[ACMP investiupgradifications] All things PREAMP

[clintrubber]

Has the topic of transistor types been brought up yet ?

The BC184 NPNs are there, but for the BC214 (PNP) the BC557 is substituted, which per datasheet has poorer noise performance. Might not matter that much, but since it has been reported that engaging the EQ increases the noisefloor one might be better safe than sorry.

Bye,

  Peter

I realized this is only relevant for the '81-preamps.

 

[Steve Hogan]

"I believe that a more practical choice for resistor replacements would be Roederstein MK-2 resistors. I wouldn't wholesale replace the resistors, though, just the ones that show signs of physical deterioration and the ones that must be changed to fix the gain switch pop, etc."

Thanks for the guidance on the resistor replacement. Anyone know a good source for these resistors. I can't seem to find them.

Tom B.

TAW Electronics in Burbank, CA stocks many, but not all Roederstein MK-2 Values. What they don't have in Roederstein they have in RGM which are decent imports, better than the ones on the Clone amps now.  They are also factory authorized Wima Distributors and have an excellent selection of FKP-2 polypropylene foil and film caps.

GAIN SWITCH MODIFICATION UPDATE

Well, I just finished working out the new resistor values for the gain switch. My proposed cure does involve cutting the offending trace between section 1 and section 2 on switch position 6.  That position will become the "OFF" position in the reworked attenuator.

The rework will involve replacement of almost every resistor in the attenuator.  A few are unchanged, one is moved, and the rest are all new values.  The original gain switch had 11 positions with 6 dB nominal gain steps.  My new version has 10 active gain steps and an OFF position which prevents the disastrous "pop" resulting from the connection of output to input of the first gain stage during one of the switch transitions.

My new proposed Nominal Gain steps from max gain are:
-0dB
-6.67
-13.33
-20.00
-26.67
OFF
-33.33
-40.0
-46.67
-53.33
-60.0

In the first five steps the input transformer is loaded directly by the first 30 dB gain stage.  The output of the first stage is attenuated by the first stage output pad and fed to the second stage with 28 dB gain.

In the "OFF" position the 240 Ohm resistor (1R51 or 8R23) at the bottom of the pad is moved to replace the original 12K resistor (1R45 or 8R17) This terminates the input of the second gain stage with 240 Ohms in the "OFF" position.

The last 5 gain steps load the mic input transformer directly with a series of 5 separate voltage dividers. The lower leg (resistor) of each divider is loaded by the rather low 4938 Ohm input Z of the 2nd stage.  This results in a lot of spreadsheet calculations to select pad values that simultaneously give the correct voltage loss and still hit the target load Z for the transformer secondary.  In the factory version of this preamp, (and also the Neve versions, for that matter), the values of the pad resistors result in the terminating load on the input transformer making a big jump, from about 2700 Ohms in the high gain steps, to about 4500 Ohms in the low gain steps.  This change in load will cause an almost x2 change in the input impedance of the preamp, loading down microphones much more in high gain than low gain settings.  It also makes it near impossible to properly select a proper damping network to tame the overshoot and ringing of the input transformer at all gain settings.  The lightly loaded low gain settings will have more extended HF response than high gain settings.

All this goes away with the values I have worked out for the gain switch.

The new load on the transformer averages 3633 Ohms at all gain steps, both low and high gain, with a minimum load of 3589 at -53.33 step and a maximum of 3662 at the -40 step.  Those two are sort of outliers, since almost all the other steps are very, very close to the average.  Although this load value is still lower than optimum for the mic input tranformer, there is nothing that can be done since the input impedance of the first gain stage is only 3700 Ohms, there is no way to lighten the load.  The consistent loading on the transformer will enable a good RC network value to be selected that will work for all gain settings.  Overall I consider it a good fix.  It does require changing out a bucketload of resistors, however.
I don't have all the required resistor values in stock, so I will test the new setup as soon as the new resistors arrive.
More Later . . .

 

[Paul G]

Thanks, Steve for your continued work on this project.  Very impressive!  I'm getting anxious to get started and can't wait to hear the new pre.  Do you have an approximate time when kits will start shipping?  Thanks again.

Paul ;D

 

[Steve Hogan]

Has the topic of transistor types been brought up yet ?

The BC184 NPNs are there, but for the BC214 (PNP) the BC557 is substituted, which per datasheet has poorer noise performance. Might not matter that much, but since it has been reported that engaging the EQ increases the noisefloor one might be better safe than sorry.

There's also of course the 2N3055-thing, but might be a topic of its own & personal preference
(Motorola, RCA, NOS, etc etc...)

After studying the data sheets on the BC214 and the BC557 et al, it appears that the BC557 is what Horowitz and Hill call a "jelly-bean" part.  General purpose transistor with wide specs.  The BC214 is at end of life status but still available. It has a 2 dB Max noise figure where the BC557 noise figure is 2 dB typical and 10 dB max.  Good design requires that one use the "max" figure when designing for worst case performance in production.  Depending on the BC 557 to be low noise is folly, even though in most cases, there won't be a problem.  That's what "typical" means.

The BC214 hFE@2mA is 140 to 400.
BC557 comes in graded hFE of A=110-220, B=200-450, C=420-800
In the same family as the BC557 is the BC560  which is graded for low noise -- NF = 1.2 dB typical, 2 dB max
What's more important is the frequency range of the specification extends from 30 Hz to 15,000 Hz (the MHz in the data sheet must be a typo -- this is not a microwave transistor).  The low frequency specified means that the 1/f noise should be pretty good.

To sum up, this means that the BC560, IMHO, would have been a better choice than the BC557.  The BC560 is available in both B and C grades.  The B grade is a close gain match to the (Neve) BC214, but the BC560 has better specs and would be a superior replacement for the BC214.  These transistors are cheap, but there are an awful lot of them in the '81 EQ PCBs.  I honestly don't know whether or not changing out those transistors will affect the level of the hiss (white noise) of the EQ or not, since the EQ operates at nominal line level and the buffer amplifiers are all essentially unity gain compound voltage followers.  Some of the BC214's were used as input amplifiers and others were used as current sources, in which the noise may have even less relevence.
Careful control of the hFE may be important to prevent instability, which makes the pre-graded gains of the BC560 attractive.

I count 12 pcs of BC557 being used in place of the BC214 in the ACMP 81.

I suppose that I can add it to the list of potential mods.  For the DIY person ordering parts kits, the cost will be negligible, maybe $2 or $3.  But the labor time keeps rising for those who are having me do the mods, and that causes me concern.

I haven't done any work on the output stage yet -- still working my way thru the block diagram, so I have no input at this time on 2N3055 transistor types and their effect on performance.

 

[Steve Hogan]

Thanks, Steve for your continued work on this project.  Very impressive!  I'm getting anxious to get started and can't wait to hear the new pre.  Do you have an approximate time when kits will start shipping?  Thanks again.

Paul ;D

If I can quit finding and fixing problems with these preamps so I can finalize the mods, I will be ready to start shipping kits and fixing units.  When I started this whole thing, my friend told me that there were these group buy preamps that had hum and maybe you could provide a new power transformer . . .  The list of changes has gotten really long now.  I will have tranformers next week, and I hope to have everything finalized by the end of the month, unless the inductors are still a problem.

BTW, I intend to replace the 22uF output caps in both preamp amplifiers with 100uF.  The loading gets down to less than 4K on the first stage and less than 5K on the second stage when driving the output direct (EQ out).  22uF is just way too small to preserve the LF phase response with such low impedance loads and it becomes a LF bottleneck.  Normally I would recommend 220uF for loads in that range, but I am wanting to preserve the Neve-like British character which includes caps that are too small.  100uF is a good compromise.

 

[tommypiper]

Steve, thanks.  The more info the better for those of us doing DIY fixes.  Anything that comes to mind will be helpful for some of us.  The clients sending you units could have a range or a couple options whether to replace everything, or just the critical fixes, etc. Thanks for sharing your findings.

Since you're going over them with a fine toothed comb...  (BTW, I definitely understand this expression now, after my daughter's friends got head lice at school, a common problem...yikes     )  Any comments about the values and quality of the electrolytics?  And likewise for the film/whatever caps in the EQ circuits? 

The caps are doubtless mediocre quality... how important do you think those are for imparting sonic personality?  In my own Neve 1272 experiments, I found changing all the electrolytics significantly changed the character.  (I also replaced all the tants with electrolytics, probly the biggest influence.)

And... it's interested to see you writing that the Neve circuits have too-small caps which restrict the low end.  As they are famous for a robust low end.  All part of the sonic balance I suppose?

 

[okgb]

i bought two 73's & two 81's
my plan  [ so far ]
is to make the 73's as neve as possible [ barring spending
as much money as making one ]
and make the 81 as good as possible

 

[Steve Hogan]

. . .
Any comments about the values and quality of the electrolytics?  And likewise for the film/whatever caps in the EQ circuits? 

The caps are doubtless mediocre quality... how important do you think those are for imparting sonic personality?  In my own Neve 1272 experiments, I found changing all the electrolytics significantly changed the character.  (I also replaced all the tants with electrolytics, probably the biggest influence.)

And... it's interested to see you writing that the Neve circuits have too-small caps which restrict the low end.  As they are famous for a robust low end.  All part of the sonic balance I suppose?

Let me qualify my comments below with the following general observation:
The units we are improving are very low cost.  These preamps are not High-end, high-cost items.  That being said, most of the parts are of higher quality than one would expect for the money spent.  It makes no sense, IMHO to replace every resistor with PRP9372 resistors, or Caddock or Vishay bulk metal film, or replace the decent mylar caps with ill-fitting polypropylene (many of the caps are superb Wima FKP-2 polypropylene film/foil 2.5%) The build quality of the switches and other mechanical parts limit how far one should go and how much one should spend to be reasonable and not be "polishing a turd" to diamond brilliance.

I believe that these preamps with appropriate, reasonable modifications can perform at a very satisfactory level.  They will end up costing more than twice what you paid, but even twice what you paid is a good deal for the performance they are capable of when "hot rodded".

So let's start with the easy question first.  It seems that the film caps are fine. They are of sufficient quality to not require replacement except for a DIY type who wants to really push the limits on his own dime and time.

The supplied resistors appear to be metal film instead of carbon film, and the mods that I propose change values, and therefore there will be an automatic quality upgrade for many of the critical resistors -- like those in the input attenuators.

Electrolytic caps.  The supplied Rubycons are 105C and decent.  Not top of the line, but not junk.  All electrolytic caps have non-linear capacitive reactance.  All things being equal, the larger the capacitor value, the lower the capacitive reactance (impedance) at a given frequency, and the less influence (distortion) it will create in a circuit.

At 20 Hz a 470uF cap is 16.9 Ohms, a 100uF cap is 79.6 Ohms, a 47uF cap is 169 Ohms, a 22uF cap is 361 Ohms, and a 10uF cap is 796 Ohms.

The load on the first stage of the mic preamp is about 3100 Ohms.  When fed by a 22uF cap the non-linearity at 20 Hz is 361/3461 or  roughly -20 dB.  A 100uF cap is 80/3180 or about -32dB.  The lower the capacitive reactance of a coupling cap is compared to the resistive impedances in the circuit, the less damage to the signal will be done by whatever non-lineariities come from the crappy electrolytic cap.  Tantalum caps are extremely non-linear if not operated with a DC bias, so they should never be used as coupling caps on circuits with bi-polar supplies unless bias circuitry is added.  At that point, a non-polar electrolytic is much simpler and sounds much better.

I understand that the original Neve preamps used Tantalum caps. These have a limited practical size/voltage value to keep the cost down to reasonable.  22uF/35V is a big tantalum.  A 220uF/35V electrolytic is pretty average and low cost.  It is my opinion that making the coupling caps bigger has much more influence in making a piece of audio gear sound good, than the particular brand of electrolytic.  If you make them big enough, their impedance becomes so low that they only represent a very small portion of the signal so they add little distortion.  In circuits without DC bias, I use non-polar electrolytics (usually Panasonic SU series) caps because they have less distortion than unbiased polar caps.

In the ACMP preamps at hand, most of the caps are OK.  I would just replace the ones that are bottlenecking the LF with bigger ones.  The LF response of any piece of audio gear can be tested easily with low frequency square waves.
100 Hz should have very little tilt, 20 Hz maybe 10%.  Many audio devices are way worse. 

I am curious, when you replaced the tantalum caps with electrolytics, did you keep the same values?  What differences did you hear?

 

[tommypiper]

I am curious, when you replaced the tantalum caps with electrolytics, did you keep the same values?  What differences did you hear?

Thanks for the thorough explanations, Steve. 

On the BA283 card, I swapped these tants out and replaced with electrolytics: increasing C1 and C8 from 10uf to 22uf, and C14 and C15 I kept at 22uf on one card and increased them to 33uf on another, depending what I had.  Both sounded clearer, more open, particularly the top end.  I used the polarized cheaper Black Gates on one card and Panasonic FCs on other.  I also increased off-board caps C5 from 680 to 4700uf and C3 from 470 to 1000uf.  Most of the remaining electrolytics I kept at original values or upped depending what I had.  I find they sound wonderful.

 

[crazydoc]

ACMP73 gain switch position 6 mod:

Crazydoc,

Have you thought of using a low voltage photo optical sensor switch with flag arm paired with a miniature relay, used in many laser printers. Takes the mechanical idea to the next level. no parts come into physical contact.

Here are the results of my latest turd-polishing (more like grinding than polishing in my case.)

Parts were a slot interrupter, reed relay, resistor, scrap of drywall corner bead, scrap of perfboard, a few machine screw and nuts, and a piece of cardboard - less than $4. And I borrowed about 30ma of current from the +12v supply.

Again, the hard part was drilling and tapping the switch shaft. This time I broke the bit off inside the shaft, and had to drill from the side to get it out.

 

[crazydoc]

Let me qualify my comments below with the following general observation:
...The build quality of the switches and other mechanical parts limit how far one should go and how much one should spend to be reasonable and not be "polishing a turd" to diamond brilliance...

Steve

We all greatly appreciate the time and trouble you've gone to to improve these, and the helpful and instructive explanations you've provided. It's been a great learning experience.

However, as you've alluded, I think you may be starting to get into the area of diminishing returns on improvements for time spent, particularly from a fiscal standpoint. I hate to see you investing much more time and effort into these. With your current improvements we should have fairly robust, long lasting units with pleasing sonic characteristics.

The question still remains as to the influence of the new power transformer on the inductor hum. After the eq transistor oscillation, this to me was the most offensive characteristic of these preamps, making the eq section unusable at low signal levels. I guess you haven't gotten to the point of testing this yet, but I look forward (with hope) to hearing your findings.

 

[Steve Hogan]

ACMP73 gain switch position 6 mod:

Here are the results of my latest turd-polishing (more like grinding than polishing in my case.)

Parts were a slot interrupter, reed relay, resistor, scrap of drywall corner bead, scrap of perfboard, a few machine screw and nuts, and a piece of cardboard - less than $4. And I borrowed about 30ma of current from the +12v supply.

Again, the hard part was drilling and tapping the switch shaft. This time I broke the bit off inside the shaft, and had to drill from the side to get it out.

Peter,

Your ingenuity continues to impress me. I salute your solution. Too bad it's not very easily implemented in quantity.

Steve

 

[Steve Hogan]

Steve

We all greatly appreciate the time and trouble you've gone to to improve these, and the helpful and instructive explanations you've provided. It's been a great learning experience.

However, as you've alluded, I think you may be starting to get into the area of diminishing returns on improvements for time spent, particularly from a fiscal standpoint. I hate to see you investing much more time and effort into these. With your current improvements we should have fairly robust, long lasting units with pleasing sonic characteristics.

The question still remains as to the influence of the new power transformer on the inductor hum. After the eq transistor oscillation, this to me was the most offensive characteristic of these preamps, making the eq section unusable at low signal levels. I guess you haven't gotten to the point of testing this yet, but I look forward (with hope) to hearing your findings.

One of my regular clients has compared me to a bulldog in the way I am tenacious in finding the best practical solutions to optimizing audio gear. Even when the hum is fixed, the white noise level may prove to be higher than necessary with a few well-placed better transistors and reworked, lower impedance pads.  I am seriously considering adding the 12 BC560 transistors to the '81 list of mods.  In for a penny in for a pound. I already have to remove and replace the cards anyway.

I have spent some time with LTSpice and some measurements looking at the possibility of raising the input impedance of the first stage, and by lowering the gain 1 dB on that stage (105 Ohms instead of 91) the input Z goes from 3700 to 3950 Ohms.  The gain can be adjusted up on the 2nd stage by lowering the 120 Ohm gain set resistor to 105 as well.  Overall gain stays the same, but it is distributed 29dB+ 29dB instead of 30 dB + 28 dB.

It is important to realize that these gain resistors (and thus the amplifier gains) are switched in the original Neve front end, so the ACMP clones are simplifying the switching and the gain structure.  Unless one starts all over again with the 22 position gain switch of a real Neve, which is not a practical option here, I believe that by carefully reworking and balancing the gains of the first and second stage, I may be able to get better performance and still maintain the character of the original.   The three transistor Gain stage used in the preamp stage has multiple nested feedback loops, both AC and DC coupled.  I have modeled the frequency response curves as I have played with various capacitor values and I can see how they interact in a synergistic way.  At the present time, I have finalized making the output caps 120uF/25V which is the biggest cap that will fit in the original space.  This appears to have only positive benefits.  Changing some other cap values appears to not be such a good idea, due to the complex interaction of the feedback loops making bumps in the frequency and phase response, especially at the low end, where the capacitive reactance comes into play.

By the way, I keep finding errors in the schematics. A careful comparison of the 73 and 84 schematics show differences that are not real in the output section of the '84 preamp. They both use the same card and values. The '84 schematic doesn't accurately reflect how the clip light circuit is fed thru 2C30.  Note the duplication of component designations on the '84 schematic. I haven't figured that schematic out, completely, but it's wrong with regards to the preamp card.  I'm up to at least a dozen errors so far, including the fact that the schematic shows 0.1 uF bypass caps on the TL072 opamp on the preamp cards, and they are not there in any of the preamp cards.

For those who wish to disable that particular clip light, just remove the coupling cap 1C24 on the 81 and 2C30 on the 73/84 preamp card.  This simply prevents the audio from reaching the clip light circuit which happily monitors nothing.
For those who wish to keep the clip light, add two 0.1uF caps to the back of the card in the appropriate locations (I will have pics later), Replace the 10K resistor (1R62 in the 81, 2R11 in the 73/84) with 100K to lighten the load on the second stage output, which already has 51K and 5K in parallel.
 
Back to the low first preamp stage input Z.  In addition to the overall gain changing the input z, it appears that the biggest change is caused by the value of one of the internal feedback loop resistors.  15K in the first stage (1R14) and 18K in the second stage (2R14 in 73/84; 1R27 in '81).  This resistor feeds an out of phase signal into the input of the amplifier, thus providing negative feedback to the amplifier. Changing this resistor doesn't change the overall gain, just an internal feedback path.  Because it partially "cancels out" the signal by adding an opposite polarity replica to the incoming signal, it simultaneously lowers the input Z by a lot.  The difference between 18K and 15K is over 1K difference in input Z between the first and second stage with the same gain!  I'm still trying to find the best balance of values here.

In my reworked low-gain mic pads I have attempted to scale the resistor values down to the minimum necessary to simultaneously load the  mic tranformer with the right load and output the right voltage to the second preamp stage.  This means less noise generated by the pads.  It might even make sense to use a lower noise transistor in place of the first BC184C in order to make the preamps as low noise as possible without changing their sonic character.  Something like an MPSA18 might work even better than the BC184C in that first transistor position.

And last but not least, in answer to your original query, I believe I will have some production transformers later this week, so I can give better answers to what needs to be done to the inductors with regards to shielding, etc. soon.

More Later . . .

 

[crazydoc]

Peter,

Your ingenuity continues to impress me. I salute your solution. Too bad it's not very easily implemented in quantity.

Steve

Thanks, Steve. This would be pretty easy to implement, if it weren't for having to couple to the switch rotation. Too bad there's not a glue I could use with confidence to fix the head of a machine screw or other extender to the end of the switch shaft.

Back to the low first preamp stage input Z.  In addition to the overall gain changing the input z, it appears that the biggest change is caused by the value of one of the internal feedback loop resistors.  15K in the first stage (1R14) and 18K in the second stage (2R14 in 73/84; 1R27 in '81).  This resistor feeds an out of phase signal into the input of the amplifier, thus providing negative feedback to the amplifier. Changing this resistor doesn't change the overall gain, just an internal feedback path.  Because it partially "cancels out" the signal by adding an opposite polarity replica to the incoming signal, it simultaneously lowers the input Z by a lot.  The difference between 18K and 15K is over 1K difference in input Z between the first and second stage with the same gain!  I'm still trying to find the best balance of values here.

In my reworked low-gain mic pads I have attempted to scale the resistor values down to the minimum necessary to simultaneously load the  mic tranformer with the right load and output the right voltage to the second preamp stage.  This means less noise generated by the pads.  It might even make sense to use a lower noise transistor in place of the first BC184C in order to make the preamps as low noise as possible without changing their sonic character.  Something like an MPSA18 might work even better than the BC184C in that first transistor position.

I have not been happy with the noise in the preamp at high gain. I bought 100 each of the MPSA18, BC550 and more BC184C's, thinking I could hand select lowest noise transistors for the first preamp stage. I tried 3 or 4 of each, without any notable difference when listening, so I gave up on it. Do you think this or any other mod might be worth pursuing to lower the noise?

 

[Steve Hogan]

I have not been happy with the noise in the preamp at high gain. I bought 100 each of the MPSA18, BC550 and more BC184C's, thinking I could hand select lowest noise transistors for the first preamp stage. I tried 3 or 4 of each, without any notable difference when listening, so I gave up on it. Do you think this or any other mod might be worth pursuing to lower the noise?

You are good! You may have saved me some trouble if you already tried the MPSA18 -- that's my favorite low-noise, high gain transistor and I have used it successfully to seriously improve German mic preamps especially.  It is unlikely that more selection from the units you had would give you lower noise.

Microphone preamp noise. . .  that's a big subject.  If you don't want to read technical, I will summarize that the Neve gain structure and consequently the clone ACMP preamps are not optimized for lowest noise, and there really isn't a lot that can be done about it. In the following tech discussion I will try to be accurate but not mathematical, so forgive me if I simplify a bit to try to give the general idea.

The goal is to amplify a 50 to 200 Ohm microphone with as little added noise as possible. I will use 150 Ohms in the following discussion.

There is an inherent noise generated by any given resistance that is a function of its value and absolute temperature.
Low-noise (like wirewound) resistors generate a noise very close to theoretical. Noisy resistors have "excess noise" which adds to the theoretical.

E.I.N. (Equivalent Input Noise) is usually how mic preamps are specified because of their adjustable gain. It allows one to compare different preamps with different gains by giving a number that represents the equivalent noise at the input of the preamp that will increase by the gain of the preamp.

One must always specify the bandwidth of the noise or one cannot make meaningful comparisons. A preamp that is 3 dB down (signal wise) at 20 kHz might measure better than my 990 Jensen Twin Servo mic preamp which has just under 200 kHz bandwidth.  There is way more noise power between 20 kHz and 200 kHz than in the entire 0-20kHz.
to make meaningful audio band noise measurements an accurate brickwall 20 kHz low-pass filter should be used or the measurements are bogus.  The 990 will measure way worse, and yet be much quieter to the ear without the 20 kHz filter.

If a preamp has -132 dBu 20 kHz equivalent input noise, then the measured noise level at the output of the preamp is the equivalent input noise + the gain of the preamp.  If I have 60 dB of gain I would measure -132 dBu E.I.N.+ 60 dB gain = -72 dBu measured noise at the output of the preamp.  In reality one measures it backwards.  Measure the noise and subtract the gain to get the E.I.N. By the way, -132 dBu E.I.N. is spectacularly good -- almost theoretical from a 150 Ohm source.  When measuring noise in a mic preamp one must always resistively terminate the input with the same impedance as the microphone.  I use a 150 Ohm resistor between pin 2 and 3 to measure noise.  Shorting the input is cheating and will give an unrealistically low number.  Unterminated inputs can give you very audible hiss -- as much as 10 or 20 dB worse noise than a terminated input because the resulting high impedance at the amplifier input is far away from the impedance "sweet spot" that gives the amplifier its lowest NF.

Every amplifier has a Noise Figure (NF). This is how much noise the amplifier adds to the signal it is amplifying compared to the signal's own noise theoretically made bigger with no added noise.  The Noise Figure numbers are in the transistor data sheets we discussed a few posts back.  For example, let's say that the equivalent input noise of a 150 Ohm microphone is -132dBu.  If we amplify the microphone 20 dB with a noiseless theoretical amplifier, the signal from the microphone increases 20 dB and the noise goes up 20 dB as well.  No noise is added by the amplifier circuitry, so the signal to noise ratio remains unchanged.  However, let's say we use a mic preamp that has noise figure of 3dB (that's really good, by the way).  The signal goes up 20 dB but the noise goes up 20 dB+3 dB = 23 dB due to the Noise Figure of the amplifier. The signal to noise ratio got worse by 3 dB in this preamp.

Increasing Noise Figure in mic preamps comes from two main sources, Voltage losses and electronic noise.  The DC resistances in the microphone transformer are losses.  The signal doesn't go up as much as it would if there were no losses, but the impedance does go up, which makes more noise.

All amplifiers, whether discrete or IC opamps, have a noise figure that is related to the impedance that the amplifier is working with.  There are two kinds of noise generated by active circuits.  Voltage noise is the noise the circuit makes at very low impedances (shorted input). Current noise is the result of noisy currents being pulled through high value resistances and is the noise that predominates in high impedance circuits.  "Low noise" opamps have generally low voltage noise, so they may be very noisy in a high impedance circuit.  You would not want to use an AD797 super low noise opamp, for example, to DI interface to a high impedance passive guitar pickup, because it would be horribly noisy.

A 990 or an AD797 or other sub 1 nanovolt/rootHz voltage noise amplifier will have a spectacularly low noise figure when used in low impedance circuits, but have terrible noise when used in high impedance circuits.  Every transistor and every amplifier circuit has an impedance "sweet spot" where the transistor or amplifier contributes the least amount of noise (has the lowest NF).  The trick to making a low noise mic preamp is to make that 150 Ohm microphone look like the "sweet spot" lowest noise impedance to the amplifer.  To do that, one must either design an amplifier with a 150 Ohm "sweet spot", or use a transformer to step up the 50 to 200 Ohm microphone to the higher impedance "sweet spot" of the amplifier at hand.

The low ratio transformer used (1:2) or (1:4 in high gain mode) makes a 150 Ohm microphone look like 600 Ohms or (2400 Ohms) to the preamp input.  An optimum amplifier for a 1:2 ratio transformer would need to be very low voltage noise -- like a 990 or AD797 in order to have a decent noise figure.  The Neve transistor amplifier will have lowest NF at a much higher impedance.  A higher ratio transformer would have been a better noise match to the amplifier, all other things being equal.  (They weren't equal, though, because the bandwidth was better on the low ratio transformer so noise was traded away for other sonic benefits).

The heavy loading on the transformer secondary also is bad for noise.  For best signal to noise ratio, a microphone should be lightly loaded with a bridging impedance.  150 Ohm mics should be loaded with about 1500 Ohms. This is because we want maximum voltage output from the microphone not maximum power transfer which one obtains from a terminated source.  There is a 3 dB noise improvement between terminating a source (150 Ohm source/150 Ohm Load) and bridging the load.  Terminating the source drops both the noise and the signal 6dB, unterminated, the signal goes up 6 dB but the noise only increases 3 dB.  Rule of thumb, don't terminate microphones if you are trying for lowest noise.

The heavily terminated transformer loads down the microphone so the result is less signal at the secondary of the tranformer than the impedance would dictate if there were no losses.  Noise goes up with the impedance which includes resistive losses, but the signal doesn't go up as much, so there is a loss of signal to noise ratio and an increase in NF for the preamp (which includes the transformer).  By the way, Jensen transformer data sheets have the Noise Figure of the tranformer included.  This is a measure of the the extra noise that results due to resistive losses in the transformer, compared to a theoretical lossless transformer with Zero Ohms resistance in its windings.  Because of the resistances in the windings of a real transformr, the output impedance of the transformer goes up faster than the signal compared to a theoretically perfect transformer.

Transformerless designs btw are really bad because the noise generated by the feedback resistor network is part of the noise contribution.  That means that for the feedback network to not contribute significant noise, it must be much smaller than 150 Ohms.  The amplifier must be capable of driving a sub 150 Ohm feedback network.   Real transformerless mic preamps usually only have low equivalent input noise at maximum gain, because that's the only gain setting where the feedback resistor network has low impedance. Their E.I.N. gets really bad at low gains, but in that situation the microphone output is really hot, so it doesn't matter so much.

Because of the transformer's resistive losses, and because we do not want to load down the microphone, it will advantage the noise to make the transformer secondary load as high as possible.  The Jensen JT-16-A transformer (which was originally developed as a Neve replacement) is a 1:2 ratio transformer and its optimum load is 6.19K. That's one of the reasons to work on raising the input Z of the first stage amplifier in the ACMP preamps.

Because the first stage has a fixed gain of 30 dB, whatever noise it adds to the signal will be there, and the signal to noise ratio will decline permanently as it goes through the signal path.  That's why it's usually a lower noise approach to make the amplifiers variable gain instead of fixed gain.  The Neve preamps don't lend themselves to that.

I know that Avedis has done remarkable work in his "Neve" colored preamp.  He has reworked the gain structure significantly to fix both the microphone loading problem, the transformer loading problem, and the noise problem.
None of the clones do that.  They just copied what Neve did, both the good and the bad.

Well this post is already ridiculously long.  Hope it is helpful to clarify why designing really good audio gear is more complicated than it may first appear.

 

[clintrubber]

Has the topic of transistor types been brought up yet ?

The BC184 NPNs are there, but for the BC214 (PNP) the BC557 is substituted, which per datasheet has poorer noise performance. Might not matter that much, but since it has been reported that engaging the EQ increases the noisefloor one might be better safe than sorry.

There's also of course the 2N3055-thing, but might be a topic of its own & personal preference
(Motorola, RCA, NOS, etc etc...)

After studying the data sheets on the BC214 and the BC557 et al, it appears that the BC557 is what Horowitz and Hill call a "jelly-bean" part.

Hi Steve,

Thanks for your response.  I fully agree with the assumption that these BJTs shouldn't be that relevant anymore since they're after the 'real' preamp, but since people have reported (white) noise increase by engaging the EQ, I got the thought that it might be possible to improve upon that a bit by swapping the BC557-PNPs for better ones.

But who knows, maybe on original Neve units it might be the same behaviour, despite the original types. It'd surprise me though, since I just like you would like to keep living under the assumption that in an amplifier the first pre-amp stage makes lots of gain & sets noise figure and the next stages must then mess things up very badly if they want to noticably spoil that noise-performance.


For circuits where noise might be an issue and BC547/557 etc are specified I always switch to using BC550/560, they're dirt cheap & while it may not be noticable in the end still better be safe than sorry.

Or put another way, and while being too lazy now to get the datasheets out, I can't easily imagine a situation where '550/'560 would do worse in any respect than '546/'547/...'557/'558 (given we don't need for instance the higher voltage rating of the '546 etc)

Said yet another way, I figure/assume '550/'560 are simply the low-noise selection from the "jelly-bean" production, just like '109 was for '107/'108.


So yep, using the BC560 (with the correct hFE-range) for these preamps has crossed my mind, but I'm 'afraid' other reasons have mingled     Likewise for the 'power-devices' in the amp-stages: most likely BD139/140 (with proper hFE range) could have been used as well but 'cos the original types could be had we went for those.

Likewise for the BC184 but the Chinese people have already used those. Obviously '550 would work fine there as well etc.

As far as I understood the 're-creations from for instance Joe Malone & Great River don't use the 'original types' but without any noticable effect (IIRC the latter uses MPSA-something).
But since most DIY-choices will often be based on a combination of (1) technical facts & (2) some rational or non-rational preferences despite knowing otherwise  & also a bit on (3) 'just because', one might end up with using '214 i.s.o. a better type.
I realize of course that the three mentioned reasons above are less suited for an upgrade-kit, there the facts are to be weighted the most.

[quote author=Steve Hogan]I count 12 pcs of BC557 being used in place of the BC214 in the ACMP 81.

I suppose that I can add it to the list of potential mods.  For the DIY person ordering parts kits, the cost will be negligible, maybe $2 or $3.  But the labor time keeps rising for those who are having me do the mods, and that causes me concern.[/quote]

Yep, 12. Only changing them here 'because we can' and those boards need to get out anyway. Can imagine it's less attractive for 'ordered mods'. It could perhaps be economized by figuring out which swaps are relevant and which not, but that would require some more checking, but then on a single unit (or by sim). 

I haven't done any work on the output stage yet -- still working my way thru the block diagram, so I have no input at this time on 2N3055 transistor types and their effect on performance.

This choice might probably be one that mainly goes per (2) & (3) of the tastes of rationale mentioned above. But on a good day, (2) might be based on listening tests (let's all promise we do these then under condition of an adjusted bias-trimmer  8) - otherwise a proper comparison will become pretty wild).

Finding reasons for a '3055-choice based on (1) seem difficult, anybody knows a few ? 

Best regards,

  Peter

 

[crazydoc]

Steve

Thanks for the great post on noise, though I'm going to have to chew on it for awhile (even without the math.)

Though it doesn't look like much can be done for the noise as the circuit stands, do you think it would be possible to build an alternative low noise first stage of the preamp to substitute for the original? It looks like it would be easy to break into the traces at C1 and C3 to insert a different preamp stage.

I've been thinking of doing this for a while now, possibly using an IC op amp such as the TI OPA211.

http://focus.ti.com/lit/ds/symlink/opa211a.pdf

I'm sure there are impedance/noise considerations I'm not aware of, as you outlined in your post above.

 

[Rader Ranch]

Well, I recieved 4 84s instead of 81s and I've been trying to get him to send me the right pres since January. He can't even pull that off...I realize that this is somewhat of a blessing in disguise, as my '84s are at least somewhat useable out of the box.  But i would love that extra eq on the 81s. Anyone wanna trade?

Don't know if the messages here are working for me, but I'd be into this if you haven't had a taker yet. I only have 2 81's though, not 4.

 

[Tasa]

Great job all!!!

It has been hard work on these KungFu ChNeves!
I just hope that the end will result in something usable.
I feel that all is going that way....

BIG THANX to all of you....so many smart people working on this project.
I am a bit embarrassed cause my electronics knowledge is not zero, lets say 1.5 (out of 10), compared to all of you.
I will share my experiance in hope we can all get this to where it should be.
I am sure that I will learn so much from all of you.

I have 2 81's (bought from one of the members), on first look and some testing....one is with bad (wrinkled) resistors, like some of you have and some more, bent IC (6U2) pins on bass pcb, big pop on switch, a lot of noise, hiss and hum with or without EQ (toroid rotation does not help at all). Some Q switches and phase are hard to press out (from what I can see, too much paint in holes and also hole angle might be a problem).This one will stay off until fixed.
The other one has all good resistors, a very small pop (its more like small click than pop) on the gain switch and it has very little noise, hiss and hum. No problems with Q and phase switch. Not bad at all compared to bad one.
What is wierd about "good" one is that level is too hot on Mic.
Bass(NS Stick-active) - Radial J48 MK2 DI(pad in) - MIC IN(+48V ON) - out to line in RME FF400(unity gain)....
81 Input Gain on 6 o'clock, output gain on 8 o'clock and it clips and distorts with or w/o EQ!! My mistake? Impedance mismatch?? or an 81 issue???
When set not to distort (lower In and Out gain)....sounds "nice" as it can be for now at that level...EQ is usable on this one with some minor problems...
...same connection but through Focusrite Red Pre...all works good, nice levels.

I did test Line in....and levels are way low than they should...
I will test more and give some more details with photos of what looks like "very good" QC.

Maybe I missed it, but can anybody tell me info about current consumption per winding for ACMP81?

For those looking for:
AA144
http://www.banzaieffects.com/AA144-pr-16438.html
BAX13
http://www.cricklewoodelectronics.com/Cricklewood/product.php?printable=Y&productid=17260&cat=275&page=6&js=n

Any US source?

Thanx for all help.

 

[rodabod]

What is wierd about "good" one is that level is too hot on Mic.
Bass(NS Stick-active) - Radial J48 MK2 DI(pad in) - MIC IN(+48V ON) - out to line in RME FF400(unity gain)....
81 Input Gain on 6 o'clock, output gain on 8 o'clock and it clips and distorts with or w/o EQ!! My mistake? Impedance mismatch?? or an 81 issue???
When set not to distort (lower In and Out gain)....sounds "nice" as it can be for now at that level...EQ is usable on this one with some minor problems...
...same connection but through Focusrite Red Pre...all works good, nice level

Remember that oscillation will reduce headroom, although there may be further issues too.