|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
69.12.241.189
Hi, DIYers. I've spent considerable time now reading many posts regarding PSU design in this forum. Aside from the personality conflicts, there's a lot of good info. I still wish to "rest" some
conflicting or confusing points."low DCR" is "always better" and high DCR chokes and large caps are "bad". Yet, especially with SE , low ripple or very low ripple is usually "cleaner sounding" and is much more necessary in SE circuits, because of the lack of cancellation occurring in PP circuits.....yes/no???
This makes me think why not a series regulated supply, with a "medium" size "audio quality" cap at the end, sorta like JD Verdier's (French)
designs or later creations by Thortston L. Especially if you are using an "ultrapath" cap on the cathode of the final tube, which "some" say
bypasses the PSU (and it's imperfections) ??? yes/no??? But, yet there must be some reasons why this angle is frowned upon ??? what is it?Now, there seems to be a recent and interesting discussion about using extremely low DCR chokes, with carefully chosen small value caps to create an "organic non regulated" PSU with ultra low DCR, like 20 ohms! But how could we also have "sufficiently low ripple"??? What is sufficiently low ripple?
I've also noticed a relatively common view that specific relationships between choke and cap values as they "stack up" can be "better" than others that seem to be just a version of overkill. For example: A popular SE amp uses diodes 470uf - 10H - 470uf, which seems sort of dangerous and overkill to me, yet is an easy way to get low ripple.
Another view likes for example: 4uf - 10H - 8uf - 10H - 16 uf, which has a "progressive" filtering sequence that apparently has "better"
wave form or other desirable features, but kinda has to have higher DCR, unless you're using 10H/700ma transmitter chokes (with low dcr and "normal inductance") weighing 40 lbs each to run a 3W amp. ??? is this the clue?To make things worse, I read on an interesting Japanese website, about using an equal size choke or chokes on the bottom (negative) side of a typical Pi filter, as you did on the top. Claims were made as to why this was "much" better, improving bass accuracy and "quality". I haven't tried that. There seems to be something worth looking at here.
yes/no? ??I guess we're all just trying to find a "do-able" way to elegantly support a simple SE circuit with a PSU that will give it the best chance to show what it can do. Without getting into a way higher and complex technology to support a relatively low-tech circuit, what boundaries should be considered "the best you can do" , and what is an example of this "least compromised as possible within reason" PS design?
Follow Ups:
A lot of questions which have been asked, I'll just focus on the ultrapath topology. In theory, it's a shorter signal path in that the top of the output transformer is coupled to the output tube cathode via one capacitor, whereas in the more common topology, it is coupled via two capacitors: 1) the last power supply capacitor, which goes from the top of the transformer (B+) to ground, then from ground to the cathode via 2) the cathode bypass capacitor in parallel with the cathode bias resistor. Often, the cathode capacitor is an electrolytic in order to have a large enought value to provide bypass down to subsonic frequencies. However, electrolytic caps are considered to be less than optimal sonically. Remember that the signal has to have a complete circuit so in the standard single ended topology with output transformer, the signal must pass through the power supply.With the ultrapath topology, the signal can pass through the ultrapath capacitor, which is generally chosen to be one high quality film or oil cap, presumably with less coloration, so the ultrapath capacitor provides a bypass to the power supply capacitor, and thus in theory also bypasses power supply colorations. Lynn Olsen and Jack Elliano say that this is indeed the case in their experience.
OTOH, the ultrapth capacitor also provides a low impedance path from the power supply to the cathode, so any noise or hum in the power supply is actually injected into the circuit at the cathode and thus amplified by the output tube. By comparison, in the standard topology without the ultrapath capacitor, power supply noise and hum are injected into the circuit at the top of the transformer and so are not amplified by the output tube. So the ultrapath topology needs to have a quieter power supply than the standard topology. One does have to wonder whether PS colorations might also be injected into the cathode and thus potentially worsen the coloration problem - I don't know the answer to this question.
If you search the archives, you will find some comments from DIY builders that power supply capacitor choice in ultrapath preamps can make a significant difference so I think it is still an open question whether this circuit is superior with respect to isolating power supply issues.
Dear Jim,What I always wanted to know... can you use the ultrapath approach with fixed bias i.e if no cathode resistor/cap are present?
I vaguely remember that I read somewhere that this is not possible, but it never became clear to me.
Sure - just use a good cap as the last PSU cap.Fixed bias does half of "ultrapath" - it removes the cathode resistor bypass cap from the output signal current path. The other half is to run the signal current that goes through the power supply, through a GOOD cap.
I'lll just add one point which is often skipped over. You've already pointed out that "...the ultrapath capacitor provides a bypass to the power supply capacitor...". The bypass action works because the bypass capacitor impedance is smaller than the power supply impedance, so the signal current goes preferentially through the bypass instead. For this to work, the power supply impedance can't be too low. Look at the circuit below, and note the 100 ohm resistor between the power supply and the amplifier circuit. The 40uF ultrapath capacitor has an impedance of less than 100 ohms at frequencies above 40Hz.
Good point. In fact another way of looking at the ultrapath capacitor is that it IS a local power supply capacitor and a bypass for the cathode capacitor.
Could you use a choke to feed the OPT (sans final cap) therefore
creating a high AC resistance to "contain" the AC loop? And still have a low DCR supply behind the choke?
....is not science. A subjective bandwagon of opinion can be difficult to stop. When someone claims 'better' sound and is tenacious about it, it's hard to dismiss it like dust under the rug. Sometimes there are technical discrepancies that should be investigated. For example, if you model those low H input supplies, you'll notice that the peak current demands on the rectifier are about the same as those that are cap input. A conventional H supply has about 1/3rd of this. And subjectively speaking, I regularly use very large H with low C, and they sound fine to me (150h,3500r;.42c), they certainly are not 'bad'.I compared once 2 simlilar chokes that only differed in age and R, and ended up liking the one better with 3 times higher R (it was also the older made in the 50's one), but in principal, I prefer the idea of lower R if it would actually sound better.
So the answer is, there is no superior silver bullet design, and only experimentation will decide which is liked best.
> > > To make things worse, I read on an interesting Japanese website, about using an equal size choke or chokes on the bottom (negative) side of a typical Pi filter, as you did on the top. < < <Any links to the site? I would be interested to have a look at this - similar to common mode connection?..
Kind regards
Raymond
Ultra-consumers: Spending money they do not have to buy things they do not need to impress people they do not like.
Most of these confusing points are over-interpretations of small bits of data. For example, Jeff Medwin found that when he replaced chokes and transformers in his amps with much larger units, his system sounded better to him. That's an actual fact. Extending that to claims like "it was the low resistance that made the difference" or "it would be the same with a different amplifier" is just a theory. Theories gain support when a sufficiently large variety of other experiments can eliminate the other possible theories. It's a legitimate reason for pursuing this in the forums - with enough enthusiasm, more experiments will be done and will get reported, and we will all learn more. But the fact is, actual controlled and useful experiments are a lot of work, while coming up with - and posting - alternate theories that might explain the few actual reported facts is easy. So the theories greatly outnumber the facts. To add to the confusion, we humans have a tendency to express our theories as if they were facts.If you want real answers to the questions posed, you'll have to do the experiments. As you have observed, there are a great many opinions or speculations about what the answers might be. Once you give up the illusion that the answers are known, and accept that opinions and guesses are not facts, your design life becomes much less fraught with angst - at least, it works for me! :^)
I'd like to make a final observation which I find helpful to remember: amplifiers and their power supplies are a system - i.e. the combination is greater than either part by itself. Their interaction is complex, and the important parameters of one only make sense in the context of the other. You cannot simply dismiss the amplifier as a "load of such-and-such current" when designing a power supply, nor can you dismiss the power supply as a "low-impedance voltage source" when designing the amplifier.
Well, that's my humble opinion on the subject. Not a fact, just an opinion!
Thank you Paul.I appreciate your well spoken and reasoned commentary. I've never participated in a forum really, so I'm just getting up to speed on understanding the boundaries and depths that other participants have
established. As a guitar player (also) I know that theoretically "bad"
quality materials and even designs can still sound "great", but that world has a lot more voodoo than hi-fi, mainly because of the one HUGE variable: The player.
I'll admit that having been a player for 30 years (several instruments) helps me "hear" and constructively critique hi-fi reproduction, because I really know in my cellular memory "what it's supposed to sound like". I also believe that hearing is one of the senses that has a direct correlation with speed of brain, and an ability to complexly anticipate, because of understanding well the roles of the players who are playing what you are listening to.
In the end, it's all a lie anyway, because it's still "reproduced"
sound. But not getting stuck on that, we can enjoy many players who are, dead for example. So, it's all still worth "most" of the trouble...
Well thought out, balanced and expressed.
Thanks for the "feedack". I guess I knew I wouldn't get any (or many) hard facts by just asking for them, and placing the responsibility back upon myself, and my experiments, is appropriate. Honestly.Although I can "like" a variety of amplifier circuits, I always end up
preferring (or "coming home" to) a single 45 or 2A3, AC filament, and just a single low or medium mu triode in front. To me it's really just splitting hairs between RC coupled(choke or resistively loaded), IT coupled or direct coupled. I'll admit that a DRD direct coupled version, has a little "something" extra. I have found when direct coupling anything, lower (or very low) ripple in the PSU is "required". I achieve this the standard way with sizable chokes and modest cap values. Perhaps because the amp circuit doesn't have huge voltage or current demands (in a relative sense) the "highish" DCR
supply I make, "performs pretty well", at least it sounds very good.
I guess because I feel lucky in a sense to have "committed" to this
simple amp type, I don't seem to need to confuse myself (at the moment) by searching for another. So, still being a busy brain, I guess the place to focus, is on seeing whether I can "support" this
simple SE 2A3 "better" by investigating PSU "improvements". When I look at Gary Pimm's SE 300B schematic and the commentary, and because I somehow like and respect him (even though I don't know him) I start wondering what elements of his design could I apply to my lower power version. Because I guestimate that I have good output iron already,
am OK with the AC filament hum level (which I can't hear when tuned out) the frontier is about tighter bass, speed and darkness of silence or background. These fringe characteristics are probably about loading and/or PSU. Again, because of the "relatively" low demands of the little amp, improvements are most likely to be in a world beyond anything I can devise out of purely organic "combinations" and enter into, unfortunately, things I don't quite understand. There are obvious "losses" occurring in conventional loading, as small as they may be. This seems like a place to investigate. Any further comments directed to the more specific case I've described would be greatly welcomed. thanks
We audio geeks are a funny breed. Somebody else would read your post and say the guy knows what he likes so why does he want to change? Sadly I understand why:)Notice that Gary has no power supply shown. Hmmm, could it be that the power supply isn't the important part? Instead I'd point to a few big items. Number one, just about everything has an active load. Number two, pains are taken to keep the voltages where they should be on stages that operate with constant current. Number three and the biggie IMHO, the circuit strives to keep audio circuits in a tight loop and return them locally rather than through a common ground and/or power supply. I think this is key and I see it happening time and again on better designs. Moreso in push pull than SET but still there is a common thread.
A choke, transformer, or a CCS helps "isolates" the circuit from the power supply. Ditto on the parafeed with the cap to cathode on 300B. I suggest you look more into these areas. I think parafeed may bring you some of the improvments in bass you seek.
With regards to the low dcr stuff....I have looked into it a bit and in fact dcr has little if anything to do with it. The main interesting point is the voltage on the first cap is greater than 1.4 times the secondary voltage. Circuit values are critical. I would not advise the design for anyone other than an expert who was looking for a challenge.
A little time spent with Duncan's PSUD will show a very low DCR supply rings like crazy. If you are interested in the low dcr thing and understanding it better search on "flywheel" over at tube diy.
Thanks Russ.After checking back in on Gary's schematic, I realize I already ground this way. Isn't this called "star" ground? I think I'll park the ultra low dcr concept for the moment, and try to take the circuit a step further.
Since the primary of an SE output designed for standing current
is in itself "a choke", the advantage of parafeed's "isolation" is then to A: "use a bigger choke" (or CCS) and B: to loose the "problems" of standing DC current in the OPT , is that rightish? Gary's cap seems to be in the plate circuit (rev. 3). I guess I've seen it done both ways, what's the difference? I wouldn't know how to design a CCS, but I'd like to learn.
On a tangent, having played around briefly with a 26 once, it had a profound "darkness" to it, that I thought was kinda refreshing, yet after a little while, I realized (perhaps the Rp) it was just "missing" a lot of HF noise and music simultaneously.
Gary's comment about his amp "not loosing composure even when things get wild" is, I guess what I'm after. Most of my listening is less "wild" stuff, so what I have already "stays composed", mostly, but sometimes when wild stuff comes, it does seem to "not quite keep up". Not distort exactly, just, I'm not sure what. This is one of the things I'd like to "fix". Maybe it's also partially a power issue, but I'm not ready to go dc filaments to get more power, although I would if...
I would prefer to exhaust the potential of a bottle that I can operate with AC filaments first. Has someone developed a CCS plate load for a 2A3? that's "public" knowledge or is there a book on designing CCS's? or?
Is this the right forum for this or is tube/ DIY more appropriate?
The advantage of parafeed is allowing you to close the loop locally. The output tube's current is confined to the primary of the transformer and it returned directly to the tube's cathode. In a normal connection the return must be through the mian power supply cap. Hopefully the parafeed cap would be smaller and of higher quality.Putting the cap "on top" keeps the DC voltage out of the transformer. Not having to have the transformer handle either DC current, so no air gap required, as well as no high voltage DC, so not near the insulation required, can allow for a better and cheaper transformer. There are trade offs but considering your comments I think it is an area for you to consider.
Gary and Bottlehead sell CCS boards. The manual alone for the Bottlehead CCS is good reading and will explain a lot.
DC filaments wouldn't give you more power. You might be thinking of fixed bias compared to cathode bias. I would stay with AC as long as hum wasn't a problem.
I have decided it is best to use a solid state amp for the low bass and bi-amp with an electronic crossover. Not asking a couple/few watt amp to handle the low bass cleans up the mids/highs and helps the amp "keep up" on more complex music. Of course YMMV.
Russ
PS Yes tube diy is a good place to ask. Just takes awhile to learn who's advice to follow:)
I like the idea of better cheaper transformer, and wonder what the trade offs are? When the plate is loaded only by the OPT and an "ultrapath" cap is in place from top of cathode to B+ side of primary, does this actually remove the signal from the PSU? When I say DC filaments for more power I meant the use of more powerful triodes that would have to have dc to be quiet enough, like 300B, not thinking that dc alone would do anything for power, I realize I wasn't clear. I will look into the CCS boards you recommend, thanks.
I saw an interesting circuit on page 888 of the red book RCA radiotron designers handbook, that used a single input with two outputs that seemed like an elegant simple way to do a biamped/crossover thing.
I have single full range speakers so I can't do that.
This post is made possible by the generous support of people like you and our sponsors: