THAT1512 with a side of...

[JoshuaUnitt]

Hey folks. I'm working on my first preamp build and wanted to throw some stuff out to the experts here and see what kind of help I could get (doing a search through the forum has been helpful but I wanted to address some specifics). I've got some soldering/build work under my belt but I'm still lacking in the theory department.

Anywho. here is a quick block diagram I drew up for a THAT1512 pre:

-Of course there are some glaring omissions from the 1512 circuit here, but I wanted to cover the basics. This is what I'm more or less basing the actual circuit off of (from the chip's datasheet):

-the "black box" in my block diagram would be comprised of a JLM preamp go-between kit, for simplification purposes. However, I'm not exactly sure where I should add it to the 1512 circuit- should it be after C1-3, where THAT recommends that phantom power should be inserted?

-I plan on using this circuit for phantom power- http://sound.westhost.com/project96.htm any complications?

-As you can see I wanted to have a switchable output balancing stage- active balancing or transformer balancing. Something of a lemon/lime flavor choicefor the end user, I suppose. However I've read that there may or may not be certain things that need to be done with output transformer- adding a single resistor to the secondary, adding a Zobel network to the secondary, etc- but nothing definitive. And as far as Zobel networks go, the only information I have seen for calculating the values relates directly to when the next stage is a speaker- not a +4 line input. Any help here?

-finally, the T-pad 600ohm attenuator, which I figured would be a good way to help drive the pre into that fabled land of saturation and analog happiness. reasonable estimation? (I'm gonna assume in advance that it's not that simple)


Thanks in advance for any and all help, it's appreciated!

Peace,
Josh

 

[Meathands]

-the "black box" in my block diagram would be comprised of a JLM preamp go-between kit, for simplification purposes. However, I'm not exactly sure where I should add it to the 1512 circuit- should it be after C1-3, where THAT recommends that phantom power should be inserted?

Nothing should be before C1-C3. The idea there is to be as close as physically possible to the input pins and the chassis connection.

-As you can see I wanted to have a switchable output balancing stage- active balancing or transformer balancing. Something of a lemon/lime flavor choicefor the end user, I suppose. However I've read that there may or may not be certain things that need to be done with output transformer- adding a single resistor to the secondary, adding a Zobel network to the secondary, etc- but nothing definitive. And as far as Zobel networks go, the only information I have seen for calculating the values relates directly to when the next stage is a speaker- not a +4 line input. Any help here?

Totally depends on the transformer. Zobel networks are added to dampen transformer ringing, with the R and C values being based on the particular transformer's resonant frequency. Same with a simple loading resistor. 600 Ohms is generally safe, which would be provided by the T-pad in either case.

One thought on the way you've got the lemon/lime switching configured. I wouldn't worry about doing "true bypass" the way you've got now. You can just split the single-ended output signal to the transformer and balancing board. In the worst case--THAT1646 in the signal path with output transformer "hanging off"--you've got a 600R load (assuming 1:1 transformer) in parallel with the THAT1646 input Z. Signal loss and loading will be negligible.

-finally, the T-pad 600ohm attenuator, which I figured would be a good way to help drive the pre into that fabled land of saturation and analog happiness. reasonable estimation? (I'm gonna assume in advance that it's not that simple)

I'd put the t-pad before the lemon/lime switch since there's no need for it in the "clean" circuit. But another thing to consider is, have you ever tried to overdrive an EA2503? It's not that easy to get much noticeable THD out of it at reasonable line levels! Maybe the T-pad works well in the API circuits because it makes the DOA work harder, I dunno haven't played around with one.

 

[JoshuaUnitt]

-the "black box" in my block diagram would be comprised of a JLM preamp go-between kit, for simplification purposes. However, I'm not exactly sure where I should add it to the 1512 circuit- should it be after C1-3, where THAT recommends that phantom power should be inserted?

Nothing should be before C1-C3. The idea there is to be as close as physically possible to the input pins and the chassis connection.

Copy that. Thanks

[quote author=Meathands]Totally depends on the transformer. Zobel networks are added to dampen transformer ringing, with the R and C values being based on the particular transformer's resonant frequency. Same with a simple loading resistor. 600 Ohms is generally safe, which would be provided by the T-pad in either case.[/quote]
I see. how does one go about finding the resonant freq of the transformer?

One thought on the way you've got the lemon/lime switching configured. I wouldn't worry about doing "true bypass" the way you've got now. You can just split the single-ended output signal to the transformer and balancing board. In the worst case--THAT1646 in the signal path with output transformer "hanging off"--you've got a 600R load (assuming 1:1 transformer) in parallel with the THAT1646 input Z. Signal loss and loading will be negligible.

So in that situation, we'd be switching between the two dual-ended outputs of the board (or 1646) and transformer, correct?

[quote author=MeatHands]
I'd put the t-pad before the lemon/lime switch since there's no need for it in the "clean" circuit. But another thing to consider is, have you ever tried to overdrive an EA2503? It's not that easy to get much noticeable THD out of it at reasonable line levels! Maybe the T-pad works well in the API circuits because it makes the DOA work harder, I dunno haven't played around with one.
[/quote]
That does make sense, and reminds me of something I had been looking at before I settled with the 2503- there seems to be an abundance of ancient 1:1 UTC transformers and others of the same sort of era on the ol' eBay. Might it be a safe bet that these older transformers have "poorer" peformance specs, which may or may not translate into what I'm actually looking for in terms of distortion?

 

[Meathands]

I see. how does one go about finding the resonant freq of the transformer?

Most will give you recommended loads and/or Zobel networks in their datasheets. Otherwise, if it's something commonly used like the 2503, see what people do with it in context.

So in that situation, we'd be switching between the two dual-ended outputs of the board (or 1646) and transformer, correct?

Yep.

That does make sense, and reminds me of something I had been looking at before I settled with the 2503- there seems to be an abundance of ancient 1:1 UTC transformers and others of the same sort of era on the ol' eBay. Might it be a safe bet that these older transformers have "poorer" peformance specs, which may or may not translate into what I'm actually looking for in terms of distortion?

The 2503's performance specs are plenty bad, it's just designed to handle a whole lot of dBus before saturating. At least that was my experience with trying to get some harmonics out of it. In my experience of playing around with transformer saturation for Colour you need some DC accross the primaries or a gain stage with it's own harmonic mojo to get much warmth out of a big line output transformer. This is why I'm using a "consumer audio" output transformer for the transformer saturation Colour stage, the Cinemag CM-10100.

 

[Harpo]

-As you can see I wanted to have a switchable output balancing stage- active balancing or transformer balancing. Something of a lemon/lime flavor choicefor the end user, I suppose. However I've read that there may or may not be certain things that need to be done with output transformer- adding a single resistor to the secondary, adding a Zobel network to the secondary, etc- but nothing definitive. And as far as Zobel networks go, the only information I have seen for calculating the values relates directly to when the next stage is a speaker- not a +4 line input. Any help here?

Totally depends on the transformer. Zobel networks are added to dampen transformer ringing, with the R and C values being based on the particular transformer's resonant frequency. Same with a simple loading resistor. 600 Ohms is generally safe, which would be provided by the T-pad in either case.

One thought on the way you've got the lemon/lime switching configured. I wouldn't worry about doing "true bypass" the way you've got now. You can just split the single-ended output signal to the transformer and balancing board. In the worst case--THAT1646 in the signal path with output transformer "hanging off"--you've got a 600R load (assuming 1:1 transformer) in parallel with the THAT1646 input Z. Signal loss and loading will be negligible.

The minimum resistive load for the That151x is 2K, so it probably won't like a direct connection to a 600R reflecting impedance. Maybe have the That1646 always connected as balancing and/or transformer driving stage.

 

[Meathands]

-As you can see I wanted to have a switchable output balancing stage- active balancing or transformer balancing. Something of a lemon/lime flavor choicefor the end user, I suppose. However I've read that there may or may not be certain things that need to be done with output transformer- adding a single resistor to the secondary, adding a Zobel network to the secondary, etc- but nothing definitive. And as far as Zobel networks go, the only information I have seen for calculating the values relates directly to when the next stage is a speaker- not a +4 line input. Any help here?

Totally depends on the transformer. Zobel networks are added to dampen transformer ringing, with the R and C values being based on the particular transformer's resonant frequency. Same with a simple loading resistor. 600 Ohms is generally safe, which would be provided by the T-pad in either case.

One thought on the way you've got the lemon/lime switching configured. I wouldn't worry about doing "true bypass" the way you've got now. You can just split the single-ended output signal to the transformer and balancing board. In the worst case--THAT1646 in the signal path with output transformer "hanging off"--you've got a 600R load (assuming 1:1 transformer) in parallel with the THAT1646 input Z. Signal loss and loading will be negligible.

The minimum resistive load for the That151x is 2K, so it probably won't like a direct connection to a 600R reflecting impedance. Maybe have the That1646 always connected as balancing and/or transformer driving stage.

Good call, I didn't check the driving capability of the 1512.

 

[JoshuaUnitt]

-As you can see I wanted to have a switchable output balancing stage- active balancing or transformer balancing. Something of a lemon/lime flavor choicefor the end user, I suppose. However I've read that there may or may not be certain things that need to be done with output transformer- adding a single resistor to the secondary, adding a Zobel network to the secondary, etc- but nothing definitive. And as far as Zobel networks go, the only information I have seen for calculating the values relates directly to when the next stage is a speaker- not a +4 line input. Any help here?

Totally depends on the transformer. Zobel networks are added to dampen transformer ringing, with the R and C values being based on the particular transformer's resonant frequency. Same with a simple loading resistor. 600 Ohms is generally safe, which would be provided by the T-pad in either case.

One thought on the way you've got the lemon/lime switching configured. I wouldn't worry about doing "true bypass" the way you've got now. You can just split the single-ended output signal to the transformer and balancing board. In the worst case--THAT1646 in the signal path with output transformer "hanging off"--you've got a 600R load (assuming 1:1 transformer) in parallel with the THAT1646 input Z. Signal loss and loading will be negligible.

The minimum resistive load for the That151x is 2K, so it probably won't like a direct connection to a 600R reflecting impedance. Maybe have the That1646 always connected as balancing and/or transformer driving stage.

I like this idea as it seems to kill a couple birds with one stone.
Forgive my unfamiliarity with the 16x6... does it actually work as a variable gain stage? Or does it just function the way a transformer would? in other words, would it make the transformer redundant, or could I actually use it to variably drive the transformer?

 

[Meathands]

Forgive my unfamiliarity with the 16x6... does it actually work as a variable gain stage? Or does it just function the way a transformer would? in other words, would it make the transformer redundant, or could I actually use it to variably drive the transformer?

The former--it kind of behaves like a transformer. The 1646 has a 1:2 voltage ratio and the 1606 has a 1:1 ratio. But you've got your variable voltage gain, the 16x6 will provide current gain (and voltage gain with 1646).

 

[JoshuaUnitt]

Forgive my unfamiliarity with the 16x6... does it actually work as a variable gain stage? Or does it just function the way a transformer would? in other words, would it make the transformer redundant, or could I actually use it to variably drive the transformer?

The former--it kind of behaves like a transformer. The 1646 has a 1:2 voltage ratio and the 1606 has a 1:1 ratio. But you've got your variable voltage gain, the 16x6 will provide current gain (and voltage gain with 1646).

I'm not sure I follow... can't seem to find this on the datasheet. (sorry I feel like this is a silly thing to get lost over haha.)

 

[dfuruta]

in your sketch you omitted CG;  you probably will want either that or a dc servo, as you're likely to have substantial dc offset going into the transformer/output stage otherwise.

 

[Meathands]

I'm not sure I follow... can't seem to find this on the datasheet. (sorry I feel like this is a silly thing to get lost over haha.)

Oops, you're right, both 16x6 chips have a gain of 2. I was thinking of the 124x series of balanced input chips that have different gains.

What I meant was the 1512 is giving you your voltage gain, so the 1646 doesn't need to be adjustable.

 

[JoshuaUnitt]

in your sketch you omitted CG;  you probably will want either that or a dc servo, as you're likely to have substantial dc offset going into the transformer/output stage otherwise.

I appreciate you pointing this out- from what I've read, CG needs to be 6800uf to get no low-freq attenuation at the lowest RG setting, correct?
This does bring up another point that I forgot to mention initially...the actual value of RG. There's a lot in the datasheet I'm having trouble grasping, but as far as I can tell it does say that at 10k (in my case using a stepped discrete switch) I can go from 0-60, no problem. Would that be correct?

I'm not sure I follow... can't seem to find this on the datasheet. (sorry I feel like this is a silly thing to get lost over haha.)

Oops, you're right, both 16x6 chips have a gain of 2. I was thinking of the 124x series of balanced input chips that have different gains.

What I meant was the 1512 is giving you your voltage gain, so the 1646 doesn't need to be adjustable.

Ahh. Understood.

 

[Harpo]

This does bring up another point that I forgot to mention initially...the actual value of RG. There's a lot in the datasheet I'm having trouble grasping, but as far as I can tell it does say that at 10k (in my case using a stepped discrete switch) I can go from 0-60, no problem. Would that be correct?

RG practically consist of two series connected resistors. 5 ohms for the 60dB max.gain setting in series to a 10K (rev.log taper pot/rheostat) for the variable part between 0dB gain (rheostat dialed in to its max.resistance) and 60dB (rheostat at min.resistance/shorted out, only 5 ohms fixed parts value remaining).

 

[tubegeek]

C1, C2, C3: use them as shown on data sheet, close to input jack, they are there to shunt radio frequencies and higher, to ground. High quality ceramic caps are best there. Note that the two input caps you show on your block diagram are there to block the phantom power from hitting the chip and they correspond to C4 and C5 on the data sheet. The large value makes electrolytic caps the most likely choice there, and also for CG. Voltage rating should be >48V (because of the phantom power) for C4 and C5. 63V or 100V are your most common choices. These caps will be a bit larger physically than most other signal-level components, so watch out for that in your layout. If they are polarized types (They most likely will be) the + end connects to the input jack.

 

[ricardo]

The place for your 6dB pad is at the INPUT of the 1512.  Download THAT DN140 & their other datasheets & notes which have loadsa good stuff.

 

[Gustav]

I have nothing real to add, but I have built both a few 1 channels and an 8 channel versions with this layout, and they both perform really well in the real world. 1 channel versions are doing rounds in different studios with engineering friends who all end up asking for keepers. Very quiet and plenty of gain.

It is more or less the THAT application notes with some very minor changes, and you should be able to print for self etch with the document attached (I am not 100% sure I did the image output right, but if I didn't you can probably convert it!).

C21 can be set to high pass frequency of choice.

https://pcbgrinder.com/download/Pre_Grinder_doc.pdf

Gustav

Heres the 8 channel version. It is dead quiet, but just for good order, I will run some shielded cable (at least for the mic input) and take another peek at the grounding scheme to make sure its 100% before I call it completely done.

 

[ruairioflaherty]

Who's making all of these beautiful panels you've been showing us lately?

 

[JoshuaUnitt]

Incredibly thankful for all the pointers, folks!
I'm especially eager to dig into THAT's design note sheets. Looks a little thick but I'm stoked.

Gustav, your documenting the progress on your Pre Grinder via facebook was what got me researching the 1512 circuit in the first place! I definitely love the look of the your case for the 8-ch unit, very elegant.
I noticed C21 while looking over your schem, I'll definitely be including a filter myself now that I'm reminded of that.

Should be buying up a couple chips and start breadboarding soon, I'm sure plenty more questions will manifest themselves along the way.

 

[ruairioflaherty]

Not sure is it's been posted above but the 16XX output drivers from THAT want to be driven from a very low output impedance.

 

[Gustav]

Who's making all of these beautiful panels you've been showing us lately?

I made the design in Fronpaneldesigner and ordered direct for those prototypes. You can get the same much cheaper from Frank - even from the same file - I just made the mistake of not asking him if he could do powder coating first.

Goodspeed on the 1512, Joshua! You will be surprised at how good it sounds Im sure.

Gustav