All the best,

---

[Quote]

Did you experience any problem?

---

To my big surprise, no! Did you register as a private person or through a company name?

Regards

Many thanks,

All the best,

I plan to make some prototype of this one and the other one for BUF634. I should have 4 of the intersil one and 6 of the Burr-Brown one. If everything is ok, maybe I will put my surplus on the marketplace. Price will follow my costs.

All the best,

I was just thinking everything was dead here, but apparently, there are still fresh will to continue

I will first test them as soon as possible (I still check my designs to be sure everything is ok, and hope those tiny traces will be ok for production). I will keep you all informed, and if it's enthusiasm enough, I could make a group buy, but still, I'm not sure because it involves a lot of work it seems.

Currently, I plan to produce boards enough to fill a 10 by 10 cm board, so not that much, and with other stuff on it. This costs already around 80 euros with masks and upper silkscreen. Maybe I didn't choose the cheapest, and surely, the price can go down with quantity. I also have no idea how much should it costs to ship this internationaly (mainly to the States for sure).

I also made a very tiny epsilon12 simplified layout just for me (don't want to spoil AMB's work on this) because my plans are to put everything in a cramped space. I think (dBel84 comments) that such kind of device is mandatory.

I would like some help to compute values for Bandwidth resistors (for the BUF634 board solution), and small bypass cap for stability aswell (same design). And the same kind of question of the limiting resistors for the Intersil design; and what if one stacks two intersil by channel, just twice the resistors values?

Just a question, the LM78/79 in TO220 package are 1A capable of?

Many thanks again for your enthusiasm.

All the best,

GregVDS

Regards

* Single Opamps w/BUF634
* Dual Opamp w/BUF634
* Single Opamps w/Intersils
* Dual Opamp w/Intersils
* Digi's discrete buffers

We have GregVDS, Gewa, and Digi all looking at these boards, or at least expressing interest. Shoot - I'll buy some from each of you if that will help.

Greg:
About the BUF634, from TI's data sheet:

---

[Quote]

BANDWIDTH CONTROL PIN
The –3dB bandwidth of the BUF634 is approximately 30MHz
in the low quiescent current mode (1.5mA typical). To select
this mode, leave the bandwidth control pin open (no connection).
Bandwidth can be extended to approximately 180MHz by
connecting the bandwidth control pin to V–. This increases
the quiescent current to approximately 15mA. Intermediate
bandwidths can be set by connecting a resistor in series with
the bandwidth control pin—see typical curve "Quiescent
Current vs Resistance" for resistor selection. Characteristics
of the bandwidth control pin can be seen in the simplified
circuit diagram, Figure 1.
The rated output current and slew rate are not affected by the
bandwidth control, but the current limit value changes slightly.
Output voltage swing is somewhat improved in the wide
bandwidth mode. The increased quiescent current when in
wide bandwidth mode produces greater power dissipation
during low output current conditions. This quiescent power
is equal to the total supply voltage, (V+) + |(V–)|, times the
quiescent current.

---

As I understand it, and this can be found in Tangent's parts strategy for the PIMETA, quiescent current increases raise bandwidth for better performance. However, this has a usable limit of about 220 ohms per BUF (with +or- 24V, I believe). Stack two buffers, and this becomes 110 ohms. What happens is that distortion increases so that diminishing returns occur. I believe the resistor sizes and distortion measurements were performed by KurtW during the PPA/META42 days. He is actually a professional in semi-conductor testing, as I understand it. His measurements became the basis for much of what we know in headphone amp design with opamps and buffers, I think (including stacking, which is not mentioned by TI or Intersil, either one)

I have a Millett with 50 ohms on a stack, based on a recommendation by Steinchen, and it works fine. It could be that tube circuits are more tolerant of the increased distortion where a PIMETA would show it up quickly. In either event, Runeight proposed the BUF without any bandwidth control. As noted earlier, we're all in uncharted waters here, but I think it's prudent to put a position for the resistor on the boards. That gives us the choice to use it or not.

To answer your specific question, Greg - the bandwidth is adjusted by the quiescent current, so if two are in parallel, the resistor must be halved to maintain the same current to each of two buffers.

As for the capacitor/etc., here's a diagram from the TI data sheet that pretty much explains the capacitor needs, etc. - which is not needed in most cases:

Finally, about the Intersil - leaving the output resistors in the SOHA may be the best thing. From what I've read in the data sheet, as little as 10 ohms may be enough to keep them stable in a parallel config (stacked). That's exactly what I had in the Millett output with the stacked Intersil's, so maybe that's why it worked:

---

[Quote]

The HA-5002 can become unstable with small capacitive loads (50pF) if certain precautions are not taken. Stability is
enhanced by any one of the following: a source resistance in
series with the input of 50Ω to 1kΩ; increasing capacitive load
to 150pF or greater; decreasing CLOAD to 20pF or less; adding
an output resistor of 10Ω to 50Ω; or adding feedback
capacitance of 50pF or greater. Adding source resistance
generally yields the best results.

---

The diagram for sizing the output protection resistors was posted a couple of times by Gewa already - Intersil recommends R= V/Imaxout, where Imaxout cannot attain 200ma, and V is either the positive or negative voltage swing. For instance, the data sheet says the Intersil will work with as little as +or-5V, so Rlim=25 ohms. For +or-12V, Rlim=60ohms. Obviously, if 200ma is a hard limit, you'd want some safety factor probably.

The Intersil has no bandwidth adjustment, so nothing to do there. However, just as with the BUF's, the output protection resistors would halve with paralleled buffers, since their total potential current could double (400ma).

Sorry for another monster post , but Greg keeps asking these questions.

[Edited by tomb on 10-30-2006 at 09:36 PM.]

Do you still have SOHA V3 boards left? Tried to PM you, but my profile shows no sent messages so don't know if you got my message and thus I'm posting public post (sorry!). I could buy two if you have any left.

Ps. Yeah, tried to design single-sided home etchable pcb, but lost my nerves. Post office probably will bring me faster two-sided pro PCB's than I can get single-sided one to work :)

Best regards,
Jari Eskelinen

Gewa, I tried to put the two resistors in parallel to the back under the pcb, but the traces become really contorted.

All the best,

GregVDS

I'm also losing track with all the different boards I came up with!

GregVDS
I don't know what you mean here?

Regards

All the best,

GregVDS

ps, I went and ordered prototype boards from PCB-POOL, just enough to fill 1dm² (4 buffer with BUF634, 6 buffer with HA3-5002, and some other stuff). Now, wait and see...

as soon as I have things, you'll have pictures

---

[Quote]

I tried to put the two limiting resistors backside under the HA3-5002, but the traces became difficult to draw. I will maybe try again. That's just that, nothing else.

---

Ah, O.K. What I ment was, to solder the SCP resistors physically to the pins of the IC socket for the Intersil buffer so you didn't need to draw the traces and pads on the PCB.
I hope youre proto boards come out the way you want, I'm also curious about the result. Never heard of PCB-Pool though.

Regards

They intelligently provide a dru file (rules to respect for production with them in eagle) with which you can check your board, and correct clearance and dimensions. I of course checked and send designs that fully passed these requirements.

I'm quite anxious at the result. that's quite exciting to know that I can design whatever I want, and have a board produced with it. I always wanted more integrated solutions (with epsilon12, linkwitz switched by relays, so the signals do not have to come from the back to the switch and come back, but only driving current, stuff like that).

I'm quite attached to symmetrical design, and in this respect, the Millett Hybrid implementation was a mess, now it seems there is a new layout in discussion, hint hint (I'm not involved, just making noise about because it's simply a very nice design, with discrete buffer integrated, and regulation, like the SOHA board has). I become lazy to connect my STEP to everything .

I still should take some pics of my SOHA board...

I currently tried it only with OPA2134, and really hope the buffer will help, for the sound quality is at the moment not comparable to the Millett Hybrid with NOS RCA 12FM6 and Steinchen DB. In the Soha, I tried JJ, which was quite nice, and also two RCA5963, one of which was awfully microphonic, the other one only very faintly. Sound is nice, but I restrain to judge now because the buffer isn't fair enough. all the resistors and caps are better in the SOHA (BlackGate and Silmic II vs Nichicon UHE, PRP and Kiwame vs vishay-Dale). The linkwitz followed the same upgrade except the orange drops. It's quite curious the PRP/Kiwame one seems to produce practically indistinguishable effect in comparison with Vishay-Dale resistors... but the microswitch I used in the SOHA are not easy to manipulate, hence I didn't compare a lot the three settings of the filter in the SOHA

GEWA, Ok, I didn't understood your point with the Rlims. This is indeed the simplest way to go. I prefered the normal way, since someone can fit testpoint plugs in the holes so as to exchange resistors, to fiddle around values, or if one wants to try first one intersil/burrbrown, then go stacking, you can replace resistors easily.

I hope my move to go further and begin prototyping these boards is not misinterpreted, I don't want to block anyone developping other boards, nor I want to subsitute myself to Digi01 in his will for groupbuy. as I said, my designs are free, if anyone is interested, GEWA, or Digi01, my brd files are available at request (they are so simple that anyone thinking about it should arrive practically at the same result, it is not a very personal work). But even before sending physical board to anyone, I wait for production to arrive, test it in my SOHA (all went ok with the Intersil samples request, my 4 intersils are on their way home), and maybe then they will be labelized ok for groupbuy, we should not go the wrong way, or a lot of us will be disapointed, or even angry with board not tested.

Could someone point me some good eagle libraries for tube sockets, or do I have to make my owns? If it's the way to go, I would like to know precise pads positions, if someone can help on this...

And again TOMB, a question, what should be the current requirement for a grado (32 ohms)? I want to use your long post to calculate resistor values.

All the best,

[Edited by GregVDS on 10-31-2006 at 02:39 PM.]

Check your inbox

Regards

Your post is much appreciated and very timely. I think the buffers have always been needed on the SOHA. With low impedance phones, the opamp may be controlling almost everything with a little bit of tube "flavoring" the signal. That is most likely an unjustified simplification, but a partial explanation. High impedance phones may react much better because voltage leverage is far superior even without the buffers. This was confirmed somewhat by Vixr's report at the "power" and much more sensitive volume control - the gain is noticeably higher.

If you go far enough back in the threads, you will find TheDoctor's post stating his preference for the 200ma OPA5551 over any other opamp he tried. Also, you can note runeight's numerous admonitions about investing too much in part selection. That's probably a nice way of saying the above. It may also be reasonable to conclude that the output resistors are not just an afterthought, but perhaps an essential component for all but the highest impedance phones.

BTW, I would love to purchase some of your buffer boards if you have a couple.

As for the buffer current, it's essentially tied to the bandwidth resistor in a relationship that's illustrated with this graph:

You guys mark this link:
www6.head-fi.or...distortion+test
As best I can tell, this was the original thread that baselined buffers for headphone amps, and also started stacking buffers. If you read closely, there is no mention of a bandwidth resistor - only regular (open pin on bandwidth) vs. WB (loop back w/o resistor - 15ma) The documentation for the PIMETA is where it is mentioned that lower values for the BW resistor raises distortion: "According to tests by KurtW and others, distortion drops as bandwidth goes up, though once R11 falls below about 200 Ω, distortion starts rising again." You will note that the middle of that graph is close to 200ohms, so that might have been a prudent choice.

Anyway, that thread says it all about stacking, and as near as I can tell, the resistor may not be needed. That's because all of the BUF634 distortion measurements are lower for all-out WB, and for all-out wideband x 2. I'm confused now.

[Edited by tomb on 11-01-2006 at 06:10 AM.]

Regards

---

[Quoting digi01]

another SV3 with classAB buffer.

---

Digi,
What does someone need to do to purchase your buffers?

Gewa: Cool board clamp!