• WARNING: Tube/Valve amplifiers use potentially LETHAL HIGH VOLTAGES.
    Building, troubleshooting and testing of these amplifiers should only be
    performed by someone who is thoroughly familiar with
    the safety precautions around high voltages.

Killer OTLs

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
IMO biasing procedure can be very simple ,

- Insert 32 ohm dummy load(power resistor) instead of loudspeaker
- Turn the amp in on position and wait about 15-20 minute until all tubes are fully warm
- by adjusting bias pot (R44)set for -20V on point J101
- press and hold switch S1
- adjust balance pot R46 to get close as possible zero reading on meter A1
 
IMO biasing procedure can be very simple, ...

That would appear to be the standard procedure for this amplifier.

Most amplifiers where triodes are used, it would be important to preserve the triode "signature" by matching up all the tube curvatures. But OTL amplifiers appear to be happy with any gm they can get, as long as it is enough to drive the low Z load.

I would at least go through the tubes with the tube tester and eliminate the weak ones, purely for drive capability. Any shorted or gassy ones get binned too. Then do the output voltage zeroing procedure. Balance the two tube banks (for idle current) by swapping around tubes as mentioned.

This amplifier appears to get its "triode" sound (2nd Harmonic) by mis-matched bank gains and N Fdbk.

Which makes it mysterious as to why they used a bunch of low power tubes in the first place.

Two big output Mosfets, or pentodes, would give exactly the same results (since the triodes are loaded down till they -look- like pentodes on the steep load line, at least low rp still), without tube aging problems (for the Fets at least). And less heat. (about 2/3 the total heat with pentodes, and 1/3 the heat with Mosfets)

Roughly guessing, each amplifier is running around 400 or 500 Watts consumption at idle, and 700 Watts at the full 50 Watts output.
 
Last edited:
Forgot to mention , before starting any biasing procedure input connector(J17) need to be shorted !, just to avoid any interference from voltage gain stage on OPS-IQ readings .

---------------------------------------------------------------------------------
smoking amp

In OTL amps during the time in some strange way output power tubes tend to equalize their mutual characteristics ,

`I would at least go through the tubes with the tube tester and eliminate the weak ones, purely for drive capability. Any shorted or gassy ones get binned too. Then do the output voltage zeroing procedure. Balance the two tube banks (for idle current) by swapping around tubes as mentioned.`
- Absolutely I agree with you !

`This amplifier appears to get its "triode" sound (2nd Harmonic) by mis-matched bank gains and N Fdbk. `
-Or maybe Mirko P. at that time don`t want to pay for patent royalties to Julius F. so his amps ends with that ``mis-matched triode`` sound :)

.
 
Last edited:

PRR

Member
Joined 2003
Paid Member
You say one amp dial-in easy. The basic concept of the amp is fine.

You say another amp would not dial-in. I suspect you have tube variations larger than the dials can accommodate. Wear, age, bad.... with 30 tubes, hard to say.

The "set bias to -20V" IS the old standard way of doing it. It assumes all tubes are alike in prototype and production, and have not drifted with age. Also that line voltage is "nominal" for that design.

That's how guitar amps used to be built. We never even biased them. Old tubes would eventually either quit or burn-up, we replaced them (and whatever damage they did in dying). But the world changed. Tubes got expensive and we struggled to use the ones we could get. Tubes got cheap when made in faraway places, without the fine control of parameters we used to have. Wall voltage has risen since 1961. G-amp workers are shifting to measuring cathode current, not trusting a grid bias voltage to give the right current with the tubes in hand.

All your 12B4 are Golden Age so you don't have the new-made tube uncertainties, good. But they are all 55+ years old. Tubes often show no large drift over decades, but a few will. Some may have gone way off their design parameters.

With two power tubes, you can sort this. With 30, it gets daunting.

Yes, I know about the zero-center meter in the amp. This is a final trim, after the balancing is roughed-in with a full set of good tubes. I believe it will be far too sensitive for roughing-in. (Also- does it work with no speaker? We would like to be close no-load.) We could tinker that, but you *have* a fine voltmeter/DMM.

"CP" comes from transistor amp jargon (sorry). This is a Totem Pole output, one stacked above another. (Or 15 over 15.) The load is taken at the center of the totem-pole, the Center Point. The voltage here normally should be roughly "centered" between the supply rails; for capacitorless output with a center-tap power supply it should be really close to power "zero" so no big DC flows in the speaker.

I believe sorting 30 tubes, even 2 at a time, is bound to get tedious. You have two variables in each test: upper tube(s) and lower tube(s), and an awful lot of tests.

The nominal operating point seems to be 150V P-K at 22mA (which is -20Vgk plotted on the sheet curves). Test one "upper" tube against, not another tube, but a Known resistor equal to what the lower tube should act-like. 150V/22mA is 6,818 Ohms.

If Cunningham would come back from beyond and hand you a "Perfectly Boogie" 12B4, you would bias for that tube to give Zero volts at CP (and speaker jack). Then check other tubes at that same bias and select for not-too-far off zero output.

Since we don't have a Known-Boogie tube for reference, you set bias for a small handful of tubes and take an average that is close for most, discarding any way-out tubes. Then leaving the bias pot fixed, run all 30-60 tubes through and note their CP voltage at that bias.

I was thinking to number the tubes and keep the notes in a notebook/spreadsheet. We could find reason for more notes. (Even if only to note brands and provenances.)

I didn't defrost my abacus, but I think the Bias pot has about +/-4V swing or +/-20% around nominal 20V. So I think it was expected that all good 12B4 would be in a +/-4V range at around 22mA.

With a set of 30 tubes all +/-4V on this test, remove the "lower" resistor, load the tubes, and I think it will bias-up fine. If you can show that wall-voltage is significantly different for you now than it was when/where the amps were sold, the "-20V" should be adjusted about in proportion. (I would speculate 117 to 125V, so "-20" should be -21.3V (-21 or -22)). DMM on CP may show many volts but you can dial that in with Balance (instead of poking the meter button).

If you resist even tacking-in that resistor: you can build an off-amp tester. Make some 170V DC (120VAC transformer through rectifier and >50uFd). That to plate pin, grid through 100K to zero V, cathode through 909r (1K) to zero. Voltmeter to cathode. Expect about +20V. The exact value does not matter; note it, test another, and sort for +/-20%. But also note the grid voltage. Nominally zero, a part-Volt may be OK, but anything more says the tube is gassy and could run-away.

BTW: idle plate power is 100W per channel. Heater power about 114 Watts/ch. Misc say 20W. Total 125W/ch at idle. In a Full Power Bench Test, the plate input power will rise; we could figure that if we cared. In unclipped speech/music work, the plate power will hardly rise enough to care.
 

PRR

Member
Joined 2003
Paid Member
Some notes on the plans shown in #22:

R41's value is unclear on schematic; parts list shows 1.8K.

The small caps are noted as "u" but can't be microFarads. They would be mmf or uuf in those days. I assume a note (not found) or shop convention says "all caps in uf"; then another u on the cap makes uuf or what we now call pFd.

Sheet 3 shows "Gain Chart" with audio voltages at three frequencies. 2.7V in makes 10V out. Observed gain of 3.7. However R41 R42 imply a gain of 9.18 max. If NFB is large we would expect gain near 9. Why the difference?

6AU6 with the super high 1.8Meg plate load should have gain over 400 over the lower audio band. The rest of the amp should be around unity gain. Cathodyne is shown with gain of ~~0.8. Upper power bottles have gain <1. Lower power bottles may have stand-alone gain over 2. The combined gain may be 0.9-1.0. So the forward gain is over 280, the gain-set is 9.18, we have 31:1 excess gain and expect gain error of like 3%. The amp working gain should be 8.8, needing 1.12V in for 10V out.

Gain difference not clear here.

Each set of fifteen 12B4 parallel @22mA each makes a triode with Mu still near 6.5. Gm is near 94uMho or 0.094S, equivalent to 10.6 Ohms effective cathode resistance. rp is about 87 Ohms.

So there is 43 Ohms of plate resistance shunting the 32 Ohm load. A pretty fair part of the audio output happens as plate-heat rather than voice-coil heat (and sound).

The output load is 43||32 or 18.4 Ohms. Disregarding active combining, the upper bottles have gain of 18.4/(10.4+18.4) or 0.63. The lower bottles have gain of 18.4/10.6 or 1.7. Output stage gain may be near 1.18, as a very approximate tangent on a significantly bent gain curve.

About loading vs THD: yes, in single ended triode amps a 5*rp load gives low THD without large loss of power. When the customer wants power too, we often go to 2*rp. The higher power may play low-power with THD similar to the 5X load, but won't go splatt so soon.

But in Push-Pull all that curvature of rp that we were trying to mitigate with high RL "cancels". The THD does not vary much with different loads, only the available power.
 

Attachments

  • MPE-annotated-1.gif
    MPE-annotated-1.gif
    29.9 KB · Views: 163
Last edited:
Disabled Account
Joined 2013
A lot of comment writings from you guys, how much a SIM for a change :)

For a pair of 12BH4A, the amp is quite well behaved but performed very badly with 14 pairs: a huge DC offset and much lower power output (<20W), high distortion than expected (what do think is the actual power out?), thank you for sharing.
 

Attachments

  • 12b4a OTL-1.png
    12b4a OTL-1.png
    82.1 KB · Views: 166
  • 12b4a OTL-2.png
    12b4a OTL-2.png
    78 KB · Views: 157
  • 12b4a OTL-3.png
    12b4a OTL-3.png
    47.1 KB · Views: 659
  • 12B4 OTL-64.asc
    20.3 KB · Views: 30
Simulation tubes should be -perfectly- matched. Seems odd that more of them would perform worse? (Oh, I see, the 2nd plot is for the 15 and 15 tubes with some matching, 15X the # of tubes makes for 15X the loop gain for the global N Fdbk) The 2nd harmonic distortion is pretty obvious for the 2 tube output without balanced gains. Looks like more than 0.9%. 30V + peaks versus 24V - peaks.

I notice that the gain equalization is not the same however as the Futterman scheme. Might try that instead, since it involves some local N Fdbk that might change the picture across low signal to high signal. Particularly for the DC offset. The present gain balancing may have DC matching different from AC matching. I think Futterman's scheme should get the two (AC & DC) to match together at the same setting.
 
Last edited:
A quick estimate: 27 V peak 1st harmonic, + - 3 V peak 2nd harmonic would give 24V and 30V peaks, which looks more like 11% dist by voltage, or 1.2% by power. OK, not that far from 0.9%.

Then 15X the tubes for 15X the loop gain (gm) should give 0.08% dist. However that includes 15X more gm that is mis-matched, not quite sure how that computes without some serious math.

I think I estimated 50 Watts output (with matched gains), depends on how far the 105 mA peak rating was "pushed", 38 Watts seems reasonable enough. Mis-matched gains should give less power output. Maybe something like the sim'd 1.5 Watts X 15 = 22.5 Watts out. That would be a pretty heavy "hit" for not matching the bank gains. Or significantly hotter tubes if pushed to the same 38 watts out. The tube curves for 0 V grid1 limit how far one can go, unless pushed into positive grid territory.
 
Last edited:
Disabled Account
Joined 2013
"Seems odd that more of them would perform worse?"
More tubes more Gm (you notice that the output stage gain of 14 pairs is higher than 1 pair?), higher output voltage, totem-pole is sensitive to load and output level, higher voltage required additional gain drive for the top bank and lesser gain drive for bottom bank with equal Gm added. In the Tim Mellow's OTL, it has 100% NFB bootstraped for lower bank and 100% PFB for upper bank (via LTP), it's least sensitive to load change and tube numbers.


“Looks like more than 0.9%. 30V + peaks versus 24V - peaks. ”
Not you should read the plot V(out) not V(top).

If you have AC balance, you automatically get DC balance provided the PSU is equally rigid, and tube equally matched for AC and DC, but the driver can compensate somewhat for output tubes.
 
Last edited:
WOW.
I never expected so much insightful knowledge to be generated from a couple of photos and a story!
So much info and so many ideas - making my head spin. I love it!
This makes me want to pull theses amps off the shelf and try a few things out. But not right now. I have a couple of things to finish up and get out of the way.

Thanks for all the great ideas. I have no where near the knowledge and experience that some of you guys have so, I’ll have a ton question for the group once I get them on the bench.

Thanks,
Joe Y
 
A lot of comment writings from you guys, how much a SIM for a change :)

For a pair of 12BH4A, the amp is quite well behaved but performed very badly with 14 pairs: a huge DC offset and much lower power output (<20W), high distortion than expected (what do think is the actual power out?), thank you for sharing.


That model is AWESOME! I wish I better understood it but very cool that you took the time to build it.

Is your reference to "a pair of 12BH4A" a typo? Tubes are 12B4A.
Also just curious about this: why 14 pair instead of 15? Just wondering.

As for the power out, in the lower, right corner of the gain chart (see post #22) there is a note: MAX OUTPUT (1Kc) - 35.
Seems low for all the hardware involved, but it may be so.

Thanks,
Joe Y
 
Kasanay,
Thanks for sharing this piece of OTL audio history. The music for millionaires article is fascinating. When your time is right you must get these things restored and working as originally designed and then report your impressions of the sound. As an OTL enthusiast, I have learned a lot from this thread and appreciate the insights provided by all.
 
Disabled Account
Joined 2013
“Is your reference to "a pair of 12BH4A" a typo? Tubes are 12B4A.”
Yes, it's a typo, sorry.

"Also just curious about this: why 14 pair instead of 15? Just wondering."
Yes I missed count: V3-V17=15 not 14, I will edit for 15 pairs and AC Bal is moded again so it'll use NFB to better control the gain of top and bottom bank, after this mode the output has increased to 63W (14 pairs), I post asap I edited it.
 
Disabled Account
Joined 2013
Here is the edited version, now 15 pairs, 70W. Included also the current waveform for top and bottom bank, shows a peak of 2.1A, or 140mA each tube (look like it has exceed the max current of 105mA), if momentary should be ok.

Athough this is not my own project, but doing it helps a lot to improve my own OTL amp, so worth the effort.
 

Attachments

  • 12B4 OTL-641.asc
    21.8 KB · Views: 78
  • 12b4a OTL-5.png
    12b4a OTL-5.png
    49.4 KB · Views: 247
  • 12b4a OTL-4.png
    12b4a OTL-4.png
    80.8 KB · Views: 255
Very nice! Staying within the 105 mA tube limit should then give 70W x (105/140)^2 = 40 Watts. Apparently the amp is rated 35 Watts (without the balancing fix in place).

Did you notice the distortion level at 40 Watts output?

Another interesting question would be how -inverted- Futterman drive compares.
(positive Fdbk to the top bank instead of neg. Fdbk to the lower bank) I would guess they come out fairly similar for the same total tube gain available.

Oh, and the output Z's too.
 
Last edited:
Here is the edited version, now 15 pairs, 70W. Included also the current waveform for top and bottom bank, shows a peak of 2.1A, or 140mA each tube (look like it has exceed the max current of 105mA), if momentary should be ok.

Athough this is not my own project, but doing it helps a lot to improve my own OTL amp, so worth the effort.

Very Cool!
I think I spotted one more error;
As PRR pointed out in an earlier post, R41 is not very well identified in the original schematic, however, it is shown in the parts list as being 1.8K. You show this resistor as R14 with a value of 8K.

Might make a difference.

Thanks,
Joe Y
 
Disabled Account
Joined 2013
Very nice! Staying within the 105 mA tube limit should then give 70W x (105/140)^2 = 40 Watts. Apparently the amp is rated 35 Watts (without the balancing fix in place).

Did you notice the distortion level at 40 Watts output?

Another interesting question would be how -inverted- Futterman drive compares.
(positive Fdbk to the top bank instead of neg. Fdbk to the lower bank) I would guess they come out fairly similar for the same total tube gain available.

Oh, and the output Z's too.

THD@40W is about 0.17%

I'll make the comparision asap, no much difference so long the amount of NFB used does it job (optimized) and not overdo it (can it?)
Here is the output Z:-

Output Z: 15 pairs

The input resistor 1Meg is shorted to aviod interference
1) Output stage only 17.7 Ohms
2) All stages in but no NFB 8 Ohms
3) Connect GNFB but remove NFB from phase splitter 0.4 Ohms
4) Connect GNFB, NFB to phase spiltter 0.35 Ohms

Very Cool! <snip>

In schematic it's 8k but in part list it's 1.8k.
I try to use 1.8k but the sensitive is way down, at 8v in, still not full power out and clipping set in. So it's more likely to be 8k, the bias is about 1V, the sensitive is 1.9v to full power.

Output Z, input short, sig gen removed, GNFB removed, only NFB to phase splitter:

MP/Inv-Futterman 14.1/5.25 Ohms

Edit: Original MP has no NFB in phase splitter, output Z is 8 Ohms, while for inv-Futterman whether NFB connected or not doesn't make much differences. (5+ Ohms), the raise of Z means there is an increased in overall gain, and if GNFB is finally applied the overall Z would be lower than inv-Futterman. You think so also?
 
Here is the edited version, now 15 pairs, 70W. Included also the current waveform for top and bottom bank, shows a peak of 2.1A, or 140mA each tube (look like it has exceed the max current of 105mA), if momentary should be ok.


Koonw,
I am now in the very early stages of learning LT Spice.
I loaded your file and it looks like a couple of pieces are missing - generating this error:

Couldn't find symbol(s):
pentode3
potentiometer


Can you help me out with this?

Thanks,
 
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.