Overkill tube pre/EQ called Drive-1. Maintenance update.

[Kingston]

First, the PSU heaters filter bank caps values are very large. I've browsed through several PSU circuits for tube gear and I've never found heater regulater filters to be that large. Is there any special purpose for that?

Two channels draw 3-4A of current. That's a lot of work for a cap to smooth out the ripple to a level that is flat enough for the regulator. 10000uF minimum for this reason (C16-17). I learned later that the post-regulator cap doesn't actually have to be this big. It has negligible ripple to smooth out and only really acts as an unnecessary reservoir. 1000uF should be enough. 22000 is overkill for all of them, I admit.

There is little to no PSRR in this Drive-1 design. This DC heater rig was a precaution to not insert any unnecessary hum to the system.

Second, the heaters supply is floated to the B+. What's the benefit of this?

Tubes have a maximum heater-cathode voltage difference. See data sheets. Stage V2A is cathode follower where the cathode is at +100V potential (I forget the exact value). Heater needs to be biased closer to this value for long and reliable tube life.

Third, is there any reason (sound wise) to make a tube regulated PSU instead of solid state? I remember Winston O'B speaking regarding that subject - that some semiconductors throw out a lot of hash and that we're essentially listening to the PSU. Any thoughts on that matter are highly appreciated.

There is one major source for noise in this zener referenced pass transistor regulator. That's the zener itself. This noise is completely eradicated with two separate RC networks on the preamp side (R33/C15 and R36/17).

There is some truth to us "listening to a PSU". If the PSU is not ideal, has not so much capacitance and no regulator (or some inferior tube regulator rig), the PSU will audibly "dip" on large current surges, and even let through significant ripple. This is very common on guitar amps because of the power stages. It can happen on plain preamps as well, especially with extremely simple PSU's like a plain CRC with very little capacitance. The practical effect is more distortion.

Drive-1 PSU on the other hand is WAY too stiff. It will always deliver solid B+. It's so clean it's boring. Think mastering equipment transparency. That was not my design goal, but an over eager design "accident". I didn't think it through. Next time I will attempt something designed for distortion I will make a really sh*tty PSU.

 

[Kingston]

This project is not dead. But after some years of steady usage the two channel unit I built began to develop some problems. There was random and very unpredictably behaving noise first on one channel, then after a while on the other channel as well. Sometimes it would go away, sometimes it would be louder.



First issue:

VR1 was working into far too high impedance and the 100k log rotary switch was making the randomly behaving noise shape change. I'm not qualified to deal with this bootstrapped cathode follower input ~1Ghz impedance (or whatever), and neither are my tools. Kind of like mic capsule where you can't let even an extra hair in the circuit. Having far too high impedance dangling in rotary switch and its wires is just bad design so I reduced it significantly and isolated it to a tiny area. No more Ghz in the rotary switch. No more changing noise shapes, and noise significantly reduced in general.

Further, the original VR1 configuration right at the bootstrap only allowed about 8dB of attenuation range making it almost useless. Now it works like any other basic attenuator, full attenuation range.



Second issue:

There's still noise. The V2 tubes I have been using are super-cheap CCCP 6N9S and it turned out these can't be trusted. They will all develop noise after a while. I don't have any 6SL7 around and I don't feel like making adapters for 12AX7 so I quickly swapped in a 6N8S (a CCCP 6SN7 clone that I was already using as V1). Turns out it works fine, even if by design this NFB around a virtual earth filter configuration should require high mu tubes. Turns out the difference between 6N9S and 6N8S is about 1dB less of bass boost. Treble part and bass cut stayed the same. I can live with that. There's still about 12dB of bass boost. And I had extra juice in the heater PSU for the almost doubled heater current.

Some guitar guy might come here and say sound also "improves" just by swapping a 6SL7 to a 6SN7, but really, no spec changes (THD/IMD/freq response).

[edit]
well, there was one change I forgot because it's a non-issue: http://www.groupdiy.com/index.php?topic=41509.msg672293#msg672293
[/edit]


Additional maintenance stuff:

Added (and finally tested) correct zobel network for lundahl (minor freq. response improvement). Added the output stage grid/cathode cap discussed earlier in the thread. Changed output "dummy" load to something more sensible.



New schematic version 2.0 added to the first post of the thread.

 

[plumsolly]

This is a very cool project and the discoveries/revelations that come as it progresses are incredibly valuable.

I want to have a crack at a Kingston Drive for myself.

I am going to build it an a old Gates portable mic amp chasis (model # GR-90) from ~1949. When I got it, it had been hacked up and modified beyond recognition and since I don't have a schematic to put it back together again it will live on as a Kingston Drive.

Mine will have some differences. I will use the original external PS, which is tube rectified and has AC heaters. R/L/C filtered, non-regulated  B+. I will float the heaters to ~85VDC as per the schematic. My first question is in regards to floating the heaters: Can you think of a reason why it would make a difference if I physically place the divider on the B+ and the connection to the fake center tap on the heater windings in the amp box vs. the PS box?

I will do six frequencies per band on rotary switches for the eq. Was thinking 2.7khz, 4.7khz, 5.8hz, 8.5khz, 10.6khz, and 12.7khz for the high band and 19hz, 29hz, 64hz, 94hz, 136hz, 194hz for the low band as these are standard cap values for frequencies above and below the example frequencies. Any thoughts on those choices? Should I go higher for the high band? They are obviously not set in stone, but curious if anyone had initial thoughts. On a Pultec I often use the 3khz high boost, but not sure how that compares.

I was going to try to use the transformers from the Gates. They are custom UTC's. I think the input is an O-2 (50, 250, 600: 50k) in a  different form factor. At least this is what makes sense based on DCR measurements and usage. If this doesn't provide the performance I'm looking for, I have a Cinemag 150, 600:15k which is a nice transformer. I also may experiment with using a dedicated bridging input transformer for the line input as opposed to the pad before the mic in.

The output transformer from the Gates is some kind of open frame plate-to-line thing that will take DC on the primary. This raises another question: What are the relative merits of powering the output stage through the output transformer primary vs. using the 15k plate resistor and cap coupling the plate to the primary? One thing to consider in this is that the UTC plate-to-line transformers that take DC on the primary are only rated for 8ma (true for LS, HA, and A series) - so that might be an issue. Another thing is, is the frequency response of this transformer likely compromised in order for it to handle DC on the primary? Check out the UTC LS-50 vs Ls-27 in the attached pic. I'd like to use the transformers I have - this is supposed to be a low-cost, parts-box summer project.

Thanks to Kingston for all your hard work on this. Iam excited to hear it!

Best,

Ben

 

[lassoharp]

are only rated for 8ma

3.5 - 4 will drive 600r to good enough levels for DAW. 

 

[emrr]

dude, if you search the interwebs that Gates schematic is out there, multiple times, it's here, redrawn at least once.  Go find it.  Smart people made that stuff, it's up to you to sort out what they made it for. 

 

[plumsolly]

Hey Lasso,

I meant they are rated for 8ma DC in the primary, which might be a problem if I run B+ through there. Kingston lists the B+ current consumption for V3b as 10ma.

Best,

Ben

 

[plumsolly]

dude, if you search the interwebs that Gates schematic is out there, multiple times, it's here, redrawn at least once.  Go find it.  Smart people made that stuff, it's up to you to sort out what they made it for. 

I searched for an hour plus. Looked for the Dynamote too, which is supposed to be similar, and no luck. In fact the only reference to the GR-90 specifically was in some catalog and even that is gone since I went looking a couple of years ago. Anyways, the front panel was drilled, and the whole thing was hacked up pretty significantly. Whatever it was, it's not anymore. But she lives on in spirit. PS - you know anything about those transformers? 0-2 sound right for the input?

EDIT: Found that reference: http://gates-harris-history.com/1946-09%20October.pdf

Best,

Ben

 

[lassoharp]

Hey Lasso,

I meant they are rated for 8ma DC in the primary, which might be a problem if I run B+ through there. Kingston lists the B+ current consumption for V3b as 10ma.

Best,

Ben

Yeah that would be tough on the UTCs .  Sowter makes a couple SEs rated for 20ma I think. 

 

[emrr]

Sorry to harsh; I feel harsh; it's not you.  Look up the Remote Combiner I'm sure I posted.  All that era Gates is more or less the same.  They usually have a rebranded A-11 as input along with the open frame output. 

 

[Kingston]

I completely forgot one thing with my 6SL7 to a 6SN7 swap on the filter network: gain dropped 3dB. It's very close to unity gain stage with high mu (6SL7/12AX7) tubes. The gain drop is non issue, I don't think I've ever needed the full +80dB available and this is why I forgot to mention about it earlier.

I will get some proper 6SL7 one day so consider this a kind of a hack that falls neatly into the original design.

Can you think of a reason why it would make a difference if I physically place the divider on the B+ and the connection to the fake center tap on the heater windings in the amp box vs. the PS box?

I will do six frequencies per band on rotary switches for the eq. Was thinking 2.7khz, 4.7khz, 5.8hz, 8.5khz, 10.6khz, and 12.7khz for the high band and 19hz, 29hz, 64hz, 94hz, 136hz, 194hz for the low band as these are standard cap values for frequencies above and below the example frequencies. Any thoughts on those choices?

It won't make a difference where the heater float reference is physically. I like the fact you might use the old PSU. My PSU design is way too stiff. Sure it makes the performance very predictable and kind of high end. But it's not a "vintage" sound.

With the EQ bands wire some test caps first because it tends to be very transparent, this might influence your choices over any calculations. It's a different sound from pultec-style passive networks. I sure would like more low mid selections on the bass and high mid selections on the treble. Those relays are awfully limiting and probably too paranoid school of design.

 

[Kingston]

Hey Lasso,

I meant they are rated for 8ma DC in the primary, which might be a problem if I run B+ through there. Kingston lists the B+ current consumption for V3b as 10ma.

Best,

Ben

Yeah that would be tough on the UTCs .  Sowter makes a couple SEs rated for 20ma I think.

That 10ma is the value I penciled in when playing with the tube load line and it's not constant current either (moves with the load line).

You need to increase the cathode resistor if this becomes an issue. Most likely goes unnoticed. You can double or triple the value from the schematic and you will only hear a difference on very high output stage distortion. You will be able to measure the change in THD. But very mild audible difference.

That was one of the main lessons I learned from this project, just how little one can influence the way tubes distort with simple loading changes or B+ voltages. Even with drastic value changes. And how any part value in a schematic becomes non-critical. 30% change in any direction will go unnoticed unless you have a reference channel for comparisons. And even then you won't care!

 

[millzners]

Hey Lasso,

I meant they are rated for 8ma DC in the primary, which might be a problem if I run B+ through there. Kingston lists the B+ current consumption for V3b as 10ma.

Best,

Ben

Yeah that would be tough on the UTCs .  Sowter makes a couple SEs rated for 20ma I think.

That 10ma is the value I penciled in when playing with the tube load line and it's not constant current either (moves with the load line).

You need to increase the cathode resistor if this becomes an issue. Most likely goes unnoticed. You can double or triple the value from the schematic and you will only hear a difference on very high output stage distortion. You will be able to measure the change in THD. But very mild audible difference.

That was one of the main lessons I learned from this project, just how little one can influence the way tubes distort with simple loading changes or B+ voltages. Even with drastic value changes. And how any part value in a schematic becomes non-critical. 30% change in any direction will go unnoticed unless you have a reference channel for comparisons. And even then you won't care!

To your point, when I implemented a VVR B+ knob like you've used on other projects I was shocked by how subtle the difference between 300VDC and 100VDC is.  Tubes are really resilient devices, whereas a 0.1V here or a few extra ohms there in a transistor circuit and you're oscillating out of control.

 

[Kingston]

To your point, when I implemented a VVR B+ knob like you've used on other projects I was shocked by how subtle the difference between 300VDC and 100VDC is.

The front panel for the switch says "clean" (300V) and "crunch" (150V if I remember correctly). Really it should say "clean" and "also clean".

Makes me feel really stupid when I built my first tube project long a go and the design said "250V". I spent days hunting that exact value when the PSU measured about 240V.

 

[plumsolly]

Still plugging away and gathering parts. Got the case repainted and have done a lot of planning. Any reason not to use a 6SN7 on the output if I wanted all octal?

Thanks,

Ben

 

[Kingston]

Still plugging away and gathering parts. Got the case repainted and have done a lot of planning. Any reason not to use a 6SN7 on the output if I wanted all octal?

6SN7 has way too high plate resistance. It won't drive the output transformer adequately. You'll have a bass cut, especially bad on the non-bypassed cathode cap mode. ECC99/6N6P was selected for this exact reason and the plate and cathode resistors were tuned to their datasheets.

[edit]

just realised, this information might or might not apply to your SE output transformer plans, but 6SN7 is likely too weak there as well.

The 6N6P output was chosen so one can use 2:1 (into 600-ohm line) output transformers. I used a 4:1 Cinemag which was actually not even necessary.

 

[plumsolly]

Thanks Kingston. I think I am going to stick with the 6N6P. For octal it looks like the 6BL7 would be the tube: Plate R = 2150Ω, µ=15. I will definitely experiment with transformer options. Not sure what the ratio of that Gates output is: should probably test. UTC plate-to-lines are 15K:600 so thats what this guy probably is, so 5:1. I wonder if a 600Ω line repeater wired 600:150 might work on the output? I have a couple of nice ones in the junk bin.

Best,

Ben

 

[Kingston]

I wonder if a 600Ω line repeater wired 600:150 might work on the output? I have a couple of nice ones in the junk bin.

It won't work here. It's a Saturday night and I'm feeling a bit lazy to find the exact calculated value of the output impedance of 6N6P/ECC99 configured this way. Nevertheless it's way above the ability to drive a 600 ohm directly. Especially with the non-bypassed cathode feedback configuration.

But worry not, I would have plenty of use for those elsewhere, should you ever happen to feel the need to get rid of them.

 

[plumsolly]

But worry not, I would have plenty of use for those elsewhere, should you ever happen to feel the need to get rid of them.

I'll keep you in mind should they become a burdon. ;D

I don’t quite understand the difference between two transformers with identical turns ratios and different nominal impedances, like say a 600Ω:600Ω transformer and a 10KΩ:10kΩ transformer. The 10K:10k has higher inductance because it is expecting higher driving impedances? If I wired that 600Ω repeater as a 2:1, it would reflect the same impedance to the tube as the Cinemag 9589 wired 2:1 and if the inductance of my repeater was high enough I might get away with it? Or are there other factors at work? Anyways, I’ll likely use my Gates output and won’t have to worry about it, but I might hook up a UTC HA108X wired 2:1 and maybe a WE 111c wired the same way and see how they do.

BTW, I know you are a fan of the oil caps - I am going to use several 1940’s U.S. Navy surplus bathtub caps in my build. They are new and I tested them for capacitance and leakage and they look great.

Best,

Ben

 

[Kingston]

First of all, it's not possible to wire CM-9598 as 2:1. It only does 4:1 (at 10k/600 to be exact).

You seem to have somehow confused the output transformer ratio as the main design factor. It's not. We are only interested in the primary impedance. The ECC99/6N6P output configuration in Drive-1 has an output impedance of about 2k, maybe a bit less. That can be thought of as the absolute lowest load it can drive. We need to select the output transformer primary based on that.

The second most important design decision is what the output transformer should see in the outside world. I have selected a 600-ohm line.

Now we can choose the ideal output transformer, which is about 2k/600. That won't be found on the shelf, but 2k4/600 (2:1 ratio) is common, we can use that. But there's even too much gain available in Drive-1 so it really doesn't matter if you decide to use 10k/600 instead (4:1 ratio, 6dB more gain loss). Clearly the output tube will happily drive that tiny load of 10k.

But we can't use a 600/150ohm line repeaters. The tube can't drive a 600-ohm primary. Again, ratio has nothing to do with it.

 

[plumsolly]

My understanding of transformers (which is limited!) was that a transformer does not have an impedance of its own as such, but rather reflects the impedance connected to the opposite winding multiplied by the turns ratio squared. So a 600:600 transformer loaded with 10k is going to look like 10K*12=10k at the primary. And a 600:150 transformer loaded with 600Ω is going to look like 600*22=2.4K at the primary. The nominal impedance, ie what separates a 10K:10K from a 600Ω:600Ω, is about recommended use, which is dictated by inductance and possibly other properties.

Here is a quote from Samuel Groner from this thread: http://www.groupdiy.com/index.php?topic=52667.msg671739#msg671739

"A transformer reflects the impedance at its secondary, not the number quoted in the datasheet. If you load your transformers with 10k, that's what's the input impedance will be (to first order, of course). As noted above frequency response/distortion might suffer if the transformer is not used with its nominal source/load impedances, though."

Best,

Ben