Hi,
This seemed simple enough to me, but the outcome wasn't what I expected, so I'm hoping someone here can tell me what I did wrong.
My amp's driver stage is a CCS'd ECC99 with a NiMH battery under the cathode. I used 2 ECC99 tubes, since I had 2 sockets in my amp. I wired the sockets to use different triode halves (including the filament supply), and was swapping tubes back and forth so that both got even wear on both triode sections. So far, so good.
This saturday I decided to try paralleling the triode halves to get some more gain. I know, increased Miller C will mean slower sound, but I wanted to see for myself. And here's where I did something wrong. I figured that I could share the CCS load and the battery bias across the triode halves. So all I did was short pins - short 4-5 to turn on the filament to both halves, short the cathodes to each other and the plates to each other. I also added 100 ohm grid stopper resistors to the tubes (resistors up against the socket pins) - I've always seen some HF hash on the scope, everyone said the 6C45pi needs grid stoppers but opinions were divided on the ECC99, and I said why not, and now that I'm paralleling things maybe stability becomes more important.
So, those are the changes I made - added grid stoppers on both grids (with their other end going to the grid leak and the input RCA), and shorted 4-5, cathode-cathode, and plate-plate. I didn't change the CCS, so each triode section is now seeing 1/2 the current it was seeing before. In my case, this means 10mA per triode. Not the best operating region for the ECC99, but a region where it should function OK, according to the curves.
This should have given me double the voltage output for the same voltage input, right. Both triode halves are now amplifying the signal. That's 2x voltage gain, so 6dB output gain. 2 channels, so 12dB overall increase in volume if the linestage's volume control is kept at the same level. I lent my SPL meter to a friend for the weekend, but I'm pretty sure I would be able to hear a 12dB increase in volume. But, that didn't happen. I'm pretty sure I didn't get any increase in volume at all.
So, what did I do wrong?
* My basic assumption is wrong, paralleling triode halves won't get me any increase in gain? That doesn't sound right, this is like a PSE output stage, triodes in parallel should have their gain add up, I think.
* The idea is right, but the implementation is wrong? Do both triodes need separate cathode bias batteries? Separate loads? Did that halving of the bias current from 20mA to 10mA/triode cause the gain to be halved too?
* My math is wrong, and something is getting halved that I haven't realized and that's cancelling out the voltage gain?
* Maybe my idea is correct, and I just mis-wired something on both channels?
* Something else?
On a side note, it does seem like the sound softened a little. I'm not sure if this is due to the increased Miller capacitance, or because the grid stoppers took care of some HF oscillation. With my speakers, this softening isn't really unwelcome, but I need to listen some more before I decide if I like it or not.
But, I'm confused about what I did wrong as far as the gain goes. If anyone could shed some light here, that would be most appreciated.
Thanks,
Saurav
This seemed simple enough to me, but the outcome wasn't what I expected, so I'm hoping someone here can tell me what I did wrong.
My amp's driver stage is a CCS'd ECC99 with a NiMH battery under the cathode. I used 2 ECC99 tubes, since I had 2 sockets in my amp. I wired the sockets to use different triode halves (including the filament supply), and was swapping tubes back and forth so that both got even wear on both triode sections. So far, so good.
This saturday I decided to try paralleling the triode halves to get some more gain. I know, increased Miller C will mean slower sound, but I wanted to see for myself. And here's where I did something wrong. I figured that I could share the CCS load and the battery bias across the triode halves. So all I did was short pins - short 4-5 to turn on the filament to both halves, short the cathodes to each other and the plates to each other. I also added 100 ohm grid stopper resistors to the tubes (resistors up against the socket pins) - I've always seen some HF hash on the scope, everyone said the 6C45pi needs grid stoppers but opinions were divided on the ECC99, and I said why not, and now that I'm paralleling things maybe stability becomes more important.
So, those are the changes I made - added grid stoppers on both grids (with their other end going to the grid leak and the input RCA), and shorted 4-5, cathode-cathode, and plate-plate. I didn't change the CCS, so each triode section is now seeing 1/2 the current it was seeing before. In my case, this means 10mA per triode. Not the best operating region for the ECC99, but a region where it should function OK, according to the curves.
This should have given me double the voltage output for the same voltage input, right. Both triode halves are now amplifying the signal. That's 2x voltage gain, so 6dB output gain. 2 channels, so 12dB overall increase in volume if the linestage's volume control is kept at the same level. I lent my SPL meter to a friend for the weekend, but I'm pretty sure I would be able to hear a 12dB increase in volume. But, that didn't happen. I'm pretty sure I didn't get any increase in volume at all.
So, what did I do wrong?
* My basic assumption is wrong, paralleling triode halves won't get me any increase in gain? That doesn't sound right, this is like a PSE output stage, triodes in parallel should have their gain add up, I think.
* The idea is right, but the implementation is wrong? Do both triodes need separate cathode bias batteries? Separate loads? Did that halving of the bias current from 20mA to 10mA/triode cause the gain to be halved too?
* My math is wrong, and something is getting halved that I haven't realized and that's cancelling out the voltage gain?
* Maybe my idea is correct, and I just mis-wired something on both channels?
* Something else?
On a side note, it does seem like the sound softened a little. I'm not sure if this is due to the increased Miller capacitance, or because the grid stoppers took care of some HF oscillation. With my speakers, this softening isn't really unwelcome, but I need to listen some more before I decide if I like it or not.
But, I'm confused about what I did wrong as far as the gain goes. If anyone could shed some light here, that would be most appreciated.
Thanks,
Saurav
yep, if you want more gain, then pick a operating point with more gain, or drive one section with the other.
if you parallel two sections, you uh.. lets see. assuming Rk and Ra are paralleled, and not just the same value shared (what you did.. well the CCS is being shared), you get double the current, and double the S.. that makes sense right.. yeah, thats it.
if you parallel two sections, you uh.. lets see. assuming Rk and Ra are paralleled, and not just the same value shared (what you did.. well the CCS is being shared), you get double the current, and double the S.. that makes sense right.. yeah, thats it.
Darn it. Now that you mention the battery analogy, I see what you're saying.
With these tubes, that would give me too much gain I think.
Oh well, looks like I'll have to try something else then. Maybe use 2 halves of one ECC99, and put a different tube in the other socket and use that for an input stage. Can anyone suggest a good 9-pin twin triode tube to use for say... 5x to 10x voltage gain? I feel like I'm *just* a little low in my overall system gain setup. Hmm... that'll mean another cap in the signal path, unless I try direct coupling the first 2 stages.
I think so. I changed a couple of resistors on my phono stage to get a couple of extra dB there, so now that and the DAC have about the same output level. Right now, I usually listen with my autoformers at about -9dB. If I crank it up to -3dB (my highest volume control position, I didn't want to go 0dB) it gets loud, but not crazy loud, especially on the LPs with lower recorded volume. I just have this irrational idea that with the volume control wide open, my system should play insanely loud. Or, to put it another way, my system should be able to go louder than I'd ever want it to, so that I can play it as loud as I want. Or something like that.
You say 'if not' - how much would this tube's gain differ from the ECC99? Let's see... the ECC99 has a mu of about 20ish, right, so this would be... 2.5x the voltage gain? So that's 6.25x power gain, which is... somewhere between 6 and 9dB, right, so between 12 and 18dB for both channels (I hope I haven't got even that math wrong). Actually, that might work.
Hmm... I see the drawback. This tube has a higher output impedance than the ECC99. Maybe this is a case where paralleling the triodes could help.
Saurav said:
If you want crazy loud, get some higher sensitivity speakers and/or a more powerful amp.
I agree that with the VC wide open, on the source with the lowest recorded volume, it should be loud enough for you. But, are you simply driving the poweramp into clipping ang getting compression?
I've also never understood people's aversion to winding the VC completely out of circuit. If it doesn't overload, why not?
Yes.
No
Gain difference is 20 x LOG(47/22) = 6.6dB
Rp = 7k2 vs 2k2 for the ECC99. Probably not a problem even as a single.
I only suggested this tube, as with a CCS load it has enough gain and some current drive, and because it's a 9pin dual. Ideally a pair or C3g as triodes, or low loaded pentodes would be better. It also keeps it in 2 stages. But both of these options require more holes in your casework.
If you really want to add another stage, add a 12B4A at the front.
I still think your amp is clipping and you really need more power. Have you measured it?
PARALLELS...
Hi,
As it seems some people have difficulty understanding what goes on when two equal triodes are wired in paralel:
Here goes:
Amplification factor stays the same as for a single triode.
Transconductance is doubled.
Internal resistance is halved.
Noise is halved.
This only applies for perfectly matched triodes and you can only hope both age the same way, something that, in practice actually does happen and is perfectly logical.
Cheers,
Hi,
As it seems some people have difficulty understanding what goes on when two equal triodes are wired in paralel:
Here goes:
Amplification factor stays the same as for a single triode.
Transconductance is doubled.
Internal resistance is halved.
Noise is halved.
This only applies for perfectly matched triodes and you can only hope both age the same way, something that, in practice actually does happen and is perfectly logical.
Cheers,
Eventually. I'll probably try a PP amp next, but for now I want to hold on to my SETheadness
I don't think I'm even driving my amp to its full output.
Not with power resistors. OK, to be honest, not really, no. I tried some very rudimentary checks with music playing fairly loudly, and I don't think I'm getting 2V at my amp's inputs.
My TT is off on one side of the room, the phono stage is right under that, and I have a 12' interconnect going to my VC. I just thought it's better to keep some current gain even at the highest volume control setting. I know, I should try it wide open and see if my phono stage can drive my amp through 15' of interconnects, maybe it can do it with no problems.
But if it's at a voltage amplification stage, doesn't it double when you get to power output? I need to think this through, I'm probably confusing gain and SPL levels. Oh, I see, you've already factored that in, that's where the 20 comes from, right?
I think I was doing the same calculation, just the other way round - square it first to go from voltage to power, then take the log to get dB.
So, 47/22 = 2.14x increase in voltage gain (I was using 2.5 here)
2.14 squared = 4.58x power gain
10 log (4.58) = 6.6dB
Anyway, that's not an issue. So 6.6dB increase in gain. Shouldn't this multiply by the number of channels, i.e. at the same level on the volume control, shouldn't I have 13.2DB more total SPL at my listening position?
That is something I would like.
My holes were originally sized for 6SL7s (I'm using hole shrinkers for the 9-pin sockets), and I have 2 holes for driver tubes. So, I could take a 9-pin pentode or something on an octal base. What would you recommend? My only constraint would be something that doesn't need more than 20mA per channel, so that I can eep my current CCS (I don't think it'll do much more than 20mA).
And it goes without saying, thanks a lot for helping me out.
Saurav
Yup, I see now that I had that wrong. I was thinking of it in terms of 2V AC in = 45V AC out, so if I put two of them together, that'll give me 90V out, or 1V in for 45V out, i.e. twice the sensitivity. Obviously, that's not how that works.
So you produce the same voltage output, but at twice the current which means half the resistance. That seems pretty obvious now that my mistake has been pointed out
Just saw a picture of a C3G:
http://w1.871.telia.com/~u87127079/pentodf/c3g.htm
I don't think I've seen that base before, and it looks bigger than an octal.
In terms of the circuit requirements the triode mode looks similar to the ECC99, but with higher mu. I think I can handle the higher plate voltage. I'm not 100% sure how to read the pentode data.
How much larger is that compared to an octal based tube?
http://w1.871.telia.com/~u87127079/pentodf/c3g.htm
I don't think I've seen that base before, and it looks bigger than an octal.
In terms of the circuit requirements the triode mode looks similar to the ECC99, but with higher mu. I think I can handle the higher plate voltage. I'm not 100% sure how to read the pentode data.
How much larger is that compared to an octal based tube?
Understood, but I don't think I'm reaching my amp's peak output. I do need to test that assumption though. I probably get close with my hotter recordings, like the bad 80's pop/rock that I own. But with the music I actually listen to, I get about mid-70s average with the VC at -9dB. Which means that when I want to really crank it up, the VC wide open is *just* a little low.
My speakers are 96-97dB, and my 2A3 SET should be good for, what, 3.5W? I'm running the Sovteks a little hot, so probably a little more than that. My room isn't too large (20x15x8 or so), and the speakers are along a long wall, which puts me pretty close to them.
My DAC is Chris Own's dAck, which has a relatively low output voltage (2V p-p, IIRC). I asked him about gain and he said it's set by a feedback resistor, so I could mod this to give me a higher output voltage if that was the better solution. On the plus side, this means that my CDs aren't a lot louder than my LPs (which used to be the case). My phono stage (Pass Pearl) has about 40dB of gain, with a 10x MC transformer ahead of it (so that's 60dB total), cartridge's output is in the .3mV range. I don't want to go 20x on the transformer because that'll cut the current available to drive the phono stage, and I don't know if that'll get too close to the input overload limit. I changed one load resistor to get more gain out of it, but I don't want to deviate from the design too much because I don't understand what affects or does not affect the RIAA response.
Basically, all of my system is set up for modest gain in an active linestage, except I don't have an active linestage
Well, it seems that (once again) my assumptions were wrong, and you guys were right
I pulled my scope out and took some measurements tonight. On a relatively loud CD, I get about 2.5-2.6V peaks at the output of the DAC. My linestage is set for a peak of -3dB, which gives me 1.8V peaks at its output (2.5/sqrt(2)). This gives me 5-5.5V peaks at the amp's output with a 7.5 ohm resistor. That's about 4W, which is what I'd expect. I don't have a signal generator, so I couldn't check for exactly when the amp starts to clip. With my test tone CD, I got about .7V input to the amp, and 2.2V out of it. That more or less matches what I saw with the music CD, which means my amp's gain is roughly 20 log (3), or 9.5dB. Does that sound about right for an ECC99->2A3? The problem is with LPs. A loud LP produced about half the voltage of a loud CD. I'll wait till my new cartridge gets here, and then experiment with the phono stage some more. I can try the 20x setting on the MC transformer and see if it has any negative effect on the sound.
Hmm... I guess I wasn't *completely* off though. It looks like I can max out my amp with my loudest music selections. It would be nice if I could do that with the music that's recorded at lower levels too. And adding gain somewhere in the chain would be one way to do that. But like you said, if I want to get any louder, I need a more powerful amp.
Once again, thanks for all the help.
Saurav
I pulled my scope out and took some measurements tonight. On a relatively loud CD, I get about 2.5-2.6V peaks at the output of the DAC. My linestage is set for a peak of -3dB, which gives me 1.8V peaks at its output (2.5/sqrt(2)). This gives me 5-5.5V peaks at the amp's output with a 7.5 ohm resistor. That's about 4W, which is what I'd expect. I don't have a signal generator, so I couldn't check for exactly when the amp starts to clip. With my test tone CD, I got about .7V input to the amp, and 2.2V out of it. That more or less matches what I saw with the music CD, which means my amp's gain is roughly 20 log (3), or 9.5dB. Does that sound about right for an ECC99->2A3? The problem is with LPs. A loud LP produced about half the voltage of a loud CD. I'll wait till my new cartridge gets here, and then experiment with the phono stage some more. I can try the 20x setting on the MC transformer and see if it has any negative effect on the sound.
Hmm... I guess I wasn't *completely* off though. It looks like I can max out my amp with my loudest music selections. It would be nice if I could do that with the music that's recorded at lower levels too. And adding gain somewhere in the chain would be one way to do that. But like you said, if I want to get any louder, I need a more powerful amp.
Once again, thanks for all the help.
Saurav
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.