| Kashmire |
I've often wondered if the electrolytic output capacitor of the Zen amplifiers will ultimately limit the amplifier - i.e. the Zen 9 is better than the preceding versions, how long until new innovations no longer help, because the design is held back by the output coupling capacitor?
I've searched the forums a few times over the past months looking for this topic, but haven't found a serous discussion (other than occasional ramblings about bypassing with film types).
How much sonic contribution does the output capacitor make? It is my opinion that it is low, and there are much more important things to address. However, in a new amplifier design, should efforts be made to remove the output capacitor, or accept it as a mechanism that allows for other improvments?
(the tube guys fight over this regularly, especially with the parafeed configuration) |
|
|
| Kashmire |
Wouldn't you know it? More searching and I found what I was looking for:
http://www.tubecad.com/2004/blog0014.htm
Comments welcome. Especially this question about push-pull amplifiers:
Op-amp servo loop to control DC offset or output coupling capacitor? |
|
|
| h_a |
If you follow Nelson Pass and Erno Borbely, then you don't need to care too much about an output cap.
If you don't, feel free to investigate the topic - preferably with nice lab equipment - and please come back and tell us!
Cheers, Hannes
PS: oh and don't forget, the same applies to the tons of caps in your crossover. True there are mainly PP caps used there, but things like non-polar elcos do exist and are used. |
|
|
| analog_sa |
| quote: | Originally posted by h_a
If you don't, feel free to investigate the topic - preferably with nice lab equipment - and please come back and tell us!
|
And what could the "nice" lab equipment possibly show? That capacitors have leakage? :)
I used to be terrified by large coupling electrolytes. After listening to a Pass preamp for a while i am now merely frightened. |
|
|
| ctong |
| I am also interested in the subject. I have read at different places that one should bypass a electrolytic capacitor with a small film capacitor for high frequencies when bypassing a power supply but not when coupling a signal. Can't think of an explanation for the latter claim. Is there any merit to it? |
|
|
| h_a |
| quote: | | And what could the "nice" lab equipment possibly show? |
THD+N, Intermodulation distortion...just as Douglas Self has shown that bad caps can cause distortion in certain places. And he proofed it with measurements.
@ctong: the idea is to lower the impedance at higher frequencies.
By the way: guys, don't start to argue with me, please. I just wanted to state my thoughts on this and I'm really not keen on flame wars about merits and the pure evil of elcos. With argueing I mean subjective flaming, in case somebody has some links to articles with technical background - you're welcome!
We all know that everybody has a different opinion on this, really.
Cheers, Hannes |
|
|
| ctong |
| quote: | Originally posted by h_a
@ctong: the idea is to lower the impedance at higher frequencies.
Cheers, Hannes |
Hannes,
What I mean is that is there any reason to believe that bypassing a coupling electrolytic capacitor with a smaller film one is a bad idea. |
|
|
| Kashmire |
No flames here.
Nelson Pass wrote an article about bypassing large electrolytic capacitors.
http://www.hometheaterhifi.com/volume_4_2/nelpass.html
Regarding Doug Self, you can read about his theory of where distortion comes from here.
I agree with Broskie that PSU caps are commonly in the output signal path. Below is his example of the signal path for a push-pull amplifier with a bipolar power supply (+/- voltages):
Notice during the up or down cycle of the output waveform, different parts of the power supply (and filter capacitors) are being stressed.
Below is Broskie's other example. The amplifier on the left is the push-pull with bipolar power supply with the standard + and - filter capacitors. The amplifier on the right is push-pull with monopolar power supply, using a singular filter capacitor and an output coupling capacitor.
Now, since the monopolar PSU design uses the same capacitor for both up and down cycles of the waveform, instead of commutating between PSU rails (from plus to minus), shouldn't this be considered a benefit? |
|
|
| Kashmire |
**** please don't let this post-script divert from the original intention of this thread, which was using an output capacitor on push-pull amplifiers ****
Here's the hidden part of the Broskie analysis: in the Zen, the PSU filter caps really don't see much of the AC waveform. The amplifier acts mostly as a constant-current load, so the PSU is rarely being modulated by an AC signal. Therefore, in a true class-A amplifier, the filter caps really aren't part of the signal path.
But an output capacitor is. But the output capacitor is also needed to function, so it plays the role as a bulk storage capacitor. How is this? Picture the Zen amplifier. During positive waveforms, the active MOSFET pinches off current, forcing the CCS to re-route its current through the load. During negative waveforms, the active MOSFET goes low impedance. All of the CCS current shoots right through it. So how does the load see a negative signal? The output capacitor discharges into the active MOSFET, adding its current the CCS's current. Therefore, the Zen is truly a special case!
**** please don't let this post-script divert from the original intention of this thread, which was using an output capacitor on push-pull amplifiers **** |
|
|
| GRollins |
| quote: | Originally posted by Kashmire
I've often wondered if the electrolytic output capacitor of the Zen amplifiers will ultimately limit the amplifier...how long until new innovations no longer help, because the design is held back by the output coupling capacitor?
|
Take five colored lights. Shine them on a piece of paper. Reduce four of the five by random amounts, but not to zero. Call the fifth one an electrolytic capacitor. Call the other ones anything you want...resistors, gain devices, wire, power supply...anything. Assume that the goal is zero coloration. i.e. all lights completely out, but accept that this is an unattainable goal.
Given that you will never get the rest of the circuit to such a state of perfection that the electrolytic capacitor is the only thing holding the circuit back, it's pointless to beat yourself over the head with this.
No reasonable person argues that electrolytics are sonically transparent. The trick is to minimize the influence of the cap on the sound.
The best thing to do is to get rid of the cap entirely, but given the topology of the amps in question, that's not going to work. Yes, there are DC-coupled amps, but then you don't have a cap-coupled Zen any longer, you have something else.
The next best thing is to try to minimize the burden the cap places on the sound. Options include:
--Bypass the electrolytic with a film cap
--Drop the electrolytic entirely and use a film cap by itself; accept the loss of lower frequencies...perhaps even turning this into a benefit in the case of a bi-amp system
--Enclose the cap in the feedback loop
--Put two of the amps back to back and feed them a balanced signal--the speaker floats and no longer sees the DC. This comes close to changing the amplifier topology. You can argue this either way.
There are some who claim that film bypasses are bad for any of a dozen reasons. I'm not among them.
Those who want to argue based on the price of film caps are rather missing the point. Either it sounds better or it doesn't. Only after that question is settled can you reasonably try to look at a cost/benefit ratio. That will be up to the individual. But be careful to separate the benefit and cost questions.
Grey |
|
|
| Kashmire |
Thank you, G. I agree wholeheartedly. That's my my prior post concentrated on the necessity of the coupling capacitor in the Zen. I'm afraid I did myself a disservice, though. The Zen discussion is a distraction from what I'm really wanting to find out:
| quote: | Originally posted by Kashmire
Comments welcome. Especially this question about push-pull amplifiers:
Op-amp servo loop to control DC offset or output coupling capacitor?
|
If I can rephrase it again,
Is a push-pull amplifier best served with:
1. a bipolar power supply (and symmetrical PSU filter capacitors), possibly a DC servo to maintain low DC offset, and potential hazard to your speakers? (i.e. if a signal-level bias transistor fails, the output MOSFETs can swing to full current, although there are ways to mitigate this, which increases complexity again.)
2. a monopolar power supply with a single filter capacitor and an output coupling capacitor?
It seems that you can trade an output capacitor for PSU complexity (bipolar vs. monopolar), robustness (hazard), and output DC offset (servo loop). You also trade the commutation between the two (biploar) filter capacitors as the waveform crosses the zero axis. |
|
|
| Kashmire |
Consider this quote from Doug Self
| quote: | The Doug Self website
Distortion Mechanism #8
... however, in AC-coupled designs the output capacitor may contribute significant distortion. |
Doug Self maintains that an output capacitor belongs on his list of Eight Distortion Mechanisms of a Generic Power Amplifier.
But also consider this countering discussion by our Zen Master Nelson Pass.
| quote: | Originally posted by Nelson Pass
It is my opinion that the role of capacitor is overstated as a source of distortion and sonic degradation.
|
I've never built a solid-state push-pull amplifier before. I have built tube SE, tube PP, and solid-state SE (Zen). I'm trying to understand why the solid-state guys use bipolar power supplies. Is there a good reason (i.e. is Doug Self correct)? If Nelson Pass is correct, why does he use bipolar supplies?
p.s. the famous audiophile quote, If one of these is right, the other must be wrong! |
|
|
| GRollins |
You're taking the Zen amps out of context. Once you do so, you're going to have conceptual problems.
They are not intended as state of the art contenders. They are projects, rendered simply and foolproof-ly (look, Ma, I just made up a new word!). If they meet your needs...great. If not...great.
You're trying to read more into the circuits than is intended.
If you want to alter the circuits, then have at it. But you're spending a lot of time and effort criticizing a rock for not being a cloud. The rock doesn't care. Neither does the cloud. You're only going to frustrate yourself. Take a deep breath, back up, and reassess the Zen amps and their place in the universe.
Grey |
|
|
| Kashmire |
| quote: | Originally posted by Kashmire
Comments welcome. Especially this question about push-pull amplifiers:
Op-amp servo loop to control DC offset or output coupling capacitor? |
| quote: | Originally posted by Kashmire
I'm trying to understand why the solid-state guys use bipolar power supplies. |
|
|
|
| Kashmire |
| quote: | The Doug Self Site
The distortion generated by an AC-coupled amplifier's output capacitor is more serious, as it is not confined to low frequencies. A 6800uF output capacitor driving 40 W into an 8-Ohm load gives mid-band third-harmonic distortion at .0025%, as shown in Fig 32. This is five times more than a Blameless amplifier generates mid-band. Also, the LF THD rise is much steeper than in the small-signal case. |
According to measurements performed by Doug Self, using an output capacitor on a power amplifier increases his amplifier's distortion by a factor of five.
I don't fully subscribe to all of Self's pontifications, although, I have never built one of his "blameless" amplifiers, either. I suppose I should build one and listen before I make any more specific commentaries on his theories.
Nevertheless, the fact remains that Self measured that a capacitor imparted fives times the distortion than the active devices. This doesn't make any sense to me, because according to the illustrations above (previous posts), the PSU capacitors are in the output signal ground-loop path in push-pull amplifiers (such as the Self blameless amplifier).
So here's a starter list:
1. Self measured distortion of a small signal capacitor (into a 680 Ohm load), with and without a DC bias. The signal got worse with DC bias applied. In his power amplifier measurements (8 Ohm load), no DC bias was used, as his push-pull output rests at zero volts.
2. Since we know the PSU capacitors are part of the output signal ground-path return loop, shouldn't the PSU capacitors impart similar sonic contributions?
3. The PSU capacitors are held a large DC bias (full PSU voltage), all of the time.
4. The blameless amplifier has a lot of feedback, so everything was operating within the feedback loop, with the exception of the output capacitor in his tests.
Since he was measuring something outside of the feedback look, I contend his measurements of 5x distortion due to an output capacitor was an "apples to oranges" comparison.
The question I am posing may be altering by tiny bits, but remains the same: "why use a bipolar PSU with Class-A push-pull amplifiers, when a monopolar PSU can be used with an output capacitor?" From an engineering standpoint, I can't see why the Pass F4 in Class-A mode would be significantly different in either configuration.
Anyone care to comment? |
|
|
| bananatuan |
Dear all
Now I'm doing a Alpeh P but I only have 4 capacitor GE 5uF 400V for input cap. I wonder can 5uF use in input cap instead of 10uF as original design ?
Thank for your advises / Banana :confused: |
|
|
| Pan |
Kashmire,
| quote: | | Therefore, in a true class-A amplifier, the filter caps really aren't part of the signal path. |
The filter caps are very much in the signal path in a single ended Pass style amp as well as in a typical complementary class A amp.
You need to go bridged to get dynamic influence of caps out of the equation.
Or you can use Andrea Ciuffoli's capacitance multiplier solution as used in the power follower of his. Still you have that cap left for DC blocking on the output though.
/Peter |
|
|
| KSTR |
"Therefore, in a true class-A amplifier, the filter caps really aren't part of the signal path"
A class-A topology is only one of a set of pre-requisites for that. "Out ouf the signal path" translates to "don't see any signal current" in the end, and that's is the only thing that counts.
Thus:
- the output stage must draw constant current from the supply, regardless of signal and load (class-A is not necessarily equivalent to constant current, it just says that no output device ever gets turned off)
- if split supplies are used, this must be valid for both of the supply halves. With split supplies a bridge output topology is needed to fulfill that requirement, otherwise only the sum of currents is constant (while the indiviual currents are copies of the output current). Peter wrote that already, I like to emphasize a certain point again: With a constant current bridge, the center tap is freed from load currents, and that allow the use of a single supply (you must still provide a "GND" reference point, of course).
With a single supply and output cap, there also is bridge topology, albeit one of the halves is passive and not actively controlled with a signal. Redraw the schem and you see it... without constant current draw there is no benefit, the supply cap sees a full-wave rectified copy of the output current, whereas in the split supply each cap sees different half-wave rectified currents. Depending on the location where you set the split point (for AC), any combination of these variants are possible.
The rectifying action is mainy what causes the problems, forcing the supply to handle signal current harmonics up into the hundredths of kHz.
- Klaus |
|
|
| GRollins |
| quote: | Originally posted by Kashmire
The question I am posing may be altering by tiny bits, but remains the same: "why use a bipolar PSU with Class-A push-pull amplifiers, when a monopolar PSU can be used with an output capacitor?" From an engineering standpoint, I can't see why the Pass F4 in Class-A mode would be significantly different in either configuration.
|
You're making the assumption that the power supply caps don't play a part in the distortion. They do. Substitution of film caps for electrolytics in smaller circuits with lower current demands can make an improvement. It's impractical to use film caps for power amplifiers in a commercial setting, but it's not undoable for the DIYer.
I've done it with a tube amp, which is cheating because the Joules of storage are related more closely to the rail voltage than to the number of uF, but the results were very nice indeed.
| quote: | Originally posted by bananatuan
Dear all
Now I'm doing a Alpeh P but I only have 4 capacitor GE 5uF 400V for input cap. I wonder can 5uF use in input cap instead of 10uF as original design ?
Thank for your advises / Banana :confused: |
Using a 5uF cap instead of a 10uF cap for coupling will raise the low frequency cutoff point; you'll have somewhat less bass. (I don't have a schematic within easy reach at the moment--will try to scratch one up later.) How much it will change depends on the impedance the cap sees. Whether this is a problem is up to you.
| quote: | Originally posted by KSTR
The rectifying action is mainy what causes the problems, forcing the supply to handle signal current harmonics up into the hundredths of kHz.
|
There's a lot of different ways this could go but I'm out of time. For the moment, I'll just say that it depends on the design and parts choices of your power supply.
Grey |
|
|
| 4fun |
Hi,
Mr Renardson has some interesting designs that include output capacitor. Clearly he does not see the cap as a major distortion contributor if enclosed by the feedback loop.
Renardson Audio Pages
MJR-6 and later MJR-7 are interesting, look at the distortion numbers. |
|
|
| Conrad Hoffman |
| These discussions are always filled with can, may, might, could and a bunch of other weasel words. Everybody wants a generalization to make life simple, and it just doesn't work that way. Circuits differ and the effect of a cap thus differs as well. About the only generalization one can make is, if you measure across the cap and find no ac voltage, the cap isn't affecting the ac signal. If you do find an ac voltage, the cap is affecting the ac signal, and it's doubtful it's acting as a pure attenuator ;-) As for the magnitude of what's acceptable, that's entirely up to you. |
|
|
| GRollins |
There are also philosophical differences to account for when deciding what is or is not in the signal path. Nelson feels that the power supply isn't in the signal path. As I understand it, his view is that the signal path consists of the shortest trace between the input and the output. My view is closer to Conrad Hoffman's, but perhaps more extreme--I think everything matters, although some things matter more than others. I'm not sure I'd go to the expense of using Caddock resistors in a power supply regulator, for instance. And...okay...if you pressed me against a wall and held a gun to my head, I'd probably admit that I don't get all that worried about what kind of resistor I use to bias the power LED for the front panel.
I now hang my head in shame. My secret is out. I'm not sufficiently tweak. My membership in the Lily Guilder's Society hangs by a thread.
Now, as to coupling caps and how to go about that sort of thing, I think most everyone would agree that ideally the best thing is to have no coupling cap at all. Some topologies don't allow that, however. If you choose a circuit that needs a coupling cap, my feeling is to avoid an electrolytic if at all possible (yes, this includes Black Gate or whoever is regarded as making the best electrolytics this week) and if you can't avoid the use of an electrolytic, then at least bypass it with a film cap.
The film cap need not be hideously expensive to give you some benefit. Digikey sells Panasonic polyester film caps at least as large as 10uF 100V. They're very reasonably priced. Granted, polyester is pretty much at the bottom of the film cap range, sonically, but it's still better than an electrolytic by a country mile. Again, it all depends on how many uF you need to get a decent low frequency rolloff point.
And, yes, the low frequency rolloff point applies even to power supplies. Most people think of the power supply capacitance as a storage depot for electrons. That's 100% correct. But there are other ways to look at power supplies. One of them is as a short to ground for audio frequencies. The bigger the capacitor, the lower the frequency you can get rid of and the purer your rails will be. Conversely, you also want to get rid of high frequencies, so it's good to use a cap that has good high frequency response...which leads us back to film caps.
In a perfect world, you'd be able to buy a 10000uF Teflon film and foil cap for about $5. The last time I looked, we weren't living in a perfect world.
Rats.
Grey |
|
|
| bananatuan |
Noted all your comments with tks
:) |
|
|
| Babowana |
| quote: | Originally posted by Pan
The filter caps are very much in the signal path in a single ended Pass style amp as well as in a typical complementary class A amp.
|
In between the rail(s) and the ground, if there is a means that
makes the signal current idle or makes the signal voltages stop,
they will not be injected to the amplifier--i.e. we could say that
PSU is not a part of signal path.
As far as I know, all Papa's amps (incl. Zen sisters) provide such
means. You know? Papa is not that stupid . . .
Just my thought . . .
:darkside: |
|
|
| GRollins |
The various Zens with the Aleph current source and the Zen 5, for instance, don't feature a constant current draw. They may be others as well, but I can't seem to keep all of them in my mind at the same time.
Grey |
|
|
| Babowana |
The Zen 5 has the rail regulators and the big caps at the sources
of the regulator mosfets. You would see these from the whole picture.
:darkside: |
|
|
| GRollins |
Regulation does not make an audio circuit draw constant current. It serves as a synthetic capacitor, if you will. Capacitor multipliers don't cause a circuit to draw constant current, either. They act somewhat more explicitly as a synthetic capacitor.
I confess that I am of two minds as to whether a capacitor multiplier counts as a regulator. Part of me argues that to count as a regulator it should have an absolute voltage reference. The other part of me figures that it serves to smooth shorter term variations and that is sorta what regulators do--but only sorta. I tend to put cap multipliers in a separate category.
Grey |
|
|
| Babowana |
Hi Grey,
If the rail voltages are perfect dc, the PSU could not be considered
as a signal path. Zen 5 has the rail voltage regulator. But, as Papa
says in the paper, inside of the regulator MOSFET, there is the Rds
of about 0.25, which is a concern with respect to the ac modulation.
Therefore, the large caps (30,000uF) are added behind the regulator
MOSFET sources to squash down the ac modulation.
Anyhow, I understand what you are talking about . . .
I’d like to return to the original subject . . .
:darkside: |
|
|
| KSTR |
| quote: | Originally posted by GRollins
I confess that I am of two minds as to whether a capacitor multiplier counts as a regulator. Part of me argues that to count as a regulator it should have an absolute voltage reference. The other part of me figures that it serves to smooth shorter term variations and that is sorta what regulators do--but only sorta. I tend to put cap multipliers in a separate category. |
Yes, that's quite dependent on the underlying definitions. My take: "Cap multipliers" are not regulators in the classic sense of the word, there is no global feedback and the local (that is, intrinsic device~) feedback is low, the "error amplifier" is effectivly only the gain device as is. It's a weak regulator, one might say. Wether the reference voltage is fixed or floating is not a point in the consideration, from my point of view. "Cap Multiplier" is quite a misnomer also (real capacitance multplier circuits do look a little different, see here, pg.24f ), after all it is nothing more than a simple source/emitter follower. The term refers to the fact that with an emitter follower the emitter current is a multiple of the base current, which "loads" the RC-filter. With MOSFETs there is no load current on the RC or whatever filter one uses (ok, there are some small transitional gate charge currents), thus no "multiplication factor" can be assigned to it (other than infinity). That's why it is a misnomer and should be best avoided, IMHO.
Er, sorry for that off-topic excursion...
- Klaus |
|
|
|
