White Cathode Follower sound

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
... Jensen's application schematic uses a 12BH7 in its "totem-pole" output driver....

Yes that's a very good, well designed White follower.

It has the correct 1/Gm for the top anode resistor.
Frankly, I think the folks who build LA-2A copies would do well to use that Jensen scheme over the incorrectly balanced White follower in the original Teletronix circuit.

Is that your design Bill?
 
Yes, thank you. Over the years I owned Jensen (1989-2014), we had lots of customers who used that circuit and loved the results. The 12BH7, originally designed for TV vertical deflection yoke driver, is very well suited for the job - it's plate current and dissipation ratings are substantially higher than the small-signal twin triodes like the 12AU .. AT ... and AX7. The most frequent question about the circuit was "Can I use a smaller coupling capacitor to drive the transformer?" Don't - smaller caps will raise the resonant frequency of the transformer primary and coupling cap sufficiently that a "bump" will appear in the LF response - which will be worst when the output load is very light (over about 1.5 kΩ).
 
Well, again, in my opinion it's a very good one 👍

Yep those 12BH7's are beefy little chaps and well suited for the job.

My guess on why folks ask about a smaller cap is that they're wanting to put some dielectric other than an electrolytic in there based on, fashion? preconceived notions or... ? A 22uF poly cap at the voltage needed is big and/or expensive hence.. Just a guess.


I imagine one of the reasons your 10K61 output transformer works really well here is the combination of the low driving source impedance from the totem pole, along with that 22uF cap. The resonance in that whole RCL loop now being able to be parked well below audio.

Neat trick on the input pentode too with sourcing extra current via the 30K for raising the input overload capability (I think?) by the way.
Edit: Actually, I'd not looked at it in a while, but your negative feedback scheme is really quite a bit more sophisticated than first meets the eye!

The 30K looks to be ensuring the correct 2.5V cathode bias on the pentode.
But I'll need to do a couple of sums to see exactly what (how) you did (things) regarding the gain. 30K and 3K3 is only 20dB, but your amp has... 28dB (less the 10K61 transformer's -14dB and a bit) for, 14.5dB total in to out so...

Anyway, nice circuit all round :)
 
Last edited:
I've been looking at other designs, and I've double-checked the pics I have of the ML-1, with it's 809 ohm anode resistor, on an ecc88 or similar tube.

What advantages, overall, could there be in making the anode resistor a bigger value, than the ideal?

Maybe current draw?
 
What advantages, overall, could there be in making the anode resistor a bigger value, than the ideal?

None. That's why it's ideal.


Edit: It occurred to me that there may be a communication error on my part or misunderstanding regarding your question.

I don't know, but let's assume the anode resistor in the Mastering Labs pre at 809R has been chosen well, and it's correct for that tube at that operating point.

But you've looked at the Jensen which has a value of 680R which is lower.
Is your question along the lines of: Why is the Mastering Labs value higher?

If that is what you're curious about, then it's because you choose the value based on the Gm of the particular tube type you're using, at that operating point you've chosen.
Once you find that out (Gm) by looking at the tube data sheets and having drawn your load line etc., then the value of the resistor would simply be 1/Gm
It doesn't have to be absolutely exact. Closest thing in standard E24 resistor values is OK. If you decide that's not what you want, then you pick another operating point, find the Gm, and try again.

If you knew all this, then, sorry, just thought I'd make sure.

Ciao :)
 
Last edited:
Thanks Bill. Your emphasis on vacuum tubes having wide variations is absolutely something to keep in mind when plotting things out. . It'll drive you crazy if you go searching for absolutes.
They're very tolerant of variations in B+/component values within a reasonable window of variance.

The main thing with a White/totem-pole follower is, we want both parts of the topology to be contributing equally to the available drive current, and that, when they are balanced, the output impedance will be half that of a standard follower, which can only mean good things for our output transformer.

I see no advantage, for instance, with the White/totem-pole in the old Teletronix LA-2A circuit with its top 12BH7 anode resistor of a value that's some 2-3 times greater than an optimum value. Getting that value even 50% closer to ideal is better than doing nothing.
 
If I were to design from scratch I would certainly go for ideal as described above.

That doesn’t explain why the LA2A or ML-1 have higher values.

Surely they must have been thinking of some kind of trade-off?
 
Housework: comp-ed 2 posts, erased another that was irrelevant

Max,
Back when the LA-2A was designed, I doubt that the knowledge of how to optimize a White follower was known, or at the least wasn't widely known or common knowledge. James Lawrence of Teletronix designed the LA-2A and may simply have been unaware of the finer details of White's topology.
However, I don't even think White's original patent talks about this detail, I could be wrong as I haven't read it in decades.

On the ML-1,


doesn’t explain why the LA2A or ML-1 have higher values.

Higher value than what?
Are you sure the Mastering Labs White follower isn't optimized? In other words, have you looked at the component values and operating point used in that output, drawn a corresponding load line using the date sheet curves for that valve, and then determined the transconductance?

To clear up the point, maybe post on here what you came up with so we can see how much higher the ML-1 scheme's value is compared to where you think it should be.
 
Last edited:
Tracing some pics of an ML-1 PCB, I found they used 809 ohms on the anode, with cathode joined to the bottom anode, and the bottom cathode to what appears to be ground, via a jumper.

My calculation was based on an article from Pearl Hi-fi dot com, and it was something like 130 ohms or something, but nowhere near the 809 ohms. If mutual conductance is someting around 12 mA/V then the ideal anode resistor should be around 80R?

I assumed that B+ was 300V and the load 1KOhm
 
Tracing some pics of an ML-1 PCB, I found they used 809 ohms on the anode, with cathode joined to the bottom anode, and the bottom cathode to what appears to be ground, via a jumper.
That suggests using tubes with Gm about 1.2 mho (typical of a weak 12AX7), which somehow defeats the purpose of a WCF.
Some mfgrs would use 12AX7 everywhere, in order to simplify inventory/purchasing, but this is not in accordance with the image Mastering Lab tended to project. And anyway the output tube is 2 sections of E188CC in parallels, for a Gm of 25 mho, which suggests the use of a 40 ohm upper anode resistor.
There seems to be a big incoherence here.
 
So the Mastering Labs is 2 X E188CC in parallel?

For one section, the 12.5mA/V is only true for 90V anode at 15mA of course.
If the supply is higher, such as the 300V as MaxDM is assuming above, then 10mA would be a safer and more typical anode current - with 150V anode to cathode per top and bottom sections.
Transconductance would be lower then but, without more circuit details, it's a guessing game really.

With 2 sections in parallel, maybe the current per is low?

I don't see how the bottom cathode can be tied/jumpered to ground.

If this were supposed to drive high impedance loads only (not likely), then the load has to be taken in account (and that top resistor would end up a different value (not drastic if the load isn't significantly higher than ra). Then the 1/Gm becomes: Ra = (rp + 2RL)/mu

A quote from PRR:
John Broskie and Alex Cavalli have thrashed the WCF far better than I could.
...
1/gm is good for low-Z loads. High-Z loads should correct for load impedance. This happens when load is above half of plate resistance. So 6SN7/12AU7 driving 600r even 2.4k, correction is small; put a 6080 in there and you should work it out.

Broskie also notes "Plate resistance and transconductance vary."

(For more fun: if you push a WCF to MAXimum output, gm varies from near zero (say 0.5X) to about 1.4X the idle value...)

My tedious sims only show that a WCF's THD will null, only if *exactly* balanced. In practice this suggests trimming two +/-20% tubes to equal current and 2% equality of Vpk. Since real tubes are not even constant one minute to the next, you get a very low distortion residual which changes minute to minute. I wonder(!!) if that could be more distressing than a higher THD which stays about the same all the time.

Overload of a tube WCF (assuming near-correct Ra) is almost always about not running enough idle current.
 
Last edited:
Is there a complete schematic anywhere of the Mastering Labs pre? Or is it one that we're not supposed to see?

Edit: How old is the ML? I think Broskie might have been the first to "go public" with the "Optimal White.." and I think that was around 1999.
I certainly had no idea about optimizing those things before then.
I've seen a J-Fet White follower from Erno Borbely circa 1995 that isn't optimized. No one would say Erno doesn't know his stuff, and yet...
 
Last edited:
So the Mastering Labs is 2 X E188CC in parallel?
Seemingly.
https://groupdiy.com/threads/gear-porn-inside-ml-1-tube-pre.37180/
For one section, the 12.5mA/V is only true for 90V anode at 15mA of course.
If the supply is higher, such as the 300V as MaxDM is assuming above, then 10mA would be a safer and more typical anode current - with 150V anode to cathode per top and bottom sections.
Whatever, even with 10mho, the 809 res is 700% out of whack.
I don't see how the bottom cathode can be tied/jumpered to ground.
Fixed bias? Grid current bias? Mmmh... Schematic error more likely. Still a mystery.
 
Thanks for the pictures link Abbey.

Seems to be two tubes in total, a 5751 and an E188CC.

Whatever, even with 10mho, the 809 res is 700% out of whack.
Oh yeah, I wasn't suggesting the transconductance would come down anywhere near to validating an 809 Ra.

Looking online it seems these are from the '90's so, quite possible the reality is that it's not optimal.
Who knows?


Fixed bias? Grid current bias? Mmmh... Schematic error more likely. Still a mystery.

Gotta be, otherwise there wouldn't be enough current to drive a paper airoplane.
 
I know the E188CC is used in my Studer C37 extensively , its a saucy price to replace them , luckily they have a 10,000 hour lifespan like many special quality tubes . I'm wondering is it just rarity that has pushed the price of these through the roof or do they have special advantages over other more moderately priced high gain high gm double triodes ?
 
Seemingly.
https://groupdiy.com/threads/gear-porn-inside-ml-1-tube-pre.37180/
Whatever, even with 10mho, the 809 res is 700% out of whack.

Fixed bias? Grid current bias? Mmmh... Schematic error more likely. Still a mystery.

Thanks for the pictures link Abbey.

Seems to be two tubes in total, a 5751 and an E188CC.


Oh yeah, I wasn't suggesting the transconductance would come down anywhere near to validating an 809 Ra.

Looking online it seems these are from the '90's so, quite possible the reality is that it's not optimal.
Who knows?




Gotta be, otherwise there wouldn't be enough current to drive a paper airoplane.
from these two photos, the resistor seems to be 809 (R19), and the cathode seems to be connected to ground, via a jumper (J1)

I am assuming the tube is an e188cc
 

Attachments

  • ML1 a.png
    ML1 a.png
    782.9 KB · Views: 30
  • ml1e.png
    ml1e.png
    1.1 MB · Views: 30
Tracing some pics of an ML-1 PCB, I found they used 809 ohms on the anode, with cathode joined to the bottom anode, and the bottom cathode to what appears to be ground, via a jumper.
No. The jumper is to pin 9, which is the isolation screen. The cathode at pin 8 goes to the power resistor mounted vertically. I can't read its value.
Indeed the resistor appears to be 809r. Could be an intern with approximate perception of colour code...
 
No. The jumper is to pin 9, which is the isolation screen. The cathode at pin 8 goes to the power resistor mounted vertically. I can't read its value.
Indeed the resistor appears to be 809r. Could be an intern with approximate perception of colour code...

Oh yes! How did I miss that... must have been late in the evening.

Great! So, the biasing of the bottom tube is not such a mystery anymore, apart from the resistor value...
 
No. The jumper is to pin 9, which is the isolation screen. The cathode at pin 8 goes to the power resistor mounted vertically. I can't read its value.
Indeed the resistor appears to be 809r. Could be an intern with approximate perception of colour code...

🤓

I suppose one way to look at the consequences of a wrong top resistor value might be - even if the voltage across that naked, output transformerless E88CC totem-pole were only a conservative 200V - maximum output voltage swing would still be about 12dB greater than would ever be necessary.

Run the thing at 15mA standing current and a lot of those unnecessary consequences from the sub-optimal value get swept under the carpet to some extent.

I've seen worse mistakes...
 
Last edited:
Back
Top