DC/DC converters: how good they are?

[Igor]

Hi to all!
The topic name says all. I can agree with use of dc/dc working on frequency over 200 khz, even in this case,
regulation and output impedance isn't as good as canonic power supply's.
For example.
I measure -88db noise rev 1V AC without any weighting filters on my Sontec power supply +/-24V done with LM317/337 and loaded 80 Ohm/5W resistors (300 ma).
I can measure in best case -67 db weighted on output of 1W 5->+/-15V $12 dc/dc converter.
Soooo....
I proposed an easy test here:

@Balijon: Sounds kind of rude?,

It was not my intention to offend anyone, if I do, I excuse.
@Igor: I have a great respect for your work, craftmanship and contributions! I hope you take my humor mildly.

On the subject:
I try to think different, as an engineer. I look at the 500/51x platform as a very cumbersome way to hook-up 2 XLR's and a power-feed. (Again no disrespect!)
Especially the exotic connector we use and the fight to support different voltages (like 24V and tubes) within the infrastructure.
I always look for simplicity, it would be a great flexibility if would just convert on the print what we need to power a specific design and use a common available (consumer-priced) widely available adapter, that is not critical in the performance. This is what Rupert Neve does in his Portico designs.
Look at your mobile phone and the adapter-nightmare we had to put up with for years! I am very happy that this has been regulated in Europe now. I am extremely happy that I do not have to rebuild my car when I switch phones...

DC-DC converters have another advantage: Classic power-lines are sewers of the audio signal, transmitting garbage from module to module, a super highway for dirt transport. The effort we put into isolation through input and output transformers is for keeping our signals clean from polluting dirt. DC-DC converters are transformer Isolated and do just this for the power-lines.
Yes, they are switching power-supplies that can introduce harmonics because they switch on a high (in-audiable) frequency. But you will notice in a bad design (same as a bad grounding design). I is a mature and affordable technology that is applied in many things that surround us and we use on a daily basis.
I think that we should be open for the merits it can bring us, if we understand how to apply it well.
I think it is too simple to only have 'religion' and tradition disregard this.

grT

He! Cannot pass it silent. Great source of amusement, aye.
Cumbersome or not, matter of personal taste.
DC/DC converters: open your own thread and discuss it. You will find same mind experienced people of course.
For the start.
Find clean dc/dc converter for less than $20/watt, running at over 200kHz. A?
Usually they use 60-100kHz. Not as easy to filter out.
BTW in case if more than 1 converter used they should be synchronized to avoid interference.
After all. Why not than use high efficiency Ge photoelements and ultrabright led's.

Classic power-lines are sewers of the audio signal, transmitting garbage from module to module
as an engineer

Yes, of course.

Do next. Use 9k+1k in attenuator to get -20 db of audio signal, than, bring back the gain with single opamp say ne5534 and 9k+1k feedback r's, 22pf for stability.
Use all possible filters and feed the power from $10 5->+/-15 dc/dc converter; measure the s/n with soundcard; than feed the power from simple +/-15V supply done on 78/7915;
measure the noise again. Than do it with analyzer with and without weighting filters. You will discover new awesome horizons to discuss. But do it separately from here please.

So, let's open the discussion.

 

[ruffrecords]

As I see it the basics are:

SMPSUs have the advantages of:

1.Good supply regulation
2. High efficiency
3. Small size

And their disadvantages are:

1. Poor load regulation
2. High output noise
3. Special layout care and possibly screening necessary to avoid EMC issues

Like everything to do with engineering they are a compromise.

Cheers

Ian

 

[ricardo]

I came across a commercial audio product with a blatant EMI problem.  When I contacted the company to kindly notify them that they should investigate the issue, they brushed me off as a quack because they didn't understand why anyone in audio would care about a problem that wasn't "audible".

Who was the manufacturer?

 

[Balijon]

Igor, very sporting from you to start this thread, thank you.

I agree that there are cost/flexibility trade-off's to be made and that quality DC/DC converters do not come for free.
Of course the current classic power-supplies do not come for free too.
So it is a good thing to create a transparent comparison.
Maybe a new level of flexibility is worth a few euro's/dollar's too.

The same applies to testing the power-supply qualities.
Certainly s/n  performance is a important factor, likely one can design to optimize this.
In computer systems s/n performance is important too, we work with very small signals at extreme frequencies over high impedance data-lines, so pollution from a power-supply is a problem there too.
I notice all the effort we do in mitigating problems with designing ground-planes/bus-systems to reduce the effect from ground problems and interference through the power-lines.
DC/DC converters create an isolated floating segment, in the ground plane and the power-lines. This is a plus, dirt from the input power-lines do not end-up in your output feeding the electronics.
DC/DC converters can operate in a wide input voltage range, so your primary supply is far less critical, as are your mains-supply quality.

Somehow I get the distinct feeling that it can be done. We see multiple console manufactures use this technology in their power-supplies, some bad and for primarily cost reasons. Some good, for other reasons like power-efficency and the ability to absorb mains-fluctuations better (and of course weight and size)...
I have seen high-quality tube based designs based on DC/DC converter technology here, to create the High-voltage. likely s/n quality will matter here too. The modules looked quite compact and not too difficult to integrate.

I really got triggered by the Rupert Neve Portico series. I do not think mr. Neve is taking quality shortcuts, apparently he masters the technology well enough to apply it in his high-quality-level designs. Best signal: I do not hear/see any customer complaining.

I am not the expert in this technology field, I simply ask:
- What did he do to make it good enough for his highest quality products?
- What can we learn from this?

Theo

 

[JohnRoberts]

A switcher is just another way to skin the PS cat... modern technology makes these easier to implement.

My first technician job was working on a switching PS for a Navy submersible project, some 43 years ago.  They're cheaper and more attractive to use these days.

JR

 

[shabtek]

I recall a manley box 10 years ago with dac\adc option that had hash spraying around inside--not smps though iirc.
I am a linear ps devotee and know nothing of the switch mode.
Ian mentioned load and supply regulation, I am not 100% sure what these things mean.
I do recall reading an article about the non-linear load smps present to the 'grid' :in some instances when an office building all comes on at the same time it looks like a fault condition to the supervisory harware at the power company...
don't know if there are any parallels here but it seems that one should not only consider radiated noise and downstream noise, but effects 'upstream' to dc rail busses, as well as compounded effects of multiple dc\dc modules (crosstalk, sum\difference interference, etc.)

 

[Bill Wilson]

I recentlky retired from the electrical engineering dept. of a manufacturer of TV inspection systems for sewers, storm drain lines & well inspections. We designed & manufactured all our electronics in house (the exception being the camera heads). We used a lot of DC/DC converters; Vicor was the best, quality consistent & rather expensive.

Over the years we were forced to change manufacturers for one specific system. We would be supplied w samples that would measure well but as time passed they would develop noise that affected the video image. In 13 years we used product from eight different manufacturers. It is a proven fact that what happens outside the working bandwidth of equipment does have some effect on signals within the bandwidth. I for one choose not to use DC converters for this reason.
Cosel does manufacture high quality converters that are used by Gary Gilmore in some of his portable headphone amplifier designs.

 

[bruce0]

SMPSUs have the advantages of:

1.Good supply regulation
2. High efficiency
3. Small size

With respect to DC-DC converters (feeding modules with DC, each having isolated DC-DC converter).
Isn't it true that one other advantage of using DC/DC converters is that you can create isolated grounds and potentially eliminate the age old hassle of "ground loop" noise?  I mean can amplifiers be grounded through audio ground alone, separately from the case safety ground?  It is all low voltage (setting tube circuits aside).  And with Audio ground truly isolated, then there is not a path other than the audio ground to add to noise.

It does sound like Igor has done some testing of the use of cheap DC-DC converters ($10).  And he brings up the $/watt issues, and he brings up frequency issues  and synch issues (multiple converters in the same module with slightly different frequencies might introduce low frequency beat frequencies or interference patterns, harmonics, like Moire patterns in image processing)

Igor, as someone already asked, it would be great to know what unit you tested out.  Good data on DC/DC converters is hard to find.

These are in use in the new Neve portico modules - so they have addressed, or are ignoring, these issues.

I have found it very difficult to find information on DC-DC converters, the spec sheets are brief, and read (to my untrained eye) like jibberish, and do not seem to address the issues that I (and I think we) am/are interested in. 

 

[Tubemooley]

We use 1W dc/dc converters in the products we manufacture where I work. We manufacture industrial instrumentation and control equipment. I've been at this company for almost 12 years now. In 1999, we paid $10-12. each in quantity (1k+) for a 1-watt dc/dc converter. We always use +5v on the primary and either +/-9v unregulated or +15v unregulated on the secondary. Those are the models we need for what we do. Since 1999, we have evaluated and qualified at least 10 different parts, all from different manufacturers. That 1W converter that cost $10. back in 1999, we got it down to just under $4 each about two years ago. And we just found a new one which is under $2 each. Back in 1999 we were using C&D Components and Newport dc/dc's. Those companies have changed hands at least twice in the last 10 years. The new dc/dc is from overseas. In fact, they have all been from overseas for way over 5 years now.

Our circuits are making DC measurements and the input signals change very slowly. Over a period of hours sometimes. So high-frequency noise is not that much of a concern to us as long as it doesn't effect the DC measurement or the circuits which handle it. We have never done extensive measuring of noise so I don't have any noise figures to contribute here.

I do kinda' know this. There is a little bit of black magic associated with using dc/dc converters. But I don't know what that black magic is yet. And no, I'm not just looking into my crystal ball and conjuring this up. We recently hired a new EE at work and I'm looking forward to seeing how he uses the dc/dc's.

My Amek Purepath Driver-in-a-Box has a SMPS built by Amek on the circuit board. Not a purchased SMPS. Homegrown. Rupert Neve designed that box. Purple uses a switcher in his Sweet Ten 500-series rack box. His may also be homegrown. Tonelux uses purchased SMPS's for his V-Racks. These are just examples offered here as additional information. Switchers and dc/dc's can be used in audio gear without increasing noise levels. I just don't know how to do that yet. This is one topic I have been trying to learn about for years now and I plan to keep on doing so. DW.

 

[mitsos]

I'm glad this thread came up.  Reluctance to use DC-DC converters because of all the noise talk... Isn't the fact that they are implemented (silently) by many companies proof that noise is more related to correct implementation??

One specific mfg (GR - Dan Kennedy) gave away his implementation of a 24V converter (+/-16V -> +24V) which I believe is being used in Peter Purpose's Neve EQ? 

There is also the 500 Neve preamp (although I don't think the converter schematic was published), haven't heard of anyone having problems with either of these modules creating noise but I don't have experience with either, just the only things I can think of right now.

To me a slight modification to the topic would be more productive and could lead us to new ground: something like "proper DC-DC converter implementation."

I'll also add that it would be nice if we could pool resources and come up with an open-source GDIY endorsed DC-DC converter schematic for +24V and one for valve voltages and put this topic to rest.

cheers!

 

[JohnRoberts]

Anything that is suitably complex is considered magic by some.

I have watched the technology evolve over the years and the technology has dramatically reduced the size and weight and cost of high power audio amps for professional use. For a long time it just reduced the size and weight, but the cost is slowly following it down as the smaller power transformers and heat sinks reduce material costs and the technology matures so cost premiums evaporate. Another area where switchers are attractive to manufacturers is to provide universal input voltage to help simplify selling into world markets.  When I first worked with with switchers there was not one single IC designed to make a switcher. Now there are too many to count. 

The main strategy for managing switching noise is to contain it at the source. PCB layout matters and for high switching frequencies PCB trace inductance matters so layouts are kept tight and runs short with consideration for keeping loop area small where large currents are flowing (a good practice for analog design too). Capacitive shunts and series inductance on input and output lines short circuit the high frequency noise and help keep it contained locally.

For precision audio circuitry, I have even used pass regulators in series after a switcher to double regulate, while I can't really justify the expense. (It seemed like a good idea at the time).  8)

JR

PS: wrt phantom power,  I think I've shared this trick before, but one switcher I did needed a low current +48V supply for a single mic input. rather than add another switching supply, I made a charge pump, a variant on the old cap doubler/tripler to step up a normal audio rail to 48V+. Instead if doubling at 60 Hz, I was charging and discharging the boost cap at the switching PS rate (in this case a couple hundred kHz).  I was able to generate the few mA I needed with a small .01uF boost cap, and the charge pump didn't make another magnetic field or interfere with the basic switcher. Just like from a normal cap doubler this output was unregulated so needs to be regulated after, or at least filtered and clamped.

PPS: Another area where switchers have been creatively applied is to convert the high voltage low current output of a phantom power drop, to higher current at lower voltage to run a DSP or whatever. There are any number of other examples where DC/DC convertors act like a DC transformer to convert power at one DC voltage to similar power at a different DC voltage.

 

[Andy Peters]

Just a couple of general comments ...

There are a lot of old-school DC-DC converter designs (the Vicors already mentioned, various and sundry others) which have low-ish switching frequencies and as such large inductors and less than great ripple.

But there's really no reason to go with an off-the-shelf module, especially for low-power designs like the kind of stuff discussed here (excluding consoles, perhaps). Spend an evening perusing the Linear Technology website and their app notes. Check out the National and TI data sheets, too. There's a ton of info out there about how to build all sorts of supplies out of a switching controller chip (which uses external MOSFET switches) or switching regulators (which integrate the power switch on die). You get recommendations for passives, MOSFETs and even recommended layouts.

Modern parts switch at frequencies waaaaay above audio (meaning easier to filter and they use small inductors) and some of them are even synchronizable, so one of your audio converter clocks can also drive the power supply switching.

I don't advocate building an off-the-line switcher (it's a good way to die if you're not careful) but a standard step-down transformer to get a reasonable DC voltage from which to derive what you need a good place to start.

-a

 

[abbey road d enfer]

The use of DC/DC converters in many Neumann, AKG,...microphones is proof enough that the concet is not uncompatible with low-level audio applications.
Over the years I have used several different off-the-shelf SMPS. There's always a concern that the sample you've thoroughly evaluated doesn't guarantee the ones you will be delivered will be similar.
Don't just believe the manufacturer's specs. You must evaluate them.
I've measured more than 20dB difference in actual noise performance of the finished product, between PSU's that were supposed to be very similar.

 

[ppa]

Just to add my two cents, I have this year designed a DC/DC con.  for a tube pre on 500 rack format for a new start up company using an inedit topology, this DC/DC converter has -92dB of noise before the regulator, EMI are pratically near to zero.
The matter is that a DC/DC conv. must be designed for audio to be used for audio.

 

[Brian Roth]

Other 2 cents....

I'm working with a friend/client who recently bought a Neve Capricorn desk (actually a "system" with a control surface and lots o' rack mounted boxes which all comprise the entire system).

I was chasing a "whine" in the analog monitoring outs, and after lots of card swapping (friend/client also got a second "donor desk" for parts) I finally swapped out the entire monitoring card frame with the "donor" card cage.

Aha...noise gone. 

The entire Capricorn system has a jillion switch-mode power supply modules...most of which power the jillions of digital cards.  But, several frames have +/- 16 V rails for analog purposes, such as feeding the CR monitors, analog outboards, etc.

Bottom line...while this desk originally sold for 6-figures, and Neve had obviously "done their homework" re. the power supplies, after some years have passed a switcher PSU had gotten "tired" and was injecting a low level whine into all of the monitor outputs.

That is something to ponder when adding switchers into an analog system.  Linear PSUs may over/undervolt, add 120 Hz noise or pops/crackles...but never (?) add a low-level whining noise ...as the PSU unit becomes older.

Best,

Bri

 

[Andy Peters]

The entire Capricorn system has a jillion switch-mode power supply modules...most of which power the jillions of digital cards.  But, several frames have +/- 16 V rails for analog purposes, such as feeding the CR monitors, analog outboards, etc.

Bottom line...while this desk originally sold for 6-figures, and Neve had obviously "done their homework" re. the power supplies, after some years have passed a switcher PSU had gotten "tired" and was injecting a low level whine into all of the monitor outputs.

My guess is that there is an output capacitor hanging on a supply that's gone bad; as the ESR goes up and/or the capacitance goes down, the ripple increases until it becomes objectionable.

Does that sound familiar?

That is something to ponder when adding switchers into an analog system.  Linear PSUs may over/undervolt, add 120 Hz noise or pops/crackles...but never (?) add a low-level whining noise ...as the PSU unit becomes older.

To answer my rhetorical question: if the cap hanging off of the output of a linear regulator, or as part of the smoothing after the rectifier, goes bad (lower capacitance), the ripple increases until it becomes objectionable.

In the switcher case: if the switching frequency is high enough, the ripple may be at supersonic frequencies, and may beat down into the audio band, or do other evil things. If it's low enough -- and older switchers didn't switch all that fast -- the noise could already be in the audio band.

In the linear case: the line noise you mention gets worse.

-a

 

[gemini86]

Most of the time it's the large electros failing at high frequencies... cheap computer PSUs are prone to this...they become audibly noisy, one could imagine what they're injecting into the rails at that point...

 

[Harpo]

Some reading on DC/DC converter design
http://www.simonbramble.co.uk/dc_dc_converter_design/dc_dc_converter_design.htm

 

[ppa]

Some reading on DC/DC converter design
http://www.simonbramble.co.uk/dc_dc_converter_design/dc_dc_converter_design.htm

nice