Marsh headphone amp from Linear Audio

Disabled Account
Joined 2012
When testing near the limits of the best test equipment, such as AP, we need to know the instruments contribution to the test results. For example, I asked for a loop-thru self test of the AP and one harmonic (3rd) was at -112dB. This is almost the same as the amp test data showed. therefore, the real amp 3rd harmonic is much lower than shown. We need more extream test measurments -- maybe with comb/notch filters on the source harmonics to get them all below the noise floor. Where would I find such equipment? Anyone?
 
When testing near the limits of the best test equipment, such as AP, we need to know the instruments contribution to the test results. For example, I asked for a loop-thru self test of the AP and one harmonic (3rd) was at -112dB. This is almost the same as the amp test data showed. therefore, the real amp 3rd harmonic is much lower than shown. We need more extream test measurments -- maybe with comb/notch filters on the source harmonics to get them all below the noise floor. Where would I find such equipment? Anyone?

Cyril Bateman presented a very high performance1 kHz sine generator and distometer in british Electronic World. Here are links :
Pro Audio Design Forum • View topic - Cyril Bateman Capacitor Sound Archive
 
Where would I find such equipment? Anyone?

Get a newer AP :)
The SYS-2722 analog generator seems to be in the 0.00003% - or better - THD range @ 1 kHz/2 Vrms out:
AP SYS-2722.PNG
Did you manage to try Bob Cordell's THD magnifier? Still waiting for LA issue describing it (and the one with your amplifier, of course ;)), but it seems that with a pre-nulling circuit you can easily measure THD well below -120 dBc - I've quickly tried a resolution enhancing setup by Scott Wurcer and Walt Jung from an old AD application note, and it works very well indeed: with both the EMU internal generator and my old HP239A 3rd harmonic from my amp sits at about -115 dBc (@ 1 kHz and 7 kHz, 1 Vrms into 30 ohms).

L.
 
Last edited:
Disabled Account
Joined 2012
use indirect IMD test - I had a noise floor of ~ 160 dB just by using ESI JULI@ soundcard DAC/ADC - one DAC channel for each of 10,11 kHz sine, resistively/summing amp added - then looked for the 1 kHz difference product with 30 dB gain amp/filter that rejects the test signal pair

ultra low harmonic distortion of the individual test signals isn't required - only the summing method has to avoid IMD at the level you want to measure the difference - can be easier than ultra low distortion osc/filters

3rd order nonlinearity can be probed with different freq pair 10,21 kHz gives 1 kHz 3rd order diff

the lower IMD difference products are more audible than harmonic distortion - at higher SPL levels masking is less
 
Last edited:
Disabled Account
Joined 2012
Get a newer AP :)
The SYS-2722 analog generator seems to be in the 0.00003% - or better - THD range @ 1 kHz/2 Vrms out:
View attachment 285260
Did you manage to try Bob Cordell's THD magnifier? Still waiting for LA issue describing it (and the one with your amplifier, of course ;)), but it seems that with a pre-nulling circuit you can easily measure THD well below -120 dBc - I've quickly tried a resolution enhancing setup by Scott Wurcer and Walt Jung from an old AD application note, and it works very well indeed: with both the EMU internal generator and my old HP239A 3rd harmonic from my amp sits at about -115 dBc (@ 1 kHz and 7 kHz, 1 Vrms into 30 ohms).

L.

I did build the R.Cordell circuit - requires X10 gain for the DUT circuit or higher. So this cant be used I guess (?). [I havent tried it on this amp at this time] Thx I need better than -115 -> about -130 or more would be nice. I could boost the headamp gain and then measure to get an idea, I suppose.

BTW - what amp do you have? What is the topology?
 
Last edited:
Actually the test I like is to compare the circuit under test to itself. I use two of them one is resistively loaded the other has a real transducer load.

I use an AP system 2 and find it somewhat limiting. Anything interesting requires additional techniques to get at the heart of the testing.
 
Disabled Account
Joined 2012
use indirect IMD test - I had a noise floor of ~ 160 dB just by using ESI JULI@ soundcard DAC/ADC - one DAC channel for each of 10,11 kHz sine, resistively/summing amp added - then looked for the 1 kHz difference product with 30 dB gain amp/filter that rejects the test signal pair

ultra low harmonic distortion of the individual test signals isn't required - only the summing method has to avoid IMD at the level you want to measure the difference - can be easier than ultra low distortion osc/filters

3rd order nonlinearity can be probed with different freq pair 10,21 kHz gives 1 kHz 3rd order diff

the lower IMD difference products are more audible than harmonic distortion - at higher SPL levels masking is less

Another good idea to try that cost less than a full blown new AP. At some point could need thd just for apples to apples comparison purposes.
 
Disabled Account
Joined 2012
Get a newer AP :)
The SYS-2722 analog generator seems to be in the 0.00003% - or better - THD range @ 1 kHz/2 Vrms out:
View attachment 285260
Did you manage to try Bob Cordell's THD magnifier? Still waiting for LA issue describing it (and the one with your amplifier, of course ;)), but it seems that with a pre-nulling circuit you can easily measure THD well below -120 dBc - I've quickly tried a resolution enhancing setup by Scott Wurcer and Walt Jung from an old AD application note, and it works very well indeed: with both the EMU internal generator and my old HP239A 3rd harmonic from my amp sits at about -115 dBc (@ 1 kHz and 7 kHz, 1 Vrms into 30 ohms).

L.

SYS2722 is about -136...... that might be Ok for awhile. But, then there is that nagging question... how much is curiosity worth for a one time test? Time or money? Hmmmm.
 
BTW - what amp do you have? What is the topology?

Basically it's a pretty standard discrete open-loop CF topology- I've attached a simplified schematics (well, actually an over-simplified one - output CFPs are not going to work that way :) ) and a sample FFT run (@ 1 Vrms into 30 ohms). It works very well - in principle you could further improve performances by selecting bjts, but with 18-20 trannies/channel it could be a real pain...The interesting point is that you can easily convert it to closed-loop if you're interested in a CL vs OL comparison (and I was pretty curious indeed, but heard no appreciable differencies...), but I know it is by no means a simple circuit, nor it was that easy to have it running smoothly (@ about 0.2 ohm Zout, at least).

L.
 

Attachments

  • HPAmp.PNG
    HPAmp.PNG
    17.8 KB · Views: 1,075
  • CMOLAmpLoZ - THD 02.png
    CMOLAmpLoZ - THD 02.png
    23.3 KB · Views: 1,064
Last edited:
Disabled Account
Joined 2012
Basically it's a pretty standard discrete open-loop CF topology- I've attached a simplified schematics (well, actually an over-simplified one - output CFPs are not going to work that way :) ) and a sample FFT run (@ 1 Vrms into 30 ohms). It works very well - in principle you could further improve performances by selecting bjts, but with 18-20 trannies/channel it could be a real pain...The interesting point is that you can easily convert it to closed-loop if you're interested in a CL vs OL comparison (and I was pretty curious indeed, but heard no appreciable differencies...), but I know it is by no means a simple circuit, nor it was that easy to have it running smoothly (@ about 0.2 ohm Zout, at least).

L.

That's a nice variation on what has become what might be called the better topologies in use today .... current-mode, too. There are a million varieties and this one seems good too. However, as you noted, the complexity is much greater. And, is it dc stable in offset and drift?

For myself, the lowest thd is not the goal. I just want to measure it accurately. It seems, IMHO, that a more elegant solution/circuit is the one which gets the job done best with the fewest of parts and the lowest cost.

An Ic with external power output stage comes close to meeting the goal.... but usually still needs a dc servo. And, the better IC's are not cheap.

Thx,
Richard
 
Last edited:
don't understand the expense argument - do you get free shipping? count every minute of hobby time as free even for the boring bits?

have to buy batches of discrete anyway for matching - I count my time doing repetitive, boring measurements/sorting as a cost

TPA6120 dual DSL driver wraped with OPA6142 ultra low distortion fet dual in a multiloop composite amp gives absurdly great active device performance for ~ $5 per channel at Digikey's overinflated singles price, ~ $2/ch in qty
 
Last edited:
Disabled Account
Joined 2012
agreed though, I dont understand it either when people are so focused on saving a few dollars while blowing hundreds of dollars in man hours

It is in the context of commercial app. Yes, i spend a lot of time and money on this sort of thing. 'cause i like it. You would be surprised (maybe you wouldnt be) that to make a mass produced item - 2-5 per dollars is way too much. Its the total package i am talking about - how good, small, cheap etc can you make it. Sorry, thats just my own thing.... the challenge to myself to keep bordom at bay during the winter when I cant travel.

But for a single DIY project, $5 worth of IC's per channel can hardly be considered expensive. Certainly is fast and convenient to make for oneself. This circuit is an alternative. yep, a little more time is involved. Before this project I went that route and bought a top rated Headroom headphone amp. It didnt sound as good as what i knew the sound should be That is what got me to try my own discrete design. BTW -- did I mention that it Sounds Great!?

[Besides, I am retired and have lots of time to blow until I'm dead. Meanwhile, sharing the journey/story.] I'm leaving for Alaska today. See you next time in Thailand and Nepal! Enjoy!

Thx,
RNM
 
Last edited:
Disabled Account
Joined 2006
Basically it's a pretty standard discrete open-loop CF topology- I've attached a simplified schematics (well, actually an over-simplified one - output CFPs are not going to work that way :) ) and a sample FFT run (@ 1 Vrms into 30 ohms). It works very well - in principle you could further improve performances by selecting bjts, but with 18-20 trannies/channel it could be a real pain...The interesting point is that you can easily convert it to closed-loop if you're interested in a CL vs OL comparison (and I was pretty curious indeed, but heard no appreciable differencies...), but I know it is by no means a simple circuit, nor it was that easy to have it running smoothly (@ about 0.2 ohm Zout, at least).

L.

DC servo would be needed with that, Im also building something similar over the weekend, just curious.
 
And, is it dc stable in offset and drift?

Output drift is less than about |5| mV over 2 hours or so @ about 100 mA Iq (fairly stable, and this was somewhat the real challenge with such a low Zout design goal), en plein air and with a far-from-optimum layout: not a precision DC amplifier at all :rolleyes:, but seems fine for audio - actually a DC servo coluld be not necessary at all, although I've already tested one, just to be on the safe side.

It seems, IMHO, that a more elegant solution/circuit is the one which gets the job done best with the fewest of parts and the lowest cost.

Agree - that's why I'm so curious about your amplifier.

L.