Comparing modern electrolytic caps

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Conrad,

The instrument is a HP3577A, and the corresponding S-Parameter test set.
I connect between Port1 and Port2 a test jig, which is a small PCB with lemo connectors and 50ohm track, and I short the trace with the DUT to the ground plane. The graph shown is the magnitude of S21, the transmission parameter.
In some places I had put also the phase.

In that jig the value of S21 depends on the: source and load impedances, the freq and the impedance of the DUT itself. The magnitude that one sees is practically the deviation from the basic attenuation [6dB] of a 50ohm - 50ohm series att. pad. An ideal short would give a line at the bottom of the screen for all freq. Open ct gives no att, 0dB at top.
inductance produces a 20dB / dec ascending line, C produces a 20db/dec descending line with freq, R is flat. Ref line for 1 ohm is at -27.95 dB, ~28dB.

By the way, this is the factory standard way of testing caps by producers, there are many instances of these graphs in [better] cap data sheets.
I have made ~hundreds of posts with these tests here, and at Diyhifi. There is also a very good pdf from mrjam in that diyhifi thread. In the diya thread posted above, there is also some pic with the jig. The quality of the jig is important, but mainly when you arrive at smd territoy. For testing big elcos, even a simple jig works nice.

Ciao, George
 
Re - performance of big elcos:

As You could notice from the many tests, yes, it' s absolutely true: even big size elcos give surprisingly good results. They are definitely good out to ~hundred kHz range, if not MHz.

But. every mm counts, each mm of lead contributes ~ 1nH stray inductance, and their size does not permit to place them very close to the circuit.
 
I have to say the difference between Panasonic FC and Rubycon ZL is quite large.

The Panasonic is subjectively much stronger in bass. The Rubycon is more neutral but seems to have a rising response. I have read some articles about electrolytic capacitors and their constructions, etc. The sonic differences do not make sense to me in theory.

But if you dare to spend 10 minutes and swap in/out your regulator ADJ cap with a Panasonic and a Rubycon (try same values and ratings), you will also note the great differences.

I can get the sound balance from Panasonic FC caps, but not from Rubycon ZL caps. However, after many comparisons, I do like the sound characters of the Rubycon ZL much better.

Next I will try to use all Rubycon ZL and add a shelving low pass filter between 2k to 15k.
 
Sometimes math is like a tree root in the woods- something to trip over in the dark. Still, I look for two things from a capacitor at high frequencies, low impedance and some reasonable phase shift, say -80 degrees or better. Those two parameters can be converted to any other two parameters like capacitance and dissipation factor, or whatever system one likes to work in. Not to mention a series or parallel model.

Here's the point I'm not too confident about. Let's say a big electrolytic approaches 0 degrees phase shift fairly quickly, while maintaining a low impedance. At zero degrees it would be a pure resistor, but let's say we haven't quite reached zero. Using the series model, you end up with a huge capacitance, but with a significant series resistance. In the parallel model, you end up with a tiny capacitance and a slightly larger resistance in parallel. In either case, the ability of the cap to deliver energy is compromised because of either resistance or lack of storage capacity. Or, for the case where the phase shift is zero, what's the energy storage of a resistor? IMO, there better be some good HF bypass capacitors downstream!
 
I forgot to emphasize that these tests notwithstanding, I don't feel to be an expert in capacitor quality. Quite obviously there is no direct connection between these electrical parameters [cap ESR, ESL, capacity] and sound quality. And unfortunately I don't believe that snubbers would help, either.
{though this does not mean, that there are a lot of purely technical issues in bypassing, which instead are much better mapped by the help of a network analyzer}

On the other hand, I would subscribe strongly to the findings of HiFiNutNut. I had been playing / swapping caps, for example, in the feedback arm of a MyRef amp - which is also a very sensitive position for differences. And had found very similar things like HiFi - Fc -s ~okay, Pana FM for me much like he describes the ZL's, very clean but with a sting, Philips 037 very smooth open but less resolution etc.. OScon are, for me, something similar to that ZL description again - very very clean & transparent, but with a metallic edge.. though I had only SH SC

If the classic electrical parameters coincide poorly with sound, I start to feel that DF instead may have a better convergence. [Dissipation Factor]
For example, I feel like more and more convinced, that high voltage ratings [which consequently has less DF] are sounding ~always ? better.. For example now I'm using an Elna Cerafine 100V cap in that feedback point. And this same experience I had with my PCM dac, also.

George - awaiting for the interesting findings of you all.. :)
 
Re: Comparing modern electrolytic capacitors

sandyK said:
"Sometimes you worry too much. Most colouration will your speakers contribute with."

There is nothing wrong with trying to obtain the most neutral response from your system before the loudspeakers.
The majority of preamplifier and amplifier designs seem to suffer from mid range colouration due to the inability of normal bypass capacitors to have a very low impedance in this critical area of the spectrum. FWIW, I also disagree with Douglas Self's normal use of a 10 ohm resistor and a 1,000uF capacitor inserted in the -VE rail of his amplifier's supply to the front end in order to improve PSRR, for the same reason. It is far better to supply the front end from additional voltage regulators or gyrators.

SandyK

Hi. You might be interested to know that my most recent designs have a slight variation on that plan; the input stage and VAS are fed from a separate supply a few volts below the main V- rail; this gives more output and allows enough voltage headroom for the VAS circuitry. This aux supply is RC filtered but not regulated in any way. The ripple figures are very good and there is certainly no "mid range colouration".
 
Eric Juaneda said:
Hi HiFiNutNut,


I wrote an another article about LM317 sounding.

LM317, understanding and listening

Each capacitors around LM317 greatly improve or degrade sound.

Eric

May I just wonder one issue.
Now Lm317 is equipped with bandgap reference, which is known to be low noise reference type, Zener reference has much more noice.

And putting two zeners in reference pin does not help.
I think that using zeners around Lm317 will degrade it's noise performance

:eek:
 
Eric Juaneda said:
About 'snubberization', you have related article written by Jim Hagerman : Calculating optimum snubbers
Problem is more complex when snubber is in a complete power supply.

Eric
You misunderstand. Snubbers are normally for switching elements and for reducing voltage spikes and reduce ringings. This "snubber" is for lower a certain resonance.
 
Hi jpanu,
jpanu said:


May I just wonder one issue.
Now Lm317 is equipped with bandgap reference, which is known to be low noise reference type, Zener reference has much more noice.

And putting two zeners in reference pin does not help.
I think that using zeners around Lm317 will degrade it's noise performance

:eek:
Good remark. Datasheet mention that LM317 is stable without external capacitors. Datasheet don't mention that LM317 is VERY noisy without capacitors. Reducing impedance on Ref pin greatly reduce noise. For noise reduction capacitor is better solution.

Adding zener, like you can see in other schema, greatly improve sound. For this reason I adopt zener. Zener + capacitor reduce noise and improve sounding.

Eric
 
Comparing modern electrolytic capacitors

"Hi. You might be interested to know that my most recent designs have a slight variation on that plan; the input stage and VAS are fed from a separate supply a few volts below the main V- rail; this gives more output and allows enough voltage headroom for the VAS circuitry. This aux supply is RC filtered but not regulated in any way. The ripple figures are very good and there is certainly no "mid range colouration"."

Hi Douglas
Thanks for chipping in. All the recent amplifiers I have built in recent years have been based around your designs. Silicon Chip magazine uses your topology in it's recent 20W Class A design.
(and others) I have also learned much from reading your books. Having said that, have you actually tried using a regulated supply for the front end with a VERY low noise and VERY low supply impedance?
A slightly modified version of the JLH "ripple eater" shown in a recent post in this thread has been very successful with several designs based on your work. I realise that you are an opponent of subjectivity, but it really does make a very worthwhile improvement, as does further striving for closer balance of the input pair. A point which you stress the importance of in your books, but don't take quite far enough. If you are interested , send me an email, and I will forward details. AKSA has already kindly offered me a technical explanation for why there is a marked improvement with close balancing of the input pair currents. You can get part way there by replacing the tied base and collector junction of the CM with a 1N4148 diode, IF the input pair is closely matched for both HFE and VBE. Incidentally, Halcro has patented something along those lines in the late 90s, but that is not quite what several of us have been doing in Sydney for much longer.
Regards
SandyK
 
I spent the past weekend to get my system to sound balanced and finally got some positive results. The system now sounds correct at macro level. Only adjustments in micro level are required to get it sound optimally. It will still be a couple of week's work.

I installed a line level passive shelving low pass filter before the volume pot. It was modeled using LTSpice so the modeling should be very accurate.

There are 4 trimpots used in the low pass filter to allow adjustments to the corner frequency as well as the gain (loss). Only a single 0.01uF polystyrene capacitor and the 4 trimpots are used in the filter.

The result seems to confirm that the impedance of the Rubycon ZL does go very low at high frequencies, while the ESR at lower frequencies is much higher, which results in very peaky sound, i.e. exaggerated at high frequencies. The Panasonic FC is similar, but the impedance does not go as low as the Rubycon ZL at high frequencies, therefore sound more balanced.

I don't know the exact figures (I have not had the time to measure them) since I turned the 4 trimpots many times before I got a balanced sound, but the rough figures would be like these: a response that is about 1dB down at 5k gradually moving down to -5dB at 20kHz would suit the all Rubyzon ZL configuration (1000uF || 0.01uF at the adj pin of the LM317/LM337, 470uF + 1R after the regulator. 0.47R in series at the output of the LM317/LM337 then 2,200uF || 2,200uF at the circuit board, then for each opamp 100uF || 0.1uF per rail).

The above shelving low pass filter corrected the response from 1kHz up but at lower frequencies the Rubyzon ZL still did not sound balanced. So I replaced the main reservoir capacitors 2,200uF || 2,200uF with a Panasonic FC 4,700uF. Below the 1k region, the Panasonic sounded more balanced. The shelving low pass filter was still needed, with the corner frequency shifted down possibly half to one octave.

What I have learnt:

(1) Capacitors do sound different. The differences can be very large. Note that I say different, not better or worse. Yes, for sure, some capacitors sound better than others, but I guess for quality capacitors like Rubycon ZL and Panasonic FC, they can all sound good or bad depending on how they are used.

(2) When used in power supply bypass or others (such as at the adjustment pin of the VRs), if the main circuit does not have infinite power supply rejection, the impedance of the power supply can heavily influence the sound by altering the frequency response / phase of the main circuit. In that case, the impedance of the power supply should be considered as part of the main circuit design in order to get the response right (in reality, this is rarely done). In many circuits, the impedance of the power supply heavily depends on the capacitors used.

(3) Many people are happy about using high ESR capacitors after the LM317/LM337 to prevent oscillation. I am not happy with that. I want better bypassing so I use higher value, low ESR/ESL caps near the opamps. To prevent oscillation, I had to use some serious resistance after the regulators. Therefore, in my circuit the sound is far more subject to the capacitors than a normal implementation. People have their individual priorities. I stand by my own choice. I look at it this way: stability comes first, good opamp bypassing (low impedance and good transient response) comes second, then the rest. I rather deal with the capacitor colorations to ensure optimal stability and bypassing.

(4) Opamps do not have infinite power supply rejection. The rejection can be poor towards higher frequencies.

(5) One problem with circuit design is that many designers assume capacitors are ideal electronic devices - a pure, idealistic capacitor with no internal resistance and inductance. Of course, there are difficulties. The ESR/ESL varies with frequencies and electrolytic capacitors have low tolerance and change tolerance during their life. If I build a circuit designed by somebody else, I would make sure I also build the power supply with the same capacitors used by the designer based on which the circuit was designed and measured, or I could possibly not get the same result, unless the circuit has very high PSRR.

(6) A lot of time it is meaningless to talk about if one brand of capacitor is better than the other, or how a capacitor sounds without considering the underlying context. In my initial post I gave descriptions to various capacitors. Those descriptions were correct based on the circuit I used. Their own individual sound was possibly due to their different impedance more than anything else, which can be corrected. The Rubycon ZL sounds very "clean" probably due to its rising frequency response which emphasizes the harmonic contents of the music. During my experiments I corrected the Rubycon ZL's response, and felt the Rubycon ZL did not sound like a Rubycon ZL anymore. Initially I found the Panasonic FC sounded a bit "metallic", but after some impedance manipulation the Panasonic FC could sound very much a Rubycon ZL! Once the impedance is corrected, I found no subjective difference between the Rubycon ZL and the Panasonic FC.

(7) Panasonic FC is easier to use. Use larger values to lower the ESR and impedance, if applicable.

(8) Subjectively, when using larger values of the Rubycon ZL or Panasonic FC, it is possibly unnecessary to use 0.1uF or I would create an impedance dip somewhere. Use 0.01uF is often enough.

Thank you for every one who contributed to this thread.

I have learnt heaps from you.

Regards,
Bill
 
The problems I have had with my LM317/337 have finally been solved!

Yesterday, SandyK brought in his JLH "ripple eater" to my place. We first listened to my system without the JLH for 30 minutes then plugged in the JLH then repeated the same music tracks. The JLH was used in the PSU that powers up the preamp/line level active XO.

The JLH was "brand new" with 4 lowly Jaycar capacitors without any run-in.

We immediately noticed the sound improvements in accuracy, clarity and details. The bass was obviously stronger and better defined. However, the sound was slightly "edgy". Later we added a pair of 2,200uF Rubycon ZL at the output of the JLH. The edginess was then gone.

The improvement was significant. Note that the LM317/337 regulator has probably been the worst part in the audio chain of my system and this makes the difference very pronounced. The difference is no less than upgrading from a 2,000 CD player / amplifier to a 5,000 CD player / amplifier.

So I could not be happier. I did not expect it would make such a difference but it did.

I highly recommend the use of the JLH! I will be building a few including one for the CD player analogue stage.

Thanks to SandyK who is so keen to share his JLH to everyone in the forum.

Regards,
Bill
 
COMMENT.

The initial slight edginess referred to by Bill, will normally disappear after the larger capacitors have had a chance to form . This will normally take at least several days of extensive use.
When the so called "Audio Grade" capacitors are used, in virtually every audio circuit, they will cause cycles of varying sound quality over a considerable period of time before fully forming. This often results in an initial warmth, with recessed sounding treble. After leaving on overnight, they usually appear to lose much of the initial warmth, combined with a more prominent treble. This may persist over a week or more, with the SQ cycles becoming less evident, before completely stabilising. This is also the experience of many members from Rock Grotto Audio Forum.
In many cases, after this period of electrolytic capacitor forming, there is often a reduction in the amount of sibilance previously evident before fitting the JLH "Ripple Eater"

SandyK
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.