KLe said:
Hey Chris
Are you enjoying the extra speed, dynamics, articulation, detail, bass ectension, etc, that this approach is providing? Sounds like it is both good and bad at the momment?
Sounds like you might need to try the 0.022uF rectifier caps. What do you think? I would suspect that because this is allowing the amp to be more revealing that controlling Noise becomes even more critical. Do you think that the 4 pole PS caps have run in enough?
[...
Hi Chris: with the 4 poles in parallel do you notice the sound that Lars has discribed ... the midrange is hard and cold, and the soundstage is blurry ???
Hi,
Well before I'd end up wanting to play around with the DSP to have it up for where the amp left off and since the upgrade I haven't touched it can't see myself wanting to again. Everything has plenty of body to it now, even the highs to my amazement, body on the highs = big air.
It can certainly have the impression to sound thin at times and rough around the edges at others but I've found them both likely flaws of poor recordings showing through almost offensively in some cases but I mostly listen to MP3 or WAVs now anyway.
The thin sensation can also be attributed to what I mentioned in another thread as being a very black background, the slightest detail shines through even on the upper end of the frequency response, sometimes portraying a silky shimmer and other a harmonically rich metallica ring. This renders everything from symbols, acoustic strings, horns, etc to all have an ambundance of air where the slightest flaw in the recording does show throw, and it seems that alot of recordings don't do at least something well.
I can't blame the caps for it though because on some recordings there couldn't be more richness, you could probably give it a richer feel by having it a little smeared and of slower response which is what I found I had with the Cerafines, but that sort of coloration isn't a recorded effect.
I don't think I'd trade this in for an effect though because everything now sounds perfectly real, if you listen to a live recording you're right there in the audience and can tell which people around you aren't signing, listening to orchestral /symphony I used to have the viewpoint of sitting in the audience, no longer possible, it slams you right into the middle of all the horns.
I believe they're burnt in by now but may continue to improve still, if they do it's that last little bit that's hardly noticeable.
I dont' think woppling as it were can occure since they're not at all being starved at the source, and currents still have what little ESL and ESR there is in the wires to help balance them, as always, and I've wired it with that in mind.
They're still not ideal components and I think a snubber could be beneficial to tweak the Q of the supply so that it sounds to your liking, true as with any.
I'll also likely snub across the secondaries right at the bridge rectifiers and see what happens with that, which is about the extent of my plans for the supply.
Regards,
Chris
classd4sure said:
Hi Chris,
Thanks for sharing all this. So I have to throw my Cerafines out of the window it seems, they were quite expensive so I will experiment as well. I may get some of those Jensen 4 poles and try it on my tweeter amps first, these caps are even more expensive than the cerafines. By the way, did you try them on ZAPpulses or on UcDs or on both?
Best regards
Gertjan
Hi,
It's been a few days since I first powered up the 4 poles. On a good recording I find the sound lacks for little to nothing, but it's time I gave them a bit of a work out and so I plan to do so today. I think my biggest concern is in crosstalk between the wires I used to parallel the caps with. I used the upmost care in layout however it's less than ideal, PCB traces would be better.
I'd like to just point out that this isn't the best measure of a paralleled/non paralleled experiment, as it wasn't exactly what I'd originally set out to do.
I'd also like to point out the very opbvious, the previous tests were with 5000uF per amp per rail @4 ohm load.
Going from the Cerafine to the BHC T-network was a brilliant change, in fact showing me how much the Cerafine was hiding, but the Cerafine wasn't all bad.
Basically I think one of the reasons for the extended response is the reduced ripple due to having enough capacitance for the load which I'm driving. The originally plan was OK for an 8 ohm stereo load which I never used, it's always been 4ohms.
I was going to make up for this paralleling the Cerafines with the T-networks as some of you may recall, but after hearing the difference between the two, I really didn't want anything more of the Cerafines. That's not at all to say they can't make a great amp, it's just not the sound I was after, also their immence size is greatly prohibitive. The amp and toroid have a silky kind of effect to them enough as it is, add the cerafines to that and it becomes unnaturally smooth and overly relaxed, also in that added softness lies the microdetail which is smeared out. A snubber is likely to tug the PSU's Q to sound like you want it to regardless of caps used, so the best would probably be a properly snubbed PSU with the cheapest industrial caps you can find that will do the job.
Another thought is that a set of bleeders may also be a welcome addition at this point in relation to output noise.
Speaking of cheap caps, these 4 poles are pretty costly, but given the holographic sound I'm now experiencing, with a supply so very basic thus far as toroidal xformer, dual FRED bridges, and 4 caps with not a single thing else, unless you want to count my lightbulb, the better parts really do make a difference, and I can easily leave it as is, even though the options for it remain open.
They're still also but 1/6 to 1/10 of what the "equivalent" in Black Gates would cost, and I feel there's nothing better.
Regards,
Chris
It's been a few days since I first powered up the 4 poles. On a good recording I find the sound lacks for little to nothing, but it's time I gave them a bit of a work out and so I plan to do so today. I think my biggest concern is in crosstalk between the wires I used to parallel the caps with. I used the upmost care in layout however it's less than ideal, PCB traces would be better.
I'd like to just point out that this isn't the best measure of a paralleled/non paralleled experiment, as it wasn't exactly what I'd originally set out to do.
I'd also like to point out the very opbvious, the previous tests were with 5000uF per amp per rail @4 ohm load.
Going from the Cerafine to the BHC T-network was a brilliant change, in fact showing me how much the Cerafine was hiding, but the Cerafine wasn't all bad.
Basically I think one of the reasons for the extended response is the reduced ripple due to having enough capacitance for the load which I'm driving. The originally plan was OK for an 8 ohm stereo load which I never used, it's always been 4ohms.
I was going to make up for this paralleling the Cerafines with the T-networks as some of you may recall, but after hearing the difference between the two, I really didn't want anything more of the Cerafines. That's not at all to say they can't make a great amp, it's just not the sound I was after, also their immence size is greatly prohibitive. The amp and toroid have a silky kind of effect to them enough as it is, add the cerafines to that and it becomes unnaturally smooth and overly relaxed, also in that added softness lies the microdetail which is smeared out. A snubber is likely to tug the PSU's Q to sound like you want it to regardless of caps used, so the best would probably be a properly snubbed PSU with the cheapest industrial caps you can find that will do the job.
Another thought is that a set of bleeders may also be a welcome addition at this point in relation to output noise.
Speaking of cheap caps, these 4 poles are pretty costly, but given the holographic sound I'm now experiencing, with a supply so very basic thus far as toroidal xformer, dual FRED bridges, and 4 caps with not a single thing else, unless you want to count my lightbulb, the better parts really do make a difference, and I can easily leave it as is, even though the options for it remain open.
They're still also but 1/6 to 1/10 of what the "equivalent" in Black Gates would cost, and I feel there's nothing better.
Regards,
Chris
ghemink said:
Hi Gerjan,
They were the ST models which actually still include the 5532 on the 180 watt modules, only with the 100k input impedance and the 12V zener already from factory so it makes a real easy op amp swap now. On the 400 the ST has the opa's.
So I changed the op amps for AD8620BRZ
Replaced the coupling caps for a wire. Nothing else has been done to the modules yet, I'm working on that now.Power supply is again a 500VA toroidal from plitron with mu-metal shielding and an electrostatic shield. I can wrap unshielded input wires around it and not get hum.
Rectifiers are IXYS FRED 68amp 600 PIV.
Jensen 4 pole caps, 2 15 000uF 63V per rail.
Amps are wired in bi-phase for better supply sharing.
Throw in two switches and a lightbulb, there you have the whole thing.
BTW after hearing how revealing my new caps are I was forced to cut about 10' off the microphone cable based interconnects. They're now something like 15' long
So I have is pretty simple.
Regards,
Chris
classd4sure said:
Hi Gerjan,
They were the ST models which actually still include the 5532 on the 180 watt modules, only with the 100k input impedance and the 12V zener already from factory so it makes a real easy op amp swap now. On the 400 the ST has the opa's.
So I changed the op amps for AD8620BRZ
Power supply is again a 500VA toroidal from plitron with mu-metal shielding and an electrostatic shield. I can wrap unshielded input wires around it and not get hum.
Rectifiers are IXYS FRED 68amp 600 PIV.
Jensen 4 pole caps, 2 15 000uF 63V per rail.
Amps are wired in bi-phase for better supply sharing.
Throw in two switches and a lightbulb, there you have the whole thing.
BTW after hearing how revealing my new caps are I was forced to cut about 10' off the microphone cable based interconnects. They're now something like 15' long
So I have is pretty simple.
Regards,
Chris
Hi Chris,
Thanks for the additional info, very helpful. I have the very early UcD180s with the NE5532, using black gates non-polars as coupling caps. I have UcD400 AD8620s to replace them but have not installed the UcD400s yet (just for the woofers in an experimental setup).
If the power supply caps make such a big difference, I wonder what difference the 470uF caps on the UcD400 module would make. Plan to replace those with 680uF 100V panasonic FC, although not everybody thinks that is a good choice, anyway, I will try them sometime.
Thanks and best regards
Gertjan
ghemink said:
Hi,
You're extremely welcome.
I plan on changing those as well, likely the same choice as you, big as I can fit/what everyone else did.
I'll likely also change the output filter cap for one just slightly smaller, since I'm using 4 ohm loads it can't hurt.
My final mod will be to load the input stage
The aux supply for the input stage will remain an outstanding long term project, but with the way it is right now it's hard to believe it can still get better. It's already a quality unlike anything I've heard.
Best,
Chris
classd4sure said:
Hi,
You're extremely welcome.
I plan on changing those as well, likely the same choice as you, big as I can fit/what everyone else did.
I'll likely also change the output filter cap for one just slightly smaller, since I'm using 4 ohm loads it can't hurt.
My final mod will be to load the input stage
The aux supply for the input stage will remain an outstanding long term project, but with the way it is right now it's hard to believe it can still get better. It's already a quality unlike anything I've heard.
Best,
Chris
Loading the output stages is also one of my agenda points, however, progress is slowly due to large workload in my daytime job.
The jensen web-site talks about the transfer impedance defined as Vout/Iin, this impedance is extremely low (mOhm order) which basically says that input current variations which you get when the caps get charged from the diode bridge (large current every 10ms if you have 50Hz mains) only result in a small voltage change on the output terminal. So very good isolation between what happens on the input and output. As also suggested on the Jensen site, this makes them very good caps for SMPS applications with much better HF suppression than any other CAP.
Wondering what the DC resistance between input and output pins are, could not find any spec for that, could just find that when paralling them in the "wrong" way, DC series resistance could increase to "Ohms". So I guess that the DC resistance is relatively high. Do you have any data on that?
Thanks and best regards
Gertjan
ghemink said:
Hi,
I am curious if they did not mean impedance? Resistance from Input to Output is not measurable on my DMM, nor should it be as it's just the foil.
Even paralleled the usual way they say not to use more than 3 or the impedance will rise, diminished returns I guess.
I haven't tried it the other way, no plans to either.
Regards,
Chris
classd4sure said:
Hi Chris
I have been away, but, it was nice to see your reply(s) when I returned. It would appear that parralleling 4 pole PS caps is now a mandatory requirement?
Chris, by your comment, "I dont' think woppling as it were can occur since they're not at all being starved at the source, and currents still have what little ESL and ESR there is in the wires to help balance them, as always, and I've wired it with that in mind.", it would appear that you think that there is enough resistance in the wire to control the woppling effect that Lars is talking about. Is that a correct assumption?
Sounds terrific
thanks
Hi,
If the woppling effect is the first cap in the chain taking "charging priority" and sucessively passing on less and less charge as the chain grows in length, all I can tell you is I have no chain. I've basically extended the inputs and had them meet in the middle, dead center of that wire, essentially giving each cap it's own feed, looks like a "T". You'd have to then redefine this woppling thing as something kin to current hogging to keep this argument going.
What would that be based on, time constants and tolerances? If you're going to somehow monitor each individual cap to produce an error signal and throw them in a feedback loop so each one is servo controlled OK but.... realistically..
Where I think the series resistors is actually advantageous is in decoupling one cap from the other to damp any parasitic resonances, but that has zip A. all to do with "woppling" as it's been defined. I don't believe woppling is actually an issue, and if it really is, I'd consider it a wiring one, like having too thin a trace or something.
So that's my theory on it and based on the scientific evidence provided on it thus far it's as valide as anyone else's.
Here's the caps wired:
http://www.diyaudio.com/forums/attachment.php?s=&postid=786554&stamp=1134231190
I had to reduce the pic alot to be able to post it but you'll get the idea I guess. Also I seperated the power rails going to the modules, had too much interference that way at higher output(never heard it before), but it seems good now. It's _very_ hard to tell. I'm still searching for that perfect recording that will let me know definitively. But I guess if I have to try that hard it can't be too bad huh?
Too bad those caps dont' fit in my case unless I lay them down. I'll have to be inventive there, find some case to put them or a rubber mat. could use a shelf but still .. I'll see.
Regards,
Chris
If the woppling effect is the first cap in the chain taking "charging priority" and sucessively passing on less and less charge as the chain grows in length, all I can tell you is I have no chain. I've basically extended the inputs and had them meet in the middle, dead center of that wire, essentially giving each cap it's own feed, looks like a "T". You'd have to then redefine this woppling thing as something kin to current hogging to keep this argument going.
What would that be based on, time constants and tolerances? If you're going to somehow monitor each individual cap to produce an error signal and throw them in a feedback loop so each one is servo controlled OK but.... realistically..
Where I think the series resistors is actually advantageous is in decoupling one cap from the other to damp any parasitic resonances, but that has zip A. all to do with "woppling" as it's been defined. I don't believe woppling is actually an issue, and if it really is, I'd consider it a wiring one, like having too thin a trace or something.
So that's my theory on it and based on the scientific evidence provided on it thus far it's as valide as anyone else's.
Here's the caps wired:
http://www.diyaudio.com/forums/attachment.php?s=&postid=786554&stamp=1134231190
I had to reduce the pic alot to be able to post it but you'll get the idea I guess. Also I seperated the power rails going to the modules, had too much interference that way at higher output(never heard it before), but it seems good now. It's _very_ hard to tell. I'm still searching for that perfect recording that will let me know definitively. But I guess if I have to try that hard it can't be too bad huh?
Too bad those caps dont' fit in my case unless I lay them down. I'll have to be inventive there, find some case to put them or a rubber mat. could use a shelf but still .. I'll see.
Regards,
Chris
classd4sure said:
Hey Chris
Why not, you are probably right. Interesting, that you had to seperate the power rails more than usual. Almost, indicates that you are getting more current delivery to each amp module.
You have done it with the Jensen's, I wonder if I can do it with the BHC T-Net caps?
But, Jensen 4 pole caps work so .... Well, thanks again
IVX said:Lars, if you don't work for lcaudio anymore, so maybe "700xe" developed not by you personally? BTW, "Lowest THD in the industry!" etc, what about -3db(350W@4ohm) THD? If claimed only 1,10 or 100w level, then i'll (sure that Bruno can better else) win the "700xe" too easy (.00075%@15w/4ohm .0014%@25w THD/Freq=Flat), with the TDA8939 in the any variations - UcD or hystersis based modulator.
Sorry for being so late, but I've not paid attn to class D until recently.
Am I misunderstanding something or are the thd figures referred logarithmically better?
http://www.dynamicprecision.no/Pages_UK/00_index_UK.html
BerntR, it's a class D forum.. for linear amplifiers (A, AB) less THD isn't so difficult, actually good opamp with biased followers should bit Dynamic Precision at 15W too.
PS:my point was is, that huge loop gain aren't wise way to build good amp, and vice versa 30db loop gain with THD =.02% (BTW UcD400 .005%@10W, AFAIR) at the rated power is pretty cool. I didn't see THD less then .0003%@10W with any loop gain from TDA8939, but i tend to suspect my EMU1212 and measurement setup at all, which was not enough precision.
PS:my point was is, that huge loop gain aren't wise way to build good amp, and vice versa 30db loop gain with THD =.02% (BTW UcD400 .005%@10W, AFAIR) at the rated power is pretty cool. I didn't see THD less then .0003%@10W with any loop gain from TDA8939, but i tend to suspect my EMU1212 and measurement setup at all, which was not enough precision.
Hi Ivan!
I think all of us agree to , that an amp, be it linear or class D should be or sound as good as possible , before applying feedback.
In the case of Dynamic Precision A1 and its distortion figures and the philosophy behind it, all I can say is that this very good performence doesn´t depend on of a lot of feedback.
Unfortunately the website seem not to work at the moment, it used to have a discussion about distortion.However,here is another link, unfortunately whithout the discussion;
Dynamic Precision AI , you have to click on Dynamic Precision at lower left, then AI S,- no luck with links at the moment.
Maybe the race for very low distortion figures leeds some constructors to apply too much nfb.But the most clever of them stay away from this.
At a hifi-show , a salesman said about a well-know and expensive and in my opinion over-rated amp ;" it has 0.0002 % THD, so now you may understand just how good it is". I couldn´t help to ask if the low figure really matters , when it comes to music.
Low distortion figures can be nice, and if they are obtained in a clever and intelligent way, why not?
I think all of us agree to , that an amp, be it linear or class D should be or sound as good as possible , before applying feedback.
In the case of Dynamic Precision A1 and its distortion figures and the philosophy behind it, all I can say is that this very good performence doesn´t depend on of a lot of feedback.
Unfortunately the website seem not to work at the moment, it used to have a discussion about distortion.However,here is another link, unfortunately whithout the discussion;
Dynamic Precision AI , you have to click on Dynamic Precision at lower left, then AI S,- no luck with links at the moment.
Maybe the race for very low distortion figures leeds some constructors to apply too much nfb.But the most clever of them stay away from this.
At a hifi-show , a salesman said about a well-know and expensive and in my opinion over-rated amp ;" it has 0.0002 % THD, so now you may understand just how good it is". I couldn´t help to ask if the low figure really matters , when it comes to music.
Low distortion figures can be nice, and if they are obtained in a clever and intelligent way, why not?
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.