Help with getting rid of DC ripple

[Belgianuser]

Hello everyone here.
I'm a newbie and have a question.
But first,...... My name is Rein, am 60 years young and was(am) a mechanical engineer with audio as a serious hobby.
 I don't know much about electronics , just de basics basic.
What is my question?.....
Well for my hobby (audio) I have build a power amp setting consisting of 4 amps (2 monoblocs for mid/high and 2 monoblocs for low)
It based on a kit of manufacturer in Germany (ALBS)
So all the main parts are ready to be assembled. (the only special work I have to do after assembling is biasing)
Last week I heard two of these amp's in  comparison off someone who made a modification at one off them.
Long story short,..... Amp 1 sounded as I knew it and amp 2 (modificated) sounded cleanen and  "dryer" both on the same speaker set, in the same room and with same program material . Everyone in the room was pleased and asked about the mod. The man wouldn't give information. The only thing he said was that he get rid off the ripple . After searching on the net I encountered the artikel from David Jones. I read his article and some more and understand (in great lines) how it works.But as I stated earlier I'm not an electronic guy . So .....Can /will someone make me a scheme/ layout   A which components I need, B a practical way of assembling (print/air) . In the attachment a foto of one of the 2 amps with a supply voltage of 70 Volts DC. The other two will have a supply voltage of 62 Volts.  Thank You!!!!!

[Benta]

Still waiting for the question...

[jwet]

You've left out a few key parameters.  Have you measured the ripple.  With 60,000 uF, I would think the ripple would be quite low.  It doesn't have to be zero- amplifiers, especically mono-bloc type amps have substantial power supply ripple rejection.

The formulas work as follows.  I = C dV/dT- this is how caps work.  A charged capacitor will droop at dv/dt volts with a load of I Load/Filter C- this is the ripple.   dt is 1/100 or 1/120 sec depending on where you live- power line frequency x 2.  This is 8.3 mS or 10 mS assuming full wave bridge rectifiers (if not, the dT would be doubled).  Ripple would be dT*Iload/filter C.   Looking at your 750 VA transformer, this would appear to say that you're shooting for something like 350 watts per channel.  At 50V, this is 7 amps peak.  Your cap would droop 7 *.01 / 60000 uF given you a ripple of 1.16V.  If my assumptions are wrong- plug in the correct values and recalc.  That is a little bit of ripple and it depends if the amp can reject it- some can't.  If you want lower ripple, increase the filter caps.  There are other ways to accomplish this (like regulation) but from where you are, I think this is what you're left with.  If the amp is stereo, take all the filter caps off one channel and put them on the other (test) channel.  See if this makes an audible difference.  There is a good web site for Audio home brewers called Elliot Sound labs- lots of good info.

Good Luck-

[Someone]

It doesn't have to be zero- amplifiers, especically mono-bloc type amps have substantial power supply ripple rejection.

Best guess would be some additional capacitance or regulation in the earlier stages of the amplifier where the sensitivity is higher.

There is a good web site for Audio home brewers called Elliot Sound labs- lots of good info.

So much good practical information there that is often backed up with experiments people can reproduce themselves. Amazing content that should be shared wider.

[Belgianuser]

Hello ,Thanks for your reaction.
0ne monoblock has a total capacitance of 4x60.000 uF =240.000 uF. And as stated my line frequency is 50 Hz.
I'm looking for someone who can help me implement the solution as in this Youtube short 
It''s from the founder of this blog.

[PGPG]

This is 8.3 mS or 10 mS

Do you (on the other side of the ocean) still use different units?
Please do not take my question as malice, but as a desire to understand the world (whole sentence made by Google translator as these are not technical words so I have never seen them).
For me (as I was taught) 10mS == 100Ω. For you not?

In 2014 I have selected TPS54061 to use. In parameters table I found amplifier gain specified in μMHOs. To calculate feedback elements to get correct stability margins I had to know what the hell is that. I read it as micro-mega so canceling one another so HOs left to analyse. I assumed that s was used to signal plural number of units so I left with the problem what HO is?
Google wasn't very helpful when I was trying to find HO unit meaning. At the end I found that TI used units that were over 100 years ago replaced by Siemens. As it happened many years before I was born I understood why I have never seen this unit before.
I mailed to TI and they corrected it in TPS54060, TPS54060A, TPS54061, TPR54062.

Now I see that may be you there use S (Siemens) instead of s (second).
Is it true, or you only by mistake used big letter?

[PGPG]

In the attachment a foto of one of the 2 amps with a supply voltage of 70 Volts DC.

Looking at your photos I am surprised seeing these twisted wire pairs used instead of just wider wire.
Is there any rationale behind it?

[Belgianuser]

Hello,
I don't understand what your reaction have tot do with my question.
But in Europe ms is used for millisecond (in technical way)

[Belgianuser]

no. there's no special reason .

[timeandfrequency]

Hello Belgianuser,

As 'jwet' calculated, with 60 mF capacitance, the ripple is damm low. So much capacitance is quite overkill.
Which amplifier model from ABLS do you use ?
I'm not totally conviced that getting rid of the ripple will provide a dryer/better sound. If your amplifier sounds weird, I would rather think about about a sproradic high frequency oscillation. Do NOT connect a vanilla oscilloscope to the ouput of the PA : it might create a circuit through the earth connexion and fry the ouput transistors.

If you really want to chase the last Volt of ripple, here's a thead discussing about PA rail regulation.
The 'Apex' schematic provided on 2016-08-27 7:39 pm by member 'xrk971' seems to be an interesting starting point.

[PGPG]

I don't understand what your reaction have tot do with my question.

Sorry. My reaction had nothing to your question.

But in Europe ms is used for millisecond (in technical way)

I just wanted to know if in USA they don't use ms and use mS for both milisecond and miliSiemens (as TI avoided to use mS so may be it was because there they use (big) S for seconds).

[Belgianuser]

The ALBS kits always are very "robust". That's probably is the reason that I (when I build my first one with separate powerlines for driver and powerstages) sold my DENON POA 4000 Monoblocs in the 80's.
The only amp that is more to my liking is the Accuphase.
One of the reasons of there success probably is  the overkill in  their layout.
I use the DAC 360 and DAC 240 (360 AND 240 w) but they also have a DAC 600 with separate powerlines for the driver and power FET's.
Each with a capacitance of resp. 320.000 and 360.000 uF.
But all this is beside the point .......my question stays the same .....anyone.
I heard the difference between an ALBS 360 Monobloc with 240.000 uF and the same with a "ripple modification". It wasn't a little difference.

[Belgianuser]

no problem PGPG

[TimFox]

This is 8.3 mS or 10 mS

Do you (on the other side of the ocean) still use different units?
Please do not take my question as malice, but as a desire to understand the world (whole sentence made by Google translator as these are not technical words so I have never seen them).
For me (as I was taught) 10mS == 100Ω. For you not?

In 2014 I have selected TPS54061 to use. In parameters table I found amplifier gain specified in μMHOs. To calculate feedback elements to get correct stability margins I had to know what the hell is that. I read it as micro-mega so canceling one another so HOs left to analyse. I assumed that s was used to signal plural number of units so I left with the problem what HO is?
Google wasn't very helpful when I was trying to find HO unit meaning. At the end I found that TI used units that were over 100 years ago replaced by Siemens. As it happened many years before I was born I understood why I have never seen this unit before.
I mailed to TI and they corrected it in TPS54060, TPS54060A, TPS54061, TPR54062.

Now I see that may be you there use S (Siemens) instead of s (second).
Is it true, or you only by mistake used big letter?

Within my lifetime, which includes the end of the vacuum-tube era, conductance (reciprocal of resistance) was quoted in "mhos" often denoted by an upside-down , originating as a clever backwards spelling of "ohm". 
Most vacuum-tube datasheets (e.g., https://frank.pocnet.net/sheets/049/6/6AK5.pdf  where the transconductance values spelled out "mho”) quoted transconductance in that unit.
In that era, before the SI was adopted, frequency was quoted in c/s instead of Hz.  (The suggestion to name the liter/min the "Falstaff" did not clear the committee.)
It would have been useful had the SI formally allowed "u" instead of {mu} for the prefix "micro", while our pharmacist friends use "mc", as in "mcg" for micrograms.
In this forum, many people can't be bothered with spelling:  one often sees "m" and "M" confused (what's 10⁹ between friends?).
Yes, "s" now means seconds and "S" means Siemens.

[Belgianuser]

Hello,
As I stated earlier I'm not a electronics expert so I don't know  how to explain some fenomena . I only can rely on my own ears and (audio) experience.
We (6 persons) all heard a difference in the bass range. The difference was big enough to chase the possibility to get rid of the last DC ripple.
Sometimes there are fenomena that are not obvious to explain but just give a different "sound", NOT PERSE BETTER!!!!! Just only different.
So my question stands : Is there someone who can /will help me to implement the solution stated by author   "CAPACITANCE MULTIPLIER"

[TimFox]

A capacitance multiplier is an emitter follower between the rectifier filter (collector) and the load (emitter) with a reasonably large capacitor from the base to ground (or power supply common).
There must be a resistor network from collector to base to keep the transistor’s emitter below the collector voltage (including its peak-peak ripple) and supply the base current.
Roughly, this is equivalent to adding a capacitor whose value is h_fe times the physical capacitor.
The output voltage will be lower than the input (with power dissipated in the transistor), and the circuit does not regulate against changes in the DC input, but it does reduce the output ripple.

[PGPG]

But all this is beside the point .......my question stays the same .....anyone.

Not looking any video. Just about ripple rejection you asked.
When you have 2 capacitors connected paralel they limit ripple to some Voltage.
Now when you disconnect them and use only one of them you will get 2 times higher ripple voltage at it.
But when you then connect relatively big L from this one capacitor to the second one you form pi-filter and ripple from first capacitor can go to the second one dramatically reduced.
You can assume that such LC (C is the second one capacitor) transfers ripple without reducing starting from DC till the resonant LC frequency. At that frequency ripple can be even higher at second capacitor than at the first one, but from this frequency it is reduced with slope 20dB/decade. Your ripple is at 100Hz. If resonant LC frequency will be 10Hz ripple from first capacitor will be reduced about 10 times (because 100Hz is 10 times higher than resonant frequency). So disconnecting capacitors you get 2x higher ripple at the beginning but then you readuce it 10 times so total you reduced it 5 times. If the resonant frequency will be 1Hz than LC will reduce ripple 100 times (so total is 50 times).
If you have 3 capacitors and you disconnect them you can have C-L-C-L-C hawing two stages that reduce ripple.
320.000uF = 0.32F. I don't know what are the capacitors. Let us assume you can divide it into three 0.1F. At first of them you will get 3 times higher ripple than now. If you will have 1H choke and connect it between this first capacitor and the second one you have LC filter with resonant frequency f=1/(2*pi*SQRT(L*C)). L*C=0.1 SQRT(0.1)=about 0.3. 0.3*2*pi=about 1.9. So f=1/1.9 = about 0.5Hz. So at second capacitor you will have ripple reduced about 100/0.5 times - menas 200 times smaller. Then next such stage will give you next 200 times. So your ripple is multiplied by 3 and then divided by 40 000. The problem is that 1H chokes for high current will be big. But may be 0.1H giving 3 times smaller values so not 200 but 70 and total not 40000 but 4900 will be enough for you.

[Zero999]

It's not possible to get rid of ripple, only reduce it.

The minimum required capacitance, depends on the mains frequency, maximum current and the type of load, i.e. constant current, resistive, or pulsed.

[jwet]

s vs. S

I simply made a mistake.  I often do this.  I think its clear from the context and the derivations what the meaning was.  I find that pedantry gets in the way of real discourse, don't you?

[Bud]

Maybe the change was on the amplifier side, the guy adjusted ripple rejection or something.

[TimFox]

If one looks through the audio design literature, there are various useful schemes for improving PSRR (power-supply rejection) or hum cancellation, to reduce output hum without brute-force filters or active regulation of the power supply.

[PGPG]

I find that pedantry gets in the way of real discourse, don't you?

I agree with it.
If not those my experience with TI datasheet I would not ask assuming it as just mistake.
But seeing you using S for seconds I just got an idea, that may be they avoided using S as Siemens as may be it is typical that you use S for seconds.
I just wanted to know if it can be such.
I remember from "Surely You're Joking, Mr. Feynman!" that they were using units that I was just surprised.
I've just born in the world where SI was used from my perspective 'since always' and am still surprised that other units are still used.

[jwet]

Most people have words that they routinely mis-spell.  I have more than my share.  I have these abbreviation misfires too.  I'm especially bad when responding to things online- I'm overcome with a didactic impulse and things get messy.  No offense to Germany and the contributions of Mr. Siemens, but I always like mho's as I was taught back in college in the 70's- it was clever I thought.

[Bud]

I've seen Mho units in several RF books and the books were not 100 years old.

[TimFox]

Although the IEC adopted “Siemens” for conductance in the 1930s, the SI waited until 1971.
According to Wikipedia, there was a time in the 19th century when Siemens was used for resistance, replaced by Ohms.