Mini³ is pronounced mini cubed. The 3 is HTML ³ (Unicode 00b3).
On a PC, type Alt+0179 or use Character Map.

Design Goals

Small, cheap, robust, great sound, easy for a novice to build
KISS philosophy (Keep It Simple, Stupid)

• 3 channel topology based on the AD8397 opamp
• ground channel separates high current reactive load from signal ground
• designed specifically for the Serpac H65 case
• through hole components for easy assembly (except the opamps)
• holes for soldering EPADs for maximum thermal transfer to both layers
• pcb mounted pot and jacks
• full ground plane on both sides
• ~$50 parts cost
• space saving 150mil 1/8W 1% metal film resistors
• Alps RK0971221Z05 10K pot with SPST power switch
• FET Zener current source turns off the LED when the batteries are low
• 2 9.6V NiMH Maha Powerex batteries
• drives high impedance headphones without clipping, unlike single 9V amps
• LM317L constant current source trickle charges the batteries with 25mA
• compatible with 24V wall warts such as the Elpac WM080-1950-760
• 2 resistors split the power rails to provide the reference for signal ground
• ferrite beads isolate capacitive loads without raising output impedance
• low value resistors, pot, and no input capacitor result in low DC offset
• R1 prevents DC offset caused by intermittent pot wiper contact
• 2 ICs for a total of 4 opamps: left, right, output ground, signal ground
• works with other dual opamps like the LM6172 or AD8620
• AD45048 is supposedly identical to the AD8397 but has no EPAD (boo!)

Attaching the SOIC EPAD

The only surface mount component is the AD8397ARDZ opamp. The ARDZ designation refers to the version with an SOIC EPAD, an exposed metal square on the underside of the package that conducts heat from the chip substrate to the pcb, improving power dissipation. Normally this would be reflow soldered in a factory, but we have devised a means to do this manually.

Apply a flux pen to the EPAD and EPAD landing on the pcb. Use a thin strip of adhesive tape to hold the IC to the pcb, making certain the pin alignment is correct. Turn the board over and apply solder to the hole in the center of the EPAD landing, which will flow through the hole and bond to the EPAD. Remove the tape. If the chip is soldered, qapla'! If not, tape the chip back on and heat the hole again. Once the EPAD has bonded you can solder the pins. If you solder the pins first, you will have no way of knowing if the EPAD bonds. Bonding the EPAD is not required, but will improve thermal transfer, and is recommended, although it will make removal of the chip nearly impossible.

Ferrite Beads

The AD8397 is the crankiest opamp we have ever seen. A standard circuit with power rail bypassing, compensation caps, and balanced impedance at inverting and noninverting inputs oscillates when presented with capacitive loads such as a few feet of unterminated coaxial cable. It is possible to quell these oscillations with 33Ω output resistors, but this raises the remarkably low output impedance and wastes power. Ferrite beads are the way to go, as their impedance rises with frequency, and they are lossier and thus more absorptive than ordinary inductors.

There are many different kinds of ferrite, each with a different frequency range. A trip to the local surplus shop resulted in 4 different kinds of ferrite beads, all unlabeled. All worked well at stabilizing the circuit under normal conditions, however, only 2 of the 4 ferrite beads were effective in controlling 3MHz oscillations that appeared during RMAA loopback testing with a defective cable. The most common ferrite beads are type 43 with a range of 20-400MHz and type 73 with a range of .5-50MHz. We are concerned that the more common type 43 ferrite is inadequate, and have both types on order from Mouser. We will report back on this asap.

Panasonic makes ferrite beads with an unknown and possibly proprietary ferrite composition with a range of 1-1000MHz, according to their graphs. These should work, and are available through Digikey.

Using good beads, our breadboard prototype works flawlessly.

History

As some of you know, the idea for a 3 channel cmoy style opamp project has been on my back burner for many years. The release of the AD8397 was the impetus that brought it back to life. We started working on it last summer, but went on hiatus from September to January. Some resemblence to Tangent's PINT is due to frequent communications between us and because there is only so much you can do with a couple of opamps. PINT is surface mount for mint tins, Mini³ is through hole for Serpac H65, with onboard jacks. Mini³ should appeal to those who balk at SMD assembly, want a better case, and don't mind the 2.75×5" size. If you want something really tiny, buy a Hornet or SuperMicro/Macro/Mini.

[Edited by morsel on 02-07-2006 at 04:22 AM.]

but if you must, why not use both op amps in the ground channel chip?

http://headwize....e=opamp_prj.htm

fig 12a

shows the current sense/follower op amp paralleling method, with only 1 mV typ Vos the parallel ground buffer ad8397 would only draw 2 mA more with 1 Ohm sense resistors

2 ground op amps are necessary to hold down the active ground when both ouput channels are dumping in-phase, large signal current into ground

i ask because a 1000uF 25v panasonic FC has a height of 20mm which wouldn't work vertically in a H65 with an internal height of ~19.8mm before anything is put into it.

---

[Quote by skyskraper]

morsel, is it the team's intention to have C1 lie flat over OP1 and it associated circuitry?

i ask because a 1000uF 25v panasonic FC has a height of 20mm which wouldn't work vertically in a H65 with an internal height of ~19.8mm before anything is put into it.

---

As are Nichicon UPW - but UPJ and UPM series have 18x15mm (wider than tall) 1000uf 25v that would be worth looking into.

skyskraper, there are Panasonic FC 25V 1000µF 16×15mm and 1200µF 18×15mm. As ericj says, the caps are wider than they are tall.

[Edited by morsel on 02-03-2006 at 02:28 AM.]

Parts List

1 Mini³ pcb
1 Serpac H65 case
1 Alps RK0971221Z05 10K pot
1 knob 1/2" diameter, 6mm shaft
2 1/8" stereo jacks CUI SJ1-3533, Switchcraft 35RAPC2AV4, Kycon ST-315
1 2.5mm dc power jack CUI PJ-002B, Switchcraft RAPC712,
Kycon KLD-0202, Kobiconn 163-5003
2 9V battery clips
2 OP1,OP2 AD8397ARDZ opamps
2 D1,D2 1N4001 diodes
1 VR1 LM317L
1 ZNR ~6V zener diode
1 FET Idss ~1mA J201
1 LED
1 C1 16-18mm diameter 15mm height 25V 1000µF Panasonic FC
4 C2 .1µF polyester film or multilayer ceramic
2 C3 33pF ceramic
3 L1 FB73-101 ferrite beads (we are not sure about type 73 yet)
all resistors 150mil length 1/8W 1% metal film
1 Rc 47Ω
2 RV 20K
2 R1 100K
2 R2 1K
2 R3 1.2K
2 R4 6K

2 Maha Powerex 9.6V NiMH batteries
1 24V wall wart Elpac WM080-1950-760

[Edited by morsel on 02-10-2006 at 12:55 AM.]

This is a project that will take price/performance ratio of pocket amps to a new height. The AD8397 with its 310mA output current capability and rail-to-rail swing, plus the dual 9.6V batteries raises the bar for output drive capability of "pocket" headphone amps, yet does not require the complexity of separate output buffers.

One notable aspect of the AD8397 is the lack of internal output current limiting and protection. This is a reason why this opamp has such stellar sonics, because current limiting circuitry has been shown to deteriorate the sound quality. However, the high-output current capability and lack of protection means that this chip could be damaged if an output short circuit or overload condition is allowed to persist. So, caution should be exercised.

For this project, we chose to target specifically for the Serpac H-65 case, which then allowed us to design a highly-integrated PCB. With the exception of the battery clips, the volume pot, input/output and power jacks are all board-mounted. While the H-65 isn't the smallest of the bunch it still fits in a pocket, has room for two 9V batteries, and is relatively inexpensive.

I will make PCBs available only if there is sufficient interest. So let me hear your feedback. As mentioned in the first post, there is currently no ETA on PCBs.

[Edited by amb on 02-03-2006 at 04:34 AM.]

[Edited by skyskraper on 02-03-2006 at 07:36 AM.]

Please add me to the waiting list for a PCB or three.

As ususal, well done - very neat and tidy design.

Nate

2nd question, would it be possible to use board mounted battery tabs like in the hornet and headroom amps? I'm not too familiar with the H65, but if it's possible to avoid 9V snaps, I'm all for it.

Is the 1000uF looked at as a minimum, recommended or maximum value for C1? If it's not the minimum how low would you feel comfortable going? I'm thinking of trying to slam this is a TINY case, just for kicks.

Nate

I'm interested in your results, though. I've considered developing the ghetto version of this sort of design, using jrc4556 or jrc2068 on a single-sided board that people can etch at home. Maybe just a small rectangular board requiring off-board jacks and pot. Or just make it easy to cut off / not etch the end of the board with the panel components.

No reason you couldn't use other opamps, but those two jrc parts don't sound bad (imho), and can push enough current for low-z headphones.

I just sometimes get the feeling that there's a gap in the DIY progression between building a cmoy or a47 on perfboard and when you end up buying a double-sided board from someone.

[Edited by ericj on 02-03-2006 at 11:09 AM.]

I just put a Pimeta into a Serpac H67 case with 2 Maha 9Vs. I added a built in charger. In your case, the charger is even more important since I don't think the batteries can be removed without unscrewing the case. I like your design because it fits 2 bats into the slimmer H65.

Here is my question... I considered a simple resistor to control the charge verses an LM317 and picked the resistor. With an LM317, you will get a steady current flow of 20-25 ma to achieve an 8-10 hour charge. With a resistor, my charge rate started out at 40ma at the instant of connection, and declined to 30ma within a half hour. At the end of an approximately 10 hour charge, it levels off at 12.1 - 14ma (not sure yet why I get inconsistent results between my breadboard and the real thing). This was with a 120R resistor and 25.5V regulated DC input.

I am not an expert in these things, but I have searched the forums endlessly about this subject. While Tangent, for example, insists that 9V batteries do not like rapid charging, Thomas Distributing markets these batteries only with a fast charger (which I have used up to now). I don't think there is a formal data sheet on these batteries. So I guess there is some uncertainty in any charge rate decisions.

I picked the resistor topology because of two reasons:

1) I will not always be there when the 10 hours are up, and in the event of a partial recharge, I will have no idea when the time is up, unless I bother to hook a milliameter inline with the DC power, which I am not going to do for routine charges. I know that a few excess hours or even days of charging should not kill the batteries if I understand things correctly, but more or less continual charging is not a good idea.

2) More importantly, in my Pimeta configuration I am drawing about 30ma with OPA627x3 and Buf634x3 (no stacked buffer). That gives me 6-7 hours of run time followed by 10 hours of charge time. I believe dual AD8397s will draw the same current? Schedules are such that this dictates that I run the amp off of wall power when I can. Otherwise I will run out of juice when I need it because I cannot schedule my life around a battery charging regimen. But I am not really supposed to run these batteries under a continuous charge on a regular basis....

My feeling is that I know I am going to "over charge" the batteries in an attempt to keep them topped off for portable use. Given that situation, I decided that the resistor method will be much gentler on the batteries (a lesser evil?) since they bottom out at 12-14ma rather than 20-25 ma with the LM317. On the front end, while my charge rate starts at 40ma, the charger that is sold with these batteries must be running a constant 100ma into them to get their 2 hour fast charge. Because of that, I am hoping that my initial more rapid pace will not hurt the batteries.

I would appreciate your opinions on this matter.

P.S. I want the best postable amp I can get in the smallest case I can get (performance vs case size) to use with my ER4-S, which are my portable phones. I guess most people now days buy the ER4-P for portable use. I find the ER4-S very bass deficient compared to my HD-650's and as a result, I built a bass boost into my Pimeta. For me, that is critical for my real day to day portable rig. Am I out in the wilderness on this? It was tough to get the bass boost into the H67/Pimeta and I would imagine it would be even more difficult to do the same in the slimmer H65, but I think it would be nice to at least make a well thought out provision to hack in a bass boost circuit. It may be that in order to do that, the builder would need to move up to the H67.

Regards,
Neil

Thanks for your kind words, Nate. I slept like the dead last night, after 10 days of working nearly around the clock.

Regarding C1 and tiny cases, 1000µF is our recommended value. We didn't try using lower values of C1. We considered designing this for the Hammond 1455C801 or 1455C1201 but in either case there is only room for 1 9V battery. We didn't want to compromise output voltage or resort to SMD and figured if people want something tiny they can build a PINT or buy a Hornet or SuperMicro/Macro/Mini.

On the other hand, the whole point of this thread is to make changes according to your suggestions. If there is enough demand to go SMD and make it fit into a Hammond case, we could do it.

Excellent questions, jar.

The WM080-1950-760 linear supply has been in a state of apparent deprecation for some time. They are still being made, but availability is poor. The Elpac switching power supply we recommend is the MW1224-760. For a comparison, check out Tangent's Power Supply Tests.

We were thinking about offering board mounted battery tabs, but I forgot to bring it up. The battery tabs are big. The power jack would have to be moved, which would require extending the board wing towards the front. C1 would have to be smaller, or perhaps have 2 of them instead. It seems simpler to use the classic 9V clips on leads. Opinions, everyone?

Neil, you bring up a very good point. We chose the current source trickle charger without much thought, and because 25mA can safely charge the batteries all the time, but there is no question that long term use will shorten the life of the batteries. I'm sold on using just a resistor for faster variable current charging with a lower final current. What do the rest of you think?

Bass Boost is another thing that I meant to address. Obviously, front panel space is tight. There is not going to be room for a second pot. The curved case walls make it tough to move the input jack to the side of the case. A DPST switch could provide a fixed level of bass boost, either on or off. This could be air wired to the existing R4 pads along with the appropriate resistors and capacitors.

This reminds me, do people like the jacks on the left and the pot on the right? We could put the pot in the middle and the input jack on the right, but this makes for poorer access to the volume knob because the cables will be in the way on both sides.