When I saw 70v line transformer, I immediately thought of parafeed. Then I found this: http://www.bartola.co.uk/valves/2016/12/04/headphone-parafeed-amp-part-i/
jeff
You knew there had to be a fly in the ointment...
After almost a week of playing this amp, I'm finding it difficult to get two 12HL7s with gain matched well enough to result in good (enough for me) channel balance. The mismatch in levels is most apparent in the bass.
I've noticed three different types of 12HL7s:
GE - grey plate; heater concealed by cathode; getter is on top
Raytheon, Philips ECG - shiny, smooth black plate; heater ends stick out of cathode; getter is on top
RCA 'clear top' - getter on one side
I have one GE. It's the one that plays the loudest.
The Raytheon plays the least loud.
The Philips ECG is in the middle.
I haven't tried the RCA tubes. I have a sleeve of those.
I'm hoping to find at least one good match out of my collection.
I figure this is a good argument for using NFB in a headphone amp. Either that, or a Balance control is mandatory. Any gain mismatch between channels is impossible to hide in headphones. No room acoustics to mask that.
--
After almost a week of playing this amp, I'm finding it difficult to get two 12HL7s with gain matched well enough to result in good (enough for me) channel balance. The mismatch in levels is most apparent in the bass.
I've noticed three different types of 12HL7s:
GE - grey plate; heater concealed by cathode; getter is on top
Raytheon, Philips ECG - shiny, smooth black plate; heater ends stick out of cathode; getter is on top
RCA 'clear top' - getter on one side
I have one GE. It's the one that plays the loudest.
The Raytheon plays the least loud.
The Philips ECG is in the middle.
I haven't tried the RCA tubes. I have a sleeve of those.
I'm hoping to find at least one good match out of my collection.
I figure this is a good argument for using NFB in a headphone amp. Either that, or a Balance control is mandatory. Any gain mismatch between channels is impossible to hide in headphones. No room acoustics to mask that.
--
famousmockingbird said:
I didn't have the opportunity to try those. They're expensive these days. Everybody who has a pair seems to like them a lot. They must be very good.
I tried a pair of ca. 2003 Klipsch RF-3, which are really pretty darn good speakers for the very little money I paid. But they had a zip to the upper mids that eventually ruined it for me, and a bit of a 'shout' to them. I sold them to a friend at work, who loves them.
Before that, I had a pair of KG4.5, which I should have kept, despite their somewhat lumpy midrange. They were great for small group jazz and especially electric blues. They loved the sound of electric guitars through tube amps. They didn't do so well with massed strings and had real problems with deep male voices (e.g., Joe Williams). Not very versatile speakers. So they're gone now.
I auditioned a pair of KLF20, and had to run from the room. They were so shrill.
I briefly tried a pair of old Heresy, but couldn't deal with the lack of bass. Otherwise I liked those. Not cheap, either.
Then I found a pair of Snell C for cheap, and they were a great match for my push-pull 2A3 amp. So I left it there. Somehow over time the 2A3 lost the ability to pass much bass. I have to open that up and fix it. I'm thinking the LTP needs its CCS back. I was experimenting with input transformers for phase splitting, which I think was not an improvement...
But now that's waaaaaaaaayyyy off-topic.
--
Thinking about that a bit, and then this:
So the imbalance is really only noticeable in the bass. However, it depends on which tube I install for that channel. In other words, it's not caused by one OPT having lower Lpri than the other, or something like that.
- Since low rp is necessary for good bass response from these OPTs (because of their rather paltry Lpri), and...
- Since gm of the tube is directly proportional to its rp; mu is fairly consistent between samples but gm can vary by quite a bit, and...
- If 12HL7 has just barely low enough rp to allow for decent bass response with these little Edcor OPTs (8H measured Lpri), then:
I wonder if in this amp, samples of 12HL7 with lower gm sound 'weak in the bass' but fine as frequency goes well above 150Hz? 12HL7 tubes with higher gm sound 'stronger in the bass'? Could be.
______________________
My rickety tube tester doesn't have settings for 12HL7. I'll need to see if it has one for 12HG7 or 12GN7. I'll take a look when I get home. If I can match pairs of 12HL7 on the tester, then maybe I can see if bass response from this amp correlates with tube tester readings.
Looking at the GE data sheet for 12HL7 (https://frank.pocnet.net/sheets/135/1/12HL7.pdf) I see that in pentode, with Vp = 250V and Vg2 = 150V, and with Ip+g2 at only 31mA, gm = 21mA/V.
Since I'm running my 12HL7s in triode, with Vp+g2 = 150V (kind of low) and Ip+g2 at a whopping 55mA, I figure I'm getting a bit more gm than what's on the data sheet. So let's say I'm getting 23mA/V.
If the 12HL7 mu is 30, and gm is 23mA/V (0.023A), then rp would be 1.3k ohms.
Let's say I have a very weak 12HL7 whose gm is only 15mA/V, but with that same mu of 30. Its rp would be up at 2k ohms.
Using this handy online tool -- (Sample)LR Low-pass Filter Design Tool - Result -
- If the rp of the 12HL7 is 1.3k ohms, and the Lpri of the OPT is 8H, f3 is still pretty high at 26Hz. That would be -1dB around 50Hz, right? I could live with that.
- But if the rp of a particular 12HL7 is up at 2k ohms due to lower than normal gm (a weak tube), and with that same 8H of Lpri, f3 would now be 40Hz, so -1dB way now be up near 80Hz. That could make an audible difference in perceived bass response.
Could be, right?
Another learning experience. (Be careful what you ask for!)
--
The thing is, specs for the Edcor XSE10-8K-50 transformer say the maximum current is 80mA. I'm pulling 55mA through the primary. Is that close enough to reduce the primary inductance?
EDCOR - XSE10-8K
--
EDCOR - XSE10-8K
--
I breadboarded SRPP with triode connected 6P41S and MOSFET on top, driving 100V line transformer. Sweet!
Parafeed could work well. I have some large 600v 45uF polypropylene caps to couple the triode stage to the transformer.
In case anyone is following this build...
I picked through my collection of 12HL7s and found two of identical construction that look like they're made by GE, but one is labeled GE, the other RCA. They have their getters on the top of the bottle, 'powder coated' grey plates and square 'windows' with shiny metal structural plates visible inside. I have these two in the amp and it sounds well balanced between left and right, with matching subjective sound quality. (I like the sound of this type of 12HL7 in this circuit.)
The Raytheon 12HL7 I was using before has a smooth plate of a shiny, dark brown-to-black metal. I found a Philips ECG with matching construction. I tried the two in the amp and they match in 'tone quality' but are a little mismatched in level (channel balance is off). I find the tone quality is somehow 'softer' and less 'present' than that of the powder-coat grey plate GE type. This type doesn't sound 'right' to me. I don't know what to make of that. Perhaps this type has a tendency to draw grid current at grid bias of -2V?
There are two other types of RCA 12HL7s which I haven't tried in the amp yet. One has a 'clear top' bottle with its getter on one side and a black 'carbon' plate material. The other has the getter on the top of the bottle, also has a black carbon plate, but has a large shiny metal plate on one entire side of the tube's construction. It's a pentode construction type I haven't seen before.
I'll try to listen to the clear-top RCAs and report on what those sound like in this amp.
It's interesting that different examples of this tube type 'sound' so different from each other in this circuit. I guess their characteristics vary widely for a tube.
--
I picked through my collection of 12HL7s and found two of identical construction that look like they're made by GE, but one is labeled GE, the other RCA. They have their getters on the top of the bottle, 'powder coated' grey plates and square 'windows' with shiny metal structural plates visible inside. I have these two in the amp and it sounds well balanced between left and right, with matching subjective sound quality. (I like the sound of this type of 12HL7 in this circuit.)
The Raytheon 12HL7 I was using before has a smooth plate of a shiny, dark brown-to-black metal. I found a Philips ECG with matching construction. I tried the two in the amp and they match in 'tone quality' but are a little mismatched in level (channel balance is off). I find the tone quality is somehow 'softer' and less 'present' than that of the powder-coat grey plate GE type. This type doesn't sound 'right' to me. I don't know what to make of that. Perhaps this type has a tendency to draw grid current at grid bias of -2V?
There are two other types of RCA 12HL7s which I haven't tried in the amp yet. One has a 'clear top' bottle with its getter on one side and a black 'carbon' plate material. The other has the getter on the top of the bottle, also has a black carbon plate, but has a large shiny metal plate on one entire side of the tube's construction. It's a pentode construction type I haven't seen before.
I'll try to listen to the clear-top RCAs and report on what those sound like in this amp.
It's interesting that different examples of this tube type 'sound' so different from each other in this circuit. I guess their characteristics vary widely for a tube.
--
Last edited:
What about Miller capacitance?
OK, I got the channel balance thing worked out. Like any SET power amp with no NFB loop applied, you just have to match pairs of tubes. It's a pain, but that's life.
Now for the next question. What about Miller capacitance?
There are no triode specs available for 12HL7, at least that I can find. The data sheet says the following, for pentode:
grid1 to plate = 0.15pF
Input (grid1 to h+k+g2+g3+i.s.) = 15pF
Output (plate to h+k+g2+g3+i.s.) = 6.0pF
_____________________________
To find Cmiller for a triode,
Cg-p + Cg-k * (mu+1)
_____________________________
In the 12HL7 data sheet, the C from grid1 to plate is measured with the 12HL7 in pentode. This means the screen grid shields the grid from the plate, so the capacitance is much lower. Once I've triode-strapped the 12HL7, won't that make the Cg-p go much higher, now that the grid is no longer shielded from the plate by the screen grid? By how much will the capacitance rise?
_____________________________
Looking at a 5842, which is an RF triode with similar gm to the 12HL7 (about 20mA/V), it says its grid-plate C is 1.8pF. Its grid-cathode C is 9.0pF.
So let's say the Cg-p and Cg-k of the 12HL7 in triode are the same as in the 5842.
Cg-p + Cg-k * (mu+1)
1.8pF + 9.0pF = 10.8pF (round up to 11pF)
mu = 30, mu+1 = 31
11pF * 31 = 341pF
_____________________________
Now, if I use a 100k volume pot, which has a maximum output impedance of 25k ohms, I've set up an RC LPF between the pot's Zout and the triode's Cmiller, correct?
RC = Rseries = 25k ohms, Cshunt = 340pF
F3 = 18.7kHz
Really? -3dB at 19kHz? Wow. That sucks.
_____________________________
Did I do the above correctly? Are those the correct capacitances to use for calculating Miller effect on frequency response?
--
OK, I got the channel balance thing worked out. Like any SET power amp with no NFB loop applied, you just have to match pairs of tubes. It's a pain, but that's life.
Now for the next question. What about Miller capacitance?
There are no triode specs available for 12HL7, at least that I can find. The data sheet says the following, for pentode:
grid1 to plate = 0.15pF
Input (grid1 to h+k+g2+g3+i.s.) = 15pF
Output (plate to h+k+g2+g3+i.s.) = 6.0pF
_____________________________
To find Cmiller for a triode,
Cg-p + Cg-k * (mu+1)
_____________________________
In the 12HL7 data sheet, the C from grid1 to plate is measured with the 12HL7 in pentode. This means the screen grid shields the grid from the plate, so the capacitance is much lower. Once I've triode-strapped the 12HL7, won't that make the Cg-p go much higher, now that the grid is no longer shielded from the plate by the screen grid? By how much will the capacitance rise?
_____________________________
Looking at a 5842, which is an RF triode with similar gm to the 12HL7 (about 20mA/V), it says its grid-plate C is 1.8pF. Its grid-cathode C is 9.0pF.
So let's say the Cg-p and Cg-k of the 12HL7 in triode are the same as in the 5842.
Cg-p + Cg-k * (mu+1)
1.8pF + 9.0pF = 10.8pF (round up to 11pF)
mu = 30, mu+1 = 31
11pF * 31 = 341pF
_____________________________
Now, if I use a 100k volume pot, which has a maximum output impedance of 25k ohms, I've set up an RC LPF between the pot's Zout and the triode's Cmiller, correct?
RC = Rseries = 25k ohms, Cshunt = 340pF
F3 = 18.7kHz
Really? -3dB at 19kHz? Wow. That sucks.
_____________________________
Did I do the above correctly? Are those the correct capacitances to use for calculating Miller effect on frequency response?
--
Thanks.
Unfortunately, there is no spec given for g1 to g2 in the 12HL7 data sheet. The closest I can see is g1 to p, but that is so small (0.15pF) that it hardly matters.
For 5842, the Cg-p is 1.8pF. I could add a couple pF to account for strays and get to about 4pF.
4pF(mu) for the 12HL7-triode would be 4*30 = 120pF.
Does that seem more realistic? If so...
If a 100k pot is used:
LPF f3 for Zout = 25k and Cshunt = 120pF = 53kHz. That's pretty low. Approx. -0.1dB at 5kHz. I wonder if that's audible as a slight attenuation of high audio frequencies...
If a 20k pot is used:
LPF f3 for Zout = 5k and Cshunt = 120pF = 265kHz. That's good for -0.1dB at 26.5kHz, so that's just dandy.
Do you think I'd hear the difference?
--
Last edited:
I just measured a GE 12HL7 with a digital LC-200 meter, grid1 to g2 & plate. Using difference between lead touching pin and just not touching pin, I get 6 or 7 pf. (lowest resolvable digit on meter, so average reading)
I also have a C-350 meter that resolves another digit for capacitance, and that gives 6.5 pf for grid1 to grid2 & plate.
Then an Extech LCR meter gives 6.8 pf for the same grid1 to grid2 & plate.
Those are all for a cold tube. Space charge would increase that slightly, can't recall what %.
I just checked with my old Heathkit Digital LC Bridge and I get 6.5 pf from g1 to g2 only. (fancy Kelvin clips will only grip one pin at a time) Couldn't read anything on g1 to plate.
I also have a C-350 meter that resolves another digit for capacitance, and that gives 6.5 pf for grid1 to grid2 & plate.
Then an Extech LCR meter gives 6.8 pf for the same grid1 to grid2 & plate.
Those are all for a cold tube. Space charge would increase that slightly, can't recall what %.
I just checked with my old Heathkit Digital LC Bridge and I get 6.5 pf from g1 to g2 only. (fancy Kelvin clips will only grip one pin at a time) Couldn't read anything on g1 to plate.
Last edited:
Hey smoking-amp, thank you very much for going to that trouble.
Interesting results you got there.
I have one of those cheap handheld digital LCR meters. I'll try measuring a 12HL7 tonight; see if I get something similar.
For now, it looks like an estimate of 7pF would be reasonably accurate for calculating Miller capacitance.
Let's say 7pF(30+1)for a total triode input ("Miller") capacitance of about 220pF, then. Does that sound reasonable?
If so, then a 100k pot with 25k max Zout will result in f3 of 29kHz. That's too low for comfort.
I think I'll be putting something like an 18k resistor from input to ground on the pot for each channel. That should make the pot look like 15k. 1/4 that would mean a max Zout of 3.8k, giving f3 of about 190kHz. That would put the pot's rolloff well above the OPT's rolloff.
--
Interesting results you got there.
That makes sense, since the GE data sheet says 0.15pF. That's basically zero.
I have one of those cheap handheld digital LCR meters. I'll try measuring a 12HL7 tonight; see if I get something similar.
For now, it looks like an estimate of 7pF would be reasonably accurate for calculating Miller capacitance.
Let's say 7pF(30+1)for a total triode input ("Miller") capacitance of about 220pF, then. Does that sound reasonable?
If so, then a 100k pot with 25k max Zout will result in f3 of 29kHz. That's too low for comfort.
I think I'll be putting something like an 18k resistor from input to ground on the pot for each channel. That should make the pot look like 15k. 1/4 that would mean a max Zout of 3.8k, giving f3 of about 190kHz. That would put the pot's rolloff well above the OPT's rolloff.
--
For the two lead LC meters I clip one lead to the grid2 and plate pins (adjacent for 12HL7) and clip the other lead to a tooth-pick, then just touch and un-touch the other clip to the grid1 pin (holding the other end of the tooth-pick). Take the difference in readings. That avoids disturbing the leads or hand capacitance effects.
Last edited:
rongon, the only problem with lowering your input impedance is that you might want to use a source whose output cap would prefer to see 100k. Why not put a couple of transistor followers after the pot instead (or a couple of sub-miniature tubes which don't demand much power, if you don't like semiconductors)
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.