair performs better than darlington pair for voltage regulation in lab PSUs and if yes why it's not used widely?

A power supply depends on the high gain error amplifier for voltage regulation so the added benefit of of a Sziklai pair over a Darlington is irrelevant.

Sziklai pairs require more attention to prevent oscillation.

a Sziklai pair with a big NPN acts like a PNP, with that comes all the stability issues of an LDO

How about with an NPN driver and big PNP?

In the past it was common to use a Sziklai pair to make a PNP power transistor out of an NPN because PNP power transistors were lower performance or more expensive.  This shows up in quasi complementary audio power amplifiers where the class-AB or B output stage uses power NPNs for both devices.  If you go further back, quasi complementary designs used germanium PNP output devices because unlike silicon, germanium worked better with PNP.

Although I cant easily explain why, from my experience, Darlingtons seem to have faster response.

I think Miller feedback from the collector to base explains lower performance but the overall frequency compensation is dominated by other factors like the open loop output impedance and output load.

Quote
Sziklais used in certain situations are said to be prone to self oscillation.

They are more prone to local oscillation compared to a Darlington and require some extra attention in this area.
« Last Edit: April 21, 2020, 05:14:18 pm by David Hess »
 
The following users thanked this post: ElectronSurf

Offline duak

  • Super Contributor
  • ***
  • Posts: 1048
  • Country: ca
Re: Sziklai pair for voltage regulation?
« Reply #10 on: April 21, 2020, 09:48:05 pm »
Back in school I built a few amplifiers based on Dan Meyers' Tiger amplifiers that used a Sziklai based output stage with a voltage gain of 2.  I built one that just didn't sound right.  It had a weird high frequency harshness like that from crossover distortion from a too low idle current.  It turned out to be a bad 100n cap in the Zobel network.  I didn't have a scope so I couldn't see if it was oscillating but I suspect it was.  In addition to a Zobel network, solid state power amplifiers almost always have a small inductance damped by a parallel resistor in series with the output.  I don't believe I've ever seen a similar network used in power supplies.  There are a number of differences between power supplies and power amplifiers such as push-pull vs single-ended, acceptable voltage slew rate, stability into highly capacitive loads, etc.  Has anyone seen this used?  Perhaps a current sense resistor performs a similar function.

It's neither here nor there but when looking for a part for an hp supply, I found an anomaly where the PNP of a complementary pair had a higher Ft - the MJ15015 & MJ15016: https://www.onsemi.com/products/discretes-drivers/audio-transistors/mj15015

If you download the data sheet, you'll see that it also includes the 2N3055 so I suspect that the NPN MJ15015 uses the same process as the 2N3055 (or vice versa).  Maybe 2N3055 are fallouts from MJ15015.
 
The following users thanked this post: ElectronSurf

Offline Neomys Sapiens

  • Super Contributor
  • ***
  • Posts: 3268
  • Country: de
Re: Sziklai pair for voltage regulation?
« Reply #11 on: April 22, 2020, 12:17:25 am »
I have used the Sziklai pair a long time ago in a power supply not for its gain but to convert a NPN to PNP.
It was the circuit below from the NS appnotes. I reasoned that I did lot like it because large PNP would be not protected to the same degree than the regulator. So I used a PNP and multiple LM395, as they proposed in circuit #2.
The power supply was not exactly high performance, but it was very robust. It lived long with much abuse and was only dismantled for being ugly. It had some additional diodes although and also ferrites, but I don't recall those mods exactly beyond the reverse protection.
 
The following users thanked this post: ElectronSurf

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 17428
  • Country: us
  • DavidH
Re: Sziklai pair for voltage regulation?
« Reply #12 on: April 23, 2020, 02:21:04 am »
Back in school I built a few amplifiers based on Dan Meyers' Tiger amplifiers that used a Sziklai based output stage with a voltage gain of 2.  I built one that just didn't sound right.  It had a weird high frequency harshness like that from crossover distortion from a too low idle current.  It turned out to be a bad 100n cap in the Zobel network.  I didn't have a scope so I couldn't see if it was oscillating but I suspect it was.

It was either Bob Cordell or Douglas Self, or both, who did not have anything good to  say about that configuration.  It seems like a small amount of local voltage gain in the output stage should work and increase headroom but in practice it raises distortion and causes other issues.

Quote
In addition to a Zobel network, solid state power amplifiers almost always have a small inductance damped by a parallel resistor in series with the output.  I don't believe I've ever seen a similar network used in power supplies.  There are a number of differences between power supplies and power amplifiers such as push-pull vs single-ended, acceptable voltage slew rate, stability into highly capacitive loads, etc.  Has anyone seen this used?  Perhaps a current sense resistor performs a similar function.

I have seen a damped inductor used in series with the output for exactly the same reason audio amplifiers include it but a current shunt in that location would provide the same benefit.

Power supplies which require 2 or 4 quadrant operation or fast response typically use class-AB output stages.  Single quadrant power supplies may include an actively driven pull-down.
 

Offline floobydust

  • Super Contributor
  • ***
  • Posts: 7681
  • Country: ca
Re: Sziklai pair for voltage regulation?
« Reply #13 on: April 23, 2020, 04:22:09 am »
Each configuration has advantages and disadvantages that you have to consider.
The Sziklai configuration has voltage gain, whereas the Darlington does not.
I find this is a disadvantage in that the hFE of the two transistors right out of the box can vary widely and hFE droops at high collector currents - so your voltage gain varies a lot, just for the Sziklai stage. This makes stability harder to obtain with overall loop feedback, because the gain drops under load and increases at light load. Your circuit must be carefully designed and it can work fine. It's just an extra hassle.

I no longer see them used anymore in discrete audio power amplifiers. I think because changing transistor manufacturers or part numbers can cause problems. Hobbyists can measure and select the hFE for each part but that luxury is not practical in production.

Another trap of the Szilaki is the parasitic oscillation at several MHz that can show up and ruin your day. The mechanisms for this I don't exactly know, IC's have it from the low hFE lateral PNP part. You can't have a lot of flying leads (parasitic inductance) in their construction.
Input ripple modulates the (emitter) of the driver transistor which gives a bit less PSRR. A few audio amplifier authors say the configuration is slower due to the voltage gain.
So a Darlington pass transistor has poor voltage efficiency, you always lose several volts there but it's much easier knowing its voltage gain is always around 1 and your overall loop feedback amount stays well behaved.

The Tiger and SuperTiger had some stability problem, I recall a hand written note from the the designer making a few mods to fix them, after a friend went into business selling them and almost went broke because they were unstable and cooked output transistors.
 
The following users thanked this post: markce, ElectronSurf

Offline Kleinstein

  • Super Contributor
  • ***
  • Posts: 15157
  • Country: de
Re: Sziklai pair for voltage regulation?
« Reply #14 on: April 23, 2020, 06:17:58 am »
Many lab supplies use the floating regulator circuit and thus in common emitter configuration. Here there is little to no advantage for the Sizlaki configuration. It is more the problem with oscillation of the pair that requires extra measures to slow it down that is a disadvantage.

In a more emitter follower like configuration the Sizlaki configuration can have a slight advantage from a lower output impedance and some 1 V less drop out can be an advantage. With the 723 regulator the Sizlaki configuration (with the NPN from the Chip) is relatively popular.
 
The following users thanked this post: markce, Neomys Sapiens, ElectronSurf


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf