Help with PSU for a Single Ended HoSo build

[ncusack]

Ok so im looking to mate the HoSo preamp to a single ended 6v6. So the designe would be 5879 into 12ax7 gain/CF stage driving a basic volume/tone stack into the 6V6. What I need help with is the resistor values I should use in the B+ rail. Here is the schematic with estimated current values pulled from the various HoSo schematics found on the forum.



I've fooled around with a few values using the Duncan PSU and with my estimated currents i could use 1k - 27k - 1k to get 370V - 363V - 248V - 246V from A-D. With these voltages i'd be pretty close to the numbers on the HoSo56 front end with a few minor tweaks to the resistor values.

Let me know if this sounds good or even close or if I even have my current values correct.

Cheers,

Neill

[RicharD]

Rab = (Va-Vb)/.0072

Rbc = (Vb-Vc)/.0042

Rcd = (Vc-Vd)/.0018

[jojokeo]

I'd like to take a stab at this as an excercise to see if I'm on track or not using ncusack's tranny and mA estimates. I will use resistor values btwn A-B = 470r, B-C = 10k, C-D = 10k
Tranny = 285v x 1.3 (for 5ar4) = 370v before the choke. Node A  would = 363v, Node B would = 361, Node C = 319, and Node D = 301
Using his load resistor values: the 5879 Vp = 178v, the cf plates would = 199v

[OldHouseScott]

I'll take a stab too. I think I would want to see about 250 at D, 300 at C (just to give a bit more headroom), and 360 at B. Using Richard's formulae, I get 1k, 15k, and 27k approximately.

[RicharD]

Duncan PSU Designer II is a cool tool.  It does have a bit of a learning curve, but if I can finger it out, anyone can.

http://www.duncanamps.com/psud2/index.html

Shoot.... I missed that you posted voltages below the schematic.  I say your resistor values are correct but, you have a time constant issue because of that big drop in the middle.  A lot of people don't worry about this, but proper is proper.  In a cascading power supply, you want your knee frequencies to start sub-sonic and descend as the voltages descend.  Using the equation 1/2piRC, I get 4.8Hz at point B, 0.27Hz at point C, and 7.2Hz at point D.  The way to correct this is tap point D from point B instead of point C.  This of course changes the resistor values for both Rbc and Rbd, (formerly Rcd).  I came up around Rbc = 47k and Rbd = 68k.  All this is assuming your currents are correct.  Looks close enough.  Try it, measure it, hack it out if it's wrong.

[moonbird]

Aha -- so that is why that is done -- very interesting. Which switch in PSUII do you turn on to see those freq values??

This PSU branching also makes a PA only VVR possible doesn't it? thx.

[Fresh_Start]

Duncan PSU Designer II is a cool tool.  It does have a bit of a learning curve, but if I can finger it out, anyone can.

http://www.duncanamps.com/psud2/index.html

Shoot.... I missed that you posted voltages below the schematic.  I say your resistor values are correct but, you have a time constant issue because of that big drop in the middle.  A lot of people don't worry about this, but proper is proper.  In a cascading power supply, you want your knee frequencies to start sub-sonic and descend as the voltages descend.  Using the equation 1/2piRC, I get 4.8Hz at point B, 0.27Hz at point C, and 7.2Hz at point D.  The way to correct this is tap point D from point B instead of point C.  This of course changes the resistor values for both Rbc and Rbd, (formerly Rcd).  I came up around Rbc = 47k and Rbd = 68k.  All this is assuming your currents are correct.  Looks close enough.  Try it, measure it, hack it out if it's wrong.

Did Leo Fender follow this descending "knee frequencies" approach?  I'm not trying to be difficult - just trying to understand.

Cheers,

Chip

[RicharD]

Unfortunately PSUD II does not calculate frequencies for you.  It can't do a split rail or differential power supply either but for free, who's complaining.  1/2piRC is not that difficult of an equation.  If I'm doing more than 1, XL is a double click away.  My flavorite feature is it's ability to estimate an unknown transformer based on unloaded voltage and the DC resistance of the primary and secondary.  I've used this several times and it's right as rain.

Again, a lot of people would tell you not to worry about frequencies in a DC power supply.  Rules are made to be broken when designing a geetar amp, but I tend to draw a line when it comes to power supplies.  It can be a very fine line when trying to get a saggy on the edge power supply, but a weak PSU is begging for trouble.  

[RicharD]

>Did Leo Fender follow this descending "knee frequencies" approach?

I just glanced at 2 Fender Circuits.  The Super Reverb AB763 is a yes, 8Hz into 1.7Hz, and the Bassman 5F6a is an almost 1.7Hz into 1.9Hz.   

[RicharD]

As eye understand it, the critical thang is to make sure your time constants are sufficiently sub-sonic so that audio signal does not pass erroneously through the power supply.  Using cookie cutter values we see all the time pretty much ensures this. 

[jojokeo]

I'll take a stab too. I think I would want to see about 250 at D, 300 at C (just to give a bit more headroom), and 360 at B. Using Richard's formulae, I get 1k, 15k, and 27k approximately.

It seems like you're voltages you're talking about are node voltages not plate voltages?

[ncusack]

jojo thats actually another aspect of this build i'd like some help with. I know that judging from other post about the 5879 that the happy plate voltage should be around 140-160V and like 50-70V on the screen. I can designe around that but what are peoples thoughts on the CF voltages?

[jojokeo]

As long as you play w/in a range of voltages and put efforts into component values, biasing and other items, you'll be fine. There's always final adjustments if you're lucky or complete re-do's if you're not. tubenit's a great example though his are more a matter of choice due to his personal tone he's listening and searching for.

[PRR]

> In a cascading power supply, you want your knee frequencies to start sub-sonic and descend as the voltages descend.

That's a side-effect.

Power-plates node. Next node has power screens plus three small stages. Next node has three small stages. Next node has just one small stage.

Voltage drop might be pretty much the same from each node to the next. (Neill's 2V from C to D violates this.) 

Current is much less each node down.

Therefore the resistors will get bigger as you go along.

10-pack of 22uFd caps is cheaper than ala-carte 33u 22u 6.8u assortment. Also the first stage is most buzz-sensitive, you don't mind it being "a bit large".

So the frequencies will -tend- to go down as you go along.

> make sure your time constants are sufficiently sub-sonic so that audio signal does not pass erroneously through the power supply

Right; though in short-bass work you might also consider buzz signal. (Cheap practice amp with 300Hz low end, you might still want to know how much 120Hz will arrive at the first stage.)

Some rules-o-thumb:

For cheapest motoboat and buzz rejection, let resistor drop 30% of previous node voltage. BUT when overload is an issue (often true in guitar-amps), 3 nodes of 30% drop is a BIG drop and will leave your preamp below 150V, maybe too low.

For decent filtering with modern inexpensive caps, allow at least 10% drop each node. That will leave 290V at the first node.

> to get 370V - 363V - 248V - 246V from A-D

Is there some specific reason for these voltages? 370-363 is a very-small drop, though this is Screens so may be valid. 248-246 is a silly-small drop, resistor will be too small to filter audio or buzz unless the cap is VERY large.

You seem to have 22u caps. Stage gains run about 50 and low-cuts around 50Hz so you would be thinking 1Hz R-C product in rail filter. The R should be at least 7.5K.

V1 pulls 2mA. 1K gives 2V drop, 7.5K gives 15V drop. Is there any reason V1 can't run on 233V? Or the 27K to V2V2 dropped to 15K or 22K, then 10K to V1?

In general we stock 5K 10K and 22K resistors, 10u 20u and 40u caps, and throw them around until happy.

BTW, 1.5K is an unhappy value for that second stage. The cathode-follower will distort severely. Expect to reduce that. 5F6A used 820 here.

[spacelabstudio]

BTW, 1.5K is an unhappy value for that second stage. The cathode-follower will distort severely. Expect to reduce that. 5F6A used 820 here.

Interesting.  Is there a theoretical explanation for that or just something observed empirically? 

[Geezer]

BTW, 1.5K is an unhappy value for that second stage. The cathode-follower will distort severely. Expect to reduce that. 5F6A used 820 here.

Interesting.  Is there a theoretical explanation for that or just something observed empirically?  

I had the same basic question........that seems counter-intuitive to the "normal" way of triode operation.
We would "normally" (stand alone triode) increase the Rk value to reduce gain/distortion....is there a different process at work here because the triode in question is feeding a CF?

BTW, the OP's values are from the "HoSo" preamp, and 1.5k is what I have in all of my versions of that amp, which all "sound" very good to my ears (the best amps I've ever built or played, in fact.) But if lowering that value can possibly show even more improvement, I'm all for giving it a listen. Easy enough to clip another 1.5k (or 1.8k to get closer to the 820R) in parallel w/ the existing 1.5 Rk.

G

[Fresh_Start]

Very interesting question!  Looking through the 5F6-A schematic along with a bunch of Marshalls, I focused for the first time on the fact that the 820 ohm cathode resistor on the triode preceding the cathode follower is NOT bypassed by a cap.  There is at least one Plexi circuit that does have a cathode bypass cap, but that clearly is the exception which proves the rule.  The AX84 preamps almost all use a common cathode gain stage tied to a cathode follower with an un-bypassed 820 ohm cathode resistor.

Unless I'm mistaken, an un-bypassed 100K/820r 12AX7 triode would have less gain than a bypassed 100K/1.5K triode.  Is that correct?  If so, you all might want to try either (a) bigger cathode resistors with a bypass cap, or (b) smaller cathode resistors without a bypass cap.

How could we calculate how big a bypassed cathode resistor would need to be to get voltage gain equivalent to an un-bypassed 820 ohm cathode resistor?

Another thought occurs to me.  With the un-bypassed 820r cathode triode driving the cathode follower, whatever compression the cathode follower produces applies equally across the frequency spectrum.  IOW putting a bypass cap on that triode's cathode would apply the CF compression more to frequencies above the knee of the cathode resistor/cap network than to frequencies below that knee.

Does any of that make sense or do I need more coffee?  Ok, I know I need more coffee...

Cheers,

Chip

[DummyLoad]

Interesting.  Is there a theoretical explanation for that or just something observed empirically? 

yes theoretical - nodal analysis.

1) the bias point (set by Rk,Ra and B+) of a grounded cathode (GC) gain stage sets Va of the stage at some value < B+. in the absence of Rk it is set by the values of Ra and -Vg and B+.

2) the values of Vg and Rk of a CF set the bias point of the CF at a given B+.

you are direct coupling the gain stage and the CF. so then we could say now that the value of Va of the gain stage is Vg of the CF stage. if that value is too low or high, the cathode follower will asymmetrically clip either positive or negative parts of the waveform depending on the bias condition.

3) as a rule of thumb, for maximum headroom, the value of Vg of the CF should equal ~1/2 of applied B+; center biased. - in other words; both triodes should should split B+ approx. equal since we are direct coupled.

if i were to use 1K5=Rk and 100K=Ra w/ 300V in the GC gain stage, Va is not 1/2 B+. with those values Va will be ~200V.
in your plan, with 1K5=Rk and 100K=Ra w/ 250V used in the gain stage, Va ~160V. not 1/2 B+.

with 300V for B+ i come up with 680R/100K for GC gain stage with Rk=100k for CF. each stage will bias @ ~ 1.5mA; 3mA total for the node C.

there are may other ways to make you plan work if you MUST have a 1K5 under the GC gain stage. split the B+ of the GC gain stage and CF, you could then use the 1K5 Rk for the GC gain stage. dissimilar Rk for CF and Ra for GC, even dissimilar triode types.

who knows... maybe you want asymmetrical clipping and gross distortion; is it part of tone that you seek?

back to the topic - i'd set node C to 300V, so then for the PS resistors:

node C=300V so 60V/4.8mA = 12K5 @ 300mW 
node D=250V so 50V/1.8mA = 27K7- @ 90mW

Rbc  ~12K @ 1W
Rcd  ~27K @ 500mW
Cbc 10uF @ 350V
Ccd 10uF @ 350V

respectfully,

--DL

PS inclusion or omission of a bypass cap is not going to change the bias point of a GC gain stage. it only changes gain - e.g. the usable sensitivity of the the stage before the onset of clipping.

[ncusack]

Wow talk about a lot of information to take in at once. This is why I love this forum.

So first off I think I'll address the PSU voltages by saying there is no set in stone values here my numbers were just ones I aimed for because I could get matching plate values using the same resistor values in the original HoSo56 preamp. The voltages for the 6V6 plate and screens are open to whatever will work and same goes for the CF stage. The only voltages that are even slightly set to anything specific are the V1 5879 voltages. I know from what i've read elsewhere on the forum that the 5879 like around 145V on the plate and around 55V on its screen. Aside from that tidbit of information the rest of the amp is fair game for changes.

What I would like to achieve with this one is a usable bluesy clean to buttery overdrive. I know on some other renditions of the HoSo56 the bypass cap on the CF gain stage was switched in and out as a boost switch so that is still a viable option. 

I don't have an outfitted shop so ill need to put in an order for the components for this build so im looking for a decent starting point and ill grab values a little higher and lower of that starting point. I know when it comes to doing something new it's not going to be a first time success it will be a matter of tweaking until it sounds right.

Im still processing all the PSU theory posted above but im starting to get a feel for it.

Thanks again and keep the advice coming

[RicharD]

Your PSU resistor values are certainly close enough to work.  Resistors are cheap so buy a few extras in adjacent values.  Simply tack in components in question or pick up a bag of clip leads if you're comfortable with that.  Once you're happy, make it tight.  The abortion I have going right now has seen an inordinate amount of parts swapping. 

[ncusack]

Ok so how do these values sound? Rab = 5k, Rbc = 10k and Rcd = 30k. With my estimated currents im getting 370V @ A,333V @ B,290V @ C and 234V @ D. I can adjust the plate resistor values to dial in the required plate voltages. Do these values look like they will manage any motorboating or buzz?

[jojokeo]

You'll be fine and can always adjust later. You won't get any of the issues you're worried over. Your main concern is the voltage to the tubes for the tone you want. But you need to build it and then think about tweaking after you're playing it.