Discrete opamps in the instrumentation config.

I know the instrumentation op-amp is the configuration of choice for things like the SSM2017, but how come we don't see this topology for discrete stuff as well? Wouldn't something like the JH990 be ideal for this?

This question was inspired by the George Massenberg interview someone was kind enough to post a link to recently...is this how he does it in his pre's?

Thanks,

Bjorn

Great to see that someone did listen to those interviews.
I'm still downloading one now ... Dave Hill.

George was great ... he doesn't hold back does he ? :green:
He is happy to tread on anyone's toes and ruffle all feathers.

I can make anything work
Transformers !! ... super high noise ... thousands of volts ground differential ...
DO NO HARM
Stay OUT of the way !!
An engineer is there to do a job and stay out of the way.
Unacceptable
... stay the $%^& out of the way.
He should pay the band !!

Click to expand...

:green: :green: :green:
fantastic
please listen to it
just fucking great stuff
:thumb:

Don't buy my stuff ... I don't care

Click to expand...

fantastic stuff !!!!


The instrumentation stuff ?
You know that application note at Jensen that is basically a servo 990 with an opamp back to input. Is this where he is ??
Back up one notch.
What makes an instrumentation amp ... an instrumentation amp.
definition ... know what I mean ?

Thanks Kev and Peter.

I have to say I almost started feeling like we've lost the plot building all these pre's with different colour and tone...still, I only really record my band and I can mess with it as I like! Definately the best interview out of the lot; certainly made me think.

The instrumentation op-amp configuration is as far as I know exactly what is used in the SSM2017; two 'input' op-amps that are normal non-inverting gain stages, but where the bottoms of the two feedback potential dividers are connected together rather than to ground. The third op-amp is just like a differential line receiver stage, and extracts only the differential signal. I suppose that any instrumentation amp is free to use the name and have a different topology, though. Where's that Mark Burnley chap...this is just the thing he'd dig into!

A few years ago I thought about doing such a preamp using 3 AD797's but it only got as far as a half-stuffed stripboard. If anyone has a few JH990's lying around, then it might be worth a try to see what it sounds like, or if they'll even work in that configuration. It might not be that stable in the differential receiver position as that is usually unity gain.

In the massenberg unit, I take it he still uses phantom blocking caps, or has he found some clever way of getting around that too?

Bjorn

The purpose of that 'instrumentation' amp configuration is to:
- provide (usually) a high input impedance on both input legs
- provide best in class CMRR
- provide the necessary gain
- provide low noise and distortion.

The first 2 are particular features of that 3-amp layout, and the 3rd is readily acheived. As usual, the 4th is a matter for the specific implementation....

If anyone has a few JH990's lying around, then it might be worth a try to see what it sounds like

Click to expand...

... IMHO, any high-quality amplifier (using the rather more conservative definition of high quality - ie, low noise, low distortion, no instabilities, flat frequency response to sensible limits, etc. etc.) will sound identical.

Where one amp has a different 'sound' to another one - one (or both) has some (usually very) measurable impairment (enhancement????) to the signal.

eg. You can readily measure the effect that transformers, inappropriate electrolytic caps etc., and poorly designed circuits (or at least, circuits designed with aims other than high fidelity) have on a signal. You may like what it does, you might not - but it is distortion...

<short_rant> You cannot, to my knowledge, yet measure how 'golden' someone's ears are, though </short_rant>

;~)

Alan

I looked at a few ways of dealing with the phantom blocking caps - and eliminating them - it can be done, I think, one of several ways. I haven't built any of them but here's what I came up with:

1. Using an instrumentation amp, the input impedance is quite high - which means you can use small caps, say, 0.1uF or 1uF. The load resistance goes on the phantom side of the caps. You can now pick whatever caps you want instead of large-valued electrolytics. This doesn't get rid of them but it does allow latitude in choosing a good-sounding cap.

2. Floating the instrumentation/preamp setup to +/- 15 volts around the phantom voltage, and use some kind of level shifter after the gain has been done. It'll need four power supplies - +48, (phantom+15), (phantom-15), +15, -15. You could use a high-voltage op-amp to servo the power rails to +15 and -15 around the phantom supply. Basically, the output amp needs to be an inverting configuration, and have two inputs - one is, say, a 10k resistor from the preamp stage, the other goes to the negative rail, and the resistance is adjusted to null the DC output voltage but it'll be roughly 3k - it's just a current source. This probably needs to be a Jensen-style servo to account for DC voltage shifts as the mic will load down the DC phantom supply.

3. Build a discrete diff amp that can power from +65 and -15 volt rails - and reject the DC phantom in the diff amp.

4. Phantom power the ground of the mic with -48 volts instead of +48. This is a bad thing though, if two mics touch, bam!

5. Put in coupling caps in, but do feedback around the coupling caps. The minimum gain is something like 20dB if you do this, as you are limited to the resistance values that'll work without doing something to the op-amp.

have nice look at this one:

http://www.forsselltech.com/JMP-1.htm

Three Fredd Forssell opamps arranged in a instrumentation configuration as a discreet, transformerless mic preamp!

Hi,

good topic Bjorn!!!

I'm looking at designing a instrumentation amp with OPA2604s for input buffers and an OPA604 ouput servo around a Forssell 992 at +-24V, for my monitor console.

Found a good document at Walt Jungs site last night about low distortion instrumentation amps......I'd post a link but I don't have time right now. Its easy to find - in PDF format.

I think he states that FET input opamps must be used to achieve good performance.

Kev, I listened to all those interviews and movies - I think Georges and Dave Hills were awesome! Especially Dave Hill talking about innovation in audio design.........

Cheers Tom

While it is true that instrumentation amps like the INA217 and SSM2019 have very high input impedances, for audio use, the inputs are shunted to ground with as small a resistance as possible to limit common mode noise pickup.

Click to expand...

It is true - you loose many of the benefits of that configuration in a mic amp - you simply don't need the high input impedance that you *might* need when driven from some whacky strain-gauge or fancy transducer (although the CMRR is useful).

A mic amp has other needs - not the least of which is a good noise performance when fed from a low impedance source (which is whay many mic amp designs are either entirely discrete, or have at least a couple of well-selected transistors right at the front end).

Regarding the eventual sound of the amp, it comes down to the quality of the circuit design (or the op-amp) in the end, rather than the configuration (although some configurations may be more suitable than others).

Three Fredd Forssell opamps arranged in a instrumentation configuration as a discreet, transformerless mic preamp!

Click to expand...

Actually... no. While the JMP-1 it is discrete and transformerless, it is not configured as a classic intrumentation amplifier. Its topology is very much the same as the design presented in a 1984 AES paper by Graeme J Cohen (AES preprint number 2106) but implemented using only JFETs and using a single-ended output topology. The Green preamp is similar to the Cohen, as are several other preamps being sold today as "neutral" or "accurate" mike preamps. The Green front-end seems to me to based on an implementation of Mr Cohen's orginal circuit but with some added circuitry to increase the common-mode input impedance (good idea to me). I think Rupert Neve used a similar approach in the TLA stuff sold by some company. Other implematations of Mr Cohen's circuit are less creative in their additions or modifications to Mr Cohen's circuit but do retain the inherent benefits of such a topology. To my knowledge, Mr Cohen is the first to present such a topology and I welcome an information available that might point to a similar topology dated prior to 1984. I'm just curious. I want to give credit where credit is due and with this topology, that is with Mr Cohen... as far as I know.

In his AES paper Mr Cohen goes into some detail regarding the advantages of his topology vs other conventional input stage topologies, including the typical instrumentation amplifier. It's pretty good reading and is available from the AES.

1. Using an instrumentation amp, the input impedance is quite high - which means you can use small caps, say, 0.1uF or 1uF. The load resistance goes on the phantom side of the caps. You can now pick whatever caps you want instead of large-valued electrolytics. This doesn't get rid of them but it does allow latitude in choosing a good-sounding cap.

Click to expand...

Be aware that using small caps as mentioned above needs to be carefully considered. The capacitive reactance of small value caps effectively increases the generator source impedance (Zs) and can have very significant effect on the noise performance of the front-end. Do the math. Figure out what your circuit's noise performance will likely be, especially if you use bipolar input devices. I use JFETs and I use fairly small cap sizes. Use whatever you think makes the most sense, but I recommend that you give it some serious thought so you know what the trade-offs are. To me, that's what this design stuff is all about.

[quote author="Bjorn Zetterlund"]Thanks Kev and Peter.

In the massenberg unit, I take it he still uses phantom blocking caps, or has he found some clever way of getting around that too?

Bjorn[/quote]

Hi Bjørn

Yes The GML Preamps are using Phantom blocking caps...Hmmm Thats the only thing where I would greatly disagree with GML...it says that the topology does not have any electrolytic capacitors in the signal chain hmmm I dont know..I mean..Phantom electrolytic blocking caps would be in the signal chain...soo I really dont understand that one myself..;-)

Kind regards

Peter

They may not be electros, Peter...Roger Foote (above) I think used some mighty big black polyprops or something in his.

I've still got my six Jfet 993's sitting in a box...think I might just have to give 'em a go in the in-amp configuration at some point...just for the hell of it!

Thanks for all the interesting posts!

Bjorn

:sam: :sam: :sam:

This thread is widening and I don't want to fuel that.
but
Thanks to Fred for his post.

I think it is OK to dis-agree with GM if it works for you. It is good to try to understand where he is coming from.
I personaly see no problem with having a coloured pre-amp and multiple Mic and pre-amp combos at your disposal.

There have been too many Producer/Eng/Band combinations that have made classic albums. The Band often site their fellow behind the glass as an added Member of the band.

What ever gets you there is ok by me. Lots of different guitars and lots of different amps and lots of different Mics etc etc .... if it sounds good ... do it and everyone else can go to hell.

If you can get a Grammy with One Guitar and One Mic and One GML then that's great too ....

These forums are great and these guys are free with there info.
Stop ... Look ... and Listen to what they have to say. Many of them are DIY'ers through and through. :green:

and then thank them for their time.

:shock:
shut up Kev
I am so full of it, at times :roll:

What AP and Fred said.

Small input caps can give the frequency response but will usually increase low-frequency noise. At least with bipolar transistors biased for low noise in 200 ohms. Tubes and FETs don't have the same problem. But it takes a yard of cathode to get a tube-amp's noise low compared to 200 ohms. Until recently it wasn't really a job for FETs, but FETs have improved and I'm sure Fred knows what he's doing. FET inputs have so many other advantages, maybe it is time for me to learn what hot FETs are available now.

Some of the things the classic 3-opamp instrumentation amp is good for are NOT relevant to mike-amps. The "infinite" input impedance is pointless with a 200 ohm source and now a pair of 6K resistors hung across every mike jack. It is rare to find a pre-made opamp that has low noise in 200 ohms, and if you find one it is too expensive to buy three or even two of them. Super-high CMRR is usually not really necessary or happening in mike lines: we don't run arc-welders while recording, and our mike cables are rarely so perfectly balanced that some 120dB CMRR spec can actually happen. 20dB to 40dB of CMRR is usually ample, and many times 0dB CMRR (unbalanced operation) is fine if your ground connections are really good. And in audio, we usually want to simplify the signal path. Some of use are cheap and won't pay for a lot of parts. Others want a very clean sound, and more stages means more complex and less-euphonic distortions. And the two separate input circuits of the 3-opamp IA usually mean more noise or more trouble minimizing noise.

There are simpler ways than IA to build a mike amp. Two discrete transistors can work much like the two input amps of the IA configuration, but with less noise than two 2-transistor-input opamps. The feedback input does NOT have to be high-impedance, it can go to an emitter. In fact when you do that, you get what is now fashionably called "Current Mode Feedback": closed-loop bandwidth stays fairly constant as gain is varied over a wide range (as we like to do in mike-amps). With classic voltage-mode feedback and fixed-compensated opamps, closed loop bandwidth drops as gain rises. If you demand 20KHz at 60dB gain, then you will get 2,000KHz or 2MHz response at 20dB gain, and pull in every AM radio station in town. The input impedance of the 2-transistor and an opamp mike amp, not counting the base bias resistors, is very high compared to mike impedance. I see over 30K on one such design I happen to be simulating. The base resistors want to be low to handle the fairly high base current... but if that were a problem there are ways around it.

The 3-amp IA configuration solves problems we don't have. It is still a perfectly valid concept for a mike amp, but compare carefully: it usually is not the best available concept.

It is a better plan for balanced line inputs. Way better than the standard 1-opamp "differential" input, which never works right. Lower noise than the most common 2-opamp plans. There is another way to do a 2-opamp diff-in which is hi-Z input and low noise, but I don't have a plan handy. Of course in practice the stupid 1-opamp "diff"-amp really does work fine most of the time. And none of these active inputs will handle many-volts of ground difference as well as a transformer, the gold standard of baluns.

[quote author="Bjorn Zetterlund"]They may not be electros, Peter...Bjorn :sam: :sam: :sam:[/quote]

Hi Bjørn,

Ohh yes they are..470uF ;-) I have owned GML pres for a couple of years..soo I have looked under the hood..;-)

Kind regards

Peter

Aah! You own one...lucky bugger! What brand does he go for? Does he do the poly in parallel trick?

And thanks for all the replies to this thread folks :thumb:

Bjorn

Hi Bjørn,

Yes I got four pres..the GML 8304 very nice unit..the "sound" of these baby´s..hmm you wont belive it..people used to the SSL..Neve ..focusrite (5543..chips).."standart" etc... preamps..have never imho heard a real transistor preamp..;-)

Uhh Can´t remember the cap brand..hmm..I´ll let you know next time I have the time to open the box..I´m in the middle of a session right know..dont think it would be the right time..*GGG*. And yes he uses some exsotic cap in parallel..can´t remember the type..looks like a teardrop to me..what are those..??? *S*..

Kind regards

Peter

I'd be interested to know how the GML 8304 compares with John Hardy's and Fred Foressell's pres.

Hey guys.....

I found this after a little searching on the web for some JFET transistors...looks like this preamp references the same AES design that Fred mentioned.

What do you think - seems to be a little similar to the instrumentation amp config???

http://www.proaudio2.republika.pl/pre_jet1.pdf

Any comments?

Cheers Tom

I would love to have a look at the article Fred pointed. Does anybody have a copy?

When I hear instrument amp, i just think of an opamp with a precisely fixed gain, x1, x10 x100, etc.