New, Cheap Dynalo...

[Beefy]

Forgive my skepticism, but what benefit could possibly be gained from custom caps in a fully DC-coupled amplifier?

[Maniac]

Forgive my skepticism, but what benefit could possibly be gained from custom caps in a fully DC-coupled amplifier?

Power supply of course, believe it or not, but I've found DynaLo and DynaHi family of amp to be extremely sensitive to caps. When I was first building my DynaHi a few years back (was 2006 I think), I installed two Panasonic FM caps on each DynaHi amp board... Well, DynaHi sounded every bit like the transistor amp that tube lovers hate, it was cold, gritty, harsh, and overly bright. I spent a lot of time checking out a LOT of stuff, nothing helped. At the end, I decided that the Panasonic FM caps is about the only thing that I haven't swapped out, and I bit the bullet there by swapping in some rather expensive ROE boutique electrolytic caps. The result? Well, the guy who got that DynaHi from me is happily taking that DynaHi around to his friend's place for shoot outs, and he is winning most of the time, and on rare occasions, there might be a tie or a lose, but that's usually against rivals costing 2 to 3 times the price of the DynaHi that I sold him.

After that, I did not go on build more DynaHi anymore, because it was more of a challenge to build one of those amps at least once, and learn something in the process. And boy did I learn a lot from building DynaHi.

The actual blind listening test is done with a good HDD player (FLAC files)and a PCM1794 DAC as a source, the custom cap actually contributes a lot to the sonic performance of the amp. We compared a fully upgraded HA-006++ and a similarly upgraded unit but uses the original WIMA MKS caps in the power decoupling, and the result was quite clear. I did not do the listening myself, but I did the switching around and the setup, all with the listener's back toward the equipment. I switched around randomly, and the listener is able to consistently pick out the difference between 006+, 006++ with WIMA and 006+ with FPP cap. I did try to tune it further, but it was getting to the point of diminishing returns at this point unless we redo the entire setup and go for a bigger unit with even beefier power.

[Beefy]

Power supply of course, believe it or not, but I've found DynaLo and DynaHi family of amp to be extremely sensitive to caps.

Put me in the 'or not' camp.

[Maniac]

that doesn't sound like a particularly good test protocol, especially if there was any conversation between you and the listeners.

The only thing we ever talked is volume higher or lower, and I'll set it for the listener, and also the listener indicates that I can switch to another unit after the listener have heard enough in one run. As for switching, I decide which to switch, and if I even switch at all, while listener only assumed that I switched to another unknown unit.

When I switched back to an unit that the listener have heard, I often get the comment that this unit is "surprisingly like one of the earlier one that was used", which was quite true as it was the same unit.

I'll list the routine of the test:

All unit is burned in with random music from my Rio Karma at least overnight, and all left their power on for quite a long time to keep the class A setup warm and ready to go.

1. I hand listener the headphone (Music ][ and ER4P), set amp to minimum volume and insert phone plug.
   
2. Listener indicated the volume to be turned up till listener says stop.
   
3. The music used in the tested is restarted from the beginning.
   
4. The listener indicated that the basic performance of the current setup is noted, and can go on to the next setup
   
5. I switched gear (sometimes not, but just make the same noises that switching gears normally makes)
   
6. Go back to step 2 until consistent result is achieved.
   

If listener cannot reliably identify the difference of two units, that basically means I've pretty much reached the point where it is not going to make enough of a difference anymore. (Or the listener is tired, which the test will be done at a later time.)

All the while the listener is back toward the equipment, and nothing else was talked. All idle equipment are still powered on while the session is in progress, so that the next time it is listened, it will require no warm up time. And oh, I also use the same ONE power cable from the same one socket on the wall for the unit currently being listened, so I also switch the power cable around as well as the RCA cable and phone cable. I'm not sure if the mini power cable would make enough of a difference, but switching that around is simple and quick, and with just a few seconds of power-down, the transistor won't take more than a few seconds to warm up to the original temp too.

Believe me that I want to know the TRUE result even more than you do, as good parts ARE expensive, and I don't intend to spend the budget on parts that makes little to no difference. I tried very hard in keeping the test result uncontaminated, and the listener too is quite interested in finding out if there's actually a difference. It was actually the listener who have suggested that the session will be done with back toward the equipment, and thus there will be no visual cues about which unit is being used.

[Maniac]

Put me in the 'or not' camp.

Well, you can build a DynaHi and find out if I'm kidding or not. I used to think those won't make that much of a difference, and learned the hard way by wasting a lot of time trying to see what other parts are causing the problem.

[Maniac]

the lack of real volume control completely invalidates your results.

What real volume control? The stepped volume control have extremely precise setting each step, and when compared, they ARE set at the same level. As for the original unit, the listener can reliably identify the unit day after day, the comment did not change over the days of testing that we have done here.

[grawk]

did you measure the output of each at the specified level to verify the levels match? Given the steppers, it seems really unlikely that they're exactly matched.

[Maniac]

did you measure the output of each at the specified level to verify the levels match? Given the steppers, it seems really unlikely that they're exactly matched.

Actually the steppers are made with 1% SMD resistors, and is extremely consistent at each step. Two added bonus with this stepper is that, since it uses SMD resistor, it is inherently non-inductive, and since it IS a stepper, it will have extremely good L-R balance. (unlike film type pots)

[grawk]

I'm not concerned about stepper to stepper matching, but amp volume to amp volume matching. Given the lack of fine granularity in level setting, I just can't imagine they're the same.

[Maniac]

When the listener is able to clearly identify the units day after day, comment remain the same from the first test to the last, no matter what volume (within reason of course), I was pretty convinced that the listener is not getting the placebo. All I tell the listener is "done, you can start", and all the listener tell me is "volume up/down", "I'm done". There are no visual cues, I even step back from the test area and observe the listener from a bit of distance (making sure the listener is not looking at the machines, which the listener never did look.).

Now, with the listener able to identify the 3 test units consistently day after day, with no visual, verbal or even body language cues. How does it get completely invalid? I'm not planning on writing an AES report or a science project, but to get useful information when tuning. True, it is not as air-tight as would be needed by the standard of a science report, but all I need is reliable info on tuning, not something to publish paper with.

[Maniac]

I'm not concerned about stepper to stepper matching, but amp volume to amp volume matching. Given the lack of fine granularity in level setting, I just can't imagine they're the same.

The amp's gain is set with a 1% precision resistor (0.1% on the ++ version), and in combination of a stepper, it will match to a degree that you won't believe.

Since the amp work in this fashion:

Input -> Volume control Pot -> Amplification circuit with feedback setting gain to a fixed setting -> output

Now, there are 3 variables in this equation.

Input = The same source, same cable, so there won't be any difference, since it is the same unit.

Volume control pot = Since each step is extremely consistent from unit to unit, 3 steps on one will come within about 1% of the other unit's 3 steps. Normal film pot have larger channel mismatch than that.

Amp's feedback setting/gain = set using 1% (006+) or 0.1% (006++) resistor, this also makes the amp's gain extremely precise from unit to unit.

Now, with all three variable firmly fixed, I can't imagine how a noticeable difference in volume can be achieved when two units in question are done in that fashion. Now with the original unit using film pot, I can't say for sure. However, the listener indicates that the difference between the original unit and the upgraded unit using stepper is quite clear, sound quality difference is very noticeable. The listener like the smooth volume control of the original unit, while liking the stepper's sound.

[tkam]

that's all well and good, we know how stepped attenuators work, but did you measure the outputs of the amps at a specified level?

I don't see why he'd need to, he using the same source into amps with the same gain, the amps had different gain level's I'd agree he'd have to measure to properly adjust the volume control but with the same gain seems kinda pointless to measure.

[Maniac]

that isn't what he asked.

does anybody else see the problem here?

it's invalid because you aren't controlling. the kind of well meaning test you're running is the kind of thing that gets people to believe in homeopathy. there is a big difference between "the test looks okay" and "the test is well designed and will produce reliable results," even if the differences between the two seem minor. it's fine if you believe in the results, but don't try to convince us of potentially dubious results that were obtained with potentially dubious tests.

it's entirely possible that the caps are doing what you say they are doing, but i see no evidence of it.

Well, no. The comment from listener is extremely detailed, the listener CAN and have reliably identified the 006++ unit with FPP cap and WIMA cap. The listener actually indicated the 006++ unit with WIMA cap having some traits that was heard on the original 006+ unit (which uses WIMA in the decoupling).

Now then, can you give me a hint on HOW I could have leaked any info on which machine was being listened when the listener don't even see me, don't even see the machines and was able to identify the unit out rather clearly day after day? We limited our communication to only what I have described above.

When listener is not SURE about what was being heard, the listener will say so, and have done so quite a lot of times. Sometimes the listener was tired and sometimes the difference is too small. The listener will only report when it is very clearly noticeable, and won't try to hear for something that wasn't there. We also have a habit of not revealing what was being improved and what I was hoping before or during the test.

[Maniac]

Due to popular demand, the volume of the two unit with stepper are MATCHED, as shown below with both unit set to 8 notches from silent.

Both results are given about 20 second for the scope to average, with measurement resetted when a different unit is being probed.

Average RMS 2.09 vs 2.08

Now is someone going to complain about frequency generator's slightly different frequency average of 0.0208KHz?

[Maniac]

the detailed info makes the test even more dubious. a blind test between close items (and we're talking about close items here, once the audio hyperbole is removed) typically creates simple preference results. highly detailed difference results suggest the possibility of subconsciously well-meaning confabulation.

your very presence in the room can create a perceptual bias, because the listener may know that you are expecting the expensive caps to sound better. in that kind of scenario, the listener latches on to any differences, objective or subjective, and attributes those differences (which, in this scenario will almost always be "positive differences [just look at burn in reports, for a similar phenomena]) to the expensive caps. making the test double blind doesn't completely remove this, as the person knows they are being tested, and will likely know the aim of the whole thing (something is supposed to sound better than something else), which is why every other aspect of the test has to be strictly controlled. there are hundreds of ways that non-obvious aspects of a test can invalidate the results, because humans are finicky creatures easily lead by internal subconscious bias and subtle external cues.

a habit is not a control. even if the listener doesn't know specifically what they are listening before, they are still obviously listening for something, which is why everything else has to be strongly controlled for the results to matter.

Your reasoning still does not explain how the listener is able to identify each unit by sound, as each time involves multiple listening sessions, with listener having no idea which is being listened this time around. During the entire time, the listener makes note on what is being heard, and I make note on which unit is being tested in what sequence.

When we've had 5~6 (anymore is just too taxing IMHO) sessions in a row, we stop and I compare the notes. And each time, the listener was able to identify the same unit with identical comment, sometimes even commented in the note that "did I just heard this at trial number x?". And no, the pause between sessions are under 20 secs, we don't sit down and compare note before the entire test run is done.

Now, can you explain how was the listener able to be extremely consistent?

[rwires]

that's all well and good, we know how stepped attenuators work, but did you measure the outputs of the amps at a specified level?

As others have pointed out, the amps have the same gain. Thus the stepper allows the volume to be the same.

I think you're beating a dead horse here, obviously this amp is not marketed to you. Maybe wait until you or someone else has done a comparison between the two units??

[Maniac]

no. but neither can you. showing that the volumes are matched is a good step towards a controlled double blind, though, if you want to keep going with it. i probably wouldn't bother.

I have explained how the listener is consistent, you can't explain it away as a fluke, that is your problem, not really mine. I have told you the volume are extremely consistent, and explained why, you still don't want to believe in logic and fact. Yet, as seen in the photo, the volume is extremely consistent. Now please show me some beef in proving how is it possible for the listener to be extremely consistent in identifying the amp by sound, and yet still be a fluke?

[grawk]

He's describing real scientific controls. You could just say "it's not scientifically established" and everyone can go on their way.

[Maniac]

science works the other way around, typically. you aren't gonna get it anyway, or you wouldn't be saying the same things over and over, so i'm done here.

If you don't get it, you don't get it. What's not proven is not proven, it is not wrong or non-existent. Science does not work the way you think it works. Not to mention I have proven it's consistency and accuracy while all you were really saying is FUD.

If you are not able to prove or disprove anything, I'll have to ignore further similar messages from you, sorry.

[Maniac]

listen, buddy, i'm gonna use small words. you make claim. me say, experiment not good. you say me am not like you say experiment not good. me say here part how make experiment gooder. you say me have fact! you say me am science! me say here other problems. many problems. you say me am fix one part me am science! me say face palm.

Clearly rage has overcame you...

[Hopstretch]

Clearly rage has overcame you...

It happens.

[morphsci]

This is not going to end well.

[jinp6301]

Blind experiment - Wikipedia, the free encyclopedia

[Pars]

Power supply of course, believe it or not, but I've found DynaLo and DynaHi family of amp to be extremely sensitive to caps. When I was first building my DynaHi a few years back (was 2006 I think), I installed two Panasonic FM caps on each DynaHi amp board... Well, DynaHi sounded every bit like the transistor amp that tube lovers hate, it was cold, gritty, harsh, and overly bright. I spent a lot of time checking out a LOT of stuff, nothing helped. At the end, I decided that the Panasonic FM caps is about the only thing that I haven't swapped out, and I bit the bullet there by swapping in some rather expensive ROE boutique electrolytic caps. The result? Well, the guy who got that DynaHi from me is happily taking that DynaHi around to his friend's place for shoot outs, and he is winning most of the time, and on rare occasions, there might be a tie or a lose, but that's usually against rivals costing 2 to 3 times the price of the DynaHi that I sold him.

After that, I did not go on build more DynaHi anymore, because it was more of a challenge to build one of those amps at least once, and learn something in the process. And boy did I learn a lot from building DynaHi.

The actual blind listening test is done with a good HDD player (FLAC files)and a PCM1794 DAC as a source, the custom cap actually contributes a lot to the sonic performance of the amp. We compared a fully upgraded HA-006++ and a similarly upgraded unit but uses the original WIMA MKS caps in the power decoupling, and the result was quite clear. I did not do the listening myself, but I did the switching around and the setup, all with the listener's back toward the equipment. I switched around randomly, and the listener is able to consistently pick out the difference between 006+, 006++ with WIMA and 006+ with FPP cap. I did try to tune it further, but it was getting to the point of diminishing returns at this point unless we redo the entire setup and go for a bigger unit with even beefier power.

I seem to recall reading this on the DynaHi thread on hf. I was skeptical, but I do know that Dr. Gilmore used some crazy fast Illinois Caps in his original Dynalo supply psu (last caps in the PSU). Just curious how big of FMs you were using? 470uf I would guess... I thought I would try 100uf in the one I am nearing completion on, but might try some film in it and see if I hear anything.

[digger945]

I seem to recall reading this on the DynaHi thread on hf. I was skeptical, but I do know that Dr. Gilmore used some crazy fast Illinois Caps in his original Dynalo supply psu (last caps in the PSU). Just curious how big of FMs you were using? 470uf I would guess... I thought I would try 100uf in the one I am nearing completion on, but might try some film in it and see if I hear anything.

You're right about the crazy fast Illinois caps Chris(I read it in KG's "Pure class A dynamic HPA" article on HW), and I have searched for them wanting to use them but haven't found anything that looks suitable. I would like to know if they are even still available. I have FMs to replace the originally spec'd FCs on the boards, but haven't done the work yet. I can't recall all the specifics now but some of the specs for the FM are way up there in the mhz and the ripple multiplier suffers some at audio freq's.

I'm all for squeezing every ounce out of a very good sounding amp. I think you will enjoy yours.