The Headcase Stax thread

[JimL]

1 hour ago, kevin gilmore said:

I posted this over there, will post here.

if I had time I would find to do a way to do the graph that shows frequency response relative to output voltage, maybe in a few days if people don't get it

so here is how this works

Required Slew Rate = 2 x π x Frequency x Peak Voltage  x K

we can argue about K later, different people have different ideas about the value.The load is a capacitor and it takes a particular amount of current to charge/discharge the capacitor at a particular frequency and voltage.

Lets reference the maximum peak to peak output voltage of the amplifier at 1khz as 0db.

Now lets calculate what amount of current is needed to do 20khz at 0db at the same peak to peak voltage into one standard headphone load, 120pf

Now lets figure out (and this part gets complicated) how much more current is needed to guarantee .01% thd at 20khz at 0db. taking into account the open loop gain.

Its clear that some of my earlier amplifiers with output currents in the range of 5ma to 10ma cannot do full voltage swing at 20khz. But will do a full flat frequency response at less than full voltage swing. So the question is how loud do you listen and how much actual audio content is at 20khz relative to 1khz.

It is my opinion that 20ma however is enough to do full voltage swing even with 500V power supplies.

There is no free lunch.

Absolutely agree as a technical matter that being able to swing full voltage at 20 kHz is better than not being able to.  The SRX Plus, for example, is not able to do so, and in fact, as I posted in that thread, actually measurably rolls off at lower frequencies as the output voltage increases (it's flat at 100 VRMS to over 20 kHz which is pretty loud, though).  BTW, KG's calculation is actually a minimum estimate of the current required, because a stat headphone is not a capacitor - it makes NOISE, e.g. music, which means it has to consume ADDITIONAL current over what a capacitor requires.

 

However, as Dr. Gilmore points out, the question is, how loudly do you listen and how much actual audio content is there at 20 kHz.  There is actually some data on the latter point.  A number of years ago, Peter Baxandall and Nelson Pass both measured slew rates in recorded music and both came to similar conclusions, namely that an amplifier capable of reproducing a 6 kHz sine wave (yes, 6 kHz, that is not a misprint) with low distortion is adequate for music signal, because of the natural roll-off of musical instruments.  Note that this was equivalent to the fastest music signal they found, and that everything else they measured was SLOWER than that. More recently, Baxandall published that he had found a recorded music signal that required an amp to be capable of reproducing a 15 kHz sine wave with low distortion.  That is significantly faster, and in fact is within striking distance of the 20 kHz signal at full output voltage requirement - however, it must be noted that this is a unique event, and for almost all music the lower requirement is adequate.  Is it appropriate for a state of the art amplifier to aim for the more stringent requirement?  Absolutely!  Will the lower requirement suit the needs of most? Probably.  Of course, extra headroom is always nice - but there is a cost.  For Stereophile Class A, the cost is always worth it, almost by definition.  For Stereophile Class B, the question is, how much can you cut back for as little sonic cost as possible.  

 

Finally, if you listen at lower levels, the requirements decrease accordingly.  This is why spritzer suggested to Tyll that he listen to electrostatic headphone amps at high levels to differentiate between them - because this stresses the amp the most.

[kevin gilmore]

a graph is in order here so that people would better understand whats going on.

unfortunately I can't think of generating a quick way to do this, I took 6 of the kgsshv schematic at different levels and tried to sim at the same time but the software hung.

may be able to look at this again late tomorrow.  if someone else can do it so much the better. because I think that most people are not going to understand without a picture.

[sorenb]

@kevin gilmore combine in Adobe CS?

[kevin gilmore]

are you completely out of your mind, that would take me a week.

ok, here is something quick that gives you the idea  4 x lm741 at different gain levels

why it looks like crap and the background is not black, not sure

and resulting phase

 

Edited July 23, 2016 by kevin gilmore

[spritzer]

This shows just why some of these amps sound like shit.  A combination of low current, low voltage and bad current sources will change how they sound quite dramatically. 

[PretentiousFood]

Is that the FR of with different amounts of feedback? If you keep feedback constant, I think it just clips, rather than rolls off gracefully.

Here's a sim of an ES-X sort of deal. I stepped the source amplitude at 100mV, 1V, and 10V (some people have silly DACs). The EL34 (6CA7 in sim) is at 6mA and rails are +/- 400V, 120pF load. Happy to try different parameters.

FR does not seem to change in the magical ideal world of Spice?

http://imgur.com/a/2uglC

Of course, the current requirement into the load increases with frequency and signal level. The blue line crosses at 10mA, and shows the frequency at which current limiting would occur for different amounts of voltage swing.

http://imgur.com/qh0esa3

However, output distortion gets gross well before clipping.

At 1 kHz: http://imgur.com/vJAiMyW
This shows clipping into the rails.

And at 10 kHz: http://imgur.com/u1sD8ze Waveforms: http://imgur.com/UYLhk9W
This one shows current limiting.

Edited July 23, 2016 by PretentiousFood
whoops, wrong graph

[sorenb]

1 hour ago, PretentiousFood said:

http://imgur.com/a/2uglC

Of course, the current requirement into the load increases with frequency and signal level. The blue line crosses at 10mA, and shows the frequency at which current limiting would occur for different amounts of voltage swing.

shouldn't the current I(R14) or is it I(R34) and I(20) be added to the current requirement?

[JimL]

One additional comment, although this may be getting deep into the weeds for some.  It's not just the raw standing current that matters, but also how well it's used.  The current generation of Gilmore amps, along with the SRX Plus, uses a cascode constant current load which significantly increases the efficiency with which the signal current goes to the headphone rather than heating up the load.  Few, if any, other designs do as well.  For example the original SRX design with 50k resistors uses the bulk of its signal current to drive the resistors.

[PretentiousFood]

1 hour ago, sorenb said:

shouldn't the current I(R14) or is it I(R34) and I(20) be added to the current requirement?

Absolutely. Those are feedback resistors, and they eat up a whole 1.5mA each during peaks (~400V / 250k), independant. This may be more academic than anything else for reasons Jim brought up; there's really not a whole lot going on at 20kHz, and little reason to push the output to full swing.

At the risk of being lynched, I took them out and run it open loop...

[sorenb]

48 minutes ago, JimL said:

One additional comment, although this may be getting deep into the weeds for some.  It's not just the raw standing current that matters, but also how well it's used.  The current generation of Gilmore amps, along with the SRX Plus, uses a cascode constant current load which significantly increases the efficiency with which the signal current goes to the headphone rather than heating up the load.  Few, if any, other designs do as well.  For example the original SRX design with 50k resistors uses the bulk of its signal current to drive the resistors.

with all its merits I wonder why the cascoded CCS doesn't find its way to the magic Carbon CC ...

20 minutes ago, PretentiousFood said:

Absolutely. Those are feedback resistors, and they eat up a whole 1.5mA each during peaks (~400V / 250k), independant. This may be more academic than anything else for reasons Jim brought up; there's really not a whole lot going on at 20kHz, and little reason to push the output to full swing.

At the risk of being lynched, I took them out and run it open loop...

so the amp with feedback will have less current for the load ...
I just tried about the same simulation using a KGSShv in LTSpice ... seems like I get result somehow similar to yours 

[JimL]

29 minutes ago, sorenb said:

with all its merits I wonder why the cascoded CCS doesn't find its way to the magic Carbon CC ...

I believe it is.  Here is a quote from the Headcase thread:

kevin gilmore

4/22/15

....

I have also added the cascode to the carbon

[congo5]

not sure what Carbon CC is but its defiantly on the Carbon, Grounded grid and BH, I know I've blown them up :-)

and Thanks Jim for how to set them up in advance with a battery, works great!  or in the circuit...

[sorenb]

4 minutes ago, JimL said:

I believe it is.  Here is a quote from the Headcase thread:

kevin gilmore

 

4/22/15

 

....

I have also added the cascode to the carbon

I guess you are referring to the utterly pile of crap some calls the Carbon, but is massively different from a real Carbon ... which, as far as I know, doesn't feature the cascode 

[spritzer]

Mine doesn't feature the cascode.  That was done on purpose. 

[Laowei]

Deleted 

Edited July 24, 2016 by Laowei

[arnaud]

As discussed privately with @PretentiousFood, assuming spice circuits exists for many of the amps discussed here, wouldn't it make sense to have standard results from a spice simulation for 120pF capacitive load and various voltage swings?

In terms of V swing, I don't know how load you listen to but, in my case, I seldom seem to exceed 100V PPSS. Something like 80V, 160V, 320V, 640V, 1280V PPSS would seem sufficient.

As I understood from spice simulation, the output current vs. frequency don't show actual current limitation of the amp (e.g. 6mA in example shown above) but transient simulation would?

Similarly, we could have thd vs frequency at same Vswing levels.

Seems like that would be more suitable than just talking from the plate current, rail voltage specs and topology when comparing these various amp designs.

Arnaud

[MLA]

Had it wrong, so deleted.

Edited July 25, 2016 by MLA

[Arthrimus]

I've been listening to my SRS-002 system these past few days (thanks @justin ) and I have to say these little things rock! I've put probably 20 hours of listening time in at work and I am digging them for sure. This is the best portable setup I've heard, and at the price I can't imagine you could do much better. Fit is not an issue for me thankfully! 

Now for something stupid. I couldn't leave well enough alone so I made a full size Stax adapter and sampled some of my other phones on the SRM-002. It doesn't have the guts to properly drive any of my Lambdas (obviously) but it actually does surprisingly well with my SR-X Mk3. 

What you see there is my SR-X Mk3 -> Pro Bias to Normal Bias adapter -> 5 pin Stax to mini Stax adapter -> SRM-002. This was done for kicks of course and is not to be taken seriously. 

In a few days I will be building the much more useful reverse of this adapter so I can use my SR-002 on my SRM-727.

Edited July 25, 2016 by Arthrimus

[spritzer]

Now you just need a SR-X Mk3 Pro....   

[astrostar59]

The discussions here on mA requirements and voltage swing are fascinating. Coming from a layman / audio fan not engineering perspective I think maybe a very powerful amp on tick over may sound better than one at it's limit in normal operation. Music has a huge dynamic range and fast paced signals, so I am thinking a amp that is more highly spec'ed and less stressed could sound more natural and realistic. I guess there are limits as regards heat and current tolerance. But regardless it seems KG keeps pushing out the envelope which is a gift IMO to the DIY community.

Reading a post above, my incoming Carbon has the LV boards addition. Looking forward to hearing it.

On 22 July 2016 at 6:22 PM, purk said:

Can you do that at a source end?  I am just using the Yggy into either the BHSE, and Carbon....and honestly I like them just about equally with maybe a tad to the BHSE.  I'm using NOS Holland XF4.  Note that I do like the SR009 to begin with.

Purk, I have had tube based DACs for years driving my solid state speaker amps and later my KGSShv's. I do believe that you can tube roll a DAC / Source and get a great system synergy going. I m not talking tube roll off or rosy glow, more beautiful midrange and timbre, smooth treble response. Wether it is best to have a tube DAC or tube amp I have no idea, but my system seems to work well. I dropped my tube pre-amp (but kept my tube DAC) years back to go passive on my SS speaker amp with gains in pace and dynamics.

I have no idea if it is best to tube roll in a DAC or amp. In the source the signals are tiny so 'may' be more sensitive to change there. God knows. For me it was more trail and error. When I was upgrading my DAC kit with different boards and parts, I rolled tubes a lot to get the sound where is sounded correct to me. For example when I changed the I/V transformers it was more transparent and open, so I went NOS Mullards on the driver tube. When I fitted Duelund caps on the line board, I changed the Mullards on there for Tungsols (less warm). Is this the correct way to do it? Maybe not, but it is a way an owner can tweak the sound without hacking the circuit.

I know all amps should sound the same or at least be transparent 20hz - 20k hz. But we know they sound different as cars perform different round a race track.

Edited July 25, 2016 by astrostar59

[joehpj]

On 2016/7/23 at 9:42 PM, kevin gilmore said:

The performance of an el34 at 20ma begins to effect the lifetime of the tube. Significantly.

Since it's adjustable, how much current will you recommend for the balance of performance and lifetime?

[purk]

I have been pretty happy with my Yggy and Carbon/BHSE/T2 combo...but I also have a tube CD/SACD players in the past that also work equally well in the front end of the amps.  Hence my thought about using a tube front end instead of a tube preamp in front of the amps.  I think Birgir used to have APL Denon 3910 at one point and it also have a tube output stage.

Edited July 25, 2016 by purk

[Dave R]

7 minutes ago, purk said:

I think Birgir used to have APL Denon 3910 at one point and it also have a tube output stage.

I've seen it for sale on Birgir's website.

[astrostar59]

14 minutes ago, purk said:

I have been pretty happy with my Yggy and Carbon/BHSE/T2 combo...but I also have a tube CD/SACD players in the past that also work equally well in the front end of the amps.  Hence my thought about using a tube front end instead of a tube preamp in front of the amps.  I think Birgir used to have APL Denon 3910 at one point and it also have a tube output stage.

Purk, I think tubes in a DAC can do other things besides allow tube rolling. Some amplifiers and DACs use tubes in the regulation of the power supply as well. My pre-amp I used to have had tubed regulation and tubes on the line board as well. On tubed DAC design the Audio Note DACs and some of the Lampizator DACs are examples.

[purk]

I hope you will get your KGSSHV Carbon soon.  It is quite a great sounding amplifier.  My unit has the Golden Reference PSU but the audio boards are of earlier version.