Hot-running 317 / 1086 in Tangentsoft's STEPS psu

[hollowman]

I built Tangentsoft's STEPS psu and PPA2 headphone amp based on the designs on Tangent's site; I'm using a 24vdc version of STEPS to power the PPA2.

When loaded with the PPA2 amp, the STEPS' voltage reg. runs hot, even with a decent heatsink (see image below). I've tried both 1086 and 317 regs; both run hot. STEPS was built correctly per the design on Tangent's site -- this means all test-point metrics are "okay" except, perhaps the voltage drop across the reg (see below). With no load, the reg runs cool. All this said, neither the 1086 nor 317 ever shut down due to thermal overload protection.

The PPA2 amp was also built correctly (TPx metrics are fine; "finger-temps" are okay; it sounds good), but with STEPS I noticed clipping at high volumes (same volume levels where other "lesser-quality" headphone amps are okay). This may be due to STEPS not being able to supply sufficient current/power (??)

I also have an ELPAC 24vdc 0.33A wall-wart. It performs fine (i.e, runs cool, and no clipping).

Using STEPS to power the PPA2, I noted that audio clipping didn't happen until about 5-10 mins into use. This seems to correlate with temp of the reg vs. the temp of the heatsink: the heatsink gets uncomfortably hot quite quickly, but the reg temp stays okay until about 5-10 mins of use.

Some FYI stuff:

With STEPS, I deviated from the suggested component recommendations as follows:

C7 = 100uF/50v Panasonic FM
C8 = "
C5 = 4700uf/50v Nichicon (only one large cap is used)
Heatsink: The one suggested ran hot, so I engineered what I consider a "better" heatsink using thermal tape to piece together units for more surface area (see image). But this strategy does not make a huge difference -- the reg STILL runs hot.

PPA2 draws (needs) 130mA (idle) at 24vdc. So, P = 3.18W

Tangentsoft has a Power Supply Parameter Estimator. It spat out these results:

"Unregulated voltage = 44-53 V, ripple = 0.2 V, drop across regulator = 20-29 V, regulator temp < or = 30 deg. C over ambient"

I measured the voltage drop across the regulator: 23vdc. That's, of course, way more than the recommended datasheet max of 2.5vdc for the LM317 or LT1086. The developer (Tangent) suggested that the huge drop across the reg may be the culprit. Dunno why Tangent designed it this way?

Bottom line questions:

Why does the reg run hot with a load of 3.18 W (at continuous current draw of 130mA / 24vdc)?

Is it due mostly to the large voltage drop across the reg?
TTBOMK that, by itself, doesn't explain why the reg runs cool with a small load (created by STEPS' built-in indicator LED and its current-limiting resistor).

That is, even with no load, shouldn't the 317/1086 be "burning off" the excessive voltage drop to maintain a 24VDC output?

[hollowman]

The cause of the overheating problem, as it turns out, is embarrassingly simple.

When calculating power dissipation in the regulator, power equals current through the regulator times voltage drop across the regulator (Vdo). Vdo, as noted in the orig. post, is 23v, which is clearly too high.

... even with no load, shouldn't the 317/1086 be "burning off" the excessive voltage drop to maintain a 24VDC output?

I *think* that for very small loads the reg maintains a high input resistance, only "trickling" in current. When load reaches some "threshold", it turns on *all* the way -- i.e., no scaling or no in-between. Pure speculation here.

Anyway...

For anyone interested in building the STEPS PS, please do not use Tangent's default transformer suggestion (as I unfortunately did), nor assume the validity of the "Analysis: Configuration looks sane" comment in the Power Supply Parameter Estimator.

All that said, I don't want to buy another tranformer! Does anyone know of a quick and clean way of reducing the pre-regulator rectifier voltage (Vin) other than, say, using a series of several cascading regs, or via a voltage divider using high-power resistors?

[infinia]

Hi
You have almost 2x the raw voltage needed into the Vreg. So if you drop the Vraw to half by rearranging the secondary windings and diodes (take out 2 diodes and ground the center tap). Now adjust the Vreg output voltage down (change R4) by the amount of headroom (3 V or so). Simple and elegant, should work... yes.

P.S. You also could rearrange the primary to run at 110V with a 220V setting, but you'd cut the XFMR VA by half.

[hollowman]

Originally posted by infinia
You have almost 2x the raw voltage needed into the Vreg. So if you drop the Vraw to half by rearranging the secondary windings and diodes (take out 2 diodes and ground the center tap). Now adjust the Vreg output voltage down (change R4) by the amount of headroom (3 V or so).

P.S. You also could rearrange the primary to run at 110V with a 220V setting, but you'd cut the XFMR VA by half.

Thx for you reply. For my line voltage (117VAC), Vraw x 0.5 = 23.5v. Then, subtract 2.0v for Vdo, and then there's a good volt or two of drop along the power path from PPA input to the power pins of the op-amps, so the STEPS should (ideally) output 27 or 28 V. Even if the PPA amp could live with lower voltage, I'm not sure I want to remove two diodes due to added ripple. Your second suggestion seems better: halving transformer VA (power) is still more than plenty for this application.

Rock and hard place as far as this transformer is concerned!

I started building this STEPS psu about two years ago, but tossed it into a to-be-completed drawer for various reasons. Since then, I've built several PSUs that are probably better; I only really finished the STEPS project for the sake of completeness.

Overall, I think the STEPS project is a good one for beginning DIY, and Tangent has done a good overall job in PS scholasticism. But for better performance, one should look elsewhere.

[infinia]

Hi
Quickly looked at the amp schematic and the whole power distribution scheme seems pretty dodgy overall. I'm thinking that the amp would be much better served with real +/- Vregs, w/ quality bypass caps, using good layout and grounding techniques. Re> the supply... How could someone messup a XFMR with this much V overhead is a puzzle and makes one question the other design choices?

P.S. Changing to 2 diodes and center tap return is still a full wave rectifier, so the 120Hz ripple stays the same, just uses the XFMR windings with a little less eff.
Running with 2x series primary windings would be the fastest to setup and a little less modding overall tho.

edit> This amp is also for battery operation so I wouldn't worry too much about setting Vreg a couple of volts lower.

[AndrewT]

use a 24Vac transformer for a regulated 24Vdc supply!!!!!

[hollowman]

Originally posted by infinia
Quickly looked at the amp schematic and the whole power distribution scheme seems pretty dodgy overall. I'm thinking that the amp would be much better served with real +/- Vregs, w/ quality bypass caps, using good layout and grounding techniques. Re> the supply... How could someone messup a XFMR with this much V overhead is a puzzle and makes one question the other design choices?

P.S. Changing to 2 diodes and center tap return is still a full wave rectifier, so the 120Hz ripple stays the same, just uses the XFMR windings with a little less eff.
Running with 2x series primary windings would be the fastest to setup and a little less modding overall tho.

edit> This amp is also for battery operation so I wouldn't worry too much about setting Vreg a couple of volts lower.

I just tried your suggestion WRT the primary (220 VAC). Vraw=18.5 and Vout=17.3. It seems to work fine: opamps have a bit over min. voltage and they are cascode class-A biased. Sound qual. is fine, too.

I agree with your other comments. Some of the issues you raised -- real +/- Vregs -- are covered on Tangent's site, but the site is somewhat disorganized and pages are hard to find.

Getting a bit OT, but TTBOMK, the PPA2 is the one of highest-rated DIY amps out there. If someone knows of a better one, I wanna know so I can build it!

[infinia]

Originally posted by hollowman

I just tried your suggestion WRT the primary (220 VAC). Vraw=18.5 and Vout=17.3.

It's a littler lower than I thought it would be from the 42-44 Vdc you meas. before.
Linear Data sheet shows LT1086H Dropout Voltage at approx 1.0V at 200mA, so you should be OK there, but ripple rejection improves at a 3V drop. No biggie but should be cool enough to run w/o a heatsink.

[infinia]

Originally posted by hollowman

I agree with your other comments. Some of the issues you raised -- real +/- Vregs -- are covered on Tangent's site, but the site is somewhat disorganized and pages are hard to find.

Hi
I still haven't found the discussion of the power scheme, but in retrospect on battery operation I'm sure there were tradoff's to accomadate them. In the big picture of what they (batteries) gain, I don't really know if it's worth it unless I built two versions to listen to. What's your set up and do the batteries make a difference in your experience?

PS It might be worth it and try the other mods instead, to boost Vraw a couple of volts. (better XFMR regulation due to VA derating) Pull D1A&C, wire the center tap to the largest cap ground directly. Use the DMM on VAC mV setting to verify for yourself Vripple pre and post mods.

[hollowman]

Originally posted by infinia
I still haven't found the discussion of the power scheme, but in retrospect on battery operation I'm sure there were tradoff's to accomadate them. In the big picture of what they (batteries) gain, I don't really know if it's worth it unless I built two versions to listen to. What's your set up and do the batteries make a difference in your experience?

Dunno -- although I have the battery board for the PPA, I've never used it (still have to finish assembling it!).

I have a couple of other headphone amps and PSUs, DIY or modded commercial models, and batt. vs. AC is always an interesting comparison. Batts produce smoother, more focused sound, but they lack the dynamics, bass slam and infamous PRaT of a good AC psu. Also, with AC PSUs, I can always seem to keep tweak/improving them. With batts, there are not a whole lot of options: a 2000uF electro at the batt. pack improves dynamics and bass. Some swear by batt, but they are using car and motorcycle models! Batt. power for me is all about portability.

The discussion of the power scheme is here (look for "Single Voltage or Dual Voltage Power Supply?" in this page -- sorry, not as much as I thought!)

PS It might be worth it and try the other mods instead, to boost Vraw a couple of volts. (better XFMR regulation due to VA derating) Pull D1A&C, wire the center tap to the largest cap ground directly. Use the DMM on VAC mV setting to verify for yourself Vripple pre and post mods.

I'll definitely keep it in mind. One can also snubber the regulator. But methinks the overall topology of STEPS ain't all that great -- especially at the regulator (the stuff before it is salvageable). Perhaps choosing another, completely different, but tried-and-test option is better:

http://www.pinkfishmedia.net/forum/s...ad.php?t=44928
http://www.pinkfishmedia.net/forum/s...ad.php?t=39990