• Disclaimer: This Vendor's Forum is a paid-for commercial area. Unlike the rest of diyAudio, the Vendor has complete control of what may or may not be posted in this forum. If you wish to discuss technical matters outside the bounds of what is permitted by the Vendor, please use the non-commercial areas of diyAudio to do so.

Aspen Headphone Amp

pcb and enclosure size

Is it a good idea to aim for a single pcb, keep it simple and reliable ?

If so, we'd want to pick a low profile pcb mount transformer and allow this to set the maximum height. Choice of capacitors and small heatsinks have to fit within this limit.

Can we fit this into standard 1U case height ?

e.g. Digikey pc-mount Toroid, 15VA is 26.3mm above the pcb surface.
TE70053-ND
 
The heck with the regulators , go battery. 2 small 13.6v gel-cells with associated trickle chargers would address this issue completely .

I have a home-made gel-cell lm317 charger in my workshop and it has been with me for 4 years , very reliable. I have seen this done with pre-amps (this is what I might do with it - this circuit)

OS
 
Disabled Account
Joined 2007

Attachments

  • Picture 1102.jpg
    Picture 1102.jpg
    92.3 KB · Views: 623
By aksa - The trick with loop gain is useful, thanks to GK and Andy_C, very useful way to figure stability.

Here is another one . It does work with the "older" models (2sc/a) , works even better with fairchild models (ksa/c ksp ). You drive a 2uf - 10uf cap at the OP with a "fast" square wave at 10k, also add a second -2.3v voltage source and short the input cap (attachment 1 is self explanatory). The first output plot is driving the standard resistive load. The second is with a 2uF OP cap and 10pf lead comp. (slight overshoot) , The third is with the original 5pf lead comp. (1.1mhz UGP) , here is where it really gets "ringing' but still does not break out in sustained oscillation. This little amp is REAL stable .. but this is just a single stage OP. If you do this blasphemy on a triple (with more complex models) the "cause and effect" of total loop gain /UGP AND the parasitics/higher gain in the more complex current stage (EF2 or triple) is quite evident.

One can also step (.step command) through different Cdom's , biases and do this with different device models to see the effects on distortion harmonics and stability.
I feel quite confident because when scoped in real life , you see nearly the same responses. :)
(attachment 5 is the fft , nearly similar to the 55 !! )
OS
 

Attachments

  • SWtest.gif
    SWtest.gif
    8.3 KB · Views: 632
  • aksaHFSWnormalload.gif
    aksaHFSWnormalload.gif
    13.9 KB · Views: 591
  • aksaHFSW2UF.gif
    aksaHFSW2UF.gif
    16.1 KB · Views: 565
  • aksaHFSW10UF.gif
    aksaHFSW10UF.gif
    14.9 KB · Views: 565
  • aksaHFFFT.gif
    aksaHFFFT.gif
    22 KB · Views: 181
Last edited:
IKO,
Thanks for the good explanation, I can see your point, particularly with the sensitivity to noise. Further, the very low energy levels of headphones would clearly mean transient response is not the first priority, too.
I agree about shunt regulated power supplies. I have done a few myself, including one using a 5881 pentode with a TL431 in the cathode circuit which sounded wonderful. If we use 40mA quiescent, same for the shunt element, I'd reckon since that will give just shy of 120mW (5.5Vpp) into 32R cans. Might even be possible to reduce rails to 12V.

We should try to ensure at least 15Vpp output for 2K cans, too, although that's only 14mW......

OS,

Thanks a lot for your stability input. This is certainly very stable, nice to see the sims bear it out. Can I talk privately with you on this?

Hugh
 
Hi Hugh and Bigun,

It's also a didactic projet, easily analysed and where effect of some very basic modifications could be easily experienced.

First, I would try it in the inverting mode : it mainly diminishes H2 but not very significantly harmonics higher than H3, it should be subjectively detectable.

Back to the non-inverting mode, I think of a curious way to achieve the current source for the LTP : it uses a bootstrap circuit coming from a divider of the output voltage to the splitted resistor of the tail, as it was shown some time ago for a power amplifier by a transistors manufacturer, but I do not remember the reference.

Regards
 
Thanks Forr!

This same circuit is analysed in an electronics text, 'The Art of Electronics', Horowitz/Hill, published by CUP, 2nd Edition 1989, p. 236-238. I claim no originality at all; this is a stock, easy circuit, with very good performance, something unexpected in this complicated world. The secret with most of these simple circuits is good compensation techniques.

Any circuit with lower levels of H2 is interesting, but not mandatory. H2 in a feedback amp is very low, and not strictly audible until at least 1%. A good way to do this, however, is to use a single transistor input stage, with feedback taken to the emitter.

Cheers,

Hugh
 
Disabled Account
Joined 2007
Guys,

We have the basic amp but it doesn't feel as if a decision has been made on the psu. So we're in a holding pattern...

Hugh - what does the 'boss' say here ?

Hi Gareth,
My opinion is to put the amps themselves on a single board and have it so that each amp channel can be connected to either a single or a dual supply.

The standard power supply could be a simple one, a single supply with cap multiplier as you suggested would be nice. This is easiest (while still providing great performance).
The rest of us can design and build our own or use an existing design and post the results here for comparison (and bragging rights ;) )

:)
 
Thanks John, Gareth, OS,

Good ideas all round.......

One further to muse on. If we make the amp single supply, it will be cap coupled, and that throws up the wooden cross for most here. However, there are three advantages. One, the supply is simpler, two the cans have automatic DC protection, and three, a substantial DC bias across a cap makes it sound way better.

The input cap on any bipolar supply power amp typically has around 0.1 volts across it. Only a very good cap sounds any good in this position, and that's expensive. I often use Sonicap, and they cost an arm, leg, and left kidney.

Thoughts before we start on a pcb design?

Cheers,

Hugh
 
Last edited:
Hi Tessier,

Wouldn't work. Charge suckout works to remove minority carriers from base junctions, and that's effectively done with the vbe multiplier bypass cap, which is 22uF.

In any event, if you use at least 40mA of quiescent, this headphone amp will keep both outputs on until the signal gets to 2.7Vpp with a 32R load, so charge suckout, which helps outputs to turn off gracefully, won't be an issue. With 120mA of quiescent, OS's suggestion, we have almost 7.7Vpp (231mW into 32R cans) before the inactive output device switches off. To that point, it's in Class A.

Hope this explains it,

Cheers,

Hugh
 
Last edited:
Series regulators become like inductors when in their BW limits. Shunt regulators act the same except with usually lower ESL because of design circumstances.

So I guess the question of reg vs. no reg comes down to bandwidth (Hugh's "leading edge"). So then, which has less ESL: a capacitor, a shunt reg, or a series reg? I don't see that DC precision is necessary here, which is one point down for regulators.

It should also be known that when any regulator nears it's bandwidth limit, not only will it act inductive but it will also act like a distorted inductor - an argument in favor of a simple rectifier + capacitor PSU.

There are a few different configurations of regulator+output caps, each of which has different results...

Precision reg: lower BW, lower ESR. (say -90db)
Fast reg: high BW, larger ESR. (-60dB or even worse)

1: Precision reg + perfect output decoupling.

With high regulation, it will take an EXTREMELY LARGE, LOW ESR output cap just to cancel the output inductance of the somewhat slow reg. $$ is the limit here (Can you imagine using 10,000uF caps AFTER a regulator? we use those without regs...)

2: Precision reg + minimal output decoupling.

Say you just use a 390uF cap and some HF decouplers... You take care of ESL at very high frequencies, but there is a large peak before the small decoupling begins to lower impedance... In other words, you have a nasty gain peak at say, 100KHz-1MHz.

If those decouplers don't have the same or lower ESR than the actual regulator, your regulation won't have a nasty peak but it will "step down" at a certain frequency.

3: Fast reg + perfect output decoupling.

I have preliminary simulations of a regulator that gets -60dB regulation at reasonably high BW which I'm using as a reference here.

With a 390uF output cap, there is no peaking in the output. So even though we have less regulation, we have a smooth frequency response, depending on the ESR of the decouplers.

It seems to me that for audio (and for other things as well) smooth response and good compensation is better in a regulator than pure precision. I think an amplifier will find a mostly resistive supply more livable than a mostly reactive one. (what it comes down to I think, is resistive or reactive in regulators, but you can have both with $$)

Some of this is my own opinion, which means it is somewhat subjective or I write it in context of subjective preferences. So unless you understand the technical details, I recommend you don't repeat it elsewhere.

My sources for this information is much experimenting with regulators in LTSpice.

I discuss factors that might affect the operation of an amplifier, but I have no idea what it will sound like (although I have an educated hunch that slow and steady wins the race, IE non-reactive regulators or simple rectifier+cap).

Also, how about you try bootstrapping to Q7's emitter instead of to the output? This will make the bootstrap much more accurate. One transistor works more, but this should theoretically encourage even harmonics. (no use simulating this - what I want to know is its affect on the sound).

Hope this was informative,
- keantoken
 
Last edited:

iko

Ex-Moderator
Joined 2008
Hi keantoken, and thanks for chipping in. I have to say I don't understand some of the things you say. As I said before, I don't want to go at length about regulators here, this not being the thread. Perhaps we should open a thread where only regulators can be talked about?

I have preliminary simulations of a regulator that gets -60dB regulation at reasonably high BW which I'm using as a reference here.

Line regulation or load regulation? I'm curious what number is for reasonably high BW in your experiment.

With a 390uF output cap, there is no peaking in the output. So even though we have less regulation, we have a smooth frequency response, depending on the ESR of the decouplers.

I thought one can have a smooth frequency response (over a certain range) with a good regulator as well. No?

It seems to be agreed upon that matching the sound performance of a good shunt regulator is very hard with a simple psu (RC or LC filter, a little easier with good iron in the LC).

But I think we shouldn't get stuck in choosing a psu for the HP amp. After some have build the amp they can always share their experience and subjective sound impressions relative to their particular psu and parts choice.

It's good that we have all sort of options.
 
Agree emphatically with Keantoken when he remarks that precision is NOT the requirement, bandwidth is.....

Therefore, let me suggest a simple full wave rectifier to a 1000uF cap, then passed to a second cap, bypassed with a quality 10uF MKP, also of 1000uF through a couple of 2.2 ohm resistors, one between each of the cap electrodes.

This will contain the intermodulating charge pulses in the first cap, largely quarantining them from the second. With a good second cap and bypass, we have the bandwidth, there is little inductance in the circuit, and the ripple hard edges are softened by the time they reach the second cap, to be reduced to insignificance.

This is simple, cheap, and uncritical, and performs well....

Hugh
 
Hi keantoken, and thanks for chipping in. I have to say I don't understand some of the things you say. As I said before, I don't want to go at length about regulators here, this not being the thread. Perhaps we should open a thread where only regulators can be talked about?

Line regulation or load regulation? I'm curious what number is for reasonably high BW in your experiment.

I thought one can have a smooth frequency response (over a certain range) with a good regulator as well. No?

It seems to be agreed upon that matching the sound performance of a good shunt regulator is very hard with a simple psu (RC or LC filter, a little easier with good iron in the LC).

But I think we shouldn't get stuck in choosing a psu for the HP amp. After some have build the amp they can always share their experience and subjective sound impressions relative to their particular psu and parts choice.

It's good that we have all sort of options.

I agree that this is not the place for a long drawn-out discussion. I also must say that I'm not the kind of person to get into mathematical analysis. All I have is a large number of observations from the simulator which makes sense to me; all my knowledge is relative. I also worry that such a thread would get bogged down in mathematics.

I think I made a technical error in posting this here (it came down to rounding, and I favored 1 over 0).

And sorry for being vague, I was referring to load rejection in every case in that post. In response terms, regulation for my regulator is -3db at 1.6MHz. (My apologies if my sense of scale is off - not sure whether this would qualify as good or bad to you, but I say it's "High BW" because it's the best I could squeeze out of the simulated circuit, which has a max output of around 2A)

If you want, I can explain more clearly, though not here.

- keantoken
 
Last edited: