DDDAC PSU “C-L-C Choke” Tests

DDDAC PSU “C-L-C Choke” Tests

Introduction

In my post on the DDDAC Power Supply, I have already been pointing out, that using a CLC-Filter after the rectifier is an improvement over the standard RCRC-Filter. Several DDDAC DIY enthousiasts have been reporting the positive sound impact of using CLC instead of RCRC only. The Magic Power Supply from the Audio Creative People is also based on this concept. But what Choke to use? Also, here there are many opinions, none being wrong, just different routes to Rome and different tastes may be? Or budget reasons?

I already announced a few times I wanted to dig deeper into this and report back. Well, read on… As I have been working on the MK3 versions of the DDDAC17904 the past few months and also making prototypes for the accompanying Power Supply MK3 (with more SMD and electronic rectifying), I decided to use the existing test setup for a Choke review – Technically and sound wise.

The Test candidates

This was difficult and it took me a while to come to a kind of reasonable choice. So many Chokes out there? Which ones to choose? So, I decided to pick two, from which I already had positive feedback and have a third choke made to my speciation by AE Electric: A Dutch company producing transformers and chokes for the industry, but also for Audio DIY hobbyists. In the past they have already delivered me many transformers and chokes with excellent results and proven top quality!

What minimum specs did I need? As much Henry as possible? More is better, but so much that the DC resistance would not exceed like 4 Ohm and max current to > 400mA. Why? Because the new 4 deck equivalent of the DDDAC1794MK3 draws 350mA. and 4 Ohm, so max voltage drop would be like 1.5Volt, which fits nicely in the MK3 PSU design. Others are also possible, but then the PSU would need extra tweaking to accommodate the choke. To be honest I do not believe that was really needed. OK, these were the Final Choke Candidates:

Hammond 159ZC (Iron core)

  • 60mH
  • 2A
  • 0.7 Ohm
  • ~35 Euro

Lundahl LL2733 in series
(audio C-core of Lundahl own production)

  • 400mH
  • 1.25A
  • 3.4 Ohm
  • ~100 Euro

AE DDDAC Choke (Amorphe Core)

  • 1H
  • 0.4A
  • 4 Ohm
  • ~150 Euro
AE DDDAC Choke

Measurements

No DDDAC Blog Posts without measurements!!!

Choke Comparison impedance Test - Drive with 4,7V 470k =10uA in high frequencies - Hammond 159ZC - Lundahl LL2733 - AE DDDAC

I used my AP System Two to measure the impedance versus frequency. All chokes have a maximum impedance at some point. As you can see at the left these are between 10k and 50 kHz. No surprise that the chokes with a larger Inductance (Henry) have the highest impedance. In combination with the second capacitor, this will result in the real Filter function.

But… what does that mean in terms of FILTERING the rectified signal. So, I build up the CLC with 4700uF (as in the DDDAC PSU) and measured the filtering of the ripple versus frequency. As you can see at the right, a whopping >100dB for all chokes > 5kHz. The best performing is the DDDAC Choke, as it filters the most and reaches it maximum damping already at 700Hz.

Choke Comparison Filter Test with 4700uF Hammond 159ZC - Lundahl LL2733 - AE DDDAC

There is NO WAY, you could have this damping with RC filters only… Just saying….

OK, this only means that the more Henry, the more filtering, as predicted. But what is the sonic effect? For all, in audiophile communities the Amorphe Core is normally seen as the best way to build transformers and chokes. I needed a Test Setup to make real A:B:C comparison…

Test Setup for the listening review

DDDAC Chokes Test 2

I strongly believe you need to be able to switch back and forward between different situations to be really able to give a judgement. I am using my Bluetooth Switchboard for that. From the listening position I can select (with my iPhone) any of the three chokes, while playing music.

And of course also the fourth one: no choke.

Listening

I listened to several different tracks and switched between the 4 situations. Four? Yes, also the “no-choke” option of course. Let’s start with that. Between “no-choke” and any of the three other chokes, the difference is so clear, I am asking myself why not standard in the PSU. Honestly, I would certainly recommend it as a standard tweak after these sessions. The sound stage is wider, deeper and there is more air between everything. My rating here is “significant” if you will… Martin Been described it more eloquently than I can in the DDDAC PSU Post:

The authority of the acoustic bass is the first what strikes me, less blur in the lower register. With that it seems that there is more room for details and space in 3D , so also more surrounding space of each instrument and when turn up the volume there is more punch.
Fresh in my mind was a concert of Kayhan Kalhor on kemençe ,Tristan Driessens ud ,Pouyan Ataie santur ,Tony Overwater violone/bass and Vinsent Planjer on percussie, The longest night, a concert in Amare -The Hague.
The kemençe is a spike fiddle with some times a harsh sound when bowed what goes right to the soul when Kayhan play that way and that happen again yesterday, goosebumps and wet eyes listening to Silent City, a elegy for Halabjah, a Koerdisch city razed by Saddam Hussein.
By then I know, if the sound is so close to reality, I’m done, finito, (for now hahaha)

The differences between the 3 chokes, I would describe as follows: Between Hammond and Lundahl the difference is not so large. Both gives a wide deep stage and improve low level detail, gaining on texture in voices and instruments. The Lundahl (thanks to larger inductance?) was stronger in the Bass improvement. more punch, more authority.

The absolute winner (but also most expensive) was the Amorphe Core AE DDDAC Choke. It was every time striking how much air in between voices and Instruments came and how strong the bass authority became. I must say the engagement with the music is clearly at is highest with this one. It becomes all very natural and analog low-level detail touch and feel.

Any metaphors from Real Life, in terms of how to rate this on a scale or so?
let me try. My wife and I drive Audi at home; hence I choose Audi… Disclaimer: BMW and Mercedes are fine cars too (my children drive these….) More disclaimers: Many cars are nice 😉

  • no choke – Volkswagen Golf or similar
  • Hammond – Audi A4 TFSI 2.0
  • Lundahl – Audi A6 / V6 / 3.0 TDI-Diesel
  • AE DDDAC – my tuned Audi A8 / V8 / 4.2L TDI-Diesel

Anyone who has been driving cars like that, know what I mean…
For whatever the above list is worth… (I am very biased of course with the A8, LOL)

conclusions

  • The DDDAC MK3 power supply will have standard connection options for a choke – Best PSU tweak possible.
  • If the wallet allows for it, go for an amorphe core.
  • More Henries are “better” – but as always there will be limits what is practical (Ohm and max current)
  • In my, still to build, DDDAC1794MK3 reference model, for sure, there will be two amorphe chokes implemented…

Feel free to share your experiences and favorite chokes in the comments!!! My preference does not mean it should be yours…

Update 04-03-2024

Based on some comments and mails I received on trying other chokes, here are my 2 cents

The only way to get it right is to simulate first and pick your first setup.
It is less about the operation point where constant current runs (after voltage is done right) It is more about swinging (high voltage peaks) around the resonant frequency at start up when the DAC is not on full current! Fc is normally round the 1 to 10 Hz.
in general, for all readers and all situations this is the rule of thumb:

  • More Henry is lower Fc and larger swings (more voltage peak at start – up: danger for Cap max voltage – use Zener)
  • Less Ohm is higher peaks at swinging (more voltage peak at start – up: danger for Cap max voltage – use Zener)
  • Lower second C is increasing the resonance frequency by a factor SQRT (C-new/C-old) and damping of noise get worse – the peaks stay the same in voltage level.

just adding chokes might not result in desired or expected results.
use simulations first and know what you are doing.

32 thoughts on “DDDAC PSU “C-L-C Choke” Tests

  1. Love this Doede and the comparison to the cars. I am using the magic PSU on my signature DDDAC and used Lundahl in another DAC and it is a great improvement. Edward is the choke king as you know

    All the best pal – Simon

  2. Hi Doede,

    Again what a impressive tweak and a substantiated explanation. Your investigation has a level that can be understand by the lesser gods among is. The explanation in the form of graphs and written text helps to investigate the options in different diy projects. Great.
    Has the comparison between the several coils been made with the new active rectifier? If not would that add an extra improvement in the SQ?

    What do you think: will a similar tweak in a tube power amplifier, result in a notable SQ improvement.

    Are there any restrictions to take care of when one starts to tweak power supplies. I mean would every power supply benefit from applying a choke.

    Looking forward to the MKIII DDDAC

    1. Hi Jan,

      Thanks for the flowers 🙂 Yes, the Test has been made with the MK3 PSU which has the electronic rectifying. But independently if you are using the “old” or “new” supply, the choke is always adding the same improvement. The electronic rectifying on itself was already an improvement over the “old” one. It just adds up.

      Tube amplifier? OH Yes!!! CLC in my Tube amplifiers beats any other way of filtering. Also, Tube rectifying is a must in tube amplifiers.

      Yes, there are restrictions if you “just start adding chokes” – every power supply will benefit when after the rectifying you will have CLC in stead of straight C or RC. BUT, the LC is creating a resonant system which must be dampened. Else you could have significant spikes in voltage at start up or larger load changes. It needs to be carefully calculated / simulated. I have been doing this for several DIY’s already and is a small thing for me, so in case you (or anyone else) need some help for a specific situation, just let me know.

      I Hope MK3 will be available in like 6 weeks or so. First batch is at the SMD company right now.

  3. I’ve used the Hammond chokes in my builds for around 10 years. great improvement at low cost. I also separate digital and analogue power with a choke on each

    1. Hi David,

      thanks for confirmation. You are 100% right. The Hammonds are good value for the buck. It is like couple capacitors. Ranges from like ten euro to hundreds of euros. Just pick what works for you (and your wallet 🙂 )

      And yes, if you look closely, you will see that in my Test Setup, also the digital side has a choke. I must say that the impact on the analog side is more prominent.

      1. Hi Doede,
        I missed the choke on the digital side. It seems like you are using your own DDDAC power supplies with a choke insert? I found best results were choke input straight to the regulators. This is particularly true for the Tent shunts, a bit less so for other shunts on the digital side.
        David

  4. Hello Doede,
    Hard to find info about chokes on the internet most of what i know is because i tried it.
    We know it does not like changes so when something changes it will restore the situation by using its energy so to say.
    Lundahl have more mH so will it automatically store more energy?
    Everybody who is so amazed by the impact a choke has in the power supply MUST try a choke input.
    Also there more mH will give better results .
    Will better core material makes it filter differently, store more energy?
    Greetings Eduard

  5. Great test Doede. Have you tried the Lundahl choke in common mode. Would that do any good in comparison to normal serial connection. I would like to use it for the old Dddac power supply with 8 layers.
    Regards John

    1. Hi John, thanks for the feedback. I did not test this, as it would require a significant change in the PSU PCB layout (ground plane cut to add the second Choke part) – So hard to tell if this would improve any further in this particular case…..

      1. Well, I think I will try it anyway by cutting the groundplane between c1 & c3. First I have to learn to model a common mode choke in Ltspice to find out the correct R for zeta 1. Regards, John

  6. Hello John,
    I asked Doede the same on that notorious American audio board. If you are using a board it will be easier set than done.
    I want to try a choke input with the big Lundahl i am going to send to him next week.
    This ” new way ” of rectification seems to offer advantages so i ask Doede if i can pay him to make a tiny board that will just have rectification and the first cap . This board then can be mounted close the choke and you could decide right there how you wanna use the choke.
    I have several LCLC supplies and when i asked another famous Dutch designer about how to connect the choke. He told me for the first choke it might be good but for the second choke he wouldn’t not introduce that much dcr in the ” return “. Of course this is just an opinion. This designer never used Lundahl so he could be wrong. Usually he gets his power transformers and chokes at AE company.
    Could well be that for analogue the conventional connection is the best and for digital the common mode is the one to use or the other way. Would be nice if the one that measures best will also give the best sound.
    Maybe ” new style ” rectification will make things less critical?
    Greetings,Eduard

    1. Hello Eduard,
      You mean using an ideal bridge or diode. I’m using one Saligny Universal bridge full wave for my AlephJ and I have another for the Dddac power supply. The advantage will be the almost neglectable loss and absence of diode switching. It certainly sounds better in the AlephJ in comparison to 1 or 2 normal bridges with snubbering or 2 Saligny’s. At this moment for the Dddac I use a low Vf FERD diode, a snubber designed with Quasimodo/Cheapmodo and a Hammond 28mH choke. I would like to change it to a Saligny and a 2733 in common mode between C1 & C3 with enough R to reach a damping ratio of 1 but for me the common mode is something unusual so that’s why I’m asking. I would put one side of the 2733 in place of the fuse and for the other side I would split the big chunk of copper between the neg of C1 & C3. Greetings, John

      1. Hi John, in the new DDDAC MK3 PSU this (LT4320 with MOSFETs) is exactly what is on the board for rectifying.
        In the new design it will be much harder to scratch away the ground connection between the two Cs though..

  7. Hello John,
    Because i use choke input i can use lower current diodes . sbyv 28 or 26
    when simulating in Psud they say it does not matter how you connect the choke. the total number of mH will be the same but there should be some difference measurable but software does not allow to simulate it like this.
    When using Lundahl you can ask them to change the airgap to the current you will have running and by doing so create some extra mH for the same price.
    it took Doede some time to start using serious chokes so i hope it won’t be long before he starts with choke input . it is has always been a real big step for me.
    greetings, Eduard

    1. I of course thought about this, but I want the PSU’s to be as universal as possible (for different loads). Input L means having to tune it with a bleeder to control voltage swings at first cap.
      But, if someone like to tune the PSU for personal use, this should be not a too big deal of course. Replacing the first R (before firs cap) with a L and adding the bleeder will do the trick. But better start with some serious simulations with PSUD for example

  8. Hello Doede,
    If there is a resistor between the bridge and the first cap it will be easy to add a choke the conventional way.. The choke i will send you have a 200mA rating with coils in series so have to put them in parallel so can only use them in the conventional way.
    The power supply board will be smaller than the previous one because more parts are mounted onto the dac boards or not?
    Will i really need serious psud simulations? Usually i will go for low ripple and gently rising of the voltage.
    Maybe i will take a serious look at the Grimm mu1 streamer when i am back. The tube preamp i have now is more suited for passive bi amping and i acquired lots of spare tubes and expensive parts for it so decided not to go for the mu 2.
    Greetings, Eduard

  9. Hello,
    The LL2771 i will send to Doede next wednesday has around 3500 mH and 2,8 ohm DCR. So considerable more mH and 0,6 ohm lower dcr.
    What i have seen so far in simulations is that going up in mH with also substantial rise in dcr sometimes will only go well if you reduce the cap value after the choke because with bigger dcr it will take longer to ” fill up ” the cap so to say.
    But sometimes higher mH chokes will need a bigger cap to make the curve doing start up a bit more regular.
    Of course this is more critical if there are no active circuits after the passive network.
    But i think it will not be that critical in general. Usually to big caps will do more harm than a to big choke.
    The famous Allen Wright used to say in his better days an input choke can never be to big and should be bigger than the power transformer.
    With around 380 mA current running including bleeder you already have a very hard time to find a choke that would be to big. My ll2771 is 4,9 kilogrammes and as far as i know not available with amorphous core.
    The LL 2733 with smaller airgap can give you a bit more mH than the amorphous choke.
    If the amorphous one will still be the winner it would be nice why better core material will give better results. Is it indeed something like a turbo charger in a car( if they still exist)
    Greetings Eduard
    p s today i got a funny private message on diyaudio.com.

    1. looking forward trying this one Ed.

      My general feedback on try (and error…)
      The only way to get it right is to simulate first and pick your first setup.
      It is less about the operation point where constant current runs (after voltage is done right) It is more about swinging (high voltage peaks) around the resonant frequency at start up when the DAC is not on full current! Fc is normally round the 1 to 10 Hz.
      in general, for all readers and all situations this is the rule of thumb:
      – More Henry is lower Fc and larger swings (more voltage peak at start – up: danger for Cap max voltage – use Zener)
      – Less Ohm is higher peaks ate swinging (more voltage peak at start – up: danger for Cap max voltage – use Zener)
      – Lower second C is increasing the resonance frequency by a factor SQRT (C-new/C-old) and damping of noise get worse – the peaks stay the same in voltage level.

      just adding chokes might not result in desired or expected results.
      use simulations first and know what you are doing.

  10. Hello,
    It is still hard to grasp.
    I know that during start up ” things are swinging” and i assume that with a capacitor input ( with a small resistor between rectifier and first capacitor) this will kind of continue because the charging will continue.
    With a choke input the charging current should remain the same once the capacitors are ” full of energy”
    . The pulsating current should be gone as long as you assure that there will always be a minimum current being drawn related to the inductance of the choke. Does this mean that with choke input supply we should only look at what is indicated in the graphics once the voltage is at steady level or are things not that simple?
    If we use RCLC and we change the second choke from 400mH 3,4 ohm to my big boy 3500 mH 2,8 ohm just a basic simulation will not give us enough information about what is going to happen? Soundwise ( especially with choke input) you will get more traction ( speaking of cars) and spacious sound when going up in mH.
    Jean Hiraga sometimes used a small resistor in his high voltage supplies just like Doede does to tame the solid state diodes. Using input chokes is just a little bit more tricky. The minimum current should be dealt with by means of a resistor.
    Doede’s idea of using a big zenerdiode was given to me by Guido Tent a few years ago.
    Probably will send the LL2771 to Doede this week after i have been told which Canadian boards he likes to try.
    The new boards from Canada i will ditch on a local website in my country. Expecting answer from Grimm and Doede which board to buy to ” convert” I2S input circuitry into spdif style coming from Grimm mu1
    Greetings,Eduard

  11. Hello,
    I have done some VERY basic psud simulations and it appears that solid state rectification, bigger caps and bigger currents at lower voltages cannot be compared to the data i usually have to enter like 600 volts, 100 ohm dcr secundairy windings, tube rectifier and 60 mA current.
    In general with tube circuits i will use input chokes 20H up to 60H.
    The latest dddac will have 350 mA of current running so you need a rather thick wire so even with a big seize choke you won’t even get close to 10H.
    I just simulated to find out about the ripple voltage. Did not do anything like a step load.
    I started with LCRC and then moved to LCLC. I have to run some more test but it looks like that with higher currents the input choke doesn’t have to be that big
    If you take a 400mH 4 ohm input choke and a 3500mH 2,8 ohm you will get a wobble during start up and after 7 seconds ripple at second cap will be 1/2 millivolt. After 7 seconds with 6000mH same dcr you will get more ripple but if you wait long enough it will get lower.
    There is an American company that use both choke input and a sophisticated solid state regulator because he says that a choke can filter things that this solid state thing cannot do.
    So the choke input does his job, could be helped by a second LC and then there will be Doede’s supply right on the board where it should be.
    I just discovered i have another LL2771 that could be tested in Germany 6 Henry 0,5A 11,2 ohm dcr because two coils are in series. Two coils in series could give better filtering in the higher frequencies i guess.
    Greetings,Eduard

  12. Hi Doede,

    Good to see you also have taken the choke route! As you know I am already using the choke power supply for more than 10 years in my dddac.
    Another tip: try a mosfet rectifier instead of diodes.
    And do yourself a favour, do not wait another 10 years to try this 😉

    Regards,

    Stefan

    1. Hi Stefan,

      Thanks for the reply and confirmation… now wait and see… in a few weeks the MK3 power supply will be introduced with the MK3 DAC. And yes…. It has electronic rectifying with MOSFETs…

      Best regards
      Doede

        1. Ah yes I see it. Very nice!
          Which mosfet did you choose? I have tested a bunch of them (selected as per the recommendations in the datasheet of the controller) and found that the sonic differences are quite big.
          My favorites are the IRLB3034.

          Regards,

          Stefan

  13. Hi Doede,

    Slightly off topic, but a few years ago I had to replace the transformer on a Sony cassette deck. Finding a suitable replacement was quite difficult because of the *leaking* Electromagnetic Fields that were being picked up by the playback head on the deck!

    I tried four transformers, settling eventually for one of two toroidal transformers – but even then they are also leaking EM fields! Eventually, I had to rotate the best performing toroidal for *minimal* interference – it worked well, but in my opinion not as good as the original. I didn’t need to put in any additional magnetic shielding.

    Kind Wishes
    Eric.

  14. Hello,
    Most transformers are made by companies that dont care to much about aspects audiophiles think are important. AND most audiophiles want as much VA possible for the money they are spending. A properly constructed 300 VA split bobbin/ double c core will sound better than a 500 VA toriodal one. Yes, but all these famous brands use these transformers so they must be good. NO, the sales department knows damn well that customers will be more impressed by a 500 VA than by a lower VA one with higher build quality. So you will get the 500 VA and 90% of the diy folks will use the same one.
    These are the same people that use gigantic supercaps and mount them at a serious distance from the ” consuming board ”. Doede kind of integrated the last stage of the power supply into the consuming board where it should be. THIS appears to be even more crucial than most people are thinking.
    We all remember reading the Tent shunt application notes or not? To get all the benefits from these tiny boards mount them where they should be next to the chip it is feeding, right at the input stage of your Hiraga power amp where the zenerdiode used to be. THAT is why Doede took his time to integrate these Tent boards at his dddac boards. Now he managed to develop an even better solution but the crucial parts are still where they MUST be. NO big supercaps with long wires acting as an aerial. Yes, but it sounds good. Maybe, but it can be much better. But we use thick cables. A thick cable is still a cable.
    But walk inside most brick and mortar hifi gear shops and you will hear what people are looking for. The ess dac chip sound lots of detail usually but no invitation to sit down and enjoy. If a designer has no turntable i will usually not trust him.
    A pity that Doede did not enter the streamer business . To much garbage for sale on diyaudio.com
    Greetings, eduard

  15. Excellent depiction of how inductance improves power supply design. Fully choke loaded performs even better than the π type filter or perhaps with a very small cap of less than 1mF.

    This old school method is as applicable to low voltage as it is for the vintage tube rectification which required choke loaded supplies. Nanocrystal material is even more effective than AM in the heavy AC environment of a PSU post rectification.

    Thank you for mapping the differences in measurements and sonics! I look forward to V III.

  16. Hi Doede,

    very interesting, I just found out this article of yours on chokes, which I am also planning to try out in my DAC and preamplifier.

    I wonder if you also have interest in trying other options, like LCL. I have not yet found reports of subjective comparisons to cLC (notice the small c instead of a capital “C” denoting a small “voltage tuning” value) to true LC/LCL, only some claims (theoretical) that the small cap in the first position lets the filtering work much like LC/LCL.

    From hearsay, in an LCL configuration, the first choke should be more critical, and where the amorphous core would have a larger advantage there. The mechanical construction of the choke seems to be of great importance in a choke-first power supply.

    Finally, I take note of the choke’s lesser effect on digital. I would still like to try it out on a clock supply too (A. Mori being used here too).

    Sebastian

  17. Thanks, Sebastian, for the suggestion of the small “c” could function as a first L… I will do some testing / simulations to get some insight on this. I never tested full LCL in this setup as it goes a bit beyond the practicality of updating a standard DIY kit…

    Doede

Leave a Reply

Your email address will not be published. Required fields are marked *