Recommendations For Medium Input Impedance Preamp

[ach]

Hi everyone -

I am looking to build a preamp specifically for presenting a 10k input impedance to a passive DI box. Obviously I know that most DIs are designed from the other vantage point (to work with a common ~1k mic preamp input impedance), but I have transformers that were spec'd to reflect 1M from 10k, and would like to provide the highest input impedance for the instrument side.

I am looking to keep the noise floor as low as possible, and will use a differential front-end.

My general rule of thumb has been to use BJTs for low impedance inputs and JFETs for high impedances, but I don't have a clear (scientific) intuition of where that line should be drawn. Perhaps this 10k impedance is in the grey area between the better shot noise characteristics of a JFET and the better thermal/voltage noise of the BJT?

I haven't found many discussions on 'medium' input impedances in this range, but I would really appreciate anyone's input on best practices, experiences, further reading, or reasoning for choosing an input transistor type for this purpose.

Thank you kindly - Andrew

 

[Bo Deadly]

There is a standard mic pre circuit that has been used for decades in all manner of devices. It's sometimes referred to as the Cohen circuit but THAT Corporation refined it into what they call CIFA. You can just buy a chip like THAT 1510 that is the whole shebang in one IC. Fortunately for you, what it does not include are the DC bias resistors which set the input impedance. Typically these are 1K. Just make then 10K and you're done. That's going to give you just about the best possible performance you could hope for at all gain levels. Although the 10K will generate more noise than the 1K of course. So if you really must minimize noise to pedantic levels, then use high quality metal film.

Although, I must say that if you're plugging your guitar into a transformer, then agonizing over every minutia of performance is misplaced given that a trivial JFET input buffer would outperform even the lowest noise preamp behind a transformer. Usually transformer DIs are used because there's one console with lots of mic inputs with 48V power. But nowadays there are lots of line inputs. If you're ultimate just going into a computer audio interface, I would recommend dropping all of the gobbledygook in between and just JFET buffer it straight into the puter.

 

[NOON]

I've done something very similar just using a THAT 1510 or 1512 chip with appropriate loading resistors at the front end. Works brilliantly, very simple and low noise.

 

[ach]

Typically these are 1K. Just make then 10K and you're done. That's going to give you just about the best possible performance you could hope for at all gain levels.

Thanks very kindly Bo - The THAT Corp datasheets/whitepapers are definitely one of my research sources. I appreciate the topology name, as I was just able to find this site (Graeme Cohen's Page) which seems to be an overview.

While I am certain that I won't approach theoretical limits for noise floors if designing a discrete build, I would still be interested in the questions regarding BJT/JFET. But as you said, perhaps it is foolish to think to hard about this level of noise considering the overall topology/design may be worse performing overall.

Although, I must say that if you're plugging your guitar into a transformer, then agonizing over every minutia of performance is misplaced given that a trivial JFET input buffer would outperform even the lowest noise preamp behind a transformer. Usually transformer DIs are used because there's one console with lots of mic inputs with 48V power. But nowadays there are lots of line inputs. If you're ultimate just going into a computer audio interface, I would recommend dropping all of the gobbledygook in between and just JFET buffer it straight into the puter.

That is absolutely a fair point, but In this case I am partial to the sound of this transformer from a curve and distortion perspective, but want to reduce noise where possible considering the non-optimal gain-staging. Especially considering later analog processing which might reduce signal to noise even further.

I've done something very similar just using a THAT 1510 or 1512 chip with appropriate loading resistors at the front end. Works brilliantly, very simple and low noise.

Thanks kindly for your experience with this, Noon.

 

[ach]

It's sometimes referred to as the Cohen circuit

Interestingly, from Cohen's paper I referenced above it states on page 10:

Why not FETs on the front end? Match of Vgs of depletion mode FETs is very poor, enhancement mode FETs are worse. They also have too much noise for low impedance sources. Need to use bipolar for low impedance work because of much better noise performance. Vbe is very tightly matched in multiple emitter dual transistor chips.

I wonder how this would hold up with the modern/recent super matched low noise JFETs like the JFE2140 or LSK389?

 

[NOON]

BJT vs FET for lowest noise is fairly well explained in Self's Small Signal Audio book. From memory there's some good graphs in that section.