Dbx Vca Construction

[RuudNL]

To preselect the transistors for Hfe, I used the following setup:
Emitter to ground, collector to the supply voltage (+ for NPN,  - for PNP) via a 1 K.ohms resistor (I used 10 Volts), base to the supply voltage via a variable resistor of 1M.ohms. Then I adjusted the variable resistor for 5 V. on the collector.
I used a DIL IC socket to make it easy to exchange transistors.
Now, if the Hfe is higher than the reference transistor, the collector voltage will be lower. On the opposite: if the Hfe is lower than the reference transistor, the collector voltage will be higher. This proved to be an easy way to select transistors, although it is time consuming! (Every time you have to wait until the temperature is stabilized after touching the transistor.) With a digital voltmeter you can select them with millivolt accuracy. I tested 50 PNP + 50 NPN transistors and the result was two sets of 'ideal' matched transistors and one set 'almost ideal' matched transistors. All transistors came from the same batch!

 

[sahib]

Thanks! If I am off to fight the dragon I am going to need an ultra big heat-sink!!
Also, Silent:arts It might not be worth the time to make a heat sink like the one pictured. it might be worth noting that those bc109 transistors aren't being made anymore and the only replacements i could find are not in the same rounded can design, (the TO-19 I believe?) so I was just thinking of making the flat parts of the 2 pairs of replacement transistors (bc550c) face each other and then glue and wrap some sorts of metal ring around them all. But that was before people brought up the transistor array idea. Now I think a matched transistor array is the only way to go...

Yes I am his twin but my name is Sahib.

Making a coupling for BC550 would not be difficult either but the transistor array of course is the way to go

 

[abechap024]

Yes I am his twin but my name is Sahib.

Ahhh Silly me, I just saw the logo and assumed it was silent:arts. Thanks for the calrification and glad to meet you! I too agree the matched THAT340 transistor array is the way to go. Though further inspection might be needed to see if the array can actually substitute the bc109s in the circuits.

 

[abechap024]

That 340 array. Looks like the Hfe for these is only ~100

not good considering the bc109s in the original VCA had a very high gain. ~800...

thoughts?

http://www.seas.upenn.edu/~ese319/Lab_Handouts/THAT340.pdf


Edit: Also looking at integrated darlington arrays. were they piggy back 2 transistors to increase Hfe.
http://en.wikipedia.org/wiki/Darlington_transistor

also the Sziklai pair looks interesting? where they piggy back an NPN and a PNP together.
http://en.wikipedia.org/wiki/Sziklai_pair

These are also sold in matched transistor arrays. I do not know very much about these. maybe someone with more knowledge could chime in and help?

Would having the transistors piggyback possibly help with matching? as in the small differences would even themselves out? 

Also its says the darlingtons act like a single transistor just with more current gain. but more phase shift at high frequency and saturate earlier, but that shouldn't really be an issue, i think, because were not putting these in a high power/high frequenct situation.

also would it be possible to use the Sziklai method to replace the 2 pairs of NPN and PNP transistors, for example, replacing each of the 4 matched transistors with 4 Sziklai configurations (8 total transistors)

seems maybe interesting?

 

[JohnRoberts]

That 340 array. Looks like the Hfe for these is only ~100

not good considering the bc109s in the original VCA had a very high gain. ~800...

thoughts?

Yes, there is a distortion mechanism associated with the beta of the devices inside VCAs. Since these bases are not all tied to zero impedance, any base current working into the actual termination impedance will generate a voltage at each base that is a function of the signal current it is passing divided by beta. Since each pair of devices is only handling half of the waveform, beta or termination impedance differences will cause different signals at the different bases.

Clearly the higher the beta, and the smaller the resistance termination at these bases the smaller this potential signal. I am not smart enough to tell you off the top of my head if the benefits of the array out weigh the lower beta, but it is my guess it would.

Also there is no rule saying you can't change the design, Dropping the impedance of the networks around the transistors should help reduce the sensitivity to beta and beta matching (note you will probably want to crank in an offset to account for the nominal 75 vs 100 beta of array devices.) I suspect the original dbx design impedances was influenced by the opamps and electronics they had to work with. 

Edit: Also looking at integrated darlington arrays. were they piggy back 2 transistors to increase Hfe.
http://en.wikipedia.org/wiki/Darlington_transistor

also the Sziklai pair looks interesting? where they piggy back an NPN and a PNP together.
http://en.wikipedia.org/wiki/Sziklai_pair

These are also sold in matched transistor arrays. I do not know very much about these. maybe someone with more knowledge could chime in and help?

IIRC Paul Buff did some work with diode connected devices in series to reduce some errors. The darlington connection will deliver higher beta, and higher noise. I suspect dbx investigated this and rejected it but if you have time and curiosity burning a hole in you days, go for it.  Note: two junctions in series instead of one will have a different mV/dB gain law IIRC. 

Would having the transistors piggyback possibly help with matching? as in the small differences would even themselves out? 

yes and no... Vbe differences average, but beta differences multiply..

Also its says the darlingtons act like a single transistor just with more current gain. but more phase shift at high frequency and saturate earlier, but that shouldn't really be an issue, i think, because were not putting these in a high power/high frequenct situation.

A little more complicated than that, darlingtons don't act like a single transistor because they aren't...

also would it be possible to use the Sziklai method to replace the 2 pairs of NPN and PNP transistors, for example, replacing each of the 4 matched transistors with 4 Sziklai configurations (8 total transistors)

seems maybe interesting?

Not to me, but go for it.. What's the worst thing that could happen..? Nominally the PNP beta times the NPN bets should be the same independent of their order they are used in, so at least on paper that is a good thing.. I think there are probably other issues that would dominate your results.

I usually answer such questions by building one up... But I haven't rolled my own VCA since the 80s.

JR

 

[abechap024]

Thank you so much for the information JR.

gives me a jump start on all of this! I am going to start laying the VCA out, and then hopefully answer some of my other questions and probably find me a bunch more.
Cheers!
Abe

 

[damnyankee]

I don't want this to sound like I am actually encouraging this pursuit, but here is one way to match the devices for beta and Vbe at the same time.

It requires a bunch of precision resistors, + and - power supplies, and a pile of transistors.

First connect 10 or 20 transistors together at their emitters. So this voltage will be the same by definition. Connect each individual base to ground through a precision resistor (20-100k). Then connect each collector to the appropriate supply (negative for PNP, positive rail for NPN) through individual precision resistors (say 2k). Then connect the common emitter through a low value resistor to the other supply as appropriate (plus for PNP, minus for NPN). Size this common emitter resistor, and voltage it is connected to for the test current you want times the number of devices. So if you want 1 mA in each of 20 transistors you will want 20 mA total supplied at this common emitter node.

With all the emitters tied together, they will share the current between them based on the individual differences in Vbe. Further the resistor in series with the base will give you a direct reading of beta and further influence the current sharing. With a number of transistors connected, a quick reading of the collector voltages (at the 2k resistors) will indicate best Vbe and beta matching within the group. After identifying the best collector current matches, confirm that their base resistors read the same too. Weed out any where both collectors and base voltages don't match. You can confirm matching at several different currents by varying the total current sent to all transistors.

Depending on how many transistors you have and can test at once you can either pre screen all for closest rough matches first or just weed out and replace the worst ones from the 10 or 20 being tested concurrently as you converge on the best matches within your group. It might be nice to use one of those plastic plug in prototype strips to hold a bunch of devices at once without soldering to the leads which will heat up the part.

You might wait until after getting a promising group of matches before heat sinking them together for final best match, to save time. The bad outliers are not likely to get better after being heat sunk, but you want the best matches that stay matched when held to the same temperature.   

JR

PS: The VOM diode tester may be an easy way to prescreen devices but i wouldn't trust it for final matching.

John,

Please remember this is an education for some of us - and certainly for me.  I've learned quite a bit from you regarding VCAs...so yes, please encourage us.  If nothing else, we've learned more about these old VCAs and how to hfe and Vbe match transistors.  I can say it's pretty amazing how Blackmer used common Radio Shack components yet ingeniously combined them in a very unique way and created something ground breaking in his day.

Thank you for offering your expertise!

DY

 

[abechap024]

Yes, this board forum is very fantastic. and of course when I say that it is because the people on this forum. I'm paying XXXX amount of money to go to school to learn more about EE, and even still its not like the direct kind of Q and A that can be had on these forums. Its just great that there are places like this were knowledge and learning are free flowing. Thank you for sharing what you've spent your life collecting!
AC

 

[damnyankee]

Not to me, but go for it.. What's the worst thing that could happen..? Nominally the PNP beta times the NPN bets should be the same independent of their order they are used in, so at least on paper that is a good thing.. I think there are probably other issues that would dominate your results.

I usually answer such questions by building one up... But I haven't rolled my own VCA since the 80s.

JR

John,

I need your expertise here.  Here's the 200 VCA out of a 160VU: http://www.cmaudio.net/forum/dbx%20model200%20vca%20module.pdf

I pulled the 200 VCA out of my dbx 119.  It identical in every way to the 200 VCA out of the 160VU EXCEPT:

         160VU     dbx 119
R1          13k      *not stuffed*
R2          17k      5.6k
R10       100k      27k
Q4     BC263C     BC479
Q5     BC263C     BC479

Now, the hfe for the BC263C is rated 450 to 900; the BC479 is only 200 to 500.

Can I still use the 200 VCA out of my dbx 119 for this clone project?  Should I make the resistors identical to the 200 VCA out of the 160VU?  Will the BC479 matched transistors work?

As always, I'm always grateful for your expertise!

Thank you in advance,

DY

 

[abechap024]

Hi DY,
I am not trying to answer for john by any means, but I just wanted to mention that the resistors that are different are the "select" resistors. As in they were selected after the transistors to trim out any distortion and artifacts.

Maybe someone could correct me, but I would be suprized if the 200VCAs were not all interchangeable with one another.
The fact that they used lower Hfe could be :

Cost cutting measure? Like you were saying, maybe they put the VCAs that didn't quite meet spec in the 119s (were you got your VCA from)

Or they had to replace the BC109 for some other reason.

But since the THAT  arrays are lower Hfe I think it is great that you found out your VCA has lower Hfe transistors and still worked fine in your 119!


Cheers
Abe

 

[damnyankee]

Hi DY,
I am not trying to answer for john by any means, but I just wanted to mention that the resistors that are different are the "select" resistors. As in they were selected after the transistors to trim out any distortion and artifacts.

Maybe someone could correct me, but I would be suprized if the 200VCAs were not all interchangeable with one another.
The fact that they used lower Hfe could be :

Cost cutting measure? Like you were saying, maybe they put the VCAs that didn't quite meet spec in the 119s (were you got your VCA from)

Or they had to replace the BC109 for some other reason.

But since the THAT  arrays are lower Hfe I think it is great that you found out your VCA has lower Hfe transistors and still worked fine in your 119!


Cheers
Abe

Thanks, AC!

I have to tell you, those dbx 119's are not the hype as some owners claim they are.  They are NOISY little basterds.  And the more compression that's applied, the noisier they get AND the output volume drops. 

(Yeah yeah yeah, I know Tom Scholz used the dbx 119 with an EV RE-15 mic when recording More Than A Feeling, but I'll bet he modified his 119.)

 

[JohnRoberts]

Not to me, but go for it.. What's the worst thing that could happen..? Nominally the PNP beta times the NPN bets should be the same independent of their order they are used in, so at least on paper that is a good thing.. I think there are probably other issues that would dominate your results.

I usually answer such questions by building one up... But I haven't rolled my own VCA since the 80s.

JR

John,

I need your expertise here.  Here's the 200 VCA out of a 160VU: http://www.cmaudio.net/forum/dbx%20model200%20vca%20module.pdf

I pulled the 200 VCA out of my dbx 119.  It identical in every way to the 200 VCA out of the 160VU EXCEPT:

         160VU     dbx 119
R1          13k      *not stuffed*
R2          17k      5.6k
R10       100k      27k
Q4     BC263C     BC479
Q5     BC263C     BC479

Now, the hfe for the BC263C is rated 450 to 900; the BC479 is only 200 to 500.

Can I still use the 200 VCA out of my dbx 119 for this clone project?  Should I make the resistors identical to the 200 VCA out of the 160VU?  Will the BC479 matched transistors work?

As always, I'm always grateful for your expertise!

Thank you in advance,

DY

I'm not sure what you're asking?  Resistors that are labelled as optional/select means they build up the module with the best matched parts they can find, then measure the results with a distortion analyzer and probably a scope, while they use a resistor substitution box, or perhaps trimpots to tweak the actual parts for good results. This is from a time in manufacturing when labor was relatively cheap.

I have always used a decent opamp in VCAs I've made, in place of their two transistor + JFET amp. Their design may predate the relatively inexpensive TL07x fet input opamps.  The S.A.T. (select at test) resistors in that simple amp (R1 and R10) are selected so at 0V Vgs they provide the full Idss (current draw of JFET at 0 Vgs) with the base of Q1 at nominally 2x diode drops or roughly 1V. It seems to me that if the JFET Idss varies a bunch, the SAT resistor to dial it in will change the open loop gain of that crude gain stage. So it is probably over compensated to deal with production variations. Also it will be temperature sensitive so keeping that dialed in over temperature extremes may be difficult. Also stability of 7.5V rail can chage this operating point so that voltage needs to be constant (preferably VCA dialed in with actual supplies in place).     

The beta is not a game stopper. I suspect the devices in the IC VCA are similar beta to the THAT arrays. It is useful that the beta be matched, but those early VCAs had different impedance terminations at the different bases, so even perfect beta tracking will still result in some errors.

Not to be any more discouraging but you need to have modest expectations for this exercise. As I've been saying all along, even done as well as dbx could do it, these are not the cleanest VCAs.

JR

 

 

[damnyankee]

Thank you for your continued patience, John.  Really, my questions are solely about the original dbx 200 VCA can (I'm learning).  I understand these VCAs suck camelwang and I understand there are better alternatives (THAT).  The original 200 VCA cans are found in the dbx 117, 119, 154, and 160VU.  

Web rumor is the 200 VCAs placed in the dbx 117 & 119 companders and 154 noise reduction units are supposedly "inferior" to the 200 VCAs in the dbx 160VU.  And, the "inferior" 200 VCA cans don't have all the transistors of the 160VU 200 VCA cans.  Well, the transistor rumor is not true...but I noted my dbx 119 uses matched BC479 and the dbx 160VU uses matched BC263C...and I've already noted the resistor differences between the two 200 VCA cans.  

What I'm asking is, since the matched transistors differ, can I assume that the resistors would also differ?  And can I assume the overall function between the two 200 VCAs would be the same?

Thank you in advance,

DY

 

[Biasrocks]

Web rumor is the 200 VCAs placed in the dbx 117 & 119 companders and 154 noise reduction units are supposedly "inferior" to the 200 VCAs in the dbx 160VU.  And, the "inferior" 200 VCA cans don't have all the transistors of the 160VU 200 VCA cans.  Well, the transistor rumor is not true...but I noted my dbx 119 uses matched BC479 and the dbx 160VU uses matched BC263C...and I've already noted the resistor differences between the two 200 VCA cans.  

I highly doubt that the 200 VCA's and RMS units are inferior in any way, what is definitely inferior in those units is the surrounding circuitry and features as they were not aimed at the pro market.

I suspect the example that your holding may simply be one of the later/early revisions of the VCA. DBX probably used whatever transistors they could get a large supply of at the time and as John says, the SOT resistors will be different, that's normal.

Mark

 

[damnyankee]

Mark,

What's interesting is, my 119's 208 RMS is identical to the 160VU's 209 RMS.  You may be right about the 200 VCAs: revisions occurred due to whatever resistors were available for matching at the time and the resistors used were dependent upon transistor types.

DY

 

[emrr]

Nov. 1972 ad:

202: linear dB gain versus control voltage $47
203: same, but lesser performance specs on noise and distortion $39
204: similar, but voltage gain proportional to control current, same gain and specs as 202  $60

 

[damnyankee]

Hi Doug,

Question about those VCAs: weren't those manufactured exclusively for mixing consoles?  I ask because the earliest year I've found for dbx 117/119 was 1973.  And if memory serves, the dbx 160VU didn't roll out until '76.  As far as I know, these dbx models used dbx 200 VCAs, not 202, 203, or 204.

 

[emrr]

The ads doesn't say what the intended usage is.  What I quoted was all the ad said.  It does speak a bit to the question of sorted quality. 

Nearby, they also had several NR products listed at astronomical prices. 

 

[damnyankee]

The ads doesn't say what the intended usage is.   What I quoted was all the ad said.   It does speak a bit to the question of sorted quality. 

Nearby, they also had several NR products listed at astronomical prices. 

Yes, I'm sure they look astronomical in price but remember that gear was cutting edge technology in those days AND they were made in USA.  I am mindful when the first IBM 486x computer came out with a 480 mb harddrive and 15" monitor: I paid over $3k for that PC.  Looks like a ripoff today...but not in those days.

 

[JohnRoberts]

Thank you for your continued patience, John.  Really, my questions are solely about the original dbx 200 VCA can (I'm learning).  I understand these VCAs suck camelwang and I understand there are better alternatives (THAT).  The original 200 VCA cans are found in the dbx 117, 119, 154, and 160VU.  

Web rumor is the 200 VCAs placed in the dbx 117 & 119 companders and 154 noise reduction units are supposedly "inferior" to the 200 VCAs in the dbx 160VU.  And, the "inferior" 200 VCA cans don't have all the transistors of the 160VU 200 VCA cans.  Well, the transistor rumor is not true...but I noted my dbx 119 uses matched BC479 and the dbx 160VU uses matched BC263C...and I've already noted the resistor differences between the two 200 VCA cans.  

I don't have the sundry models numbers memorized nor do I care.

i don't know for a fact that dbx used lesser VCAs in their NR but in general the VCAs used in companding noise reduction circuits don't need to be as good,  so they could have. The DBX NR used symmetrical 2:1/1:2 compression/expansion with 12 dB of HF pre/de-emphasis. So a few factors contribute the relaxed demands. First the VCA is only open and operating at full unity gain when the signal is present at full volume, so that signal would mask unity gain noise. When the signal is quiet, the VCA would also be operating as an attenuator, reducing it's own noise contribution too. Further the HF pre/de-emphasis, has the effect of rolling off the high frequencies in the VCA noise floor by 12 dB. This also reduces low-mid frequency harmonic distortion while that is probably a less import benefit than the noise floor improvement. 

What I'm asking is, since the matched transistors differ, can I assume that the resistors would also differ?  And can I assume the overall function between the two 200 VCAs would be the same?

Thank you in advance,

DY

The selected resistors will always be different, even when same part number transistors are used.

It would really be useful to find a copy of the factory procedure for setting up those VCAs.

I realize I explained what you need to accomplish when dialing in the JFET gain stage, but not exactly how. When there is no resistor at all in either R1 or R10 the top circuit is not turned on, and the VCA will be saturated low by the negative pull down resistor R8. The gate of the JFET will be pulled down to roughly -0.5V.  As you connect a resistor to R1 turning on the top stage, you will see the gate voltage of the JFET Q3 rise up toward 0V as you lower the value of the resistor. The concept is to get it as close to 0V with the first resistor, while still just below 0V, then use a much larger value second resistor at R10 to fine dial it in closer to 0V. In your example the 13k in parallel with 100k makes the same current as a single 11.5k resistor which is not a common value.

Note: I don't know if you can probe the gate of the JFET directly as that is also the input to the VCA and might make it oscillate. You should be able use a couple K resistor between the meter and the VCA input for the purpose of measuring the DC voltage at the JFET gate.

JR

PS I would expect all dbx VCAs with the same part number to be interchangeable.