AD8610, OPA134, OPA627, BUF634 SMD headphone amp

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Today I have both tested and listened to OPA627, OPA134 and AD8610 (at +- 15V, didn't die!). Electrically AD8610 is outstanding, can handle too large signals excellent, offset at the out 0.046 mV cold and 0.079 mV warm!!! The OPA627 can't handle large signals at high frequencies very well. It may be better if the signal is DC connected.

OPA134 is also excellent but have slightly higher offset, 2.5 mV at the output. I have also tested the buffer at medium speed mode and high speed mode, no visible or audioable change. Soundwise all opamps do their work very good. I (my ears) can't hear any major difference.

I notice also that groundplane pcb cools rather good. BUF634 in high speed mode gets rather hot. Also the regulators are rather well cooled.

Some guy here (forget the name right now) has elevated AD8610 to the skies and must admit, it's a good opamp!
 

Attachments

  • qrv04proto5.jpg
    qrv04proto5.jpg
    36.5 KB · Views: 5,396
Ad8610

Some guy here (forget the name right now) has elevated AD8610 to the skies and must admit, it's a good opamp!

Might have been me - it's the best I've heard, bar none.

The AD8065 is worthy of investigation too - it has even better capacitance driving capabilities than the AD8610, and lower distortion at HF.

Try them, I'd be interested to know your views (I've not compared in detail).

Wouldn't run it for long @ 15V though, the XFCB process really is restrictive, the 8065, IIRC, is +/-12 - lower than the +/-13 of the 8610.

Andy.

P.S. Are you selling those boards?

P.P.S. can I be so bold as to suggest there's something wrong if you can't hear the difference between the two bandwidth modes, BTW, it's obvious to me, the WB mode being better. Maybe it's the yellow electron clouds, clouding judgement :)
 
Christer said:
Per-Anders, did you assemble three PCBs using different op amps, or how did you compare them since you use SMD?
I have only one prototype and I switched (slowly!) between opamps. With this method you will forget how it exactly sounded, I know, but still, every opamp sounded splendid. Electrically I liked AD8610 the most. "Klockrent" performance!

Can anyone give me an english word for klockrent?, Bull eye maybe? :bulleye:
 
Re: Ad8610

ALW said:
Wouldn't run it for long @ 15V though, the XFCB process really is restrictive, the 8065, IIRC, is +/-12 - lower than the +/-13 of the 8610.
I used 15 volts only because I don't feel like fiddleling with the voltage. Off course should you use max +-13 volts if you don't want to risk anything.

ALW said:
P.S. Are you selling those boards?
Yes.
ALW said:
P.P.S. can I be so bold as to suggest there's something wrong if you can't hear the difference between the two bandwidth modes, BTW, it's obvious to me, the WB mode being better. Maybe it's the yellow electron clouds, clouding judgement :)
Note that I NOT have used low speed mode, only medium and high speed.
 
peranders,

all of the BUF634T's that I have bought or obtained as samples have had TO220-5 case with five in-line pins, that is why I use in-line mount.
I have not tried ground plane yet, but I did not get into trouble with this PCB, considerable amount of measurements was done and about 20 pcs of 2-channel circuits were assemlied and proved.

Pavel
 
Hi peranders,

I just finished designing a PCB for a BUF634 / OPA2134 based headphone amp - I sent it off to the board manufacturers last week. Here is the layout for my (single sided) PCB, if anyone's interested in seeing. I would be very interested to hear your opinion of my layout, although it is too late to change anything now :rolleyes:

See ya,
Tim.

PS - Surely using the buffer outside a feedback loop is a bad idea? Someone posted on Headwize that this topology would lead to high distortion and DC offset.
 

Attachments

  • cans.gif
    cans.gif
    32.6 KB · Views: 3,589
Tim

I would be very interested to hear your opinion of my layout, although it is too late to change anything now

The most obvious thing to me is that the PSU rails, as seen by U2, would seem to be modulated by the potentially significant currents drawn by the buffers.

It would have been better to star-wire the supplies, or considered a seperate low current supply for U2, IMVHO.

Some RC decoupling may be beneficial as a result of the above.

The grounding, via the large copper area is thinned by neccessity in many areas, and this is also likely to bring reference problems and potential error signals.

I'm sure it will work fine, but you may like to consider a double-sided artwork for a future variant, as this will simplify the artwork and allow you much more board area to play with and costs very little extra. I think if you were to consider all the potential error sources and address them, you will be surprised at the difference.

It's easy to forget the PCB is a critical part of your circuit diagram, that rarely shows in a SPICE model or simulation unless you put it there yourself.

Out of interest who is doing your PCB's and how much are they?

Hope that doesn't sound overly critical, it's not intended to be, PCB layout is very much a skill that get's better every time you do it. There's not a single PCB I've ever done that I would not change, if I ever re-visited it later on ;)

Andy.
 
Thanks for your comments, ALW.

I'm sure you're right about the supply to the opamp U2. I decoupled it with a 10uF tantalum cap to ground on both supplies next to the device (C5 and C8), and I used 40 mil tracks for all the buffer's power rails. I assumed this would be sufficient, maybe not :(

I did think about thinning of the copper ground plane, and I took some steps to minimise it. Maybe I didn't do enough, although some of the thinning looks worse on the screen shot than I remember it being close up in my layout program.

"I'm sure it will work fine, but you may like to consider a double-sided artwork for a future variant, as this will simplify the artwork and allow you much more board area to play with and costs very little extra. I think if you were to consider all the potential error sources and address them, you will be surprised at the difference."

OK, double sided it is next time... for my board manufacturer it is significantly more expensive, though. I use PCB Pool, at: http://www.pcbpool.com/uk_index.htm - they have given me good service in the past. They charge about 40 euros for a 1 dm^2 single sided PCB on a 10WD lead time, including postage.

Thanks again,
Tim.
 
Tim: In addition to Andy's comments, which I agree with, in general conducted and radiated noise levels will be lower if you _make_ the ground network have as low an impedance as possible. And this usually suggests multi-layer pcbs with a solid copper pour applied to every millimeter of the board that isn't being used for traces. Certain circuit nodes may prefer not to have capacitive coupling to ground, in which case I would broach the ground layer in the immediate vicinity of the circuit node in question, and possibly add guard rings.

The ground behavior tends to be more stable if you install the power supply filter capacitors so that the return nodes to ground are all grouped in as small a loop as possible, preferably with no roundabout routes or thinned copper between. In your design, for example, I would flip C19/C20 by 180 degrees so that the ground return nodes are facing each other, and do the same for C21/C22. If you look over your design with this concept in mind, I am sure that you will see other areas which can be refined.

Regarding Per's layout, in my opinion here too not enough effort has been paid to reducing the ground network impedance. Now if you wanted to increase the track widths as much as possible, this could be a plausible reason not to install a solid poured copper ground, but as it is, this hasn't been done, either. Also, I suggest that via's are avoided as much as possible, unless there is absolutely no other way to connect the required traces and pads together.

My two cents.

regards, jonathan carr
 
Tim

I looked at PCB-Pool for my boards, but they are, frankly, bloody expensive!

Talk (email!) Minnitron (www.minnitron.co.uk) a D/S board like yours will have a one-off NRE (tooling charge) in the order of 40ukp, but cost per board should be quite sensible (a few pounds), even for small quantities.

They offer excellent service but you'll need to send them Gerber files + NC drill data - not a problem with any of the major PCB packages.

Standard lead time is 15 working days, with modest premiums for faster service. I wouldn't use anyone else anymore, adn use them exclusively for my own, and work-related boards.

Ask for a quote for various numbers of boards, you'll be surprised how much the prices can drop for even modest quantities.

I had one batch of boards with faulty PTH holes, but these were replaced within a few days, FOC and I got to keep the faulty (but useable to me, boards.

Andy.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.