-10 unbal to +4 xfrmr balance how much gain?

[Blue Jinn]

I understand the difference is about 12dB. However, if you use a transformer for the balance output, do you want a 4x voltage gain before the transformer or something different? (assuming a 1:1 transformer)

This is kinda what I'm talking about here (this was taken from another site) the opamp circuit is representative so assuming -10dB into that stage should that have a voltage gain of 4x/12dB or ???:

 

[mshilarious]

Yes, or use a 1:4 transformer. If you don't have the power supply (volts) for +4dBu, you use a trafo for the gain, but that means you have to feed it more current for a given load.

 

[Blue Jinn]

Thanks!

BTW, I'm thinking of using this Oatley 6418 preamp, has a gain of 8x with a 2:1 xfrmer to knock that down to 4x gain and provide the balance out. Just not sure of the output impedance. I am also pretty sure it is inverting, so thinking about either a opamp buffer or swapping leads on the trans.

http://secure.oatleyelectronics.com/files/K261notes.pdf

 

[mshilarious]

I've seen this crop up a few times, and I'm not just saying this to be contrary: that amp has serious issues. It would not be usable as a mic preamp, and it would be marginal on anything else. Its output impedance is 220K; with a 2:1 trafo on the output, that would still be 55K. It will have serious problems with switching power supply noise leaking into audio. It will be totally unacceptable without R9, and still not very good with it. I would look for a different preamp kit if that's what you want to do.

If you correct the power supply problems, then you would need to add an opamp buffer to correct the output impedance problem (and as you noted, reinvert the signal). You could also add a pot in between to control level.

 

[Blue Jinn]

I've seen this crop up a few times, and I'm not just saying this to be contrary: that amp has serious issues. It would not be usable as a mic preamp, and it would be marginal on anything else. Its output impedance is 220K; with a 2:1 trafo on the output, that would still be 55K. It will have serious problems with switching power supply noise leaking into audio. It will be totally unacceptable without R9, and still not very good with it. I would look for a different preamp kit if that's what you want to do.

If you correct the power supply problems, then you would need to add an opamp buffer to correct the output impedance problem (and as you noted, reinvert the signal). You could also add a pot in between to control level.

I don't quite grasp the math for the Zout, and the figures I came up with were between 100-200k, which surprised me, I'm guessing this thing is intended for a guitar amplifier with a nice high input Z??? Also, the clear warning about "high frequency hash" due to the switching psu sounded a few bells.

They have several variants with the 6418 tube, two of them have the switching psu, and one has a low voltage plate: I missed this one at first, which has the opamp output stage: http://secure.oatleyelectronics.com/files/K272A NOTES.pdf


However, for $30 AUD (don't know what the shipping to USA is) it seemes intriguing.

The opamp supply voltage is only 5v though, which seems inadequate for this purpose, but with a few hacks ... fashioning a sep PSU with LM317s for 15v (substituting a 5532 or other opamp), 9v and 3v....then a trafo to provide the balanced output.

Too much work?

 

[mshilarious]

No, not too much work, but at that point there is no reason to get their kit. 6418 is a fun tube, but you can buy them for $2.00, and the rest of the parts are generic.

You are probably right on output impedance, I was thinking that plate resistance was higher than it probably is (420K at 22.5V, I usually used the tube at lower plate voltage/lower gain where plate resistance was high enough to ignore), but it's still an absurdly high impedance to send directly to an output. I mean, the stock kit is totally unusable to an output. Even into a guitar amp, you still have to contend with cable capacitance--you'd be limited to a 1 ft. cable.

I would sketch out a 1:4 input transformer (maybe 1:6) into a LTP of 6418s feeding an output opamp, that sounds a lot more interesting as a mic amp project. I have thought about doing a 6418 stage in my portable mic amps, but these tubes don't like to get bumped.

 

[Blue Jinn]

No, not too much work, but at that point there is no reason to get their kit. 6418 is a fun tube, but you can buy them for $2.00, and the rest of the parts are generic.

I like the convenience of a pcb though, and it looks simple enough to mod with or w/out cutting traces. The price difference for the parts, vs the time figuring out pt2pt is worth it to me the US$30 or so (incl shipping. )

Here again, my equation knowledge is limited, but do you think I could just replace the IC1 with a OPA2136 or 5532 eliminate the 5v regulator, and run this at +18 rather than +9v without any additional changes? (or maybe changing the plate resistors to something higher?) I'm not sure what R5 and R6 do for the opamp, my understanding is I have to have half the supply voltage there to bias the opamp if running a single supply.

If I understand correctly that would provide a 9v swing on teh opamp which should be enough for headroom of about +18dBu, but I intend to keep levels around 0VU anyway. (I suppose even the 2.5v stock would be enough, but the additional headroom seems wise.)

 

[mshilarious]

but do you think I could just replace the IC1 with a OPA2136 or 5532 eliminate the 5v regulator, and run this at +18 rather than +9v without any additional changes? (or maybe changing the plate resistors to something higher?) I'm not sure what R5 and R6 do for the opamp, my understanding is I have to have half the supply voltage there to bias the opamp if running a single supply.

Yes, although I wouldn't increase the plate resistance, it strikes me as way too high for 9V already. R5 and R6 form a half-rail voltage divider, which serves as a virtual ground reference for the opamp.

If I understand correctly that would provide a 9v swing on teh opamp which should be enough for headroom of about +18dBu, but I intend to keep levels around 0VU anyway. (I suppose even the 2.5v stock would be enough, but the additional headroom seems wise.)

Not quite, you have to give up some voltage to the rails. Also, ain't no way 6418 is gonna do +18dBu anyway, so you'd have to add more gain at the opamp. I don't like the 1M resistors at the opamp at all; I'd have to work through the whole circuit noise analysis, but that is not going to help. When you need a very high input impedance input, you don't use an inverting opamp.

You should be aware that the circuit as they drew is probably running about 10% THD at 0dBu.

edit: ugh, I just noticed the 1M input pots. Why? Honestly, these kits are extremely poorly engineered.

 

[Blue Jinn]

Not quite, you have to give up some voltage to the rails. Also, ain't no way 6418 is gonna do +18dBu anyway, so you'd have to add more gain at the opamp. I don't like the 1M resistors at the opamp at all; I'd have to work through the whole circuit noise analysis, but that is not going to help. When you need a very high input impedance input, you don't use an inverting opamp.

You should be aware that the circuit as they drew is probably running about 10% THD at 0dBu.

edit: ugh, I just noticed the 1M input pots. Why? Honestly, these kits are extremely poorly engineered.

I'm only looking for about 12dB of gain to take a -10 output and bring it up to a +4 balanced output, with enough headroom for perhaps a +18dBu peak. (6 volts or so?) My understanding is there needs to be an adequate supply voltage to handle that much input voltage, hence running the supply at least +18v rather than +9v. (Makes the new psu a little simpler too.)

There does seem to be a lot of 1Meg resistors. Don't see why that is necessary, unless the design headphone amp chip somehow wants them. I don't see why they can't be half or even 1/4 as much, same for teh trimmers on the grids. What do you think?

 

[mshilarious]

I'm only looking for about 12dB of gain to take a -10 output and bring it up to a +4 balanced output, with enough headroom for perhaps a +18dBu peak. (6 volts or so?) My understanding is there needs to be an adequate supply voltage to handle that much input voltage, hence running the supply at least +18v rather than +9v. (Makes the new psu a little simpler too.)

There does seem to be a lot of 1Meg resistors. Don't see why that is necessary, unless the design headphone amp chip somehow wants them. I don't see why they can't be half or even 1/4 as much, same for teh trimmers on the grids. What do you think?

The input pots have no need to be 1M, that's just a source of noise. The resistor off the plates, unfortunately, do need to be that high, because the inverting input of an opamp is a virtual ground. Input impedance needs to be high (given the high output impedance of the plate), and the input impedance of the inverting input is its series resistance.

This is why one doesn't feed a plate directly to an inverting input, you would need two stages--first into a noninverting input (or simply a buffering FET), then invert it.

I mean there is more . . . I think the grid bias is less than ideal (it's different left to right!) . . . these things are just a trainwreck. If I get time, I'll try to breadboard something better this weekend . . .

 

[Blue Jinn]

Input impedance needs to be high (given the high output impedance of the plate), and the input impedance of the inverting input is its series resistance.

This is why one doesn't feed a plate directly to an inverting input, you would need two stages--first into a noninverting input (or simply a buffering FET), then invert it.

I mean there is more . . . I think the grid bias is less than ideal (it's different left to right!) . . . these things are just a trainwreck. If I get time, I'll try to breadboard something better this weekend . . .

I forgot how high the plate impedance was on these tubes. I read a reveiw also that mentioned an issue with the uneven grid bias. It doesn't follow that VR2 would go to the heater supply and VR1 goes to ground. Why not cut a trace at VR2 and take it to ground?

(R15 is a jumper per the instructions)

 

[mshilarious]

I forgot how high the plate impedance was on these tubes. I read a reveiw also that mentioned an issue with the uneven grid bias. It doesn't follow that VR2 would go to the heater supply and VR1 goes to ground. Why not cut a trace at VR2 and take it to ground?

(R15 is a jumper per the instructions)

Well it kinda sorta does, because it's with respect to the filament, but that means it's different relative to the plate (which is a smaller problem, but still a problem).

They were trying to be efficient with their 3V battery supply--should have just wired the two batteries in parallel.

 

[Blue Jinn]

This seems to be taking on a life of its own here.

Well, how about eliminating the 5v chip, running on 18v, leave the plate resistors alone, reduce r3 through r8 to 500k, (closer to the plate resistance) reduce vr1 and vr2 at the grids to something like 100k and cut a trace to send vr2 to ground rather than filament, depending on gain select a trafo on the output for balanced out?

 

[mshilarious]

Let's do a basic noise/gain analysis. An ideal mic preamp has noise of -133dBV, which is the thermal noise of a 150 ohm resistor. Not too many preamps hit that, but many come within 6dB. So if the purpose of this amp is a mic preamp, we have to be really concerned with input noise.

You will almost never see a pot at the front end of a mic preamp for that reason, and if you do it will be more like 1K. A 100K pot, worst case scenario, has thermal noise of a 25K resistor, which is -111dBV. That will only be OK for condenser mics, and relatively hot ones at that.

I don't remember the noise of 6418, but I think it's probably in the -110 to -120dBV range. I will try to test that, but let's say it is -110dBV. That noise will sum with your input noise, so you'd have -108dBV, and be multiplied by the gain of the circuit. If you have a gain of 8, now your noise is -96dBV. But the noise of a 500K resistor is also -98dBV, those will sum to -94dBV. A nominal mic signal of -40dBV/Pa is now -28dBV, with signal to noise ratio of only 66dB.

One way to solve that problem is to put your 1:4 transformer at the front, that gives you 12dB of gain, and in a perfect world, your equivalent input noise would drop to -118dBV, which is workable.

They designed this thing as a headphone amp, so they could get away with not really caring about noise, because they are presuming feeding it a 0dBV signal. 6418 isn't going to cope very well with that after a gain of 8, but according to them that is "tube sound".

I'm afraid there is no solution for their topology that will fix the noise problem without reducing gain of the circuit, which makes the noise problem worse, which requires you to further reduce gain . . .

A functional mic preamp with 6418 has a 1:4 transformer input (invert the signal at the transformer), followed by 6418, followed by a FET (or noninverting opamp), then a pot (or put the pot in the opamp's feedback loop), maybe another 6418 if you don't use the opamp, then finally some type of output buffer, probably a noninverting opamp. Another transformer on the output if you wish. That could have reasonably low noise, a pleasant amount of THD, and any amount of total gain you require. It could be fully discrete if you want, and all tube & FET if that floats your boat. You could have two pots for a topology like a Fender Champ if you want to drive a 6418 deep into overdrive.

 

[Blue Jinn]

I don't remember the noise of 6418, but I think it's probably in the -110 to -120dBV range. I will try to test that, but let's say it is -110dBV. That noise will sum with your input noise, so you'd have -108dBV, and be multiplied by the gain of the circuit. If you have a gain of 8, now your noise is -96dBV. But the noise of a 500K resistor is also -98dBV, those will sum to -94dBV. A nominal mic signal of -40dBV/Pa is now -28dBV, with signal to noise ratio of only 66dB.

If I'm following this correctly, with a nominal input of 0dBV that would still be a 94dB S/N which would seem to be acceptable amount of noise. E.g. if I use this to take a consumer level line out to drive either a Dolby 360 series or an Otari tape deck directly or maybe a Sony DAT: the S/N on the Otari itself is only about 65dB and I think the DAT about 92dB.

A functional mic preamp with 6418 has a 1:4 transformer input (invert the signal at the transformer), followed by 6418, followed by a FET (or noninverting opamp), then a pot (or put the pot in the opamp's feedback loop), maybe another 6418 if you don't use the opamp, then finally some type of output buffer, probably a noninverting opamp. Another transformer on the output if you wish. That could have reasonably low noise, a pleasant amount of THD, and any amount of total gain you require. It could be fully discrete if you want, and all tube & FET if that floats your boat. You could have two pots for a topology like a Fender Champ if you want to drive a 6418 deep into overdrive.

This sounds cool too, though.

 

[mshilarious]

OK here is a basic functional schemo, build from here. The grid bias scheme isn't very elegant, but actually you can get an interesting range of distortion flavors by playing with it, so I stuck with the trimpot. I also played with actually wiring it as a pentode, but there is hardly any gain to begin with and that didn't help. In the end, I went with simplicity.

This has gain of 8 (4 from the tube). If you want higher gain, you'll need to go to a 30V supply. Recycling the filament current to drive a class A FET for the output saves the need to branch off the power supply, but ultimately uses more power than is strictly necessary. Still, hey, pure class A throughout . . .

EIN is about -98dBV, not too good. This will tolerate a -18dBV input for 0dBV output with just about 1% THD below, nearly all second. You can change that profile by changing the grid bias.

Add a 1:4 transformer on the input if you need more gain/lower noise/balanced input; add it on the output if you need higher levels (but be sure you are feeding a 10K input). If you use an input transformer, you can chuck the input cap and bias the transformer instead.

 

[Blue Jinn]

EIN is about -98dBV, not too good. This will tolerate a -18dBV input for 0dBV output with just about 1% THD below, nearly all second. You can change that profile by changing the grid bias.

Mmmmh. Thanks. Looks straightforward enough to put on one of those Radio Shack small pcbs.

With a 8x gain I'm thinking to eliminate the coupling caps c2 and c3 and drive the transformer from the JFET Drain. (Probabaly an EDCOR 2:1) I can live with the S/N as this is intended to drive either a tape deck or an older DAT and that is about the dynamic range of the DAT. Perhaps as the +4 stage to that PRR limiter I need to get around to building. I didn't realize you were teh MSH-4 guy until recently. (I actually emailed you directly about that a few days ago....) I was surfing the 'net for some info on the G7 (the Aurycle kit seems to be an OK place to start) and stumbled across the MSH-4 which led me to the Oatley 6418 kits.... the internet can be a oddly interconnected place....

 

[mshilarious]

If you're up to the PRR, I would skip 6418 circuits entirely. They are fun, but they aren't serious. I mean a gain of 8 is kind of a joke, and the filament really likes to ring. Maybe try a PRR-type circuit using 5840 or something (not that I've given that any thought at all, which I haven't).

The genius of the PRR is subtle; I think it's the use of the second mains coil as the high-voltage rail. A huge part of the cost of a tube circuit is the specialized transformer, but when you buy a garden-variety 12VAC to drive the filament and a solid-state output & sidechain and 120VAC (which is practically no-load and will thus rectify to 140VDC), I mean you're down to about $50 in parts.

I used that trick to build a cheap hybrid tube guitar amp years ago. Lately I've been working on an all-FET, all-discrete guitar amp, but I doubt it will see the light of day, it just costs too much for the box/power transformer/panel etc. Maybe I'll do it as a 2010 version of the Rockman (which I hated!)

 

[Blue Jinn]

If you're up to the PRR, I would skip 6418 circuits entirely. They are fun, but they aren't serious. I mean a gain of 8 is kind of a joke, and the filament really likes to ring. Maybe try a PRR-type circuit using 5840 or something (not that I've given that any thought at all, which I haven't).

I'm not so sure I'm up to the PRR just yet, although I've been acquiring parts for the Bluebird version. For the purposes here, I think the 6418 circuit should work out well as a -10 to +4 amp, don't need a whole lot of gain for that, and the relative simplicity too.

The genius of the PRR is subtle; I think it's the use of the second mains coil as the high-voltage rail. A huge part of the cost of a tube circuit is the specialized transformer, but when you buy a garden-variety 12VAC to drive the filament and a solid-state output & sidechain and 120VAC (which is practically no-load and will thus rectify to 140VDC), I mean you're down to about $50 in parts.

Never thought about it, but yeah, you could use that PSU in a lot of other circuits. Plus the forum over there, I've learned more about electronics just reading through the several hundred page support threads.

 

[mshilarious]

But 6418 can't handle a +4dBu signal level - my circuit only does +0dBV (+2dBu) as a maximum level. For +4dBu, you really need about a +22dBu max level. Even if you step up the plate voltage to 30V, you aren't going to get there. And since plate current is limited to 500uA, you can't use a step-up transformer to do it either.

Thus, you must follow it with a transistor(s) that is doing all of the work--voltage and current gain. At that point 6418 is just a harmonic distortion generator, and a noisy one at that. Replace V1 with another J107, with a common source and 10K drain resistor, and you've got pretty much the same thing.