Tube DIY Asylum

Do It Yourself (DIY) paradise for tube and SET project builders.

Return to Tube DIY Asylum


Message Sort: Post Order or Asylum Reverse Threaded

Ping deathtube 667: Smoothing LC Filters

63.246.183.22

Posted on October 14, 2016 at 19:44:16
Triode_Kingdom
Audiophile

Posts: 10012
Location: Central Texas
Joined: September 24, 2006
One aspect of my day job is the design of LC filters. It's difficult to maintain smooth response in power supplies, but over the years, I've come up with a few tricks. Most techniques aren't directly applicable to the HV supplies we work with here, but what they all have in common is dampening. So, one thing we can do with the filter you posted (first image below) is to reduce the effective Q of the last L. That's primarily responsible for the nasty 200 Ohm spike at 56 Hz. Looking at the second image, it's a duplicate of yours with one exception - a 200 Ohm resistor has been added across that last L. This reduces the 56 Hz peak to only 40 Ohms, or 1/5 the original impedance.

The primary risk in doing this is the possibility that the resistor might allow some percentage of ripple and noise to bypass the choke. We can check that by comparing the frequency response of the filter with and without the resistor. That comparison is shown in the last two plots. Ironically, 60 Hz performance actually improves with the parallel resistor. In the plot without the dampening resistor, the 60 Hz response is pulled up by the 56 Hz resonant peak to -13.5 dB. That's pretty poor IMO. Rejection at 120 Hz (perhaps the most important frequency) is -53 dB. The last plot shows response with the dampening resistor. 60 Hz rejection has improved to -25 dB due to the absence of the peak. We pay for this with a slight degradation in the 120 Hz number, now 3 dB worse than it was. That can easily be recovered with a slightly larger cap at either the last position or the one before it. 70uF in either location will do the job.












 

Hide full thread outline!
    ...
RE: Ping deathtube 667: Smoothing LC Filters, posted on October 15, 2016 at 04:26:03
TK,

I can't see the plots you posted on my PC for some reasons.

dt 667






 

I can't pick it up on my I-MAC either., posted on October 15, 2016 at 04:57:52
Michael Samra
Dealer

Posts: 36118
Location: saginaw michigan
Joined: January 30, 2005
They look like little square boxes with old desktop computers and a monitor in them.
"For every complex problem there is an answer that is clear, simple, and wrong" H. L. Mencken

 

The pictures were there yesterday, gone today. (nt), posted on October 15, 2016 at 07:45:53
Tre'
Industry Professional

Posts: 17263
Location: So. Cal.
Joined: February 9, 2002
.
Have Fun and Enjoy the Music
"Still Working the Problem"

 

They're Back! (nt), posted on October 15, 2016 at 11:24:12
Triode_Kingdom
Audiophile

Posts: 10012
Location: Central Texas
Joined: September 24, 2006
nada aqui

 

Now they're gone again. WTH? (nt), posted on October 15, 2016 at 12:14:28
Triode_Kingdom
Audiophile

Posts: 10012
Location: Central Texas
Joined: September 24, 2006
nada aqui

 

Repost, posted on October 15, 2016 at 12:19:04
Triode_Kingdom
Audiophile

Posts: 10012
Location: Central Texas
Joined: September 24, 2006
One aspect of my day job is the design of LC filters. It's difficult to maintain smooth response in power supplies, but over the years, I've come up with a few tricks. Most techniques aren't directly applicable to the HV supplies we work with here, but what they all have in common is dampening. So, one thing we can do with the filter you posted (first image below) is to reduce the effective Q of the last L. That's primarily responsible for the nasty 200 Ohm spike at 56 Hz. Looking at the second image, it's a duplicate of yours with one exception - a 200 Ohm resistor has been added across that last L. This reduces the 56 Hz peak to only 40 Ohms, or 1/5 the original impedance.

The primary risk in doing this is the possibility that the resistor might allow some percentage of ripple and noise to bypass the choke. We can check that by comparing the frequency response of the filter with and without the resistor. That comparison is shown in the last two plots. Ironically, 60 Hz performance actually improves with the parallel resistor. In the plot without the dampening resistor, the 60 Hz response is pulled up by the 56 Hz resonant peak to -13.5 dB. That's pretty poor IMO. Rejection at 120 Hz (perhaps the most important frequency) is -53 dB. The last plot shows response with the dampening resistor. 60 Hz rejection has improved to -25 dB due to the absence of the peak. We pay for this with a slight degradation in the 120 Hz number, now 3 dB worse than it was. That can easily be recovered with a slightly larger cap at either the last position or the one before it. 70uF in either location will do the job.
























 

RE: Now they're gone again. WTH? (nt), posted on October 15, 2016 at 12:19:13
kyle
Audiophile

Posts: 1838
Location: London Ontario
Joined: September 29, 1999
I can see them now.
They're taunting you.

 

Murphy's in the server, LOL (nt), posted on October 15, 2016 at 12:20:45
Triode_Kingdom
Audiophile

Posts: 10012
Location: Central Texas
Joined: September 24, 2006
nada aqui

 

RE: Repost ... damping resistor ?, posted on October 16, 2016 at 16:43:50
TK,

Thanks for posting the damping resistor idea.

I ran the spice graphs and was able to replicate the results you got both in/out.

What kind of resistor do you use to damp the 320mH choke?

Is 200 ohms the optimum value or can it be played with a bit?

dt 667

 

RE: Repost ... damping resistor ?, posted on October 16, 2016 at 21:25:38
Triode_Kingdom
Audiophile

Posts: 10012
Location: Central Texas
Joined: September 24, 2006
200 Ohms is a ballpark value that I arrived at by trial and error in SPICE. It was a compromise between reducing the impedance bump on the one hand, and rendering the 0.32H choke ineffective on the other. In this case, the filter with the resistor across the choke exhibits rejection at 1 kHz approximately 30dB better than with the choke bypassed. I concluded from this and the previous measurements that 200 Ohms is an effective value. By all means, feel free to try other values if you want, it's just a matter of achieving the right compromise.

The resistor type can be almost anything. It's inductance is relatively unimportant, and power dissipation will be well under 50 mW if the 0.32H choke is 10 Ohms DCR.


 

RE: tried the damping resistor ..., posted on October 17, 2016 at 17:39:46
Got a 220 ohm resistor (closest value to 200 ohms in my inventory) across the 320mH choke for a quick listen tonight.

There was a slight increase in AC ripple at the output from 3mV to 5mV.

The overall subjective sound is pretty smooth on my bookshelf speakers listening to some random classical music selections.

I'll need to do some bench testing with/without the damping resistor to see if I can detect any measurable changes/improvements.

In too much discomfort from a 3 hour trip to the dentist this afternoon to do much of anything tonight.

 

RE: tried the damping resistor ..., posted on October 17, 2016 at 19:29:11
Triode_Kingdom
Audiophile

Posts: 10012
Location: Central Texas
Joined: September 24, 2006
Sorry to hear you were tortured by the tooth doctor. That's never any fun. Well, I expected the 120 Hz energy to go up a little. The real problem in terms of overall ripple is the 2uF cap. It's just not large enough clean it up for a SET. Anyway, if ripple is acceptable without the 200/220 ohm but not OK when you install it, just add 20uF to C2 or C3 in your drawing.

 

RE: 2uF?, posted on October 17, 2016 at 20:15:35
TK,

Not using any 2uF caps in my filter to the finals.

C1 = 8uF
C2,C3 = 50uF

Adding a bit more filtering if needed is not a problem.

dt 667




 

Sorry, my mistake, meant the 2H choke, posted on October 17, 2016 at 21:26:25
Triode_Kingdom
Audiophile

Posts: 10012
Location: Central Texas
Joined: September 24, 2006
Just not sure that's enough to get it "SET quiet." You'll get it right over time though. You've got the tools and the smarts to steer it where you need it to go. :)



 

Page processed in 0.019 seconds.