Intersound IVP Preamp (Resistor Value)

[Rob Strand]

Maybe someone can help me here, or has some photos of the PCB around the tube-sound transformer.

I was just on this thread,
http://www.diystompboxes.com/smfforum/index.php?topic=119131.msg1110520;topicseen#msg1110520

And I realized the emitter resistor on the current source, R92, is 43ohms.  The schematic I made *ages* ago had 430ohms. I could have made a mistake - fair enough.

However, when dide some checks around the current source (Q22), the 430 ohms makes more sense:

Assume the transistor gain is 400.

Emitter         Tail            Base
Resistor        Current    Current
43 ohms:     12.5mA      31uA   
430 ohms:    3.3mA      8.3uA

The current in the divider feeding the base of Q22 is approximately
(30V-2V) / 430k  = 65uA

Given this is a professional device and designed well we wouldn't expect them to design a circuit with with a divider current of 65uA when the base current is 31uA.   To avoid the tail current from being dependent on Q22's gain, we would normally set the divider current to be say 5 to 10 times the base current.  The 430 ohm looks more reasonable from this point of view.

A side comment, the DC offset at the output is much larger with 43 ohms.   This is mostly due to the fact the base current in the diff-pair is higher and the base circuit is assymmetrical.   Sure this might be deliberate but it doesn't get around the bad design practice evident with the 43 ohm.

From a signal swing point of view both look about the same.

So it is an interesting dilemma.

[PRR]

> bad design practice

Sometimes "good distortion tone" is ABOUT bad design practice.

It would help to know the intended output level. In hi-fi more is better; in distortion less may be more.

Sometimes commercial practice is about using-up the overstock of 43 Ohm resistors the last purchasing agent left you.

[realkillercults]

glad I could help to inspire this inquire
didn't know people gave a piss 'bout this old box!

[Rob Strand]

Sometimes "good distortion tone" is ABOUT bad design practice.

It would help to know the intended output level. In hi-fi more is better; in distortion less may be more.

Sometimes commercial practice is about using-up the overstock of 43 Ohm resistors the last purchasing agent left you.

Sure.  I suppose my beef/dilemma is if they wanted the tail current the 43 ohms gives for sound purposes then why not make the tail current more deterministic by putting more current down the divider feeding the base.   It's normal practice.   As it is, it depends on the transistor gain.

[Rob Strand]

glad I could help to inspire this inquire
didn't know people gave a piss 'bout this old box!

Those things were pretty good.   I've seen pro's using them here.

The tube sound was pretty good for bass with 2x15" EV speakers.  Not sure if it would be overly buzzy with tweeters people use these days.

Maybe this guy, around 1983 to 1987, (could be wrong)
https://en.wikipedia.org/wiki/Richard_Grossman_(bassist)

[antonis]

I suppose my beef/dilemma is if they wanted the tail current the 43 ohms gives for sound purposes then why not make the tail current more deterministic by putting more current down the divider feeding the base.

Maybe their resistor purchacing agent overstock them with also 430k & 36k items... 

I totally agree with "bad design practice" of strongly depending on transistor's beta while using resistive voltage divider for bias..
(it's like introducing unfiltered PS after hum eliminating filter..)

[Rob Strand]

Maybe their resistor purchacing agent overstock them with also 430k & 36k items... 

It's a conspiracy!

[PRR]

It is an interesting blend of 2nd transitioning to 3rd harmonic bends.

Note that the input is not differential but in-phase at 2/3 level to the offside input.

Aside from production variability: the main difference 43 or 430 seems to be about 3X higher output level, which could be trimmed in some later gain stage, or to suit the saturation of the transformer.

As a 9V design you sure want the 430 or higher to keep from going into hard clipping.

[Rob Strand]

It is an interesting blend of 2nd transitioning to 3rd harmonic bends.

Note that the input is not differential but in-phase at 2/3 level to the offside input.

Aside from production variability: the main difference 43 or 430 seems to be about 3X higher output level, which could be trimmed in some later gain stage, or to suit the saturation of the transformer.

Yes, it a bit skewed isn't it!  Exactly *why* it sounded good is up for grabs.   I suspect in a lot of circuits the transformer filters out the junk and makes it sound more pleasant.    When I was playing around with this thing I used 2x15" EV's.   So it's not as if there's a lot of high-end coming off them.

The gain difference got me thinking as well.   Haven't checked it out yet.   If memory serves me right, because it must be getting on 30 years ago now,  the Tube sound  "channel" had noticeably more gain than the clean "channel".  Perhaps a factor of 2 but maybe not a factor of 4.

[Rob Strand]

Aside from production variability: the main difference 43 or 430 seems to be about 3X higher output level, which could be trimmed in some later gain stage, or to suit the saturation of the transformer.

I'm seeing  24.0dB gain for 430 ohm and 28.7dB  for 43 ohm.  That's taking the output differentially.  I also have a 500ohm in series with 3H to emulate the transformer.  So I get 1.7x gain difference not 3x.   It doesn't change that much without the transformer load.  I guess this is exactly the issue. The behaviour of the 43 ohm configuration depends on the transistors.  In this case it depends on the transistor models  (I'm using a Fairchild model).
[If there was no loading on the divider we would get a factor of x10.]

As expected, when I add the transformer load the collector DC voltages are forced to be closer.

For the *rest* of the circuit the tube sound  has about 1dB higher gain than the clean.   (Each path has two gain stages.  The tube sound stage a little more pre-distortion gain and a little less post-distortion gain.   The balance is 1dB up for the tube sound path.)

Well, that's stuffed me up.   The gain is so much higher than what I remembered.  I'm struggling to accept more than 20dB difference.

[teemuk]

Aside from production variability: the main difference 43 or 430 seems to be about 3X higher output level..

I agree, the most distict differences are gain and headroom of the output signal swing. With 430 ohms both gain and clipping headroom are lower. The resistor value doesn't seem to affect actual clipping characteristics that much; they are pretty much identical with both circuits. The other just has lower headroom and lower gain.

...which could be trimmed in some later gain stage, or to suit the saturation of the transformer.

The "Tube Voice" circuit is pretty much there as is. Things not shown in the schematic are "pre-gain" and "attenuator" stages, which show that "clean channel" has somewhat lower gain and somewhat less attenuation. Nevertheless, these differences are virtually minor and the gain difference of the circuit paths is still around 20 dB or more. So, no evident "balancing"  of the signal levels (aside separate gain controls) seems to be incorporated to the design.

Greater balance of clean and distorted signals would indeed seem sensible, and therefore the circuit with 430R tail resistor, with its lower gain and lower headroom. One obvious possibility is that the transforner ratio is something different than 1:1, in which case it could be employed to attenuate the output signal furher.

As for other transformer-induced effects... I'm pretty sure they did not want saturating effects from the transformer. First of all, they actually sound pretty horrible when occuring (which fortunately is rare) and I seriously doubt the circuit's ability to saturate the transformer at those very low voltages within usual frequency band restrictions of a guitar. We can also see clear intention to minimize DC voltage potential across the primary winding, as usual in "push-pull" circuits. I know "transformer saturation" is a nice cathphrase but overall such effect isn't usually introduced in typical designs for good reasons. Not even by accident.

I'm pretty sure they just wanted to incorporate the typical push-pull scheme, which combines the asymmetrical amplification and clipping of two amplifiers (receiving opposite phase signals) in balanced configuration. The scheme cancels even order harmonics and effectively it converts the two (fairly hard) asymmetrically clipped outputs into a single (moderately softly) symmetrically clipped waveform. Intersound designers did not want to introduce just another generic "diode clipping" circuit, instead they mimicked the circuit architecture of a genuine push-pull output stage.

Perhaps at the time when this circuit was developed a genuine audio transformer was an "easier" component choice to implement than using a differential amp for the same purpose. However, do note that these days when you see these kinds of circuits a transformer output is uncommon and you usually see a differential OpAmp handling its task instead. e.g.
- Marshall DBS 7200 (https://elektrotanya.com/PREVIEWS/63463243/23432455/marshall/marshall_dbs_200w_7200_72115_72410.pdf_1.png)
- Vox Valvereactor power amps
- Vox VBM1 (http://i284.photobucket.com/albums/ll33/bonosurf/vbm1_zps6cd81f13.jpg)
- Behringer hybrid effects
All employ conceptually identical "push-pull" distortion circuit but simply replace the transformer with a differential amp. Line 6, Peavey and Quilter have also introduced conceptually similar designs but have exploited the bi-polar nature of solid-state devices, in which case balancing is not even required for effective signal summation. No transformers or opamps needed.

In my experience, clipping characteristics do not effectively change in either of these schemes because, as I mentioned earlier, the transformer does not introduce any "special effects" of its own, such as core saturation. It's just a balancing device to extract differential and nothing more. It should be virtually "transparent" and if you ever do experience "core saturation" with them chances are you immediately want to turn the gain down to get rid of it.

I suspect in a lot of circuits the transformer filters out the junk and makes it sound more pleasant.   

I'm not sure I can agree. If its a halfway decent audio transformer with halfway decent impedance match the bandwidth will likely greatly exceed the effective bandwidth of a guitar. A mediocre audio transformer may have a bandwidth from 40 Hz to 10 kHz, however the guitar works in the range of only about 100 - 5 kHz! ...And even less if you want the ovedrive to sound differenct from farting fuzz with obnoxious overtones. One needs a very low frequency input to saturate the transformer but common guitar effecting procedures pretty much ensure such frequencies are filtered away from the signal. Speaking of filtration, there may be "filtration" effects and bandwidth reduction if the transformer is deliberately badly matched to surrounding circuit. Those effects to frequency response are what people usually mistakingly interpret as "transformer saturation".

[Rob Strand]

One obvious possibility is that the transforner ratio is something different than 1:1, in which case it could be employed to attenuate the output signal furher.

It crossed my mind.  The calculated gain of the Tube-sound circuit is so much higher than what I remember when I played the unit.    The other explanation for this is the real circuit has a 430ohm resistor and the limited signal swing makes it sound quieter than the gain implies.

'm not sure I can agree. If its a halfway decent audio transformer with halfway decent impedance match the bandwidth will likely greatly exceed the effective bandwidth of a guitar. A mediocre audio transformer may have a bandwidth from 40 Hz to 10 kHz, however the guitar works in the range of only about 100 - 5 kHz! .

Hard to know.  The load on the transformer is quite high impedance so you might be right here.  I guess it only takes a small amount of filtering to make it sound a bit smoother.   (The Vox VBM1 circuit seems to use a lot of post filtering!)

[PRR]

> I'm seeing  24.0dB gain for 430 ohm and 28.7dB  for 43 ohm.

I was looking at maximum output level, not gain. (Gain is cheap and easy, up OR down.) The ~~3X higher current leads to ~~3X higher maximum output level. In pure linear circuits we could trim that up/down as desired. But a transformer has a soft limit on maximum output level. I was wondering if the operating current might be trimmed to the specific iron which was on hand (which might have changed over the production run).

Just a thought: a 1K:8 OT was used in almost every transistor pocket radio and widely available as replacement part. The primary might phatten-up at the higher levels of this circuit, and the ~~10:1 step-down would deliver at a more guitar-friendly level, able to drive any load.

[Rob Strand]

I was looking at maximum output level, not gain. (Gain is cheap and easy, up OR down.) The ~~3X higher current leads to ~~3X higher maximum output level. In pure linear circuits we could trim that up/down as desired. But a transformer has a soft limit on maximum output level. I was wondering if the operating current might be trimmed to the specific iron which was on hand (which might have changed over the production run).

Based on the perceived gain from when I played the real thing I was thinking I could back engineer the operating current based on gain.

Interesting idea,  those transformers aren't gapped, so yes,  the DC operating current through the core might just be enough the saturate it.  The transformer is 10k:10k so it has heaps of turns making saturation easier.  That would definitely drop the apparent transfer vo/vi of the transformer.   Saturation can also increase leakage inductance which will increase HF roll-off.

Just a thought: a 1K:8 OT was used in almost every transistor pocket radio and widely available as replacement part. The primary might phatten-up at the higher levels of this circuit, and the ~~10:1 step-down would deliver at a more guitar-friendly level, able to drive any load.

Well given the part number and the physical size matches that 10k:10k transformer I was willing to give that spec the benefit of the doubt.     Nonetheless that unit was made a long time ago, perhaps before the numbers meant what they do today.

At this point I like your saturation theory.  I can try to estimate the number of turns and saturation point of that transformer based on the DC resistance and the core size.

[Rob Strand]

The estimated hinges on choosing the ur for iron to get  the guessed/estimated primary inductance.

I started with 300Hz -3dB at 10kohm to estimated 5.3H inductance.   (Someone else reported this as 3H but it looked like they misquoted the measurement of another transformer.)  That worked out as a ur=1650; fairly reasonable.

With those assumptions, at 17mA DC I get a DC B-field of 1.5T which is saturation.  We could back off from that to say 11mA and 1.0T.    I wouldn't put a lot of faith in the numbers but it's probably in the zone 0.5 to 2 times these estimates.

The diff-pair tail current is about 12mA with RE=43 ohms on the current source.

So we are in fact fairly close to the saturation point.

[Rob Strand]

I started with 300Hz -3dB at 10kohm to estimated 5.3H inductance.   (Someone else reported this as 3H but it looked like they misquoted the measurement of another transformer.) 

For the record,  it occurred to me the -3dB point for a 10kohm:10kohm  transformer should incorporate a 10k drive impedance *and a 10k load impedance*.  That means the estimate for the inductance using the -3dB frequency should be based on 5k ohms which brings the inductance down to 2.65H; more in line with the speculated 3H value.     I cross-checked this idea against inductance measurements people have done on other Xircon transformers and it appears to hold up.

Unfortunately that stuff-up my estimates.   

One problem is the mu value required to get 3H is quite low.  To compensate for that I've had to reduce my window utilization factor (ku) to quite a small value.  It is a small transformer after all.  Without knowing how sloppy the bobbin is on the core and how packed the winding is it's hard to come-up with a precise ku value.   

Anyway with ku= 0.2 and ur = 900, the inductance and resistance match the real unit, and now the current for 1T is now about 20mA.   Beyond that the ku and ur values start to get unrealistic.