Schottky diodes in diff pair in PWM amp?

[stickjam]

What is the function of the paralleled Schottky diodes highlighted in the attached circuit?  I've never encountered this before.

This is a PWM power amp from a Yamaha powered speaker.  It's generating a nasty rustling noise when warmed up -- and even then, only in a certain temperature range: it goes away as it gets warmer.  Using a freeze-sprayed swab, I've narrowed it down to one of those diodes (6107)

Don't have the exact replacements onhand.  Will anything else work as well? (well naturally I want it to work better. )

Thanks

-Bob

 

[JohnRoberts]

They form an anti-saturation clamp to prevent the bottom current mirror from saturating, one way or the other so it will recover faster. The Schottky diodes when forward biased are less than  a typical base-emiiter junction voltage.  The current mirror transistor bases never see reverse bias.

JR

 

[stickjam]

Ah.  Makes sense.  Sounds like they need to be replaced with same then.

 

[stickjam]

Well, the schottky diodes highlighted in the schematic above are now replaced.  However, the noise problem is only slightly better.  What else would be causing that circuit to produce noise* when that diode pair is within a specific temperature range?

(*It's a sputtering brown noise kind of thing.  Imagine the noise that little saliva vacuum thing the dentist leaves under your tongue when he's doing major work, pitch-shifted down an octave. )

The amplifier has a cooling fan.  For diagnostic purposes, I disconnected the fan.  The speaker emits a normal very low level white/pink noise when the amp is first turned on.  After it is powered on for several minutes, the undesirable noise comes in, fading up in volume for a minute, then fades away, leaving the tiny bit of white noise that it started with.  Touching a freeze-sprayed swab to either of the schottky diodes brings the noise back as the temperature of the diode moves through the range.  Touching any other component with the cold swab has no effect at all. 

The problem might be tolerable if the amp continued to be quiet after a brief warmup outburst.  However, with the cooling fan in place, the fan seems to keep temperature of the amp right within the "noize zone." 

Any ideas or things to try to narrow down the culprit?

 

[stickjam]

Out of grins and giggles, I tried just removing those diodes.  Noise is gone and the amp actually sounds good. 

What deleterious effects would come from just omitting them? 

 

[JohnRoberts]

That comparator will saturate and recover slower from overload (transients?). In a class D input this could add distortion or tracking errors.  I don't think it will affect the outputs switching (that could damage the amp), but the money was spent and they were put there for a good reason.

I would replace them with good working diodes if it was my amp. 

JR

 

[radardoug]

Have you talked to Yamaha about it?

 

[stickjam]

Haven't gotten a reply from anyone knowledgeable at Yamaha.  A drop-in replacement for the AMP board is out of stock.

I did order exact replacements and replaced the diodes--twice.  Thought I may have fried them the first change. The problem improved a little but never entirely went away.  The tempco resistor is functioning but actually in a very different region of the PCB.

I'm getting really afraid of over-reworking this SMT board.     Daughter needs the speaker for a theatre pit gig the next couple weeks, going into tech rehearsals tonight.

 

[ricardo]

Schottkys go VERY leaky & noisy in use.

I've had this in diodes used to protect mike preamps from being zapped by shorting phantom powered cables.

THAT changed their recommendations from Schottkys to cheapo 1n4007 diodes because of this.

http://www.aes.org/e-lib/browse.cfm?elib=15104

Dunno if this is the case with your circuit cos there isn't much (any?) voltage across your diodes in use.  And dunno if you can use normal diodes.

But this is a serious problem with MOTU preamps.  You can expect on average, 12mths use before their mike preamps go VERY noisy because of leaky Schottkys.

MOTU refuse to acknowledge there is a problem so it is likely still there on their new stuff.  If you look inside a brand new MOTU, you will see signs of the diodes being swapped .. alas for more Schottky's which may eventually go noisy too.

Their stuff appears designed by consultants and they have no technical expertise in-house.

 

[JohnRoberts]

Replacing the Schottky diodes with silicon may prevent the bottom transistor in the current mirror from saturating, and the top transistor base from being completely reverse biased, but it will increase the amount of voltage that base must increase in transition to turn on again (Shottkey vs silicon forward voltage drop so hundreds of mV).  The design could probably be changed to use silicon diodes by adding a few parts but it should be easier to just replace the original diodes with similar parts.

I am not aware of them having such a short life, but i generally avoid using them so have no first hand experience.  A microphone may be a more difficult environment, as they are more exposed to external stresses.

JR

 

[aparatusonitus]

Hi stickjam,

You may try a 1PS76SB10 schottky diode as use in grandpa of all UcD amps - Philips UM10155 as well as in numerous Philips boomboxes with PWR303 UcD power module.

http://www.nxp.com/documents/user_manual/UM10155.pdf
http://elektrotanya.com/philips_pwr303.pdf/download.html

 

[ricardo]

I am not aware of them having such a short life, but i generally avoid using them so have no first hand experience.  A microphone may be a more difficult environment, as they are more exposed to external stresses.

The leakage spec for Schottkys is an order of magnitude (or more) greater than for 1n4148 or 1n4007.

It's likely a Schottky unacceptable at the front of a mike preamp still meets its spec.  An 'acceptable' one is likely to become unacceptable with time.  This is certainly the usual case with the standard MOTU mike preamp.

About the only creditable action by MOTU in this denial is that they will replace complete PCBs for $250.  But in Oz, this requires sending it to the US and a 3 mth wait.  And the replacement is likely to go noisy too.  This happened to me and from the MOTU forums, lots of other people. 

If your MOTU is still quiet after 12 mths, it is likely to remain so.

Dunno how this affects the PWM amp. and whether cheapo Si diodes are a suitable replacement.

 

[stickjam]

Well, it turned into a rubber-meets road decision, and I removed the second set of replacement Schottkys that were almost as noisy as the original pair and buttoned it up. 

She played three dress rehearsals and the five-show gig with the amp, even ocassionally pushing it pretty hard.  It sounded great...sans Schottky pair. 

If I understand the "preventing saturation" explanation, it makes sense to minimize distortion, given the speaker's marketed purpose as a PA speaker.  But in a dedicated application as a synth/sampler instrument amp, typically reproducing solo wind/brass instrument sounds in an ensemble, is it really going to be worth the hassle of putting more new ones in there and hoping they aren't noisy?

Besides, much has been written about shot noise (clearly what this is) in Schottky diodes.

Basically, is leaving them out going to hurt anything other than the specs a little?

 

[Samuel Groner]

Thanks for pointing out the effect of Schottky diode aging, I've not been aware of this. Do you have any references which would present measurement data and details on the involved mechanisms? I quick search revealed http://scitation.aip.org/content/aip/journal/apl/28/3/10.1063/1.88677 and http://scitation.aip.org/content/aip/journal/jap/89/12/10.1063/1.1370367, but these seem to consider the manufacturing level.

My MOTU interfaces use standard rectifier diodes for input protection and have not shown increased noise even after 10+ years of use, so condemning the entire brand seems not appropriate to me.

Samuel

 

[ricardo]

My MOTU interfaces use standard rectifier diodes for input protection and have not shown increased noise even after 10+ years of use, so condemning the entire brand seems not appropriate to me.

Which MOTU have you got?  What diodes are on the Mike inputs?

My experience is with the Traveler Mk2 but if you look at the various forums including the MOTU ones, you'll see that several MOTU designs have this problem (horrendous noise on mike i/ps after a suitable period).  eg its there in Traveler Mk3s too

What is clear is that most (all?) MOTU designs are done by consultants and they have no in-house technical expertise.  Otherwise this would have been sorted out very quickly.

I am condemning the company and its attitude (refusing to admit there is a problem)  rather than the product.  eg the Traveler is really nice when its working properly.

.. meanwhile I have other criticisms of my Traveler Mk2  eg its "P48V" only gives 39V 

Again making Service Info easily available would have helped.  There is no MOTU service info in Oz.  The agents send faulty units back to the US ... only 3 mths turnaround. 

 

[Samuel Groner]

I have the first-generation 896. I don't recall the exact part number, but (unless I'm mistaken) it was a standard bridge rectifier part similar to a MxS (M1S trough M8S). Some time back I had to change some of the input XLR connectors because they got intermittent. I first thought that this is the noise issue described on certain fora, but it turned out to be much simpler.

The design of the board is indeed from another company. Not something entirely wrong from the beginning--there are many others out there which seem to get around well that way, e.g. RME.

Anyone with more info on the aging properties of Schottky leakage current? That'd interest me far more than discussing company structures. 

Samuel

 

[ricardo]

I have the first-generation 896. I don't recall the exact part number, but (unless I'm mistaken) it was a standard bridge rectifier part similar to a MxS (M1S trough M8S).

If that is the case, I don't expect problems.

The design of the board is indeed from another company. Not something entirely wrong from the beginning--there are many others out there which seem to get around well that way, e.g. RME.

The difference is that RME actually answer customer queries and try to resolve problems .. compared with the MOTU policy of quoting for a replacement PCB which might not solve the problem.

Anyone with more info on the aging properties of Schottky leakage current?

As I said earlier, unacceptable leakage for protection diodes on mike inputs is still well within the leakage spec for Schottkys.

This is what THAT say on their latest 1510/12 datasheets.

"In revisions 0 and 1 of this data sheet, we recommended using Schottky diodes (1N5819 types) to protect the 1510/1512 inputs against overloads.  Subsequently, we discovered that the leakage of these diodes could cause problems with DC at the 1510/1512 output. Upon further investigation, we believe that conventional rectifier diodes like 1N4004 provide adequate protection and do not introduce unacceptable leakage. Additionally, 1N4004 diodes are much cheaper and more readily available than the Schottky types."

The small number of Travelers I've opened up, Mk2s & 3s, all show signs that the diodes have been swapped recently, including the 'new' replacement PCB on my Mk2.

My first Mk2 had 3 channels go noisy within 3 mths.  The replacement lasted 12 mths then 1 channel went noisy.  I looked into this cos I didn't want to lose it for another 3 mths.  Replacing the SK12b 52 diodes on the offending channel with 1n4004s cured the problem. (touch wood)

I have not heard of problems with Mk1 Travelers.

It is likely that if your  Mk2/3 is quiet after more than 12 mths, it will be OK but not guaranteed.  In fact all Mk2 owners I know personally have had this bad noise problem.  Above -20dBFS 

I wouldn't replace the Schottky's on a new Traveler, just those on a noisy channel .. but then again, I don't make a living out of its use.

 

[Samuel Groner]

I've just checked the PGA2500 datasheet--they indeed recommend Schottkys for input protection. So perhaps TI is the first to blame--or the PGA2500 indeed requires lower clamping voltages.

The recommended part (MBRA120LT3) has a guaranteed leakage of 100 uA at 10 V reverse (200 uA at 20 V) and 25°C, so a typical figure under operating conditions might be 10 uA. This gives 1.8 pA/rtHz shot noise which is not enough to cause such serious noise issues. But perhaps there is significant 1/f content, which could increase the noise in the audio frequency range by several orders of magnitude.

The SK12 has 2.5x higher leakage specs. Worse, but still not enough to get -20 dBFS noise without drastic 1/f content (but that's not impossible for a power part).

This all doesn't explain why leakage (or at least 1/f noise) should increase drastically over time, particularly without excesss current/reverse voltage applied. But apparently it does!

Samuel

 

[ricardo]

The recommended part (MBRA120LT3) has a guaranteed leakage of 100 uA at 10 V reverse (200 uA at 20 V) and 25°C, so a typical figure under operating conditions might be 10 uA. This gives 1.8 pA/rtHz shot noise which is not enough to cause such serious noise issues. But perhaps there is significant 1/f content, which could increase the noise in the audio frequency range by several orders of magnitude.

The SK12 has 2.5x higher leakage specs. Worse, but still not enough to get -20 dBFS noise without drastic 1/f content (but that's not impossible for a power part).

This all doesn't explain why leakage (or at least 1/f noise) should increase drastically over time, particularly without excesss current/reverse voltage applied. But apparently it does!

I'm not sure that 'shot noise' is the relevant spec for this fault.

The numbers seem to suggest it is the leakage current itself which is 'noisy'.  Certainly the SK12's I removed from my Traveler were on the top limit of the leakage spec.  But more practical tests are required.

I'd like to know what diodes are used in Mk1 Travelers.  When I was searching desperately for a solution a few years ago, I didn't hear of any Mk1s going noisy.

MOTU almost certainly have the evidence we want ... but they may not be able to interpret it.

BTW, what phantom voltage does your early 896 deliver?

And who were the Design Consultants?

The labels in my Mk2 Traveler proclaim S&S Research, Inc Norwood MA

 

[Samuel Groner]

The design is from the same company. I have not measured phantom power so far.

Shot noise is the basic mechanism which makes leakage current "itself noisy", as far as I understand; surely it is not the root cause here because the magnitude is not sufficient to explain the gross effects.

Can you measure the noise spectrum of the leakage current?

Samuel