diyAudio - Comments
Go Back   Home > Forums > Blogs

Please consider donating to help us continue to serve you.

Ads on/off / Custom Title / More PMs / More album space / Advanced printing & mass image saving
  1. Old Comment
    abraxalito's Avatar

    Designing the best bang-for-the-buck headphone amp

    Yeah I got banned for nothing on there - on the whim of an admin (Amir I think) - presumably because he saw me as too much of a threat?

    Well it does look like you're thinking in noise terms about Johnson (thermal) noise. In an amp the signal that's subject to the most gain needs to have the lowest thermal noise because the subsequent gain amplifies the noise. I agree so far.

    In the case of an amp, the stages with the input signal level at the lowest (the LTP typically) the power supply can be insanely regulated and the LTP is in classA. So there's as low noise as you want on the PSU - for classA you can use the 'noise finessing' technique shown by Wenzel (oscillator people) if you really need to. Hence - bye bye supply noise.

    With a classAB amp the stage where you have to tolerate some load-induced supply noise is the OPS. Hence in practice that's the one which is going to give most grief, because it has the heaviest load.

    TL,DR version - noise per se isn't really a problem, but shifts in the noise floor correlated with the music ARE. Noise modulation in other words.
    Posted Yesterday at 06:03 AM by abraxalito abraxalito is online now
    Updated Yesterday at 06:07 AM by abraxalito
  2. Old Comment
    rjm's Avatar

    Designing the best bang-for-the-buck headphone amp

    You got banned?

    Okay, 'competent' was a poor word choice in my part. Still, total output noise is most likely to be dominated by the stage where the signal level is *lowest* not the stage where the signal level is highest, i.e. the power amplifier output stage.

    With a chip amp it's all in one, so the point is moot. You perceive the noise and since there is only one component you correctly assign it to that component. That and chip amps are notorious for noise coupling into the noninverting input, generating positive feedback of ripple and hum.
    Posted Yesterday at 05:36 AM by rjm rjm is offline
  3. Old Comment
    abraxalito's Avatar

    Designing the best bang-for-the-buck headphone amp

    On your first point, it must follow that I so far have only designed incompetent amplifiers as I've not heard the distortions of my output stages yet, but have heard noise from power supplies. I'm still learning though - but given what I hear, I'll concentrate on improving the things that I hear.

    On point 2, I agree and have done several designs where the output stage drives a speaker directly. I would also agree that optimizing isn't necessary, however it is desirable where the amp is classA. I'd like to build a portable classA amp and battery life most assuredly is an issue.

    Point 3 - agreement.

    I'm working on the design - yesterday I worked out the power supply so I'll post that up shortly. Today I'll continue to work on the amp stage - I listened to a single-ended version of it (with different transformers) and now I'm modifying that into balanced form.

    Thanks for the comments btw, appreciated.

    Incidentally here's a thread about a speaker amp where once again, power supply noise was the audible issue. I even show some measurements - Possibly the most frugal high-end sounding amp?
    Posted Yesterday at 12:06 AM by abraxalito abraxalito is online now
    Updated Yesterday at 12:12 AM by abraxalito
  4. Old Comment
    rjm's Avatar

    Designing the best bang-for-the-buck headphone amp

    To be clear, I'm totally behind your experiments - I just want to keep you on the straight and level, thus:

    1. What we hear as the sound of an amplifier is first and foremost the distortion of the input signal by the output stage, the background noise level meanwhile almost always originates at the point where the signal is lowest: the phono stage or the input section of the preamplifier / headphone amp. The noise, self generated or power supply, of the output stage is not usually relevant in a competent design.

    2. Tube amplifiers are transformer coupled because vacuum tube output impedance is high. In solid state you can easily make an output stage to drive 200 W into 4 ohms directly. Making a transistor output stage drive any kind of headphone 16-600 ohms to 100 mW or more is trivial. "optimizing" for a given load is not necessary: milliwatt power is dirt cheap design cost. The only issue is gain, which can be addressed at the input stage.

    3. Speaking of voltage gain - voltage gain is not free. More gain introduces distortion and noise at the input / voltage amplification stage, and attenuating it back at the transformer doesn't remove either.

    That said, I can see that for your stated topology - single ended - the transformer has advantages since the circuit is fundamentally happier running at higher voltage / lower currents, and bias current (i.e. output power) is a relatively expensive design cost, compared to push-pull.

    By the way, could you perhaps append a sketch of your proposed circuit to the post above? It's not totally clear whether the transformer is parafeed or not, and I'd be interested in seeing the other details of the intended application.
    Posted 12th October 2015 at 11:40 PM by rjm rjm is offline
  5. Old Comment
    abraxalito's Avatar

    Is neutrality really the same as accuracy?

    It remains to be demonstrated that simply measuring 0.0001% THD+N and ruler flatness in FR means an amp is 'neutral'. The onus is on the claimant to show that it is, I'm skeptical myself.
    Posted 12th October 2015 at 12:06 AM by abraxalito abraxalito is online now
  6. Old Comment

    Choosing a USB DAC

    This is a very interesting thread, perhaps the best way of preserve the quality of your CD's and/or vinil, is converted to DSD:

    In the thread say that the ifi nano idsd use this tecnique of no-dac, i don't now if the ifi micro also use the same technique:

    nano – iDSD

    Or perhaps you can think in increase your budget and buy a Chord 2Qute DAC (Hugo derived) , yes, more expensive, but I think that you are cover the next years:

    Products: 2Qute DAC
    Posted 28th September 2015 at 04:13 PM by raul_77 raul_77 is offline
  7. Old Comment
    abraxalito's Avatar

    Choosing a USB DAC

    I'd not go back to a soundcard now - they use PC power but barely any filtering between it and the opamp rails. But then I rarely use my PC as any kind of audio source, preferring dedicated players - normally battery powered ones for the isolation and convenience.

    How much of an issue the isolation is, is going to be context dependent. If you share the same mains distribution block for all the audio kit then the loops may be jolly small in practice. Coax might have the advantage of transformer isolation in practice however I've not found trafos to give enough isolation where the layout's compromised by common ground impedances between analog and digital. Given the choice I'll go for layout over isolation though.
    Posted 27th September 2015 at 04:34 AM by abraxalito abraxalito is online now
  8. Old Comment
    rjm's Avatar

    Choosing a USB DAC

    Thanks for the comment. The Denon unit has isolation. It's the only one on the list that does.

    Thing is, I've never noticed that as a problem though. Comparing COAX vs. Toslink inputs, that should hand a clear win to the optical connection but this is almost never reported to be the case even with computer as the source. (Many motherboards have those outputs as part of the on board sound features) Or, for that matter, my previous soundcard which had all the analog circuitry in the PC chassis itself and shared the power supply as well!

    While it would help in the event that a bad layout creates a ground loop, I don't think a good layout should need isolation to have low noise.
    Posted 27th September 2015 at 12:20 AM by rjm rjm is offline
    Updated 27th September 2015 at 12:25 AM by rjm
  9. Old Comment
    abraxalito's Avatar

    Choosing a USB DAC

    My advice would be - go for one that has its audio output grounds isolated from the PC's ground. Otherwise you have potential common-mode noise issues from the ground loop you'll set up, coupling the PC's SMPSU noise into your audio kit. Exasound has a couple of nice FFT plots for how much noise is reduced with isolation.
    Posted 26th September 2015 at 11:07 PM by abraxalito abraxalito is online now
  10. Old Comment
    abraxalito's Avatar

    Designing high order active filters

    I realize that gain isn't required from the resulting filter (in fact in my understanding it can attenuate too without difficulty) but it does need a gain element (an opamp for example, not a unity gain follower). That was my meaning.

    Low input impedance isn't really manageable when good SQ is a target - means current source biassing the output and running fully balanced to equalize PSU currents. That's why I'll take a pass on MFB - the BOM cost will escalate as will the power requirement. Thanks for the offer though.

    I've already built the elliptic filter - see one of my previous posts. Even when passive, elliptic filters I found have hideous component sensitivity (due as you point out to the extreme Q), which is why I gave up on elliptic passive filters. I'm curious how you arrived at your 60dB figure? I came across ICs to implement filters many years ago when in the industry but I doubt they'd be optimum for power and cost in my application.
    Posted 24th September 2015 at 11:17 PM by abraxalito abraxalito is online now
  11. Old Comment

    Designing high order active filters

    Also, if you want to design an Elliptic filter for this purpose and want more freedom in choosing the implementation you need to know the filter in terms of first and second order stages and the associated Fc and Q of each. I can send you a program (its a DOS program unfortunately) that lets you provide these as input and then spits out the parameters for the various stages. Then it is up to you to determine how to implement them, e.g. a cascade of SK or MFB or other. I have used this program to design a set of filters that have ripple in both the passband and stopband, like elliptic filters, but are meant to be used in complementary pairs as loudspeaker crossover filters. I hope to write all of that up someday...

    For an anti-aliasing filter you want something that is very steep right away. I think that typically the transition band starts around 18kHz like you mention but the stop band edge is set to Nyquist, e.g. 22k or 24k Hz and attenuation hits max there, at least 60dB of it. I have measured this exact response from commercial products, so I believe they are using an elliptical filters. These performance specs are ALOT steeper than the 9th order (the 1kHz Fc one) Elliptical filter shown on the simonbramble web page you linked to. That filter is not a fully elliptic filter, it's a Chebyshev type 1 filter - these have passband ripple but no stopband ripple. You can see these differences in filter types highlighted here:
    Having both passband AND stopband ripple (full elliptic filter) gives the steepest transition band but the stopband rebounds up to some "ceiling". This is usually a minimum of 60dB down. The Q values needed for this kind of filter are much higher (like Q=50) and there is no way you could use a Sallen Key topology for that. Even MFB is hard pressed to do such a high Q, but that is what is needed to get the required rejection over only a band that is only a few kHz wide. This is the kind of filter I had in mind when I first saw your post, and so I thought that using a SK was ludicrous. In any case, making these filters yourself will be difficult and expensive. Why not just use a monolithic solution? There are 8th order Elliptic LP filters available as ICs that are purpose built for anti-aliasing.
    Posted 24th September 2015 at 03:56 PM by CharlieLaub CharlieLaub is online now
    Updated 24th September 2015 at 04:30 PM by CharlieLaub
  12. Old Comment

    Designing high order active filters

    Gain is not required for an MFB filter. You may find low input impedance as you mentioned, but unless the Q is really high it's manageable. You do need to use an op-amp that has high open loop gain and high bandwidth for MFB, but this is not a problem these days. Just don't use a TL072 or NE5532! If you want I can send you a MFB LP filter design spreadsheet (Excel) that I put together. The user chooses Fc, gain (V/V), Q and the capacitance used, all of which are independently variable. The 'sheet automatically chooses the "best" resistor values from the E96 or E192 series for the circuit. PM me with your email address and I will send it over.
    Posted 24th September 2015 at 03:42 PM by CharlieLaub CharlieLaub is online now
  13. Old Comment
    abraxalito's Avatar

    Designing high order active filters

    Perhaps we have different requirements for how flat our passbands need to be - for me I don't worry much at all about half a dB of ripple seeing as I know my speakers in-room will be much worse than that.

    MFB is a topology I've tried before - it needs an inverting amp with gain so wouldn't work with a simple buffer stage. When I tried it with opamps one major drawback is the lowish input impedance in practical realizations.

    How do the sensitivities compare - between S-K in unity gain and MFB in (inverting) unity gain?
    Posted 24th September 2015 at 01:41 AM by abraxalito abraxalito is online now
  14. Old Comment

    Designing high order active filters

    Well I wouldn't use SK above Q=2 or so but maybe your low sensitivity version would work fine. You can still try MFB for the highest Q stages, but know (I'm sure you do) that they are phase inverting.
    Posted 24th September 2015 at 12:25 AM by CharlieLaub CharlieLaub is online now
  15. Old Comment
    abraxalito's Avatar

    Designing high order active filters

    Sallen - Key it is. I'm aware of the component sensitivities - they're set out very well in the paper I linked to. I am using the least sensitive S-K variant - with unity gain buffer.

    I have built some filters with Qs greater than 2 and done some simple measurements - 'no good' is hype.

    Which schematic do you want? I'll be posting the schematic for the discrete buffer stage in a few days but sounds like its not for you as in your perception they're 'no good'.
    Posted 22nd September 2015 at 12:57 AM by abraxalito abraxalito is online now
  16. Old Comment

    Designing high order active filters

    Are you using an MFB topology for your circuit? Leapfrog? Hopefully not Sallen-Key! AVOID SK for high Q circuits. They have HUGE component sensitivities when Q increases above about 2. Sims look nice with perfect values, but in reality with real world components and real world tolerances they are no good.

    Are you following the circuits from that web page? Those look SK to me. Can you post your schematic?
    Posted 21st September 2015 at 11:46 PM by CharlieLaub CharlieLaub is online now
    Updated 21st September 2015 at 11:51 PM by CharlieLaub
  17. Old Comment

    JLH-2005 headphone amplifier

    my mistake, have no problem now
    Posted 15th September 2015 at 04:27 AM by jambul jambul is offline
    Updated 19th September 2015 at 12:49 PM by jambul
Hide this!Advertise here!

New To Site? Need Help?

All times are GMT. The time now is 02:50 AM.

vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2015 DragonByte Technologies Ltd.
Copyright 1999-2015 diyAudio

Content Relevant URLs by vBSEO 3.3.2