What a difference a transistor can make!

[Mark Hammer]

So over the weekend, I leapt at a Kijiji ad and picked up a mid-to-late 1960's Heath TA-16 amp.  It's about as plain vanilla from that era as you can find.  Two channels with treble and bass baxandall tone controls, and reverb/tremolo on one of them.  Two 12" CTS speakers and somewhere around 25W ("60W peak" says the catalog ad from that year) feeding them.

I have to do a bit of work on it to clean it up and restore it, and I have a number of mods planned, one of which is the installation of an LFO speed idicator LED for the tremolo.

While searching around for more information, including the schematic (http://harmony.demont.net/documents/schematics/amps/heathkit_TA_16_01.gif), I stumbled on a web page where the owner of one purchased in a similar manner indicated that he had replaced the input transistor with a 2N5088 to reduce the hiss (which I was sort of willing to overlook on mine, given the price paid).  That first transistor is normally a 2N3391, labelled with a Heathkit house number as 417-91.  I figured, what the heck, may as well try a 2N5089 since a) I had some kicking around, and b) they're supposed to be even quieter.

Holy frijolé!!  Dead quiet.  I mean, like a 20db reduction in hiss (or more).  I wish my tweed Fenders were this hiss-free.  There is still the remaining matter of the hum stemming from 45 year-old electrolytics (soon to be replaced), but once that gets cured, the amp will be blessedly silent with respect to anything that isn't guitar signal.  This is all the more pleasing since I'm installing bright switches for each channel, neither of which would have done me much good if I was perpetually turning the treble down to rid myself of hiss.

The moral of the story is that sometimes a different transistor (abnd perhaps a newer one) can make all the difference in the world.

[thegnu]

awesome!  that looks like a sexy amp.

and "frijole" doesn't have an accent, btw.  the general rule in spanish is that the stress falls on the second-to-last syllable--the most notable exception being infinitives, which end in -er, -ir, or -ar--and you only need to write an accent to place the stress on another syllable. ok, i'll shut up now.

[Mark Hammer]

I have that key on my keyboard.  I mistook the spanish pronunciation for the french pronunciation.  Now it's MY turn to shut up......but thanks for the insight.

[PRR]

That is one funky output stage.

Yeah, 25 Watts steady sine, 60W instant peak.... Heathkit didn't cheat specs.

Very low input impedance, ~~10K.

Really "hardly enough transistors" for what has to get done.

> Holy frijole!!  Dead quiet.

Hmmmmm. 2N3391 had a noise-spec, and is not that different from 2N5089. Should not be any "wow" difference in hiss. My guess would be that some 1960s packages have not stayed hermetic all these decades, transistor gone leaky.

(Or _maybe_ old rosin gone hygroscopic, and you cleaned it when replacing.)

(OR... that bias circuit is very unstable. Maybe 2N3391 was off-bias and '89's Vbe just happens to work better?)

> installing bright switches

Buffer the input first, see what you hear. 10K will suck the top off most guitar pickups. (Course if driving through nearly any pedal, you already have buffered it.)

[Mark Hammer]

Yeah, I'm kinda mystified as to why the transistor change should have made that sort of difference, but boy oh boy, it sure did.

I changed a bunch of the larger value electrolytics this evening - some with larger values, like 220uf instead of 100uf - and that did a nice job of improving the hum specs, though the 4000uf/50v unit will need to wait a bit until I can score a suitable replacement.

I haven't compared guitar straight into the amp vs buffered input yet.  It's getting kind of late to try out the buffering thing via a MOSFET booster right now, but I intend to try it out tomorrow and see if the buffering makes the bypass cap more self-evident.

[R.G.]

Yeah, I'm kinda mystified as to why the transistor change should have made that sort of difference, but boy oh boy, it sure did.

No real mystery. Bipolars get noisy when their base emitters are reverse-broken, even once. Many of the stock circuits store up charge in an emitter cap and pull the input low when power goes down. There is a small, hardly noticeable breakage on the base emitter in many of these circuits. Over time, the little breakdowns degrade the input noise. I suspect a new 3391 would have been quieter too.

One really slick way to prevent this is to hook a diode up reversed across the base emitter. That makes it permanently impossible to break the base emitter, as the diode will conduct first. Ott goes through a lot of this kind of thinking in his book on low noise electronic design. Input transistor abuse is common in guitar amps, and that can make it worse.

I changed a bunch of the larger value electrolytics this evening - some with larger values, like 220uf instead of 100uf - and that did a nice job of improving the hum specs, though the 4000uf/50v unit will need to wait a bit until I can score a suitable replacement.

Electronics goldmine, Panasonic 18kuFs for $2, last I looked. I bought 20 for some old amps I'm reworking.

[petemoore]

No real mystery.
  I always wondered how literally ''solid state'', and ''transistors that never need serviced'' blurbs were.
  Bipolars get noisy when their base emitters are reverse-broken, even once. Many of the stock circuits store up charge in an emitter cap and pull the input low when power goes down.
  This could be more common than I thought. Hiss from out of the past from what sounds like old circuit inputs on tape decks etc. all of a sudden the hiss is there.
  There is a small, hardly noticeable breakage on the base emitter in many of these circuits. Over time, the little breakdowns degrade the input noise. I suspect a new 3391 would have been quieter too.
  If quiet was possible from a current device at the time, it'd make it a good choice for input component for Heathkit product.
  One really slick way to prevent this is to hook a diode up reversed across the base emitter.
  Was this a commonly known thing around the time just prior to it's production ?
   That makes it permanently impossible to break the base emitter, as the diode will conduct first.
  Ott goes through a lot of this kind of thinking in his book on low noise electronic design. Input transistor abuse is common in guitar amps, and that can make it worse.
  Seems like it would be a junction safety/convenience to offer a bundle package with the B/E protection diode built into the chip.

[PRR]

Sorry: the impedance at Q1 Base is ~~10K, but there's 22K in front, so ~~32K presented to gitar. Still quite low.

Also pushing signal down toward hiss. 100mV at jack is 33mV at Q1 base. That's "reasonable" in light of tube versus (healthy) Si transistor noise voltage. However Q1 runs at a rather rich 1.4mA so its noise OSI is more like 300 ohms; the 10K it looks into puts it way up into current-noise instead of voltage-noise. Many 1960s designers were not fluent in current-noise. (And yes, B-E breakdown will raise current-noise, and as R.G. observes, input-abuse happens.)

Buffer impedance but don't gain-up. I estimate input overload around 100mV. (And 5%-10% 2nd-harmonic in Q1 as you near input overload.)

I want to re-design it more toward modern lines. And yet the world is full of fine modern designs, and the old simple transistor amps should maybe be preserved.

[skumberg]

Sorry: the impedance at Q1 Base is ~~10K, but there's 22K in front, so ~~32K presented to gitar. Still quite low.

Also pushing signal down toward hiss. 100mV at jack is 33mV at Q1 base. That's "reasonable" in light of tube versus (healthy) Si transistor noise voltage. However Q1 runs at a rather rich 1.4mA so its noise OSI is more like 300 ohms; the 10K it looks into puts it way up into current-noise instead of voltage-noise. Many 1960s designers were not fluent in current-noise. (And yes, B-E breakdown will raise current-noise, and as R.G. observes, input-abuse happens.)

Buffer impedance but don't gain-up. I estimate input overload around 100mV. (And 5%-10% 2nd-harmonic in Q1 as you near input overload.)

I want to re-design it more toward modern lines. And yet the world is full of fine modern designs, and the old simple transistor amps should maybe be preserved.

Hi, I'm just beginning to learn about optimal source impedance(osi). How would the source impedance be calculated for the above circuit with a guitar plugged in? Is it (C1+R101+guitar)||(R1||R2)||(base impedance)?

/Nicklas

[PRR]

> source impedance ... Is it (C1+R101+guitar)||(R1||R2)||(base impedance)?

This input is just NOT optimum for noise. At that time and place, it was considered good-enough, "it's only guitar". And Mark's observation is that a happy transistor IS quiet enough.

The SOURCE that the amplifier sees is just (C1+R101+guitar)||(R1||R2).

And BTW in this case C1 is insignificant over the audio band. 10uFd is around 1K impedance at 20Hz and less for higher audio frequencies, 100 at 200Hz, 10 at 2KHz. You have 22K in series. Working rough, 1K+22K is nearly 22K, and 10+22K is dead-nuts 22K. (C1 is bigger than it "needs" to be; they used that value because it is cheaper than smaller values which tend to be film-caps.)

Similarly R1 may be neglected for practical figuring.

In figuring the effective INPUT noise, you also need the attenuation from guitar to transistor. It's awful steep. Guitar may be 5K to over 100K. Then the 22K in series. There's a 24K shunt, and hFE times 47 ohms in shunt.

hFE is a wide variable. And it affects jack to base loss, and also current-noise.

Then for more fun, the OUTput noise is affected by gain which is affected by hFE. If you are eager, work out the voltage gain jack to vol-pot for hFE varying 4:1 (the type of loose-sort favored by profit-making designers). Output noise is not itself significant: the user will adjust the volume-pot to the gain needed for sound/tone. If two "identical" TA-16s have 2:1 spread of gain, one will be used with Volume turned-down 2:1 (about one number on a 0-10 audio-taper pot). But an example of the little details you wrestle when you have "barely enough transistors for the job". (Those days ONE transistor cost as much as two gallons of gas; compare to what Mark likely paid for his "kicking around" transistors.)

[Thomeeque]

 Reverb driver seems to me pretty weak compared to other designs, how does reverb sound?

[skumberg]

PaulR: Thanks for the informative and helpful reply! I've looked in some books but one always ends up with questions.

[Mark Hammer]

Reverb driver seems to me pretty weak compared to other designs, how does reverb sound?

The reverb is actually reasonably nice-sounding, though I haven't had an amp with reverb since 1975, so I might not be the best judge.  I'm still wondering whether the caps need some changing to tailor the sound.  You will note that the reverb stage is tapped off the output of the preamp, after the tone controls, such that whatever the volume, treble and bass are set to will determine the resulting sound of the reverb.  If the volume is set low (but not off), it is technically possible to have more reverb than clean if you set the reverb control to max.

I tried the amp out last night with a Minibooster (JFET mu-amp), and boy oh boy is there a difference in treble response when you feed it a lower source impedance than a straight guitar.  So yes, the input impedance is clearly low.  Kind of makes me wonder what the impedance was of the pickups found in those Harmony guitars that would often accompany Heathkit amps.

[PRR]

> wonder what the impedance was of the pickups found in those Harmony guitars that would often accompany Heathkit amps.

They may have been low-turns to save a penny, thus lower-Z than modern boutique hot-rod pups. The body/neck may have been cheap, thus gutless and shrill. HeathKit was not LA or Birmingham, tastes in the middle midwest may have been mellower than cutting-edge. And the tone control offers considerable treble-boost (not really the same as lightly-loaded pup resonance).

> boy oh boy is there a difference in treble response when you feed it a lower source impedance than a straight guitar.  So yes, the input impedance is clearly low.

You could, with minimal mangling, insert a buffer on one input. Here's a mosh with AMZ's full-frills buffer. If you leave out all frills, it could be done on a 4-tag strip tack-soldered to the input jack, or even air-wire on the input jack. (Air-wiring won't be road-worthy but I don't think this will be a road-trip gigging amp.)

[Mark Hammer]

If air-wiring was good enough for EHX back in the day, then it's good enough for me.  Besides, there already IS some of it in the amp, on the back of the control panel.
Can't see what you posted from work (it'll have to wait until later in the day), but I was considering insertion of just such a front end.  here's the thing.  You want your unaffected guitar to be bright and clear, and you want overdrives and fuzzes to rein in that excessive top end.  The current arrangement, with low input impedance, favours the top end in distortions and fuzzes (which have buffered outputs), but NOT in the straight feed of guitar; which is sort of the opposite of what you want.  The two alternatives would appear to be trying to build in greater treble-taming capabilities in every clipping device I own, or simply giving a directguitar feed a fighting chance.  I prefer the latter.

[R.G.]

Given the simplistic mature of that front-end circuit, it's tempting to put in an MPSA18 for the transistor and break the direct connection of the base to the bias divider, subbing in a 100K to 1M metal film and bypassing the bias divider to ground. This gets a bona-fide low noise preamplifier transistor in there and converts it to both high impedance and "noiseless" biasing; and without much air-wiring. Another possibility is to integrate a J201 as a source follower into the front of the transistor to kick the input impedance up while keeping much the same circuit.

[R.G.]

Here. Like this.

[brett]

Hi
building DIY hi-fi amps back in the 1970s involved pulling them apart every few years and putting the latest greatest BJTs in the front end.  Signal:noise ratios in 1970 were barely better than 60dB, but improved to about 80dB by 1980.
I can still remember getting hold of BC107s.  They sounded "invisible".  During this period many "hi-fi" amps had 741s in them  .
cheers

[PRR]

> integrate a J201 as a source follower into the front of the transistor

I don't think your sketch works yet?

You've kept the R1 R2 470K 24K, which originally was setting base bias voltage. But you have inserted the J201 JFET and its gate-source voltage. If Vgs were zero volts, fine. For that to happen, FET current must be Idss. Which is very variable, but several mA. What is the FET current? 0.65V across 1Meg plus Ib, OTOO 0.02mA. Vgs will be a volt or two. Say 1.65V. Then voltag across R4 appears to be 1V, BJT current appears to be 1V/47= 20mA, drop in R3 6800 is 136V, from a 16V supply.... transistor is slammed, no output.

Heath's bias was pretty wobbly anyway. 0.05V across emitter resistor leaves no room for slop.

[R.G.]

> integrate a J201 as a source follower into the front of the transistor

I don't think your sketch works yet?

You've kept the R1 R2 470K 24K, which originally was setting base bias voltage. But you have inserted the J201 JFET and its gate-source voltage. If Vgs were zero volts, fine. For that to happen, FET current must be Idss. Which is very variable, but several mA. What is the FET current? 0.65V across 1Meg plus Ib, OTOO 0.02mA. Vgs will be a volt or two. Say 1.65V. Then voltag across R4 appears to be 1V, BJT current appears to be 1V/47= 20mA, drop in R3 6800 is 136V, from a 16V supply.... transistor is slammed, no output.

Heath's bias was pretty wobbly anyway. 0.05V across emitter resistor leaves no room for slop.

I did worry about that. The J201 has the lowest Vgsoff of any JFET I know of, as little as 0.1V. But I did expect some diddling of either the 24K or the emitter resistor. It was a kind of puzzle left for Mark. 

Frankly, a JFET/PNP compound with gain would be even better. That's what was used in the input of the Thomas Vox "Beatle" and its ilk.