Ripple Filters

Help Support GroupDIY:

ruffrecords

Well-known member
Joined
Nov 10, 2006
Messages
13,755
Location
Norfolk - UK
There have been many threads on voltage regulators and ripple reduction methods. The most common ripple reduction method is the misnamed capacitance multiplier that is really a LPF feeding an emitter follower. I recently came across a different approach which ac couples the input ripple to a shunt transistor in order to create a near equal and opposite ripple current that cancels the input ripple at the output. Here is a video about the technique:


The interesting bit starts around 5:49 which is the end of the emitter follower technique discussion closely followed by the shunt technique.

Cheers

Ian
 

Bo Deadly

Well-known member
Joined
Dec 22, 2015
Messages
2,626
Location
New Jersey, USA
The standard two transistor CM followed by a fat cap is pretty great for multiple amps. There's a lot to like about it's simplicity. Not sure I would get too cute with ripple filters.

But shunting is not such a bad idea. I've been looking at the TL431 which seems like it could be useful for dropping, loading and filtering an SMPS output perhaps woven into the CM circuit. That's a great part actually.

Finding a good SMPS is the best solution of all. There just so good now it's hard to justify fancy circuits.

If you do find that the SMPS isn't so great but you want to use it anyway, one of the dominant sources of ripple is common mode switching noise capacitively coupled through the SMPS transformer. For that, I think the solution is a big cap between the output 0V and earth. Then you can also follow that with an inductor before your star 0V. That directs those currents back out the local ground and not through the ground of whatever your device is connected to. Although if all audio I/O is vaguely balanced, in theory I'm not sure it should be necessary.
 

trobbins

Well-known member
Joined
Jul 1, 2011
Messages
274
Location
Australia
For valve amps the topic of B+ ripple filters has quite a few aspects to consider.

One aspect is to perhaps just target ripple during idle, or low signal levels, where any such hum caused by the ripple is most noticeable. The level of hum is often not noticeable for large signal conditions, and that is also when the power supply is being asked to be part of the substantial signal current loop, which may also be large signals at or near the hum frequency and phase.

I enjoyed preparing a hum filter based on 'topping up' the B+ level with a ripple current that filled in the gaps - it used a not uncommon spare 5VAC winding as the power supply for the topping up function. It worked quite well, but has to transition from idle conditions to signal conditions, where output DCV sag occurs as well as a change in ripple level. The aim wasn't so much about achieving mV of ripple, but rather to reduce the large V ripple for B+ supplies that use a small vintage output capacitance like 8-16uF.
 

ricardo

Well-known member
Joined
Apr 28, 2011
Messages
1,431
I enjoyed preparing a hum filter based on 'topping up' the B+ level with a ripple current that filled in the gaps -
A big series choke does this. Ian, I would have thought this would be just what you want :)

Better regulation bla bla too ... see RDH 4
 

trobbins

Well-known member
Joined
Jul 1, 2011
Messages
274
Location
Australia
Unfortunately SMPS capable of 250 to 300V at 300 or more mA do not seem to be readily available from reputable suppliers.
I have used an alternative smps approach. Mains supply of 12VDC via a commercial compliant device. Ebay has had a few vehicle 'inverter' pcb modules for 12Vdc input, with '150W' rating which I know happily deliver at least 40W, and should be low risk for 75W of B+. The user provides a simple rectifier and filter to an appropriate tapped output to give the nominal B+ required. The B+ is fairly stiffly aligned to the 12Vdc level, although the smps secondary side is not regulated. These modules are based on the 3525A control IC.
A big series choke does this. Ian, I would have thought this would be just what you want :)

Better regulation bla bla too ... see RDH 4
ricardo if you were enamoured by large choke filters then perhaps use the semiconductor based equivalent (as per E-Choke).
 

abbey road d enfer

Well-known member
Moderator
Joined
Jan 22, 2008
Messages
13,012
Location
Marcelland
The most common ripple reduction method is the misnamed capacitance multiplier that is really a LPF feeding an emitter follower.
Both denominations are correct. Some would call it a gyrator and be almost correct too. :)
The interesting bit starts around 5:49 which is the end of the emitter follower technique discussion closely followed by the shunt technique.
Shunt regulators are old moons. In that specific case it's only their AC regulation that's used, just like a "cap-multiplier" is a kind of series AC regulator.
The most interesting property of shunt regulators is that they keep the input current constant, which reduces interaction when several loads are connected. The resistance/inductance of wires between PSU and loads become almost irrelevant, so they are appreciated in large-scale systems.
The penaly is they draw more current than a series regulator, when the penalty of a series regulator is they need more voltage to strat with.
Shunt regulators are audiophools' pets.
Check this:
Second post says much, but not all.
The claimed advantages:"Because a shunt regulator is a two-stage circuit, isolation between two regulated outputs is tremendous. From one regulator to the next we see a low impedance to ground, a very high series impedance (the current source), another low impedance to ground (the unregulated supply), the other current source and lastly the other shunt part. Crosstalk from one supply to the next is unmeasurable.", are somewhat debatable, in the sense that a good series regulators presents exactly the same advantages.
 

jacomart

Well-known member
Joined
Sep 13, 2017
Messages
161
Location
Tuscany, Italy
Unfortunately SMPS capable of 250 to 300V at 300 or more mA do not seem to be readily available from reputable suppliers.

Cheers

Ian
There would be the TDK mod. PF500A-360 which costs around 155 USD from Digikey, but there are no voltages lower than 360 (and not even higher actually)

Cheers
JM
 

Bo Deadly

Well-known member
Joined
Dec 22, 2015
Messages
2,626
Location
New Jersey, USA
Unfortunately SMPS capable of 250 to 300V at 300 or more mA do not seem to be readily available from reputable suppliers.

I see. So you're rolling your own then? I suppose that makes sense anyway if you want to make all the different voltages for heaters and probably other stuff.

However, this makes me wonder if you can parallel SMPS through protection diodes and smallish resistors. You say there's nothing at 300 or more mA but if there are smaller powered units then maybe you can parallel? That might seem crude but it kinda makes sense for a modular build like a desk.
 

Bo Deadly

Well-known member
Joined
Dec 22, 2015
Messages
2,626
Location
New Jersey, USA
Ian,

What about something like this:

MeanWell ELG-240-C700

1623588584492.png

but run it in constant current mode by loading it with a dropper and shunt regulator until the voltage drops to where you want. If the required power just happens to be 200W or so, it would be quite efficient actually.

Actually searching Mouser > SMPS > LED Supplies I'm seeing a LOT of potential CC supplies that you can run in the CC region. Ripple is no doubt higher in the CC region but you're effectively picking the voltage so you can afford to drop over a resistor and use aggressive filtering.
 
Last edited:

ruffrecords

Well-known member
Joined
Nov 10, 2006
Messages
13,755
Location
Norfolk - UK
Interesting comments.The shunt version for ac ripple reduction is interesting to me for two reasons. First, unlike a dc shunt regulator, the current it needs to sink is small. It does ot need to carry the entire dc current as in a series version. Secondly, unlike a series ripple reducer, it has no time constant. A problem for the series ripple reducer is the time constant of the RC low pass filter needs to be as long as possible for good ripple reduction. However this creates a problem at switch on; until the capacitor charges up, the series pass transistor has to withstand the the entire supply voltage.

Cheers

Ian
 

JohnRoberts

Well-known member
Staff member
Moderator
Joined
Nov 30, 2006
Messages
20,983
Location
Hickory, MS
I have no problem calling them capacitance multipliers and will add another name "pre-regulator". Sometime used in systems with per module regulation where the desire is to keep ripple outside the chassis and deliver adequate (but not too much, to manage heat buildup) voltage headroom for the secondary local regulators to operate cleanly.

JR
 

ruffrecords

Well-known member
Joined
Nov 10, 2006
Messages
13,755
Location
Norfolk - UK
There would be the TDK mod. PF500A-360 which costs around 155 USD from Digikey, but there are no voltages lower than 360 (and not even higher actually)

Cheers
JM
I have seen several like this. It is a power factor corrector designed to feed a dc/dc convertor. Unfortunately its output ripple is 20V pp.

Cheers

Ian
 

jacomart

Well-known member
Joined
Sep 13, 2017
Messages
161
Location
Tuscany, Italy
I have seen several like this. It is a power factor corrector designed to feed a dc/dc convertor. Unfortunately its output ripple is 20V pp.

Cheers

Ian
You are right, my fault, I had seen a ripple of 280mVpp but it was for the PFE-F series with max Vout of 48V

Cheers
JM
 

Winston OBoogie

Well-known member
Joined
Jun 3, 2004
Messages
2,656
Location
UK.
this creates a problem at switch on; until the capacitor charges up, the series pass transistor has to withstand the the entire supply voltage.

I've been using Mosfet series "multipliers" for decades and, save for a few times when I probed in the wrong spot or something, I'm not sure that's ever been an issue - I'm talking maybe 1500 channel's worth. Not thousands, but still.
For tube B+, I even think the slow RC ramp up is a bonus. I make mine really long to allow heaters to get going. 1M & 100uF would be a decent "go to" for me.

There are some plastic packaged Mosfet's that even come with their own internal protection diodes to save you having to remember to include them.

I don't know if you're talking about one BIG series pass to supply a desk's worth? But if you're using them on a per channel basis, which is where I think they really work well, then some of them are even quite happy with a load up to 40mA/300V without a heatsink if you wanted to save a buck, providing the input to output V is within reason of course.

Each to each I guess, but for B+, they kill any ripple dead for me.
 
Last edited:

Winston OBoogie

Well-known member
Joined
Jun 3, 2004
Messages
2,656
Location
UK.
What about using smaller DC-DC switching supplies per channel rather than one big 300V one?
There's a pretty simple but effective circuit on Pete Millet's site that might be of interest:


One thing that would make me personally feel more comfortable about using switching supplies for tube B+ would be if I weren't reliant on some unit made in Asia that I bought online, one which I had no idea how to fix, modify, or replace.
 

ruffrecords

Well-known member
Joined
Nov 10, 2006
Messages
13,755
Location
Norfolk - UK
You are right, my fault, I had seen a ripple of 280mVpp but it was for the PFE-F series with max Vout of 48V

Cheers
JM
However, I did find a Meanwell product that is meant for driving long strings of LEDs I think wich outputs 300V at up to 500mA. I have made some initial tests on it and the ripple isquite high which is why I am interested in ripple reducers.

CheersIan
 

jacomart

Well-known member
Joined
Sep 13, 2017
Messages
161
Location
Tuscany, Italy
However, I did find a Meanwell product that is meant for driving long strings of LEDs I think wich outputs 300V at up to 500mA. I have made some initial tests on it and the ripple isquite high which is why I am interested in ripple reducers.

CheersIan
Considering that the working frequency of modern SMPS varies between 80kHz and 100kHz, it should not be difficult to make a simple three-cell elliptical filter capable of attenuating such frequencies up to 70dB (at least). To avoid instability it will be necessary to evaluate the output impedance of the PSU and calculate the filter accordingly. Use low ESR capacitors and other tricks we are all familiar with.

Cheers
JM
 

Attachments

  • AppNote_OutputImpedance_Stability_V1.0.pdf
    1.3 MB · Views: 15

ruffrecords

Well-known member
Joined
Nov 10, 2006
Messages
13,755
Location
Norfolk - UK
Considering that the working frequency of modern SMPS varies between 80kHz and 100kHz, it should not be difficult to make a simple three-cell elliptical filter capable of attenuating such frequencies up to 70dB (at least). To avoid instability it will be necessary to evaluate the output impedance of the PSU and calculate the filter accordingly. Use low ESR capacitors and other tricks we are all familiar with.

Cheers
JM
For many applications this may be sufficient but tube mixer, which are my main area of interset are different. The main problem area is the output stage which typically has an PSRR of no more than 20dB. If we want power supply noise to be no more than -90dBu at the output then it needs to be no more than -70dBu at the HT rail which equates to about 0.25mV. I generally aim for no more than 100uV so 70dB just about does it assuming the raw ripple is less than 200mV which, if you need 250V at 300mA it probably isn't. For example, the Meanwell ELG-150-C500A supply which is about the only supply I know capable of over 250V at 300mA, has a ripple current of 5%.

Cheers

Ian
 

ricardo

Well-known member
Joined
Apr 28, 2011
Messages
1,431
ricardo if you were enamoured by large choke filters then perhaps use the semiconductor based equivalent (as per E-Choke).
Where are the bragging rights in using EVIL silicon :) A series choke can be as large & $$ as your mains tranny and your OPT(s) on a power amp.
Besides, Ian is a vacuum guru so we should use appropriate technology.

BTW, there are good audible reasons for using a properly designed SMPS for a Class A valve power amp ... in addition to less $$$, size .. bla bla
 
Top