Head-Fi.org › Forums › Equipment Forums › Sound Science › Objectively "Good" Amps and DACs
New Posts  All Forums:Forum Nav:

Objectively "Good" Amps and DACs - Page 6

post #76 of 88

Hey all,


I thought I'd weigh in here. I know I won't convert anyone who thinks we're a bunch of ancient mouthbreathers slumped over their Tek 465s, shaking our canes at these "newfangled objective youngsters," but I did want to address a few things:


1. Why did the Asgard and Lyr not have relay protection when we first launched? I've explained this before. When we started the company two years ago, we assumed we were building the highest-possible sonic performance products for an audience looking for the same, and who knew that this could involve taking some precautions. It's very common in megabuck gear to have a "turn on sequence," for example, where if you turn on the preamp after the power amp, you might be looking at voice coils. I'm not saying this is right, I'm just saying this is how it is. Now, we understand clearly that we have to provide reasonable protection. It's why we added time-delayed startup and fast shutdown relays on Asgard and Lyr, why we offered retrofits for people who want them (for free), and it's why we created one of the most sophisticated protection systems available for Mjolnir--time-delay relay mute, coupled with DC and overcurrent sensing. Any fault lifts the outputs. 


2. Why do we have a distaste for negative feedback and op-amps? I think you might need to look at the kind of statements we make. This is a statement of fact: "We make amps without overall negative feedback." This, however, is a value judgement: "We make amps without feedback because we think all feedback is poopie." First, we don't make value judgements. We don't comment on competitive equipment, we don't call people names in public or private, we don't tear down other manufacturers or their approaches to product design. That's just how we operate. We make no-feedback amps. We make things in the USA. We typically use simple, fully discrete topologies. We use tubes in some applications. We like to pursue alternate topologies, especially in the headphone amp realm, where the power requirements are very low and invite experimentation. These are all statements of fact. But, bottom line, if someone can make an amp with 100dB of negative feedback and a Class-B output stage sound good, that's cool! More power to them. It's just not what we do.


3. Ears versus instruments. Yes, we listen to our products, and we believe this is a valuable part of the design process. I understand some people believe differently. That's cool. You can save a ton of money and angst in your quest for perfection if you simply purchase by measurements. To us, both instruments and ears are important. On my bench, the Stanford SR1 is one of my most-used pieces of gear. Same with Mike (he has the uber-low-jitter model, I have the analog-leaning one.) They're invaluable for finding out what's wrong with an early design and tweaking it for best performance. And it's not like this is new. Since the Theta days, Mike and I have invested heavily in test gear, from Stanford spectrum analyzers to interval counters for jitter. The reason we haven't talked about measurements very much is that they can easily turn into a game of pointless specsmanship (hey, I have -120dB signal to noise, how could you ever get by with -118db!)


4. Black and white and audio answers. To us, there's less black and white as time goes on. When I started designing audio gear, I was very much a meter reader and specs-hawk. And I have designed plenty of high-performing products by objective measure, including all the Sumo amps (which still used only moderate amounts of loop feedback, coupled with a Hawksford-style error correction system for the MOSFET output stage), and the analog output stage of the Theta DS Pro Gen V. The Gen V was an interesting case, because the Theta DS Pros before the V used (very expensive) op-amp I/V conversion. Mike believed it was the best way to go. I told him, "I can make a discrete I/V with better performance, any way you want to measure." He took me up on the challenge, and I designed a true current-input, very low impedance discrete I/V that measured better than the op-amp (10x less bounce at the summing junction for current input, and 20% lower overall THD with lower overall feedback.) Op-amps don't allow you to directly access a low-impedance input (such as paralleled emitters of a complementary discrete input,) so this is one case where, objectively, the best solution is discrete. As time goes on, things become less black and white for us. Maybe that means we're getting old and losing our grip on reality. Or maybe it's a case of having more years of perspective. 


Again, I'm not here to convert anyone, nor do I think this will have any effect on anyone who thinks we have slighted them, or their beliefs. But I did want to at least have a voice.


All the best,


post #77 of 88

There's a difference between a well-documented limitation—that many cheaper amps share, that is not an issue if you use sensible gain values for whatever source you're using, that can be fixed by a user with a couple minutes and a knife or scissors—and an undocumented problem.  Regarding the gain issue, I think the builders / distributors have much of the blame on themselves for not informing people about it.  This situation exists because of a design compromise, yes, getting lower noise at the expense of a little lower usability.


Personally, I think that unplugging headphones is a hassle, just like unplugging a portable amp that can't be used while charging, having to use a 6.35mm to 3.5mm adapter, and plenty of other examples.  It shouldn't have to be that way.



Originally Posted by Anaxilus View Post

We were discussing this perceived notion many have that design by measurements leads to a sonically perfect design w/o flaws or limitations.


The idea is that if you have numerical design objectives, then you can actually meet and exceed them (and others can verify if you've succeeded in those), or at least you can figure out where to compromise on them.  Furthermore, if you're careful about defining enough conditions and parameters, and if you actually meet them all, then that's when maybe you end up with something without sonic flaws or limitations, under certain operating conditions.  i.e. before it starts clipping or hits the knee, or outside a more linear operating region on certain types of tube amps, or whatever


I'm still waiting on a demonstration of sonic flaws and limitations.  There are plenty of good amps and DACs to choose from.  Show that you can tell them apart under rigorously controlled listening conditions, or grab some lab testing gear and show how badly it performs with music or audio-frequency test signals, despite meeting objective parameters A / B / C / D / ...




edit: didn't see the above post.  Thanks for stopping by and at least responding, Jason.


edit2: any comment on the "sonic compromise" represented by using a relay (what kind of relay is used, specifically?) or other protection mechanism?  Or is that a misrepresentation?

Edited by mikeaj - 8/4/12 at 9:27am
post #78 of 88

i'm beginning to like jason more and more. A rational approach.


As to the diy t2, that is the turn on power spike as a result of the

high voltage transformers kicking in.  The actual output spike

on the electrostatic headphone outputs is barely audible, and

is no more than a couple of volts. Remember that electrostatic

headphones really don't care about dc levels as long as the

resulting bias does not waver more than 10%.


The original does exactly the same thing. 45 seconds after

the tube filaments turn on, the high voltage kicks in.


The original T2 does have an output mute relay, but it

is not driven by any protection circuit.


The SRM-717 original model does in fact have output

protection. But later models have the protection removed.

Edited by kevin gilmore - 8/4/12 at 9:36am
post #79 of 88
Originally Posted by kwkarth View Post

You're evading the question, but then I'm not here to "argue" a position. If you would truly choose one piece of gear over another by the numbers, despite what your ears told you, then welcome to your decision, but it is not the decision I would make.

It seems you are assuming that good numbers mean that a piece of equipment can sound terrible. In that case there are at least two things that went wrong: the measurements, or as said before, bias creeped into the decision making process.


In the former case we can easily verify the measurements, correct them if needed or do different measurements altogether, such as null difference testing with your favorite piece of music- the beauty of science.

In the latter case do a blind test to see if you really heard it.


You're also assuming it has to be 'ears over numbers' or the other way around. I'm afraid it's not that simple. Maybe that's why it looks to you like I'm evading that catch question?

Edited by xnor - 8/4/12 at 11:24am
post #80 of 88
Originally Posted by Anaxilus View Post


Hmm, well...I'm certainly in the habit of unplugging my phones during during power on/off.  Thx to user error and bad habits/behavior, left-foot brakers like myself will likely have to suffer due to proposed NHTSA mandates about gas cut off on brake depression so forgive me if I'm a little less sympathetic to your perspective and the nanny state.  I also still find it odd that many who find the absence of the relay a design flaw and not user error while an objective amp that clips w/ anything above 2V is seen as user error rather than design flaw.  


There were zero operational guidelines, the issue was noted nowhere because once again it was unknown.  I was the first to discover it.


Plus it is not false equivalency since I was not the one making a straw man to equivocate gain w/ DC spikes.  That's your new point, not mine.  We were discussing this perceived notion many have that design by measurements leads to a sonically perfect design w/o flaws or limitations.


I didn't bring the O2 into this, you did.  You then compared them arguing about how can we call one defective but the other limited and this not making sense.


Any questions about what is or isn't audible (extrapolated from his sonic compromise comment about relay delays that 'can still bite you') should be asked to Jason directly.  I know we both agree to disagree on the merits of what is and isn't audible so no need to go down that rabbit hole.  


The Gilmore DIY T2 was the reference in question.


Check Kevin's response, electrostatics are a different ballgame.


Responses in bold.

post #81 of 88
Originally Posted by Shike View Post

Responses in bold.


The 'objective' concept/philosophy and it's champion/minions brought the O2 into the discussion not me.  Wholly relevant to the topic IMO.


I already knew Kevin's response from other sources as I did Jason's from actually staying informed, I wanted to see what yours would be.

post #82 of 88
Originally Posted by Anaxilus View Post


I already knew Kevin's response from other sources as I did Jason's from actually staying informed, I wanted to see what yours would be.


So trolling, that's great. rolleyes.gif

post #83 of 88

More like trawling.  wink.gif

post #84 of 88

Jason, a well measured response and reasonable ethic (well actually I dont think opamps are all bad, just not applicable everywhere). Might I suggest you actually include the word global in your 'no feedback' statement on your product pages as you have here?, the 'no feedback™' catchphrase is an advertising construct that does not reflect the truth on closer inspection.


Degeneration and local feedback are everywhere, not least in the dac itself; (which very likely also includes a poopy internal opamp IV w switch cap) as well as power supply, feedback from whatever forms the analogue references that form the basis of the dacs output, clocking, receiver etc etc.. Without feedback, as you know, the dac would not know which way was up … I encourage you to not dumb down or be selective with your 'no Schitt' attitude.


all in all I think you are doing a good job with this stuff, but its just little inconsistencies like this that can make it appear as if you are speaking just another brand of marketing speak, targeted at a different consumer.



BTW I agree about relays personally, i'm working on an MCU that controls the sequencing of my whole multichannel, multiamped dac/amp system , all my amps are DC coupled and my speakers are without crossover; so I need something and dont want to rely on everyone following the correct order when i'm not around. I dont like relays directly in the signal path either; relay contacts go bad through aging and arcing; through this, like pots they can become non-linear. This is perhaps more important with power amps that see a larger current than headamps and dacs, but the theory is the same.

Edited by qusp - 8/4/12 at 9:25pm
post #85 of 88

A lot of RMAA you will find are basic load (usually just a loopback), so the measurements aren't always reliable.


Amps and dac measurements have such a small impact, that close enough is regularly good enough.

There are a number of problems that plague measurements (this applies for LCDs, Headphones, Amps, Dacs, whatever). They have become a marketing tool. Not measuring to AES standards, using not LUV or LAB colour spaces, you name it. One of the key problems with AMP measurements, is that they rarely do reactive load measurements, and if they do - chances are, it's not for your headphone (and I think it would be unfair to expect this, when we consider thousands of potential combinations, including thousand dollar or more headphones). The only exception I have found is that stax, at one stage, produced reactive load measurements with the headphones their amps were actually sent with.

The best 16/44 RMAA I've ever seen was from an RME 96/8. Truly, outstanding. I'm skeptical about the results however, as they're simply a loopback. However, as it is analog external (LO to LI), the results should be pretty accurate, as far as I know. I don't particularly like RME otherwise. But as the card is now old (I believe), it's very cheap to grab and try. But then you have measurements of the same card, which don't seem up to snuff.

Like this


From memory, the best 24 bit RMAA that I trust is a Lynx L22.

RMAAs I don't trust are typically those that come with internal soundcards, particularly those which come with sound-changing software. An example are cards from EMU from the 90s. There are a few RMAA results that are 140dB+++ SNR for 24 bit. To put things into perspective; that's 20dB hgiher than I've seen for ANYTHING else. Consider that 3dB is roughly "twice the performance" (off the top of my head). It seems to me that there are soundcards that employ tricks to alter measurements (though I could be getting a little tin-foil-hat here). It just seems off to me that a company could produce something that performs so much better than everything 20 years later, without anything funny going on


FWIW: I still don't think the LCD-2 has good measurements (frequency response), but I think that it performs well on measurements designed for speakers (i.e. where a treble-shy headphone will excel over one that is FF or DF equalised). Not that it isn't a fantastic headphone (best bass in a headphone ever? Probably).

Before buying ANY dac or amp, I'd ask for a range of results, information about the load and if they were measured to AES standards.

Companies I know that measure to AES standard (certainly) are focusrite and echo.


To put things further into perspective for headphones, I've discussed with Tyll about the curves/HRTF he uses for innerfidelity. I suggested that he include options for things like freefield equalisation.

He told me that he uses an independant of direction curve to compensate the responses. Meaning that the sound is never exactly like what you would hear from a speaker, but is the average of the sounds you would hear from a speaker from a particular direction. Seems valid, no? I agree it is.

However, is it not also valid to argue that, music is typically heard face on? And therefore a freefield reproduction is ideal? Again, I agree this is also true.


Tyll mentioned that he'd like to use a curve (but is unable to set it up) where, the compensated curve would look like the standard studio monitor setup, About 30 degrees difference. Also valid surely.


But then what about headphones mixed on headphones, for headphones? Binaural recordings and purely electronic music come to mind. Where is your ideal?

The problem with the new objectivists (one in particular), is that there is so little standardisation that the objectively superior option becomes nebulous. This is true for anything audio, ESPECIALLY headphones. How do we measure our amps? Our dacs? Our headphones? Which is more ideal?

As of yet there is no answer for this - or at least no true standardisation (where as for speakers; flat = king). 
It's not just audio. Take displays for example. Is it better to have n colours that conform to the RGB colorspace, or the AdobeRGB colorspace? Some might turn around and say that RGB is the intended look for a screen, others might say that the colours in adobeRGB (assuming they are displaying the same number of colours) is closer to print. Who is wrong?

How is the contrast ratio performance of a screen, measured in a COMPLETELY black room, relevant to your use - when any screen on the market degrades rapidly in light (where most professional work is done)? Or better yet - how does contrast ratio actually tell you anything important? What good is a screen that can be as bright as the SUN, but only produce blacks as deep as the sky during the day? This is still contrast ratio in the thousands - but it is NOT at all good for use in a display.

It gets worse when you look at OLED screens that take Emission Only measurements. This would mean that an OLED screen of green apperance when off provides exactly the same contrast ratio as one printed on one that looks perfectly black. Both provide "infinite contrast ratio".
How is this anything other than misleading?



So many other problems exist with the measurements we use. A lot of them are designed for speakers, and therefore assume flat is king. A square wave on a headphone device, for example, will appear scooped, because headphones produce more treble than bass. Feel free to go and look at the very, very bassy options and then compare them to some headphones that people might consider "treble heavy". Does the square wave of the former not appear rounder?

What about measurements like CSD, when applied to headphones that are planar? How does the 'ring' of sound after it has stopped affect a bi-directional system when in motion? I personally don't see how it applies (to things like stax, LCD2, etc), since our music RARELY contains silence.




I won't argue against the idea that better measurements = Better gear, because I (and in my opinion, any informed individual) believe this. But when every single source produces something different - no matter how slight - we have to wonder; at which point do our methods for measurement fail? In my opinion, they fail almost immediately. I used to trust measurements, but now I am as skeptical of them as I am of people writing anecdotal evidence of their own gear.

I post this, not only as an objectivist, but also a realist.

Edited by MrGreen - 8/4/12 at 11:38pm
post #86 of 88

nah you mustve had the routing set up incorrectly, the loopback would have to be internal/digital to get that result. I feel RMAA has only VERY limited usefulness for testing modern high performance gear as the used ADC and input preamp limits performance, so that you are basically measuring your audio card not the intended target. at least with anything i'm building anyway

Edited by qusp - 8/4/12 at 11:35pm
post #87 of 88
Originally Posted by qusp View Post

nah you mustve had the routing set up incorrectly, the loopback would have to be internal/digital to get that result. I feel RMAA has only VERY limited usefulness for testing modern high performance gear as the used ADC and input preamp limits performance, so that you are basically measuring your audio card not the intended target. at least with anything i'm building anyway


Sorry, I was editing my post a lot, and probably wasn't clear. That is not my RMAA. 

I was comparing two different RMAAs of the same gear using the same methodology, that produced results upward of 3dB different.

Let me find the other one (will edit the post)

EDIT:: I don't have time to trawl the internet for all of them now.

But if you look up RMAA results for the RME Babyface in 16/44, you will find DNR results that are higher than the limits of 16 bit audio (~99 DNR babyface vs ~~97.9dB OTTOMH for sinusoidal waveforms).

To me, that strikes alarm bells in the reliability of many measurements you can find, and not actually do yourself for your system, in your home. But I largely don't care anymore because I am both happy with my gear (though maybe I want more speakers...) and I understand how little of a difference an amp/dac makes for basic listening with todays technology.

Edited by MrGreen - 8/5/12 at 12:36am
post #88 of 88
Originally Posted by adydula View Post

What would you do if you bought a device, amp, dac whatever and it sounded really, really good to you, makes you all tingly inside etc...and all those wonderful adjectives start flowing about it....


Then you discovered that the objective measurements were just plain off the charts good as well???


Would you throw it away????


Has this ever happened to you??


Thinking, thinking....




I would not throw it away. I would acquire more products and try to figure out a correlation between the sound and the measurements if possible. It has never happened to me, usually the numbers match up with what I hear.



Originally Posted by kwkarth View Post

So, if "science" says something measures well, but my ears tell me it sounds terrible, who wins and why? Goes back to what values I hold as an individual.


Anyone can convince themselves to go either way. I don't see why we have to limit ourselves to being objective only or subjective only. You can use as many tools as you wish, including your own senses. It's not really about winning. What's more important is being able to articulate the difference/like/dislike that is being heard and being able to reproduce it. That said, someone should know about their own biases and how it influences what they're hearing and expecting. Which is why objectivity is nice, it takes more control of the situation by not relying on a human's nonperfect senses. It takes experience from a variety of different setups to really know what a sound should sound like.

Edited by Deathwish238 - 8/5/12 at 1:09am
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Sound Science
Head-Fi.org › Forums › Equipment Forums › Sound Science › Objectively "Good" Amps and DACs