Cumulative caps..

[jeth]

Just thinking about something and managed to confuse myself...

How do you go about calculating the overall HPF effect caused by all the input/output/blocking/coupling caps in a long signal chain? Presumably, despite all the buffers etc seperating them the seperate high pass filters are in effect cascaded through the signal path and result in a cumulative filter effect. Would seem simple enough to work outif all filters were equal, but alas...

 

[JohnRoberts]

In series it is simply additive. Two -3dB @ 20 Hz HPF will add to -6dB @ 20 Hz .

JR

 

[jeth]

That's what I figured, but how do I go about working out the combined response of all the filters throughtout the signal chain, bearing in mind they may not all have same fc?

 

[Guest]

[quote author="jeth"]That's what I figured, but how do I go about working out the combined response of all the filters throughtout the signal chain, bearing in mind they may not all have same fc?[/quote]

Multiply their functions and calculate.

 

[clintrubber]

Or simply measure it ?

(check where the response has dropped -3.0103 dB :wink: )

 

[SSLtech]

Graph it, begin with the lowest fc, then overlay a subtraction of each fc, increasing in frequency (for low-pass/high-cut rolloff)

For example;

-6db/8ve @ 20kHz & -6db/8ve @ 30kHz & -6db/8ve @ 40kHz

will give you the following:

flat to 20kHz, -6db/8ve slope to 30kHz, then -12db/8ve slope to 40kHz, then -18db/8ve upwards.

The quoted example is for a low-pass; High-pass is the same but working from right-to-left, and you work DECREASING in frequency to keep things easy.

Keith

 

[JohnRoberts]

It's been a while since I had to deal with this, but if I recall correctly, the approximation I used, was if a real HPF pole was -3dB at a given tuning frequency, it would be, -1dB at 2x that pole frequency and -.1 dB at 10x that pole frequency.

So to simplify, 10 2Hz poles in series would be approximately -1 db at 20 Hz.

I used to have a computer program I wrote running on my old H-11 where I could plug in actual values but I haven't turned on that PC in 25 years.

JR

 

[jeth]

Many thanks all..

The SSL method makes most sense to me.. I'll trust my results more with a visual representation of the filters combined effects.

From the answers received it seems that this isn't necessarily a calculation that people make on a regular basis, that surprised me as I thought it was fairly crucial. Is this because most folks just make sure all filters in the signal chain have Fc significantly below audio frequency, or are you all building circuits completely free of blocking caps?

 

[SSLtech]

John's information is excellent about the CURVE at the 'corner' (i.e. the onset of rolloff).

But if you're looking for how known curves accumulate, the stuff which I babbled about will get you an idea of how it all adds-up.

Keith

 

[JohnRoberts]

[quote author="SSLtech"]John's information is excellent about the CURVE at the 'corner' (i.e. the onset of rolloff).

But if you're looking for how known curves accumulate, the stuff which I babbled about will get you an idea of how it all adds-up.

Keith[/quote]

Keith Your info is about how they act in full rolloff, but real poles don't make straight lines. My rule of thumb also suggests how your example poles would interact below cuttoff, but for LPF instead of 2x it's .5x cut off frequency, etc.

Your pole at 40kHz would also be 1 dB down at 20 Khz so add that to the 20 Khz -3dB pole making it -4 dB at tuning. Further the 30kHz pole is 1dB down at 15Khz so now adding all three, we're down -1 dB at 10 kHz (from the 20 kHz pole), between -1 and -2dB at 15 kHz, and -4 to -5 dB at 20Khz. If you want a useful 20 kHz bandpass you don't start filtering until octaves higher, especially if using multiple cascaded natural poles. I rarely set my lowest LPF pole below 40-60 kHz for audio.

But I used to sell test equipment too, so I knew my customers could and would check my work...

JR

Note: my rule of thumb is for cascaded one-pole filters. When using two-pole active filters the Q of each filter set defines behavior around tuning.

 

[Guest]

[quote author="jeth"]

From the answers received it seems that this isn't necessarily a calculation that people make on a regular basis, that surprised me as I thought it was fairly crucial. Is this because most folks just make sure all filters in the signal chain have Fc significantly below audio frequency, or are you all building circuits completely free of blocking caps?[/quote]

Xc=1/(6.28*C*F)
It is simple in case of blocking caps.
Calculate functions for all R-C networks and multiply them, then build a graph. If you've designed some special active filters you know their functions so can multiply as well, then draw a graph and see.

Some active filter design software allows you to build complex schematics then see graphs, I played with them some time ago for fun, got some intuitive feelings as the result, and returned back to calculations on a napkin.

 

[PRR]

> that this isn't necessarily a calculation that people make on a regular basis, that surprised me as I thought it was fairly crucial. Is this because most folks just make sure all filters in the signal chain have Fc significantly below audio frequency

Life is short. Most caps are cheap. Make most caps way too big/small, so their effects are way outside the band of interest. If some cap is "expensive", like a 2,200uFd speaker blocking cap, or dolphin-oil coupling caps, make your other caps even more over-size. If you need a specific simple roll-off, make most caps big and one cap just barely big enuff.

John's short-cut approximations are very handy to know. 1dB down an octave up from the 3dB corner. Around half a dB at 2 octaves up. If he sez 0.1dB a decade up, I'm sure that's true; I never stacked that many caps that I would care. Personally, I'd rather have LF response falling-for-sure than fret over 0.5dB errors at frequencies where loudspeakers are -10dB or worse.

On "units" like a power amp, tape deck, etc, you pick overall +/-3dB or +/-1dB points and meet them. Note that tape, and some kinds of power amps, and many feedback amps, are liable to rise before they fall.

Obviously a customer can collect 5 units each specced +/-3dB at 50Hz, and wind up -15dB (or +15dB!) at 50Hz. This is why the FCC used to require a test from mike input to transmitter carrier, a test to see if accumulated small errors were all leaning the same way. It's why a lot of older broadcast gear has 15KC trim caps: if your chain wound up -2dB -1db +0.5dB -1dB = -3.5dB at 15KC, you could trim your last box for +2dB@15KC and arrive at -1.5dB overall.

 

[JohnRoberts]

I agree that micro managing this isn't typically necessary. I did a lot of work with consoles and you can have a surprising number of DC blocking caps in series with a single audio path.

While it doesn't look that pretty on the schematic, you use DC blocking for pretty much every level trim or pan because customers "perceive" scratchy pots as broken or dirty, and otherwise bad product.

Finally, after my research in phono pre's I don't like to let electrolytic capacitors set any audio frequency pole anywhere in my bandpass.. so I make all the dc blocking electrolytic poles much much lower and then use one well placed film cap to reliably establish one real pole to define the LF skirt. Any electrolytic smudge will be down octaves lower, and if you establish the quality pole early the electrolytics never get exercised where they could express terminal voltage effects.

JR

 

[jeth]

Ok, thanks PRR and John for clear explanations and practical advice...

Everything in Johns last post makes perfect sense to me so i think I'll take that approach.
On that subject, I'm on the learning curve and really attempting to take in an unhealthy amount of information here and elsewhere. There are, as we know, so many differing opinions and component options that at times one doesn't know where to start selecting. Having just about worked out what's probably better bthan what when it comes to film caps I'm now looking at electros for both signal path and PS bypass and am lost... quite incredible how many capacitors are being churned out into this world..
Anyway, recommendations for sensibly priced electrolytic caps for coupling and bypass please, nothing controversial as I don't want to be yet more confused. Compact package size would be an advantage, just something known ,easy to find and reasonable quality.
Failing that advice on how to go about choosing, eg, types of cap to look out for, specs etc

 

[bcarso]

"I'm on the learning curve and really attempting to take in an unhealthy amount of information here and elsewhere. There are, as we know, so many differing opinions and component options that at times one doesn't know where to start selecting. "

This is a pretty reliable site for information. The recommended adult dose of information here is well below LD25 levels, most of the time. For LSD 25 I refer you to deoxy.org and other sites, or PM CJ.

 

[PRR]

> really attempting to take in an unhealthy amount of information

Most people would rather talk than listen. Listen to others talk, or listen critically and honestly to the dang speaker.

In digital computing there is a joke: Write-Only Memory (i.e. you can never read back what you stored). The converse would be true Read-Only Memory (reading something that has never been written to). Beware of minds which are Read Only. (There is computer "ROM" but they really mean it was written once then read-only forever.)

> sensibly priced electrolytic caps for coupling and bypass

Use the ordinary Panasonics as sold by DigiKey and others. They really are very good caps.

In some places, for some ears, there may be a "better" cap. That's something you will have to find for YOURself, because there is very little consensus about what cap is "better". In a few cases, audiophiles have selected "worse" caps to "flavor" the sound (or disguise complementary flaws).

Yes, the Panasonics are "too cheap". But I sense your position is like learning cooking, only a few hundred gourmet meals made. While a 40-year chef may use greek salt and moravian lamb, he would advise a beginner to use any plain fresh fault-free salt and lamb, and concentrate on the assembly and presentation. Panasonic (and several other name-brands like Sprague and Illinois) are fresh fault-free salt.

> incredible how many capacitors are being churned out into this world..

Everything interesting is AC.

Most fun AC stuff needs some dull DC to hold it up.

So we need lots of parts to work-different for AC or DC: caps or chokes.

Caps are cheaper than chokes.

We have LOTS of toys. (TVs, DVDs, cellfones, thermostats, PCs, cars.........)

We use LOTS of caps.

When you make one toy, it is $10 for parts and $100 for labor (if you value your time).

When you make a million toys, the labor gets a lot of practice. You can get high production rates from ignorant workers (or robots) in low-wage lands. Then parts costs dominate the budget. Then the Factory Buyer is told to look for better price. If the order is large enough, the major cap-makers will run a custom value or size or voltage; also every minor cap-maker will be pitching a lower bid to get work. So between the too-many types in a single company's line, and too-many companies fighting for market crumbs, the choice is mind-blowing.

DigiKey, Panasonic. Find a shape and size, appropriate value and voltage, check the temperature (105C near large tubes or heatsinks). Get on with life.

Yes, if you make a career of cap-shopping, you can get the "same" cap for a penny less, or a looser leakage spec for 2 cents less.... in DIY, that kind of decision making keeps you away from the Main Point (music).