HPA-1 Class A Headphone Amplifier

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
www.hifisonix.com
Joined 2003
Paid Member
Here is a very low distortion class A headphone amp.

This design is similar to the design I've used in my new Symphony preamp (still to pusblish the write up)

Top level specs:-

Distortion at 3 V peak output ~500 ppb
At 1 V peak 20 ppb

Class A operation up to 12V peak output

SR ~ 20V/us

no pcb - but I'd be happy to help anyone who wants to take up rhe challenge - the circuit is simple.
 

Attachments

  • HPA - Low distortion Class A HP Amp_V1.0.pdf
    640.6 KB · Views: 1,094
Last edited:
bad units in description

...that at 3V peak output will deliver ~400 ppb distortion at 20 kHz. On 90 dBm headphones, this corresponds to about 115 dB SPL...

dBm is 1 mW power referenced sensitivity - must specify load in Ohms too to be able to convert "3 V peak" to SPL - "90 dBm" alone is not enough info

newer headphones sensitivity are sometimes given in dBV - often manufacturers datasheets won't tell you which reference for the "dB sensitivity"


at <5 W per output Q I'd probably just use the 1.5 A BD139/140 as output rather than 8 A rated 250 V MJE15032
although most manufacturers don't give ft spec with the BD - the original Phillips spec was >3x the speed of the MJE - with the related advantage of less parasitic C for the op amp to drive


actually I did use TPA6120 as output in a multiloop, paralleled and biased into Class A push-pull with >100 mA standing current in a headphone amp
 
Last edited:
www.hifisonix.com
Joined 2003
Paid Member
These figures are for 32 Ohm phones as quoted in typical phone specs.

You could use alternative output devices - the MJE's however have good thermal performance and are relatively cheap.

Yes, measuring distortion would be difficult. On the simple SE buffer shown elsewhere here, the distortion in the 1 to 3 volt range was at or below the AP threshold.
 
Last edited:
www.hifisonix.com
Joined 2003
Paid Member
How would it perform with LME49990 substituted for the duals?

It should be ok. You do need to be able to check for stability though with a WB scope. These new opamps have very high GBWP, so loop compensation considerations are important.

The 49990 is SOIC only IIRC - but noise and distortion are much better than the 4562. Keep feedback junctions close to the inverting inputs for the best results.
 
200 mA of Iq for a headphone amp - taking no prisoners, eh? ;) It never even leaves Class A with 32 ohm cans, up to 2.5 watts of output. The opamp buffer better be able to drive 9 V into ~3 kOhms though (no problem for reasonably well-biased output stages).

Bug hunting time: Several part numbers in circuit description do not match schematic shown (e.g. R21, R15/R22, U1, U2).

The opamp buffer has a Zobel - not explained.

Dunno about R15/22 being such a good idea. For one thing, the LM4562 output stage is likely to be running 3+ mA of Iq as-is, so 0.7 mA does not seem like it would do much good. (The technique is best-suited for types with at least reasonably beefy but current-starved output stages and Pdiss to spare. MC33078 @ est. 300 µA should benefit ginormously from 3-5 mA, for example.) And is it really a bootstrap?

The biasing strikes me as clever. How'd you come up with that?
(BTW, metallic heat conduction beats that through plastic any time of the day, so for fastest response I'd go with a short, thick trace from Q3 emitter to Q2 base. Not like it matters much here.)

Finally, Kaiser-Bessel windows don't bite. ;) FFTs definitely look more purdy like that...
 
www.hifisonix.com
Joined 2003
Paid Member
Thanks for pointing out the mismatch between circuit and description - I tweaked the design after the write -up, and then pasted in the document. I've reposted the document with updates below.

U2 RC network from output to ground looks like a Zobel but its not - its just a loop compensation network. However, R22 and C9 connected from the output to GND is of course a Zobel - to compensate for cable inductance.

R15 and R22 (I think you mean R15 and R18?) - I have not seen any figures published for the output bias current on the LM4562, or the other LME4xxx family. My assumption is that it is low - I have heard it more like 300uA to 1mA in most opamps, and that's what I've assumed here. In any event, if some other opamp type is used, these resistors will guarantee they operate in class A operation.

Output Bias stage: this is what I used on my new preamp headphone stage and on the sx-Amp - also class A. I am not aware of it being used elsewhere - its also easy to configure it for EF2 by the way. The advantage of this approach is there is no adjustment (although I did provide some on the sx-Amp), and it transitions very gracefully to class AB into heavy loads.

Metalic conduction - yes - most of the coupling can be through the leads if you locate the device close to the heatsink. Better still, if you are good with SMD, a BC847C is a good choice - I use that a lot in my designs.

:cool:
 

Attachments

  • HPA - Low distortion Class A HP Amp_V1.1.pdf
    685.3 KB · Views: 451
Last edited:
U1 and U2 are still reversed. R28 ought to read R18.

What I like about your biasing is that it takes ones of those pesky BE junctions out. Of course in an AB design you'd have to be careful not to have it accidentally limit peak output current, but if in doubt that could probably be sorted out with the 100µ cap.
IMO graceful transition into AB has more to do with the low R_e values chosen - normally you'd end up with ones closer to 10 ohms. 1 ohms with a 32 ohm load is almost like 0.22 ohms with an 8 ohm load.

As far as opamp output stage currents are concerned, 0.3-1 mA seems about right for types of average current consumption, but higher-performances types may well run their output stages hotter than that. Samuel Groner's measurements are quite instructive here. Distortion tends to rise sharply as the output stage goes into AB, especially in older designs that may have significant emitter resistors. If (as an IC designer) you manage to omit these while keeping Iq stable, you win.
Old designs tend to be the most instructive here. Compare, for example, TL071 THD vs. level to the same for LF356. They're both quite old, but one runs at an Icc of 1.4 mA while the other gobbles up 5 mA while not being that much faster. The difference in distortion with a 600 ohm load, however, is tremendous. We see it rise above the noise floor at +7 dBu for the LF356, vs. more like -3 dBu in the TL071. Running Class A up to +7 dBu requires an Iq in the order of about 2 mA. (I assume that the VAS in the '071 is rather wimpy as well.)
For a laugh, look at the MC33078 with 600 ohms. Yikes. (You can see output stage distortion even with the ~9k load in the transfer linearity graph.) That's a JLH-like quasicomp output stage (they suck in AB) with very little Iq. Shame because the input stage actually is quite decent. The LM837 is quite instructive as well - its internal feedback keeps distortion at bay at low frequencies in spite of seemingly extremely low Iq, but transfer linearity at 10 kHz shows even more output stage distortion than in the MC33078. A rather quirky part in general. Finally, the trusty NE5532, another quasicomp output. That one's looking more like the LF356 again, though distortion at 0 dBu suggests somewhat less Iq; in return it does better at higher levels. Nested feedback is doing its thing here, too, I assume.
 
Hallo to all,

What do you think about D44H11 D45H11 as output couple?

Is possible to omit R3 R1 to ensure the lowest output impedance for best damping factor and distortion/frequency response control of the low loads?

Is possible to use +/- 18V DC supply (if the adopted opamp can handle this voltages) without any changes?

Thank You
 
Is possible to omit R3 R1 to ensure the lowest output impedance for best damping factor and distortion/frequency response control of the low loads?

Is possible to use +/- 18V DC supply (if the adopted opamp can handle this voltages) without any changes?
No, whatever way the output stage it is set up in this design it acted like a short when these were bypassed. a smaller value may work.

I'm interested if it's possible too because from what I read the main reason for using higher value is related to crossover distortion and for this amp my headphones would blow up before it leaves class A.
Ideally I would cut the class A bias in half to reduce heatsinking requirements and power supply load (also makes building a balanced pair less ridiculous), I still dont think it would ever come close to leaving class A if I did.


You will probably be told output impedance is basically 0 because feedback is taken from the output, in other words the error caused by slightly higher output impedance (which granted will be pretty small) is corrected by feedback.

Im pretty sure the supply can be increased to the limits of the op amp but dissipation will increase also, im running it lower at around 13V because i needed to give the regulator a bit more voltage headroom on its input, sounds fine.
 
Last edited:
www.hifisonix.com
Joined 2003
Paid Member
You must not remove the emitter resistors. They are needed for thermal stability and they also provide local feedback. The output impedance of the HPA-1 before the 3.3 ohm resistor is very, very low (milli Ohms).

I will take a look and see if I can replace it with an inductor since I realize a lot of folks want ultra- low Zout.

:)
 
Before it caused huge power draw with the emitter resistors bypassed,
the reason for trying it was to reduce output Z when using the output stage openloop.
For comparing ''zero feedback'' and global feedback within the same amp, I found that it sounds worse open loop and is an amazing sounding amp regardless.
 
www.hifisonix.com
Joined 2003
Paid Member
How to add a volume control to the HPA-1 if you are using bipolar input opamps. If using FET input devices, you can skip the capacitor.

The reason for the cap in bipolar input devices is that the input bias current causing ‘scratchy’ noise as you adjust the pot level.
 

Attachments

  • 740D7408-6428-4F04-877B-41319A58970F.jpeg
    740D7408-6428-4F04-877B-41319A58970F.jpeg
    746.1 KB · Views: 226
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.