Does HFE have to match on this headphone amp?

[djsb]

There is distortion on the output of this headphone amplifier. I have removed the output transistors and measured the hfe (using a Peak DCA75) as follows
Q3=398
Q4=301
Q5=422
Q6=303
Am I correct in assuming that the hfe of these transistors SHOULD be matched (the pnp's are quite close)? It does not mention having to match them in the service manual. The amplifier is from a Revox B261 FM tuner, and I've attached the schematic. Thanks.


PS All the resistor and capacitor values have been checked and the opamps are OK. The distortion does not come from the demodulator.

[magic]

These transistor have zero current gain (they are inactive) for output currents less than 2mA. This is much worse distortion source than pairwise mismatch (because if anything, you would match each NPN with its corresponding PNP).

That being said, this product probably wasn't meant to sound entirely like crap, so if it does sound like absolute crap, chances are that something else is broken.

So what exactly is your problem? If it's only a matter getting more zeros on a THD meter, your time is likely better spent buying/building a different amp...

[djsb]

When the tuner powers up the relay that connects the output clicks on. I hear distortion in my headphones (I've checked All the passive components in the signal path from the output of the op amps to the headphone socket and done continuity checks). I have tried another pair of headphones, distortion still present. I place (a quick tap of my finger) then remove a finger onto the tracks of the PCB near to the headphone amplifier section and the distortion disappears for a while. Then it slowly returns. I place my finger back again, and it goes away again. After a while it seems to settle down and the distortion gets less (maybe a temperature thing).
I have replaced All the ceramic with new multilayer ceramics (and Electrolytic capacitors) and put new LM301 Op Amps (socketed) in. I've checked the PNP/NPN junctions (all OK). So, as I'm running out of ideas, I thought I'd check the hfe of the transistors and the results are presented here. I am not familiar with the theory of how push/pull amplifiers work, and I  don't know if this is a class B or AB amplifier. So I'm not sure if these transistors have degraded or if the hfe values are correct. The sound quality is very good when the distortion is NOT present. I can reproduce the distortion by turning the tuner off for a while and then turning it back on (this is when the relay clicks ON and connects the headphones to the amplifier output). Just thought I'd ask here for a some help. Thanks.


PS I have a theory that back EMF from the relay coil may have caused the transistors to degrade over the years. I'll probably just replace them anyway and see what happens. Or would that be a waste of time?

[Kleinstein]

The effect of taping with the finger may indicate oscillation of the amplifier. The compensation around the LM301 looks a little like the feed forward compensation, but with extra capacitance to ground. The transistors add some phase shift, so the the compensation my be OK by design, but it it could as well be borderline to start with. RF range oscillation would not be heard directly but can caus distortion of the normal audio signal.

When replacing the capacitors used for compensation / filtering one should be careful with X7R and similar modern types: they can be highly nonlinear and are not meant for such use. The X7R and similar are mainly good for supply decoupling and not for filtering in the signal chain. 100 pF are usually C0G - C0G caps are the good MLCCs and as good to better than most film type caps.

AI agree with magic: the amplifiers are not build to low distortion - more like using the old OPs with a rather complicated compensation to get the best performance from the old OPs So it is not easy to change to better modern types, as the distortion / noise may be worse.

[magic]

Is it one channel or both simultaneously?

You are sure it has nothing to do with the volume pot R1? Maybe turn the volume all the way down and short the wipers to ground, see if the problem persists?

No oscilloscope at hand?

I don't know, just in case, test the resistors and retouch all solder joints. New transistors - what the heck, may try it too. They are common as dirt. Try to get some brand name ones, not random junk from eBay.

And by the way, if you got your LM301 from eBay, they may not be real LM301. Possible serious if that happens.

[tooki]

When the tuner powers up the relay that connects the output clicks on. I hear distortion in my headphones (I've checked All the passive components in the signal path from the output of the op amps to the headphone socket and done continuity checks). I have tried another pair of headphones, distortion still present. I place (a quick tap of my finger) then remove a finger onto the tracks of the PCB near to the headphone amplifier section and the distortion disappears for a while. Then it slowly returns. I place my finger back again, and it goes away again. After a while it seems to settle down and the distortion gets less (maybe a temperature thing).
I have replaced All the ceramic (and Electrolytic capacitors) with new multilayer ceramics and put new LM301 Op Amps (socketed) in. I've checked the PNP/NPN junctions (all OK). So, as I'm running out of ideas, I thought I'd check the hfe of the transistors and the results are presented here. I am not familiar with the theory of how push/pull amplifiers work, and I  don't know if this is a class B or AB amplifier. So I'm not sure if these transistors have degraded or if the hfe values are correct. The sound quality is very good when the distortion is NOT present. I can reproduce the distortion by turning the tuner off for a while and then turning it back on (this is when the relay clicks ON and connects the headphones to the amplifier output). Just thought I'd ask here for a some help. Thanks.


PS I have a theory that back EMF from the relay coil may have caused the transistors to degrade over the years. I'll probably just replace them anyway and see what happens. Or would that be a waste of time?

I don’t think the “replace everything” approach is a wise one.

Did you replace the components mentioned in an attempt to resolve the distortion, or did the distortion arise as the result of replacement?

Unless you know it to be faulty, there’s no reason to ever replace a ceramic cap; they don’t age in any meaningful way, and as Kleinstein said, it’s very easy to get modern multilayer caps that have very nonlinear characteristics that are unsuited to use in the signal path.

As for his suggestion that it might be oscillation: this sounds plausible. And if you replaced the op amps, especially if they are from eBay or Ali, they could be modern, faster op amps that are more prone to oscillation. Put the originals back in, maybe? 

[tooki]

There is distortion on the output of this headphone amplifier. I have removed the output transistors and measured the hfe (using a Peak DCA75) as follows
Q3=398
Q4=301
Q5=422
Q6=303
Am I correct in assuming that the hfe of these transistors SHOULD be matched (the pnp's are quite close)? It does not mention having to match them in the service manual. The amplifier is from a Revox B261 FM tuner, and I've attached the schematic. Thanks.

No, amplifier feedback (as used here) obviates the need for matched components, super clean power, etc.

[djsb]

The ceramics are Y5V dielectric from toelectroniccomponents.
LM301AN from jhcomp25.
Transistors from tcelectroniccomponents. All being sorted using Peak DCA75 so this should catch any fakes. Non found so far.
ALL resistors checked (all passives checked by desoldering one pad). Capacitors checked with BK 879B LCR meter).
The distortion (or to more accurate very fast intermittent signal or stuttering) has NOT got worse since the component replacements so far.
I agree that the "replace everything" approach is not the best. It's just that the fault is not obvious.
If it doesn't matter if the transistors are matched, I will replace the originals with matched pairs anyway.

Is there some sort of noise when there is no signal?

I will have to recheck this later, but I can only hear very low volume music.

Or is it only when you are actually listening to something?

Again, I need to double-check.

What happens if you connect your headphones to the main output of the tuner?

The sound on the variable output is nice and clean and is not affected.

Is it one channel or both simultaneously?


Ill double check this later.

You are sure it has nothing to do with the volume pot R1? Maybe turn the volume all the way down and short the wipers to ground, see if the problem persists?

The pot is not noisy, but I'll try shorting the pot to ground as an experiment.

No oscilloscope at hand?

That's the next thing I'll try, together with injecting a signal at the volume pot. Should have tried this first.  At least my desoldering/soldering has got quicker and I've found out how good Goot Wick desoldering braid is.

I'll review the posts again later and answer any questions I've missed. Thanks for all the replies so far.

[David Hess]

Transistor hfe mismatch will not matter in this circuit, but the transistors could still be bad.  Some obscure transistor problems can only be found on a curve tracer.  Given how inexpensive they are, I would have replaced them when I replaced the 301As.  There is nothing special about the BC550/BC560 in this application so something else could be used if more convenient.

The compensation around the LM301 looks a little like the feed forward compensation, but with extra capacitance to ground.

That is definitely feed-forward compensation but since they only took feedback from the transistor outputs, I wonder if they had compensation problems so added C10.  It looks to me like they started with the recommended feed-forward circuit without understanding its requirements and added parts until it worked.

I have never liked that class-B current boosted amplifier circuit but the alternative two transistor boosted output circuit shown below was pretty obscure at the time.

I place (a quick tap of my finger) then remove a finger onto the tracks of the PCB near to the headphone amplifier section and the distortion disappears for a while. Then it slowly returns. I place my finger back again, and it goes away again. After a while it seems to settle down and the distortion gets less (maybe a temperature thing).

That makes me think there is a cracked trace or solder joint which allows the biasing somewhere to drift out of range.  Try tapping different areas with an insulated tool.

I have replaced All the ceramic (and Electrolytic capacitors) with new multilayer ceramics and put new LM301 Op Amps (socketed) in.

Replacing the electrolytic capacitors with ceramic capacitors might cause oscillation problems.

[magic]

Y5V dielectric

Nonlinear rubbish
Put the original caps back in unless they tested wrong, doubly so if replacement didn't change anything.
As Kleinstein says, C0G/NP0 (same thing) is the stuff to get for "signal" capacitors.

LM301AN from jhcomp25

I don't know this seller and I'm not familiar with eBay in general. But I do know that some shops in Europe sell genuine parts and some cut corners and resell stuff imported from dodgy sources in China. Maybe others can offer more useful opinion

The distortion (or to more accurate very fast intermittent signal or stuttering)

That may possibly be bad contact somewhere or maybe bursts of oscillation.

If you have a scope, it's a no-brainer. As soon as you hear something weird, put a probe on the output of the amp and see what's going on. It seems to be DC coupled, so check for DC on the output.

[magic]

I have never liked that class-B current boosted amplifier circuit but the alternative two transistor boosted output circuit shown below was pretty obscure at the time.

Are you sure it performs any better?
It looks extremely dodgy, how is the quiescent current of the output transistors defined?
I mean, I know, it has to be zero. Sure enough, quiescent current of µA741 is specified as 1.7mA and that's 300mV at the bases of the outputs.
I think crossover distortion would be worse than for that emitter follower with resistor bypass

[David Hess]

Are you sure it performs any better?

Its frequency compensation is much easier despite the configuration of the output stage multiplying the slew rate.  It is suited for internally compensated operational amplifiers where there is no other choice, but it works just as well, or even better, with externally compensated ones like the 301.

It looks extremely dodgy, how is the quiescent current of the output transistors defined?
I mean, I know, it has to be zero. Sure enough, quiescent current of µA741 is specified as 1.7mA and that's 300mV at the bases of the outputs.

I think crossover distortion would be worse than for that emitter follower with resistor bypass

It is not shown in this example which is biased closer to class-B, but it is easy to control the quiescent current of the output transistors by configuring them as current mirrors.  I picked this one to show that this circuit configuration was as least potentially known at the time that the 301 was first available.  Walter Jung published the variation shown below which is closer to what I would do.

[floobydust]

OP, the Revox headphone amp impedance spec is 200-600ohms, something popular in the 80's and in recording studios back then.
If you are using 32ohm headphones I would expect the amp to clip if you expect decent volume because of the high losses in the output resistors R15, R16 120Ω plus R2, R3 100Ω giving maybe 12% or about 1/2 line level to the cans.

I don't consider it to be a hi-fidelity headphone amp and it doesn't need to be,  FM radio is over 0.2% THD, never mind the incredible amount of signal processing the stations use.
First I would fix the bad connection before blaming any components.

[Terry Bites]

Any sane desiner would try to stop Hfe makeing big differences to the performance and stabilty of his prodcut.. Hfe is an unrelaible spec and it changes with current and temperature.
Negative feedback both AC and DC should render hfe variations null and void. The tranistors are operating at low gain and increasing the current drive of the opamps. The Zo is 120.  I I think your distorion is commin from the excessive line level drive to the output amps. Or possible badly connected or pooped our 150pf compensation caps.
The input stage has of about 3. The output amps have a gain of about 3.

The diagram sugests that the preamp/ attenuation stage should generate a line level of 2V, you should expect 6V.6V/((Zo @120)+Zheadphones)) mA.
so for 300R- 60ma will flow giving 1W. Not bad. Will definitley damage your hearing.
So if it sounds vile, check that compensation caps on the LM301s (wow first made in 1967). COG/ NPO will be no better than the cheapest crap ceramic here. You might want to replace these transistor bufferd opamps with a pair of LM386s. So much less bother. A little daughter board operating off the 15V rail. 

[floobydust]

I don't match your numbers, and find the LM386 sounds worse than a trashy AM radio, high distortion and noise, no just don't use it.
Zo = 120R+100R = 220R, don't forget the extra 100R at the jack.
The headphone amp has a gain of -4 (1+100k/33k) edit: -3 (100k/33k), fed from Line Out.
The output stage transistors are emitter-followers, so a voltage gain of 1. The hFE is mostly irrelevant because the NFB loop corrects for any differences.
This can put out some serious signal swing into the correct load, rails are +/-15V for the op-amp and +/-22V for the output stage. But it's wasted in the output resistors and as I mentioned before, 32 ohm headphones give 1/8 the voltage.

OP can possibly improve the design by adding bias (diodes or VBE multiplier) to the output stage, lowering the output resistors (although the relay shorts them out), a better op-amp with way less oddball compensation (delete C1, C2).
Or just put in a small circuit board with any of the better headphone amp circuits that are out there. Or just the use tuner Line Out and your separate amplifier.

[magic]

I propose to compromise and settle on the gain of the HP amp being -3[/bi]

So if it sounds vile, check that compensation caps on the LM301s (wow first made in 1967). COG/ NPO will be no better than the cheapest crap ceramic here.

Already replaced
And no, crap ceramics don't belong here. This capacitor transforms global feedback into local feedback around the second stage; if it is shit than it render the second stage increasingly nonlinear at high frequencies, right as global loop gain falls down. I can't tell right now how strong the effect is in comparison with output stage nonlinearity, but why would you do that if it costs nothing to get it right?

Modding that thing is a waste of time. Get an external amp if you don't like it, use it as is if you do IMO.

[djsb]

No one is flogging a dead horse here. I will change the crap ceramics with COG/NP0 (any suggestions?). I posted here in the beginners section as I am a newbie when it comes to amplifiers and always willing to learn. Thanks for all of your input.

[magic]

I meant it to those who recommended capacitor replacement after:
a) it has already been done
b) it made no effect

BTW, disregard what I said about compensation, this configuration doesn't use Miller compensation. That being said, the feedback capacitor C5 still experiences full output swing and variation of its capacitance with output voltage changes the closed loop bandwidth of the amplifier, with some smaller effect possible in the audio band. Why take chances with crap capacitors if you don't need to?

I will change the crap ceramics with COG/NP0 (any suggestions?)

I would just put the old ones back in.

[floobydust]

Good books on learning audio, will save you a lot of time debunking myths - look at books by Douglas Self such as:
Small Signal Audio Design
Electronics for Vinyl
Power Amplifier Design Handbook

Author Bob Cordell I don't think went into headphone amps.

For small capacitors, COG/NP0, polystyrene, teflon and other exotics the size of a beer can do have lower distortion due to the low dielectric absorption. But it's usually causing ppm levels of THD depending on where it is in a circuit. Most practical is COG/NPO but in a tuner I would leave the original parts in.

In an EE job interview, they gave an exam and a couple questions were op-amp circuits and asked what the gain is. In a hurry I used intuition and apparently the wrong formula but got the correct answer. So since then I get it wrong with "1+". Exam anxiety lol.

[djsb]

Just had a look at the +/- 23v supply rails for the headphone amp. These rails are unregulated and taken straight from the bridge rectifier (filtered with a 2,200uF capacitor (which I have replaced with a new Vishay part) and 2 off 10k bleed resistors (not checked)). There is 180mV RMS of power supply ripple at approx 100Hz. Maybe this is normal and of no consequence for an unregulated supply?
I don't like this at all, even though it is as designed. I don't think this ripple is going to help with the amplifier's stability. I have attached the power supply schematic for your perusal. Any thoughts. Thanks.


PS Putting in the new transistors did not cure the problem with the distortion. Lovely sound though when it's not crackling. The search goes on for the fault. I will take all of your suggestions into account. Thanks.

[Kleinstein]

The dielectric absorbtion is usually still linear - at least for the low DA parts and polyseter caps. So the DA does not cause any THT, but it can have a small effect on the frequency response.  With electrolytic caps and X7R I would not be surprised of the DA is also nonlinear. The main problem there is howver nonlinear capacitance.

The DA gets an link to linearity in integrating ADCs - here the DA it self may be linear and can still cause and INL error for the ADC.


Some ripply at the +-22 V should not be a problem. A crckling noise before getting bad and than intermittant working / non working may indicate a cold solder joint or similar problem inside a component.

[David Hess]

Just had a look at the +/- 23v supply rails for the headphone amp. These rails are unregulated and taken straight from the bridge rectifier (filtered with a 2,200uF capacitor (which I have replaced with a new Vishay part) and 2 off 10k bleed resistors (not checked)). There is 180mV RMS of power supply ripple at approx 100Hz. Maybe this is normal and of no consequence for an unregulated supply?

I don't like this at all, even though it is as designed. I don't think this ripple is going to help with the amplifier's stability. I have attached the power supply schematic for your perusal. Any thoughts. Thanks.

Power supply ripple and power supply noise is attenuated by the power supply rejection (PSRR) of the operational amplifier which is quite high so ripple and and noise can usually be ignored.  Any amplifier design which requires a regulated supply because of low power supply rejection is suspect.

[djsb]

This is a follow up. I'm embarrassed to say that the problem all along was the relay. I finally did what I should have done from the start. I injected a sine wave into the buffer stage and traced it through. Signal rock solid until it came to the relay contacts. Removed the relay and did continuity/resistance checks. Relay contacts are measuring 20 to 70 ohms. NC contacts around 0.3 ohms.
I've made a custom pcb fitted with 2 Axicom IM07 24v relays to replace the original American Zettler AZ7-4C-24D 4 pole double throw original.
I'll post a photo when I fit the PCB. Thanks everyone for the efforts you made to help me. I have learnt a lot from the discussion here. I also need to be more structured in my fault finding approach. Thanks.

[Simon]

The hfe of a transistor should have little to no impact on the amplifier. Transistors will have different gains, that is expected. That is the whole point of the art of designing an amplifier, designing a circuit that will work as expected with a range of transistor hfe. The OP-AMP abstracts this further by producing a "block" that has such high gain that it really does not matter if it is 10 million or a billion when your circuit is going to be set up with a gain of 100.