PEV-style "passive" EQ experiments

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

etheory

Well-known member
Joined
Mar 21, 2011
Messages
604
Location
Sydney, Australia
Hi there!

I've had had an inkling for some time now to try and simulate the Neumann PEV amplifier circuit in LTSpice, mostly because A.) I'd really love to build one and B.) I wanted to verify that with currently obtainable parts, it should be possible to do one seriously sweet clone.

Well, it kinda looks like it's working.  It's certainly amplifying, and seems to do so relatively cleanly ( until you get up to like 1V RMS on the input, at which point it starts to clip, but it looks like quite a soft clipping ).  There are more details to figure out for sure, but has anyone else ever built one of these things?

If anyone has any solid information on the iron I'd be most obliged.  Especially modern replacements for the input, inter-stage and output transformers.

The actual amp itself seems to be a relatively straight forward symmetrical affair - almost Neve'ish in implementation.  I am going to mess with a few transistor replacements and see whether I can match the same performance, but it looks like a viable build ( at least to SOME degree of keeping the original vibe ).

Does anyone else have any experience they could share with me about these - especially some actual measurements of the transformers?  For the "transformers" in the sim, even just whacking a series resistance of 3ohms and 3H inductance per mutual winding, the sim chugs along OK.  Obviously with more accurate data I could do a better simulation.

BTW, here it is....  The Green trace is a test 100Hz input at +-100mV, and the blue is the amplified output.  Source and Load are both set in this simulation to 600ohms.  The "wobble" at the start is expected ( most equipment does this at power on anyway due to charging caps ) and goes away after a second of the simulation.

Neumann_PEV-style_Sim.jpg
 
The gain of Q1,Q2 and Q3 together looks to be about 20dB so the emitter of Q3 will have about 10V peak. Even if the emitter is at exactly half the 12V supply it is pretty close to clipping. Try a dc sim to find the quiescent point of Q3 emitter.

Cheers

Ian
 
gyraf said:
Looks like you're running the sim at +/-24V - PEV runs on +24 only

Hey Gyraf,

That's just me being sloppy.  For pretty much any audio work I do in LTSpice I have a default setup, which always includes a dual rail PSU.  In this particular case it is only using the +24V rail, even though there are +-24V rails.
 
ruffrecords said:
The gain of Q1,Q2 and Q3 together looks to be about 20dB so the emitter of Q3 will have about 10V peak. Even if the emitter is at exactly half the 12V supply it is pretty close to clipping. Try a dc sim to find the quiescent point of Q3 emitter.

Cheers

Ian

Hi Ian,

It's running at 24V, so 10V seems pretty reasonable ( kinda close to 12V which is half the supply ), although I'll definitely try and investigate the relationship between Q1, Q2, Q3 a little more to understand why the output insists on clipping so "early".  I might be able to extract some more headroom out of it, although it's likely that this is mostly induced by the incorrect transformer data I am using.  The original transformer specs would clear that up a bit also.  Actually I might experiment with those amplifier stages without transformers and see if I can work backwards to get reasonable transformer values instead....
 
leadbreath said:
cant really help much except to say good luck, this is a nice eq especially like the presence band.

Thanks leadbreath - it's kind of like my current "white whale" project.
It's always in the back of my mind, even though I have enough to do at the moment with my A*P*I compressor project and Calrec 1061 stuff....
But yeah, even though I haven't heard one in real life ( unfortunately ), like I've read a few times on this forum before about pieces of "mythical" equipment, no-one seems to have one bad word to say about it ( nothing but praise ), so it just seems far too cool not to investigate further and find out why it's so "special" ;-)
 
Does anyone out there have any info at all on the iron in the Neumann PEV?
All I can find are the model numbers:

Input stage: bv-166
Inter stage: bv-196
Output stage: bv-167

If anyone knows potential modern replacements I'd LOVE to know....

Also note that the PV46 line amp is practically identical for all intents and purposes to the line amp in PEV, bar slight circuit modifications for the transformer compensation on the input ( due to a different transformer ).

I'd also say based on the usage in the PEV compared to the PV46, it definitely looks like the bv-167 and bv-267 are identical, when the bv-267 has it's secondaries connected in parallel - at least based on inspection of the circuit, so a modern replacement for either of those two would do.

Anyone?

Also I'm trying to analyze the circuit a bit further ( please assume that I'm an idiot whilst reading this, since that's probably the safest assumption.  I am REALLY good at taking other people's stuff and putting it together even with a certain level of re-engineering and also some basic analysis, but I'm certainly no qualified electrical engineer so anyone with deep knowledge feel free to correct or educate me at any point, I REALLY want to learn more and more.... ), and, whilst I learn extremely fast, this isn't the most obvious circuit I've ever come across to decipher, so here goes:

Trs1 seems to be a BF110/BSX46/10 NPN transistor high-impedance buffer, going to Trs3 which seems to be the main gain section from what I can tell.  From there Trs5 seems to be another buffer to drive C12 which in turn drives a unity gain darlington output buffer of Trs7 and Trs9.  The feedback paths to my limited knowledge are "weird" and I'm still analysing them.

What I can tell though is in my sim I can't get more than 1V RMS on the interstage transformer secondary without asymmetrical clipping occurring on the output.  Since my spice models for the transistors are seemingly sound, one culprit could be using "C" transistors - i.e. BC109 is specified on the circuit diagram from Neumann, but I'm using BC109C in my sim.  What's the usual rule for that, if the letter is not specified should I assume A, B or C?.

The culprit in my sim seems to be Trs3 ( what I assume is the main gain stage ), which does some "weird shizzle" at higher base voltages and clips in a funky way.  I assume that it's quiescent point is wrong either for the transistor model I'm using, or, the real unit itself doesn't have the worlds best headroom, which I don't know cause I don't have one sitting in front of me.

I'm still investigating how to calculate the quiescent points for all the transistors in the circuit.  I wanted to do this by isolating the stages into their own spice files for individual testing, however the preceding and following stages present loads that I must introduce into the isolated stages to test them correctly which is currently stumping me but I'm getting there....

Is 2N2102 or maybe NTE175 for a BF110/BSX46/10 a good substitution to make?  I've had a look around and it's the best I can do, but I don't have a mental database of transistor parameters yet ( but I'm getting there! ) so someone else no doubt would know the best substitution to make with "modern" parts.

I'd like to stick as closely as possible to the original parts, and can get the BC109 transistors no issue in Australia, but I can't get the BF110/BSX46/10's anywhere here ( I can get the BC177A's however or 2N2102 or NTE175 no issue ).

THANKS!
 
etheory said:
I'm still investigating how to calculate the quiescent points for all the transistors in the circuit.  I wanted to do this by isolating the stages into their own spice files for individual testing, however the preceding and following stages present loads that I must introduce into the isolated stages to test them correctly which is currently stumping me but I'm getting there....
THANKS!

Yoi just need to do a .dc sim in spice which will work out the dc conditions of all the nodes in the circuit.

Cheers

Ian
 
etheory said:
Trs1 seems to be a BF110/BSX46/10 NPN transistor high-impedance buffer, going to Trs3 which seems to be the main gain section from what I can tell.  From there Trs5 seems to be another buffer to drive C12 which in turn drives a unity gain darlington output buffer of Trs7 and Trs9.  The feedback paths to my limited knowledge are "weird" and I'm still analysing them.

There are two feedback loops around Q1,2,3. The fist is a dc only loop from Q2 emitter to Q1 base via R4/R1 pot divider and the transformer secondary. it is decoupled for ac by C1.

The second, ac loop that sets the gain of Q1,2,3 is from Q3 emitter to Q1 emitter via C6 and R9. At mid band frequencies, C6 and C4 can be assumed to be ac short circuits. The ac gain is therefore determined by R9 and R9 and is approximately (R9+R8)/R8 = 9.94 or about 20dB.

Cheers

Ian
 
I made some progress on this finally.
I think I can use the Lundahl LL7903 on the output, and the Edcor XSM10K/10K or 600/600 at the inter-stage....
Simulating with the inductance/resistance values for each gives markedly better simulation results now (i.e. it's effectively a perfect amp now), so, there is a chance that this might get built one day....
 
etheory said:
I made some progress on this finally.
I think I can use the Lundahl LL7903 on the output,
This won't do. The output stage here relies on NFB applied to the emitters of the output transistors. The windings going in the emitters are simultaneously primaries and secondaries.
They are primaries because they receive juice from the emitters and they are secondaries because they reflect the signal coming from the collectors to the emitters, which creates NFB for the base signal. This type of stage is characterised by low THD, extended BW.
A trademark of great McIntosh power amps.
The gain of this type of circuit is essentially unity.
But you need 5 windings minimum to make it work...
In order to achieve maximum power transfer, the collector primary must have the same number of turns as the emitter primary. Then the output level will be determined by the turn ratio between secondary and primary.
The voltage at the emitter is the same as the base voltage, so with all inductances identical, you have unity gain. It may not be the best choice, because the intermediate stage has not enough juice to fully drive the output stage.
Early clipping is due to the emitter voltage being too low; if it's really 10V, it means that the collector of Q2 is at 10.7V, which in turn computes at 1.75mA, then the emitter voltage is 8.8v. that means that it is very close to pinching, with only about 2V Vce. That suggests you can't expect more than 1.4Vrms before clipping. In any case, if the values are right, but with a better operating point, this stage is probably clipping at about 3Vrms. Which suggests that the output stage should have some gain. Gain can only be achiebved by reducing the number of turns of the emitter windings. 0.5:1 => 6dB, 0.33:1 => 10dB, 0.1:1 => 20dB...
I suggest you use k=0.99. It shouldn't change much, but will allow seeing the effects of leakage.
 
Thanks for your insight - it's a great analysis and holds up extremely well, at least with the simulation, so yeah, that gives me much more to investigate.

abbey road d enfer said:
This won't do. The output stage here relies on NFB applied to the emitters of the output transistors. The windings going in the emitters are simultaneously primaries and secondaries.

The LL7903 was suggested on another forum as a possible BV-167 (or is it BV-157, it's hard to tell due to the photocopy schematic I'm using from the Gyraf website) replacement.
I can't find the link at this second, but allegedly the suggestion came from Lundahl after a discussion on the specs, which included measuring the original device properties.


abbey road d enfer said:
But you need 5 windings minimum to make it work...

The LL7903 has 8. It's a 1+1+1+1:2+2+2+2 transformer, so it still seems fine?

abbey road d enfer said:
That suggests you can't expect more than 1.4Vrms before clipping.

I was actually quite worried that the simulator showed this as being the case, but it definitely seems to be.
Due to the losses of the passive EQ front end, maybe it's all intentional, as it's dealing with lower signal levels.
One could only imagine the lower voltage handling topology is for noise reasons with a smaller signal, or something similar, but I can't really tell why they would choose to have such "low" power handling, or, more importantly, what seems to be lower headroom. This is something I want to investigate further, but if anyone has access to the original, I'd love to be able to verify some of these things.

abbey road d enfer said:
In any case, if the values are right, but with a better operating point, this stage is probably clipping at about 3Vrms.

Yep, almost exactly 3Vrms.

abbey road d enfer said:
Which suggests that the output stage should have some gain. Gain can only be achiebved by reducing the number of turns of the emitter windings.

With the 2+2+2+2 windings at the secondary side, there are a few options for different gain possibilities through differing parallel/serial configurations that could solve this.

abbey road d enfer said:
I suggest you use k=0.99. It shouldn't change much, but will allow seeing the effects of leakage.

Another great insight to try.

Thanks so much for such a detailed post, it's greatly appreciated for someone like me who is still learning so much!
 
etheory said:
abbey road d enfer said:
But you need 5 windings minimum to make it work...
The LL7903 has 8. It's a 1+1+1+1:2+2+2+2 transformer, so it still seems fine?
That's right. The way it is presented on the datasheet is confusing. Only one half is shown on the diagram. My mistake.
With the 2+2+2+2 windings at the secondary side, there are a few options for different gain possibilities through differing parallel/serial configurations that could solve this.
Agreed.

 
etheory said:
abbey road d enfer said:
That suggests you can't expect more than 1.4Vrms before clipping.

I was actually quite worried that the simulator showed this as being the case, but it definitely seems to be.
Due to the losses of the passive EQ front end, maybe it's all intentional, as it's dealing with lower signal levels.
Yes; reduced signal handling capacity may be acceptable there.
One could only imagine the lower voltage handling topology is for noise reasons with a smaller signal,
I don't think noise would be improved in a significant way.
or something similar, but I can't really tell why they would choose to have such "low" power handling, or, more importantly, what seems to be lower headroom.
The only advantage I see there is that the operating point is very stable, due to the high DC feedback. Neumann may have considered that, as long as it did not impair performance, it was a significant industrial advantage (no need for trimmers, selected transistors or precision resistors). I know that the period Neumann company would never have accepted compromising performance.
This is something I want to investigate further, but if anyone has access to the original, I'd love to be able to verify some of these things.
Do you have the schematic dgm of the passive EQ itself? that would give an indication of the expected signal levels there.
 
Another possible few output transformers could be the:

Lundahl

LL1660 - 1+1+1+1 : 2.5+2.5

or

LL1671 - 1+1+1+1 : 2+2

I can't find other manufacturers with 1+1+1+1 on the primaries.
 
Does anyone have actual Bv transformer info anywhere?
I swear I've scoured the entire Internet multiple times and never managed to find any concrete data on them.
 
BV is short for Bau-Vorschrift - building regulation or Recipe. Was probably part of the mythic BraunBuch. Idea was standardisation, interchangability, and thereby quality assurance of parts for German radio use. For transformers, it specifies EXACTLY how you build it, including what exact materials to use.

Never found the info though - but I once talked to some people at Technical Museum Berlin that mentioned that they had (in a warehouse somewhere) the complete AEG technical documentation that they inherited some years ago. I didn't have the time to further investigate - maybe it's a job for our (GroupDiy) Berlin branch?

Jakob E.
 

Latest posts

Back
Top